Effect of Urban and Agricultural Land Use Land Cover Change on Mesoscale Thunderstorms and Heavy precipitation. Thesis. Submitted to the Faculty

Transcription

1 Effect of Urban and Agricultural Land Use Land Cover Change on Mesoscale Thunderstorms and Heavy precipitation Thesis Submitted to the Faculty of Purdue University by Ming Lei In Partial Fulfillment of the Requirements for the Degree of Master of Science May 2008 Purdue University West Lafayette, Indiana

2 ii BIOGRAPHY Ming was born and raised in Yi Chang, Hu Bei, China to Shuanggui and Zehua Lei on November 7, Since Hu Bei province (located in central China) and Yi Chang are along the Yangtze River in the foothills of the Wu Mountains, summer monsoon precipitation and floods are frequent occurrences there. During the 1998 El Niño period, he experienced the heaviest rain in his life, when it rained that year for almost the whole summer. The whole country suffered from that event. He decided then to study the causes of heavy rains and how they happen. After graduating from the top high school in Yi Chang, he continued his education at Nanjing University as a student in the academically acclaimed Department of Atmospheric Sciences. This program provided him with a solid foundation in atmospheric sciences, and he finished his bachelor s degree with the thesis topic: Analysis of east Asian monsoon precipitation variation over the last 150 years. He then continued on to graduate school at Nanjing University, and refined his research topic to the effects of land use on climate change. After two years graduate study, Ming decided to pursue a Ph.D. in the United States, and was accepted into the Earth and Atmospheric Sciences program at Purdue University working with Dr. Dev Niyogi. Throughout his graduate work he has greatly enjoyed the study of atmospheric sciences as well as making professional contacts and good friends. During graduate school at Nanjing University he met his wife, Yue, and they are now expecting their first child. Upon this completion of his MS degree, Ming will continue working his PhD focusing on impact of land use / land cover change on weather and climate with Dr. Niyogi at Purdue.

3 iii ACKNOWLEDGMENTS I would like to recognize and thank the graduate committee Dev Niyogi, Indrajeet Chaubey and Sharon Zhong. Also thanks to the Department of Earth and Atmospheric Sciences, the Indiana State Climate Office and Land Surface Lab members for their advice and support. This study has benefited from NSF-ATM (Dr. S. Nelson), NASA THP (Dr. J. Entin NNG05GQ47G and NNG06GH17G), NASA-IDS (NNG04GL61G Drs. J. Entin and G. Gutman), NASA LCLUC Program (Dr. G. Gutman). The study also benefited from the Purdue Asian Initiative Program.

4 iv TABLE OF CONTENTS Page LIST OF TABLES...vi LIST OF FIGURES...vii ABSTRACT...xii CHAPTER 1. Introduction Background Urbanization and urban heat island (UHI) Urban observations Urban modeling Objective and hypothesis Outline of the thesis... 6 CHAPTER 2. Effect of explicit urban land surface representation on the simulation of the 26 July 2005 heavy rain event over Mumbai, India Introduction Data and methodology Results Synoptic scenario Rainfall Vertical structure/sounding Impact of coupling the urban TEB within RAMS mesoscale model Temperatures and fluxes change over the urban region (UHI simulation) Winds and Convergence Rain and cloud water mixing ratio, and precipitation Analysis of the explicit SST and the urban land-atmosphere interactions... 21

5 v 2.6. Conclusions CHAPTER 3. The effect of the Indianapolis urban area on summer thunderstorm behavior Introduction Background Indianapolis urban area Precipitation climatology for Indiana Methodology Model configuration Model initializations Results Case study Effect of urban environment on thunderstorm evolution Temperature Surface fluxes Moisture and rainwater mixing ratio Precipitation Vertical velocity, convergence, and CAPE Sensitivity experiments Effect of Roughness change over the urban area Effect of Albedo Effect of urban expansion Conclusions CHAPTER 4. Conclusions, limitations and future work Summary and conclusions Limitations and future work plan LIST OF REFERENCES

6 vi LIST OF TABLES TABLE Page TABLE 2.1 MODEL CONFIGURATION AND SETUP TABLE 2.2 KEY INPUT VARIABLES FOR TEB TABLE 2.3 TEMPORAL RAINFALL VARIATIONS AT SANTACRUZ ON 26, 27 JULY TABLE 3.1 INDIANA CLIMATE NORMALS ( ) TABLE 3.2 KEY VARIABLES TO DESCRIBE INDIANAPOLIS URBAN MOPHOLOGY TABLE 3.3 ANALYSIS OF THUNDERSTORMS FROM MAY 2000 TO JULY TABLE 3.4 SUMMARY OF THUNDERSTORM CASES

7 vii LIST OF FIGURES FIGURE Page Figure 2.1 Simulation model domains and idealized urban region over Mumbai city (inner most domain with TEB). Dashed line in third domain indicates the cross-section (at 19.15N) that is analyzed in this study Figure 2.2 Comparison of SST initial conditions: a). Default / climatological SST ( C) between May and August from 1950 to 1980; b). TRMM satellite-observed weekly averaged SST ( C) July, Figure 2.3 Comparison of NCEP reanalysis and model simulated wind circulation (m s -1 ) Figure 2.4 NCEP reanalysis data for zonal divergence (10-5 s -1 ) and zonal winds (m s -1 ) between 72.5E and 73.5E Figure 2.5 Precipitation time series for over Santacruz Airport (19.08N, 72.83E): Circles = observed rain gauge rainfall (mm), solid line with open squares=teb Run (mm), green dashed line with dots=control run (mm) Figure 2.6 Spatial precipitation (mm) captured by TRMM, CMORPH, gauge data, and model Figure 2.7 Comparison of observed data and model at the Santacruz airport weather station. Solid line with open circles=observation; green dashed line with dots=teb; yellow dashed line with open circles= control Figure hpa air temperature changes due to urbanization (using TEB)... 33

8 viii Figure 2.9 Cross-section of a). Vertical air temperature (K) and b). virtual potential temperature (K) profiles over the Mumbai urban region Figure July sensible heat fluxes (W m -2 ) and differences in both simulations (TEB control) Figure 2.11 Same as Fig e). but for latent heat flux (W m -2 ) Figure 2.12 Cross-section of 12Z 26 July vertical wind and its change (m s -1 ) after using urban model Figure 2.13 Cross-section of 12Z 26 July u and w winds (m s -1 ), and their convergence (10-5 s -1 ) Figure 2.14 Cross-section of 12Z 26 July rainwater mixing ratio (g kg -1 ) at 19.15N from 71E to 75E Figure 2.15 Cross-section of 12Z 26 July cloud water mixing ratio (g kg -1 ) arrows show the location of updrafts and downdrafts Figure 2.16 Simulated convergences (10-5 s -1 ) at 9Z 26 July: (a).various model simulations: control, TRMM SST, urban model, TRMM SST and urban model. (b). Analysis where the panels in Fig.2.15a were used to study the effects of each model parameterization Positive values (solid contours) represent areas convergence, and negative values (dashed contours) represent areas of divergence. (c). The same as (b) but for cross section at 19.15N Figure 2.17 Factor separation analysis for accumulated precipitation (mm) Figure 2.18 Time series of accumulated precipitation (mm) over Santacruz airport: control, the effect of TRMM SST, the effect of urban model and the effect of their interactions Figure 3.1 Land use land cover map showing Indianapolis urban region and central Indiana Figure year normal average annual precipitation: a). from 1961 to 1990; b). From 1971 to Figure 3.3 Model simulation domains, TEB is adopted over the inner most domain

9 ix Figure 3.4 Sample radar base reflectivity images for 4 thunderstorm cases Figure 3.5 a). Surface weather map for 22 June 2006; b). surface weather map for 23 June 2006; c). 500 hpa map for 22 June Figure 3.6 Accumulated precipitation (inch) map over Indiana from 21 to 23 June Figure 3.7 CoCoRaHS recorded precipitation (inch) on 23 June Figure 3.8 Observed surface fluxes in Bondville,Illinois, AmeriFlux site: a). sensible heat flux (W m -2 ); b). latent heat flux on 22 June 2006 (W m -2 ) Figure 3.9 Observed and simulated vertical temperature ( C) Figure hpa air temperature (K) from North American Regional Reanalysis (NARR) data and model Figure 3.11 Horizontal surface air temperature (K) averaged from 22 to 23 June Figure 3.12 Vertical cross-section of air temperature (K) averaged from 22 to 23 June Figure 3.13 Diurnal surface air temperature (K) over Indianapolis from 0Z 21 June to 0Z 24 June (K) Figure 3.14 Average sensible heat flux (W m -2 ): LEAF2, TEB and their difference (TEB control) Figure 3.15 Diurnal sensible heat flux (W m -2 ) variation over Indianapolis (short dashed line= control; long dashed line= TEB run; solid line= TEB control) Figure 3.16 Same as figure 3.15 but for latent heat flux (W m -2 ) Figure 3.17 Longitudinal heat flux (W m -2 ) gradient averaged from 39.7N to 39.9N: a). Sensible heat flux gradient in first thunderstorm; b). Sensible heat flux gradient in second thunderstorm; c). latent heat flux gradient in first thunderstorm; d). latent heat flux gradient in second thunderstorm (open circle= control run, close circles= TEB run)

10 x Figure 3.18 Rain water mixing ratio (g kg -1 ) over the Indianapolis urban area from 0Z 21 June to 0Z 24 June. (Short dashed line= control; long dashed line= TEB run; solid line = TEB control) Figure 3.19 Same as figure 3.18 but for relative humidity (%) over Indianapolis (short dashed line= control; long dashed line = TEB run; solid line = TEB control) Figure 3.20 Precipitation time series (mm) throughout the simulation period: a). Over the Indianapolis area; b). upwind of Indianapolis; c). downwind of Indianapolis. (Short dashed line= control; long dashed line=teb run; solid line= TEB control) Figure 3.21 Spatial accumulated precipitation (mm) :a). First thunderstorm b). Second thunderstorm Figure 3.22 Vertical velocities (m s -1 ): a). First thunderstorm; b) Second thunderstorm Figure 3.23 Same as figure 3.22 but for convergence (10-5 s -1 ) Figure 3.24 Precipitation (mm) for low roughness length conditions and explicit urban representations: a). First thunderstorm; b). Second thunderstorm Figure 3.25 Vertical cross section for convergence (10-5 s -1 ) averaged from 39.7N to 39.9N for low and normal roughness length condition: a). First thunderstorm; b). Second thunderstorm Figure 3.26 Accumulated precipitation (mm) in the simulation with high albedo condition, low albedo condition and their difference Figure 3.27 Accumulated precipitation (mm) for the two thunderstorms before and after the synthetic expansion of Indianapolis urban area a). First thunderstorm; b). Second thunderstorm Figure 3.28 Precipitation timeseries (mm) averaged over a). Indianapolis urban area and b). southern part of Indianapolis (long dashed line with closed circles= current explicit urban representations; short dashed

11 xi line with open circles= expanded urban area; solid line with open circles= differences)... 98

12 xii ABSTRACT Lei, Ming. MS, Purdue University, May, Effect of Urban and Agricultural Land Use Land Cover Change on Mesoscale Thunderstorms and Heavy precipitation. Major Professor: Dev Niyogi. Using in-situ and remote sensed datasets, and multi-scale models, this study seeks to investigate the potential impact of urban rural boundaries on heavy rainfall and thunderstorm events. Study objective is to understand the impact that urban landscapes can have on thunderstorm morphology and rainfall characteristics. Two contrasting regions are considered: the thunderstorm belt of Central Indiana, and the tropical monsoon region of India. The main analysis tool used in the study is a nested mesoscale modeling system - the Regional Atmospheric Modeling System (RAMS) coupled with an explicit urban model Town and Energy Budget (TEB) scheme. The first analysis was over the city of Mumbai, located in western India, for an unusually heavy rain event that occurred on 26 July Results suggested that the heavy rains were a result of a convergence zone that was positioned over Mumbai due to surface sensible heat flux gradient caused by urban heat island and sea surface temperature gradients. The second study was over Indianapolis, IN. A radar based summer storm climatology was extended until Results indicated that about 62% of the thunderstorms changed their behavior when passing over the Indianapolis urban area. One observed thunderstorm event (22 June 2006) was analyzed further with and without TEB (urban model) coupled to RAMS. Results

13 xiii indicate that the urban area significantly influences the thunderstorm and rainfall characteristics. In particular there is an increased rainfall downwind, a modest increase upwind and a reduction over the urban area. Study concludes that the urban landscape can exert significant mesoscale forcing that can affect the rainfall patterns and the rainfall/thunderstorm characteristics over both the mid-latitudes and the tropical region.

14 1 CHAPTER 1. INTRODUCTION Land surface characteristics determine surface energy partitioning by assigning the distribution of incoming solar radiative energy into sensible, latent, and ground heat fluxes (Pielke and Niyogi, 2008). Land surface also modulates the exchange of various elements with the atmosphere such as carbon dioxide, water, methane and other greenhouse gases. According to Houghton (1990), anthropogenic activities and changes in land use have contributed to about 25% of the enhanced levels of green house gases. The complex interrelationships between land use and land cover change and weather has also been widely investigated (e.g., Chase et al., 2000; Bounoua et al. 2002; Roy et al., 2003; McPherson et al., 2004; Feddema et al., 2005). Pielke et al. (2002) concluded that land use change is an important climate policy consideration beyond the radiative effects of greenhouse gasses. Also, numerous studies in the last few decades have focused on the effects of deforestation and denudation of natural landscapes and their associated impacts on the surrounding environment (Brown et al., 1991; Flint and Richards, 1991). A more recent assessment in the role of land use land cover change deals with the issue of urbanization. The changes in land surface due to urbanization areas is gaining importance since natural land cover including forests, grassland or even deserts and agricultural land are converted into built-up landscapes. There are significant difference between the thermal and hydraulic properties of the natural land and those of the urbanization materials. The anthropogenic activities and energy releases from the urban area is also different as compared to natural landscapes. Therefore, the impact of urban landscapes on local weather and climate over the next decades is expected to be significant (Shepherd, 2005).

15 2 Urban forcing can produce large changes in the thermodynamic, kinematic, and microphysical properties of the atmosphere, resulting in modifications of a full range of meteorological phenomena including thunderstorms, tornados and even large scale climate events (e.g. Shepherd, 2005; Diem and Mote, 2005; Zhong and Doran, 1995). The impact of urban environment on thunderstorms has been studied by a number of researches (e.g. Changnon et al., 1971; Arnfield, 2000; Shepherd, 2005 and Niyogi et al., 2006). Majority of these studies have relied on a relatively simply urban land surface representation when studying the coupled feedback. With the availability of new urban energy balance schemes (e.g. Masson, 2000), an opportunity now exists to develop detailed, high resolution modeling studies that can provide both a verification and possibly additional insights regarding the role of urban landscapes on thunderstorm dynamics. The following section summaries some of the urban heat island analysis including past observational and modeling studies. This is followed by a summary of the objective/hypothesis and the outline of the thesis Background Urbanization and urban heat island (UHI) The urban heat island (UHI) and its interaction with local weather have received much attention from atmospheric research. In 1833, Howard studied the temperature patterns for the city of London and found that the nighttime UHI was stronger than the daytime. During the daytime the urban land surface absorbs more radiative energy and has a higher temperature. At night, the stored heat is dissipated via long-wave cooling. Compared to the non-urban surface, a higher temperature and greater amount of heat storage occurs over the urban area leading to a heat island. Dixon and Mote (2003) in their study of New York City, found that climatologically the nocturnal UHI is

16 3 stronger than the daytime UHI due to the ability of rural vegetation surfaces to emit more radiation back to space than the urban center. However, this is also affected by cloudiness and other environmental forcings. While the effect of the urban region is on temperature is well known, the impact on rainfall is relatively poorly understood (Hafner and Kidder, 1999). UHI has been linked to the initiation or intensification of convective activities over cities by altering thermodynamic stability (Changnon et al., 1971). Shepherd et al. (2002) found more intense rainfall rates over and downwind of major cities such as Atlanta, and Dallas. Based on the analysis of data from the Metropolitan Meteorological Experiment (METROMEX) in St. Louis, Changnon (1971) concluded that urban effects lead to more thunderstorm activity as a thermodynamic feedback, as opposed to microphysical. Braham et al. (1981) later showed that a higher thermal instability (higher surface temperatures) existed over urban region compared to a nearby rural area affecting thunderstorms. A moisture deficit is also measured in the lower atmospheric layers over the city, yielding an overall thermodynamic instability that is lower over the urban area than over the surrounding rural areas. This suggests that urban areas, because of the moisture deficit, may be more thermodynamically stable than nearby rural areas despite the UHI effect. This can cause a cascading feedback on the local weather and regional climate. Rozoff et al. (2003) studied a thunderstorm event and found that UHI over St. Louis, Missouri played the largest role in initiating the deep, moist convection downwind of the city. Loose and Bornstein (1977) found that a weak UHI and the roughness effect could cause retardation in the movement of slow frontal passages over some urban areas such as the New York City urban region. Craig and Bornstein (2002) found a good correlation between the surface UHI, convergence, and vertical velocity in urban-induced convective precipitation over Atlanta, Georgia.

17 Urban observations Many observational studies over different cities have been investigated in the last few decades (Arnfield, 2003; Grimmond and Oke, 1999). Huff and Changnon (1973) found that thunderstorms downwind of urban regions might increase in occurrence and intensity during the summer months. Results from Kalnay and Cai (2003) suggest that half of the observed decrease in diurnal temperature range could be due to urban and other land use changes. Craig and Bornstein (2002) examined the thunderstorms over the Atlanta, GA region and found that convergence zones near the city center had possibly led to an increase in precipitation amounts. In southeastern China, Zhou et al. (2004) found that the estimated warming of mean surface temperature by 0.05 C per decade is attributable to rapid urbanization. However, the study by Peterson (2003) at the National Climatic Center Data Center suggested that urban landscapes do not have a statistically detectable impact on surface temperatures. Therefore the impact of land surface feedback on regional weather and climate continues to be a topic of interest for weather and climate studies. Current operational weather forecast models, for instance, still do not explicitly consider urban energy balance in the default land surface schemes (e.g. Noah land surface model, EK et al., 2003) With the satellite derived high resolution surface data are becoming available, satellite data are used not only for model initiation, but also for verification of model results. Shepherd et al. (2002) used Tropical Rainfall Measurement Mission (TRMM) data to study the urban effects on precipitation in major cities in the southern US. They found a distinct increase in average precipitation amounts downwind of the urban regions. The METROMEX study investigated many different synoptic classifications on an annual basis and found that the urban region did not necessarily cause additional precipitation events, but enhanced the events that were already occurring (Huff, 1986). Hjemfelt (1982) did a numerical simulation of the effects of the St. Louis urban area on the mesoscale boundary layer airflow

18 5 and vertical motion, and found that urban areas can increase vertical velocities downwind of urban centers. Gero and Pitman (2006) studied the storms in the Sydney Basin, and found that synoptically forced storms were unresponsive to a changed land surface, while local convective storms were highly sensitive to the triggering mechanism associated with land surface influences Urban modeling Due to the complex nature of the urban region, not only in terms of inhomogeneous urban ground surface characteristics, but also the heterogeneity between the different cities of the world, how urban landscape interact with atmosphere remains poorly understood. With increasing urbanization and the ability of models to run finer grids, the need for the reasonable representation of urban processes within weather climate models is increasing (Shepherd, 2005). The majority of mesoscale and regional climate modeling of urban land surface feedback has relied on simplified urban land surface representation (Pielke and Cotton, 2007). Recently, more detailed urban canopy models are available (e.g. Kusaka, 2001). The coupled performance of the models is found to be better than standard slab urban models (Chin et al., 2005). Moreover, the consideration of urban-rural heterogeneities is important in developing the urban induced precipitation feedbacks (Niyogi et al., 2006). Thus, while the role of urban land surface processes is increasing in importance in weather and climate studies, significant uncertainties related to these processes still exist. In particular, in order to analyze the impacts for a variety of storms over urban areas, high resolution modeling studies are needed which involve explicit urban models coupled to mesoscale models.

19 Objective and hypothesis Our main objective is to study the effect of urban landscape on thunderstorm induced convection and precipitation changes. Two contrasting geographic areas are studied: a mid-latitudinal location of Indianapolis in the central region of Indiana, and a tropical location of Mumbai in the western coastal region of India. By comparing the model results for storm events over these two areas, we expect to gain a better understanding of the role of urban areas in mesoscale convection and precipitation processes. While each of the study undertaken has specific hypotheses, a broad study hypothesis we seek to test using observations and model case studies is that urbanization can significantly modify the convection and precipitation pattern near urban areas Outline of the thesis Chapter 2 analyzes a record breaking heavy rain event that occurred on 26 July 2005 over Mumbai, India. The role of urban land and atmosphere feedback on this unusual tropical rain event is analyzed. In Chapter 3, the investigation of the thunderstorms behavior over Indianapolis, Indiana is summarized. A seven year thunderstorm climatology is used. A modeling case study is undertaken to study the possible atmospheric mechanism related to the interaction of the Indianapolis urban area and mesoscale thunderstorms and precipitation patterns. Chapter 4 provides the overall study conclusions, limitations, and brief summary of a future work plan.

20 7 CHAPTER 2. EFFECT OF EXPLICIT URBAN LAND SURFACE REPRESENTATION ON THE SIMULATION OF THE 26 JULY 2005 HEAVY RAIN EVENT OVER MUMBAI, INDIA Abstract We investigate whether explicit representation of the urban land surface improves the simulation of the record-breaking 24-h heavy rain event that occurred over Mumbai, India on 26 July 2005 as the event has been poorly simulated by operational weather forecasting models. We coupled and conducted experiments using the Regional Atmosphere modeling system (RAMS 4.3), with and without an explicit urban energy balance model-town energy budget (TEB) to study the role of urban land atmosphere interactions in modulating the heavy rain event over the Indian monsoon region. The impact of including an explicit urban energy balance on surface thermodynamic, boundary layer, and circulation changes are analyzed. The results indicate that even for this synoptically active rainfall event, the vertical wind and precipitation are significantly influenced by urbanization, and the effect is more significant during the storm initiation. Interestingly, precipitation in the upwind region of Mumbai city is increased in the simulation, possibly as a feedback from the sea breeze urban landscape convergence. We find that even with the active monsoon, the representation of urbanization contributes to local heavy precipitation and mesoscale precipitation distribution over the Indian monsoon region. Additional experiments within a statistical dynamical framework show that an urban model by itself is not the dominant factor for the enhanced rainfall for Mumbai heavy rain event; the combination of updated SST fields using Tropical Rainfall Measurement Mission (TRMM) data with the detailed representation of urban heat island (UHI) simulated by the TEB/urban model created realistic gradients that successfully maintained the convergence zone over Mumbai. Further research will require more

21 8 detailed morphology data for simulating weather events in such urban regions. The results suggest that urbanization can significantly contribute to extremes in monsoonal rain events that have been reported to be on the rise Introduction On 26 July 2005, an unexpected heavy precipitation event occurred over the Mumbai urban region and adjacent areas. A steady downpour initiated around 11 AM local time (6Z), and several regions within a 100 km area showed an unprecedented amount of rainfall over the following 24-hour period. For instance, the Santacruz India Meteorological Department (IMD) official rainfall data recorded a record-breaking 944 mm 24 h rainfall total. Storm reports recorded locations in North Mumbai that received around 1000 mm of rainfall. The event also showed remarkable variability between locations, with stations in southern Mumbai receiving from a trace to 74 mm of rain at another IMD rainfall gauge at Colaba (25 km south of Santacruz). The event caused nearly 500 deaths and is classified as a billion US Dollars natural disaster (NCDC, 2007). Most forecasting models did not predict this heavy rain event successfully. This event is being studied by a number of forecasting groups and teams to improve model performance and subsequently, forecast accuracy. Chang et al. (2008) used a MM5 and WRF model to simulate this event and found that the MM5 could reproduce about 40 60% of the Mumbai rainfall, while the WRF simulated about 60 80% of the observed precipitation. Bohra et al. (2006) documented the India National Center for Medium Range Weather Forecasting (NCMWRF) synthesis of this heavy rain event. They concluded that a high resolution model setup could significantly enhance model performance but most of the predicted rain was in the range of mm. Jenamani et al. (2006) studied the IMD analysis of the Mumbai rain event, and they found that the rain band/low pressure system positioned over the Western Ghats interacted with mesoscale processes to

22 9 create the heavy rain event. Vaidya and Kulkarni (2006) independently confirmed a similar conclusion; and found that a cloud burst phenomenon was the main reason for the heavy rain. They also concluded that the event could be reasonably simulated using a high resolution mesoscale model with a proper choice of boundary conditions and domain size (to capture the synoptic and mesoscale feedbacks). Shyamala and Bhadram (2006) extended the analysis using observational data for synoptic, thermodynamic, radar, and satellite analyses. They concluded that a cloud band over the Arabian Sea on 25 July corresponded to a strong low-level wind over Mumbai, and that mid-troposphere dryness might have contributed to the heavy rain. Chang et al. (2008) tested the impact of land surface representation within Weather Research and Forecasting (WRF) modelling system; and concluded that the results were sensitive to the choice of convection parameterizations and land surface processes. Global climate model ensemble experiments by Koster et al. (2004) have highlighted the Indian monsoon region as a land atmosphere coupling hotspot. Goswami et al. (2006) analyzed the rainfall patterns over the Indian monsoon region and concluded that the heavy rain events are significantly increasing. Kishtawal et al. (2007) analyzed the long-term rainfall data and found urban regions have more statistically significant precipitation occurrence as compared to the rural regions. Studies such as Braham and Wilson (1978), Shepherd and Burian (2003), and Niyogi et al. (2006) have shown urban land surface can influence local storm structure causing enhanced convection and increased precipitation. However, the role of urban landscape on the heavy rain simulation over the synoptically active Indian monsoon region has not been studied. The impact of urban land surface on regional meteorology is well documented. Studying the surface temperature patterns, Howard (1833) delineated the urban heat island (UHI) in London. Subsequently, a number of studies have found that the urban heat island is an important feature in regional meteorology (Oke, 1988). In recent decades, interest has grown in

23 10 the impact of urbanization on mesoscale convection and precipitation. Urban landscapes have been linked to the initiation or intensification of convective activities over cities through altered thermodynamic stability. Niyogi et al. (2006) studied urban-rural heterogeneity-based convective boundaries which affected precipitation patterns around Oklahoma City. Shepherd et al. (2002) found more intense rainfall rates over and downwind of major cities such as Atlanta and Dallas. Rozoff et al. (2003) found that the UHI during a thunderstorm event over St. Louis, Missouri had the largest role in initiating deep, moist convection downwind of the city. Therefore the urban energy balance is vital to the understanding of the importance of meteorological variables within urban areas. Gero and Pitman (2006) studied the storm climatology in Australia s Sydney Basin and found that the local convective storms were highly sensitive to the triggering mechanism associated with land surface influences. Pyle et al. (2007) analyzed the Indianapolis urban region and its storm climatology and concluded that urban land surface can change storm morphology and regional precipitation. Thus, significant background evidence suggests that the urban landscape can affect storm morphology, and that mesoscale boundaries can modify storm related precipitation. Mumbai is a major coastal city on the west coast of India and is the country s commercial capital. The population of the downtown and surrounding suburbs increased from 9.9 million in 1981 to 13.0 million in 1991 and 17.7 million in 2001 (International Institute for Population Studies). The city is experiencing rapid urbanization, and such densely populated cities with an aging infrastructure are vulnerable to natural disasters. Therefore, it is important to assess the possible impact of the Mumbai urban landscape and its representation in the July 2005 heavy rain event. To date, the 26 July 2005 heavy rain event in Mumbai has many clear features, yet many unknown causes, thus we are primarily interested in clarifying the role that landscape change including urbanization played in this event. Our hypotheses include:

24 11 1) By using an explicit 3-D urban surface energy balance model- Town Energy Budget (TEB), we can better represent the land atmosphere interactions and improve the simulation of the heavy rain event over Mumbai city in a mesoscale model; 2) Precipitation distribution and intensity will be significantly changed during this event due to the urban landscape; 3) The effect of urbanization on heavy precipitation occurs on a relatively local scale near the urban region, and the large scale precipitation will be influenced by the sea surface temperature (SST) fields Data and methodology The Regional Atmospheric Modeling System (RAMS) version 4.3 (Pielke et al., 1992, 2001; Cotton et al., 2003) was adopted for this study. RAMS is a primitive equation, non-hydrostatic mesoscale environmental modeling system; its performance to skilfully simulate convective systems has been validated in various prior studies (e.g. Nair et al., 1997; Rozoff et al., 2003), including a study over India (Mukhopadhyay et al., 2005). RAMS involves a two-way interactive grid nesting. The three nested grids in this study had an 80, 20, and 5 km grid spacing with the same center points (19.15N, 73.05E) as shown in Fig.2.1 Grid 1 had horizontal grid points and a time step of 27 s that covers the whole south Asia monsoon region. The second domain simulated the general synoptic and mesoscale setting with horizontal grid points and a time step of 9 s. The third grid has horizontal grid points, and a time step of 3 s. The grid was mostly focused over Mumbai and its adjacent Arabian Sea region. We selected 36 vertical layers with spacing from 0.05 to 1.27 km in the first 8.5 km of the atmosphere; the spacing was maintained at 1.27 km for heights above 8.5 km. The total depth of the model atmosphere is 15 km. Fig.2.1 shows the

25 12 three nested domains. The urban model coupling was carried over the innermost region that covers Mumbai. Model initialization for this study was horizontally inhomogeneous with forcing data from the National Center for Environmental Prediction (NCEP) Global Data Assimilation System (GDAS) (Kalnay et al., 1996). All simulations were performed for 120 h from 25 July to 30 July Soil moisture was initialized at 35% of soil moisture capacity (as a percentage of soil texture dependent saturation) and followed the GDAS fields. The soil moisture and soil temperature prognostically changed and quickly allowed spatial heterogeneity to the region. This study used the modified Kuo cumulus parameterization scheme for domain 1 and 2. Table 2.1 summarizes the basic model configuration. A number of model experiments were performed and the most significant impact was found for prescription of SST fields. SST was important for setting the mesoscale boundaries in the coastal regions. We used the default SST data in RAMS 4.3 as the climatological SST from 1950 to 1980 for the sensitivity study discussed below. The model results discussed in the following section, however, are based on Tropical Rainfall Measurement Mission (TRMM), weekly data from 23 July (0.25 degree resolution). The TRMM satellite travels west to east in a semi-equatorial orbit. This path produced data collected at changing local times for any given earth location between latitudes 40 S and 40 N. As compared to the climatology, the TRMM data shows a higher SST and a larger gradient along the Western Ghats (Fig.2.2 a-b). Fig.2.2 b shows a high SST zone (22 C) near the west coast of Mumbai which is noteworthy, as is shown in the model simulations. The default land surface model-land Ecosystem Atmosphere Feedback (LEAF-2) model (Walko et al., 2000) was used to account for the exchange of heat and moisture between the soil, vegetation, canopy, surface water, and atmosphere. LEAF-2 explicitly represents canopy processes following Deardorff (1978). It also represents the details of turbulent exchange and

26 13 radiative transfer, as well as physical evaporation, transpiration, precipitation, and fluxes of heat and moisture between the soil or snow and the atmosphere (Lee, 1992; Walko et al., 2000). LEAF2 vegetation types comprise a combination of Biosphere Atmosphere Transfer Scheme (BATS) and land data assimilation system (LDAS) classes (Dickinson, 1984). The land use data are based on the USGS global data set. For the innermost domain, the LEAF2 scheme was coupled with an explicit urban model: the Town Energy Budget (Masson, 2000) to study the urban area (only LEAF2 run is considered as control run). The TEB is used only for energy balance over the urban land use grids, while LEAF2 is the default. The 5 km grid spacing in the third domain was further allowed to have 4 sub-grid land use categories following a mosaic approach (Avissar and Pielke, 1989; Koster and Suarez, 1996). For the urban sub-grid land use, the TEB was used for solving the energy budget, and for other land uses, the default LEAF2 representations were employed. The total (grid-averaged) fluxes and scalar output for the grids with urban land use thus consisted of a sum of the TEB and LEAF2. For the TEB, information on the buildings, roads and average geometry of the urban area are needed. We constructed the TEB parameters for Mumbai city based on our knowledge of local geography and discussions with local experts; the final land use morphology domain was also guided by several high resolution (street scale) maps including Google Earth maps (maps.google.com). Typically for downtown areas, the average building height was 8 m, with varying building and canyon ratios reasonable for the average buildings in Mumbai. The roughness length was approximated as 0.8 m; a value considered typical for urban/suburban areas. Table 2.2 summarizes important TEB parameters. We stress that the table created was principally for driving the model and had a high degree of approximation (due to lack of publicly available morphology data).

27 Results Synoptic scenario In July 2006 when the monsoon was active, strong westerly or southwesterly wind flow developed along with the formation of an offshore trough over the Arabian Sea. A low-pressure system formed over the North Bay of Bengal and Orissa coast on 24 July and intensified as it moved inland. It brought the monsoon trough to the south of its average position along 20 22N, with a strong cross-equatorial flow over the Arabian Sea. As the system moved westward, the low-level jet continuously intensified, and strong westerly winds lashed the west coasts. Generally, this scenario leads to widespread heavy rainfall over the Western Ghats and Mumbai regions (Francis and Gadgil, 2006; Kumar et al., 2007). On 26 July, the rain belt dropped more than 100 mm of precipitation at many stations in Konkan (Western Ghats) and Goa and then moved northward. Subsequently on the 26 July, Mumbai and many other parts of Maharashtra (east of Mumbai) were devastated with heavy rains. Both the lower- and upper-level large-scale flow patterns over the Indian region can be viewed using the 850 hpa and 300 hpa wind analysis from the NCEP analyses valid at 12Z 26 July 2005 from the NCEP analyses (Fig.2.3). Fig.2.3 a shows the presence of a strong cyclonic circulation that extends upwards to the mid-troposphere over Orissa and eastern part of Madhya Pradesh associated with a well-marked low-pressure system near the surface. The 850 hpa plots also show the presence of strong westerly or north-westerly winds with speeds of 20 m s -1 (Fig.2.3 a) over the western coast of India including Mumbai. This low-level, strong westerly or northwesterly wind became easterly or northeasterly at 300 hpa (Fig.2.3 b) with speeds between 6 12 m s -1 over Mumbai. Model simulations shown in Fig.2.3 c-d, successfully captured both features remarkably well.

28 15 The divergence components of the wind provide an understanding of the atmospheric divergent circulation associated with the vertical motion field. Generally the atmospheric heating associated with convection will induce centres of divergence. Analysis of the divergence field (Fig.2.4 a-d) at various times on 26 July 2005 indicate that the upper-level divergence, overlying deep low- and mid-level convergence had been present in the Mumbai area for some time, along with a substantial increase in convergence below 700 hpa. This presence of a convergence pattern appears to be consistent with the observed changes in the rainfall intensity. The vertical cross section of zonally-averaged wind (72.5E E) for this analysis at different times is shown in Fig.2.4 e-h. The zonal wind cross sections clearly illustrate the development of a jet, which reaches a maximum of approximately 20 m s -1 near 700 hpa at 12Z and 18Z on 26 July, respectively. The convective system that produced the intense rainfall appears to have developed over Mumbai just to the north of the low-level jet. The control experiments predicted the strong westerly or northwesterly winds over the western coast of India with a maximum speed of 20 m s -1. The model showed the presence of a strong cyclonic circulation over Orissa and the eastern part of Madhya Pradesh. The control experiment simulated the upperlevel easterly or northeasterly flow reasonably well, with speeds of approximately 12 m s -1 over Mumbai. As a whole, both the upper- and lowerlevel large-scale circulation features are simulated reasonably well when compared with the corresponding NCEP analyses. The upper-level divergence and low- and mid-level convergence was simulated well in the mesoscale model run. All major features described in the analysis were satisfactorily reproduced, though some features, such as the low-level jet, were relatively weaker than the observations.

29 Rainfall According to the official IMD rainfall observation over Santacruz (Table 2.3), the nearly 950 mm total occurred from 14:30 to 20:30 26 July, The model simulated nearly 600 mm precipitation in different experiments for the same period (Fig.2.5). Simulations coupled with the TEB run showed consistently higher precipitation from noon of 26 July, indicating the significance of incorporating explicit urban model surface effects of urbanization. This feedback is discussed in more detail in the following section. Both the Tropical Rainfall Measuring Mission (TRMM) satellite and the CPC MORPHing technique (CMORPH) from the NCEP analysis over the Indian sub-continent on 26 July 2005 showed a localized hotspot over Mumbai (Fig.2.6 a,b). Fig.2.6 c depicts the spatial precipitation distribution regressed from gauge data. The RAMS simulated precipitation with, without TEB and their difference is shown in Fig.2.6 d-e. Both precipitation measures are highly concentrated over the Mumbai region. The model run that includes the TEB had some significant spatial differences which are discussed below Vertical structure/sounding One of the important sources of observation data: sounding profiles is useful in the understanding of the atmosphere s vertical structure. The prestorm sounding data (0Z 26 July), and sounding data during the storm (12Z 26 July), were available for analysis over Mumbai (Fig.2.7). At 0Z 26 July 2005 (Fig.2.7 a) from ground surface to 3000 m, the air temperature decreased with height, and air temperature in the TEB run generally matched the observation, although the control results were within 1K. Similar results were obtained for virtual potential temperature from the ground surface to 500 m. The TEB results were generally closer to

30 17 observation, being about 1 K higher than the observation, and the control run result was slightly lower. From m, the TEB results more closely matched the observations, with the control results being about 1 K higher than the observations. Relative humidity from the observations showed a decrease from 600 to 3000 m height. The TEB results captured this trend, but the control run showed a saturated condition at about 600 m. For wind speed, sounding data showed stronger values at about 1000 m. This structure was simulated well by the TEB run, but at 900 m height the maximum winds were underestimated. In the control run, the wind profile did not show a jet and reached a maximum speed at a height of about 2000 m. Thus overall, the TEB run appears to be superior than the control in simulating a vertical structure that is closer to the observations. The 12Z 26 July sounding corresponds to the time Mumbai was experiencing heavy rainfall (Fig.2.7 b). Generally the sounding data had a high uncertainty during the heavy precipitation, but these data still can serve as a reference for analysis. Air temperature and virtual potential temperature (θ v ) were nearly identical in the TEB and control runs. Both results were reasonable, and the structures for air temperature and θ v match well with the observations. Simulated relative humidity was saturated, which is reasonable considering the heavy rain, even though the sounding data does not show saturation. Wind speed during the heavy precipitation event was high at 1500 m height according to the observations. Neither TEB nor control captured this well. Above 2300 m, the TEB result was closer to the observations.

31 Impact of coupling the urban TEB within RAMS mesoscale model The simulation results discussed above indicate that the mesoscale model (RAMS), when coupled with the explicit urban model TEB, has an enhanced ability to reproduce the boundary layer over urban areas. In this section we further discuss the impact of representing an explicit urban model within the mesoscale modeling system Temperatures and fluxes change over the urban region (UHI simulation) The urban heat island (UHI) was well simulated by RAMS TEB around the Mumbai region. Fig.2.8 shows the 10 m air temperature change which occurred when using the explicit urban representation. The urban region had a significant UHI both in the daytime and the night. During the monsoon season, generally the cloud cover over west coast of India is high. This feature was well simulated by the model and the simulated cloud cover over coastal land region on the west of India is mostly close to 100% (not shown). This caused the incident solar radiation near the surface to be quite low. The low-level wind during the period was controlled by the south Asian monsoon and the strong southwest wind feature, therefore, air temperatures increased above the urban region and above downwind and northeast of Mumbai city. Fig.2.8 a-d show air temperature changes at 0Z, 6Z, 12Z and 18Z for the two runs. The corresponding diurnal time series over an urban, rural, and offshore locale is displayed in Fig.2.8 e. Throughout the day, the warming zone extended from the city center to the northeast direction of the city with a 1 2 K higher temperatures than in the surroundings. The size of this region, which experienced a significant warming due to UHI, is about km with an extended area in the northeast downwind direction. Comparing different times within one day, the UHI was apparent for most of the day except the

32 19 afternoon when heavy precipitation occurred and cloud cover was at its highest. In the upwind direction from urban region, there was a cooling from K just over the Mumbai coastal region. This cooling was probably because of the rainfall over the region and the general associated circulation. The simulation clearly showed that the UHI was generally more significant downwind of the city from 73E 73.5E. Vertically, the higher air temperature was typically noted up to a height of 400 m above the surface (Fig.2.9). The warming effect decreases with height: at 100 m TEB results were 1 K warmer; this difference decreased to nearly 0.5 K at 200 m, and then the warming effect essentially disappeared. The UHI effect was distributed downwind direction of Mumbai city, especially in the 73E to 73.5E region. One key mechanism of low-level jet formation involves horizontal differences in heat fluxes, which can produce strong, shallow baroclinicity in the boundary layer (Stensrud, 1996). The low-level jet in many other heavy precipitation events has been found to be enhanced because of the gradient in sensible and latent heat fluxes, this spatially varying flux distribution has been attributed to gradients in soil moisture (Holt et al., 2006). Consistent with the surface air temperature results discussed above, and as observed over Mumbai city, the sensible heat flux increased after urbanization (using the TEB) by approximately 50 W m -2 (Fig.2.10) During a 24 hour time period, the UHI was more significant during the night (00:00AM) and early morning (06:00AM), consistent with previous conclusions that the UHI was relatively more significant after late afternoon (Fig.2.9) (Niyogi et al., 2006). The interaction of the land-surface and the atmosphere can also produce high spatial variability in latent heat fluxes. This land-surface interaction is another important process for producing variability in tropical rainfall. Fig.2.11 shows the temporal latent heat fluxes. As expected, the TEB did not change the latent heat flux significantly due primarily to the very low evaporation over the urban region. The latent heat flux showed a slight decrease by 50 W m -2

33 20 in the upwind direction over the sea most likely due to the precipitation increase and temperature decrease over this area Winds and Convergence From the simulation results, over the city on the afternoon of 26 July an obvious vertical updraft was apparent both in the control and the TEB simulations (Fig.2.12). With the urban TEB run, the updrafts are generally in the upwind direction (Fig.2.12 c). The enhancement in the vertical velocity is 0.5 m s -1. Convergence changes associated with wind, roughness change, and thermal conditions are shown in Fig.2.13 a-d. For 12Z on 26 July, with the TEB run, the low-level offshore winds are slower. The control run shows a lower wind speed zone over the city with a speed of 10 m s -1 up to about 200 m high, and the TEB run shows the low wind zone at 10 m s -1 over 1000 m. The wind speed above the urban region below a height of 500 m decreases from 10 m s -1 to 2 m s -1 (Fig.2.13 c). This change due to urbanization indicates increased convergence in the upwind direction. As seen in Fig.2.13 d, from the ground surface to a height of 3000 m at 72E, a significant convergence zone developed over Mumbai. With the TEB run, the results indicated that the low-level convergence increased up to 3000 m and decreased thereafter. NCEP analysis also shows a similar trend with the divergence field for different times on 26 July (Fig.2.4). The analysis indicates that upper-level divergence overlying deep, low- and mid-level convergence existed over the Mumbai region and a substantial increase in convergence below 700 mb occurred that helped enhance the moisture transport.

34 Rain and cloud water mixing ratio, and precipitation Consistent with the vertical wind fields, the rain water mixing ratio in TEB run also increased upwind of the city and decreased in the center and downwind of the city compared to the control run (Fig.2.14). Above the urban region in the upwind direction, the rain water mixing ratio was modified due to urbanization. This ratio increased from 1 g kg -1 to 1.4 g kg -1 up to 4500 m in height. The cloud water mixing ratio shows a similar trend (Fig.2.15). In the afternoon of 26 July, Fig.2.15 shows the cloud water mixing ratio at 12Z. The control run shows that the cloud water ratio has increased between the heights of m with the maximum zone at 5000 m generally occurring above the 72E to 73.5E region. After urbanization, the upwind area from 71.5E to 72.5E around 19.15N had a higher cloud water mixing ratio at around 5000 m. The cloud water mixing ratio was modulated by vertical winds (Fig.2.15 c). As a result of these changes after urbanization, precipitation distribution and intensity were both significantly influenced by mesoscale dynamical feedbacks. As displayed previously in Fig.2.6, the accumulated precipitation from July 2005 for the control run indicates that the precipitation location is not well represented; the precipitation was split into two rain centres over north and south Mumbai city. After coupling with the TEB, the model precipitation appears correctly over the Mumbai bay area and was increased by a significant amount (200 mm) Analysis of the explicit SST and the urban land-atmosphere interactions To further resolve the explicit role of the urban model and satellitecaptured SST on the convection, we performed a statistical factor separation analysis (Stein and Alpert, 1993) from July Model-simulated vertical velocity, convergence, and 12 hour accumulated precipitation were analyzed. The intent of the analysis was to quantify the relative impacts of (i)

35 22 the representative urban model (f 2 ), (ii) the improved satellite SST data (f 1 ), and (iii) improvements in both the urban model and SST representations (f 12 ). According to Stein and Alpert (1993), the relative impacts f 1, f 2 and f 12 are given by the field differences between the simulations F 0, F 1, F 2 and F 12 : f 1 = F 1 -F 0 (1) f 2 = F 2 -F 0 (2) f 12 = F 12 (F 1 +F 2 ) + F 0 (3) where F 0 is the background without the TRMM SST and urban model, F1 is the TRMM SST, F 2 is the urban TEB, and F 12 is the simulation using TRMM SST and the urban TEB interaction. The analysis in the previous section showed that the convergence pattern over Mumbai is an important variable which links the atmospheric condition to the precipitation. Because the maximum intensity occurred between 10 and 11Z, we focus on the spatial convergence at 850 hpa level at 9Z on the 26 July. Fig.2.16 a shows that without TRMM SST, both the F 0 and F 2 simulations have their convergence centres northeast of Mumbai, and no convergence is seen near the urban center. F 1 simulation, with SST and no urban model, has a convergence belt along the coastal region, while its center is to the south of Mumbai. With the urban model (F 2 simulation), the convergence zone narrows and strengthens over the Mumbai peninsula. Fig.2.16 b shows the explicit contribution of each effect. The spatial plot suggests that the updated SST fields have an important effect on convergence and rainfall. Interaction of the urban scheme and the TRMM SST allows for more realistic local scale heterogeneities to be developed that alters the location of the rainfall bands. Fig.2.16 c shows the vertical convergence structure at 19.15N. Adding the TRMM SST field (f 1 ) increases the strength of the convergence at the border, but decreases the convergence above the Mumbai region, and without the SST, the urban model (f 2 ) decreases the convergence slightly above the urban area.

36 23 However, their interactions show an obviously increasing effect on the convergence above the city. Vertical velocity shows a similar effect (not shown). Without the TRMM SST data, the updraft located to the east of the city in the urban model weakens the movement. The TRMM SST provides the right gradients that displaced the updraft zone to the east of the city, and the urban model strengthens the upward motion of the air mass. The resulting impact on the accumulated rainfall is shown in Fig The rainfall simulation is enhanced due to the TRMM SST (F 1 ). The simulated locale is northeast of Mumbai. The urban model leads to changes that locate the heavy precipitation center above Mumbai, and slightly west of the observation center (F 12 ). On 27 and 28 July, the simulation showed the same trends as the 26. Fig.2.18 shows the time series of the individual simulations and interactions based on the following factor separation analysis. As discussed above, the urban model by itself did not affect the precipitation pattern over the region. However, considered with realistic SSTs, the urban model creates the right mesoscale boundaries that contribute to significantly increase rainfall over Mumbai. This interaction effect contributes to over 200 mm of the precipitation for this event, especially during the maximum intensity of this precipitation event Conclusions We simulated the 26 July 2005 heavy rain event in Mumbai using the RAMS coupled with the TEB. The following conclusions can be made: 1) The mesoscale model, RAMS had a reasonably good performance in simulating the precipitation fields when using proper schemes, reliable initial and boundary conditions.

37 24 2) Coupling the explicit urban model showed a consistently enhanced performance. 3) The urban landscape caused an UHI, which interacted with the SSTs causing a mesoscale convergence zone over Mumbai. This convergence zone appears to be a major reason for the moisture transport and a heavy rain center over Mumbai. 4). The factor separation analysis suggests that the urban model by itself is not the main reason for the enhanced rainfall; rather the combination of the updated SST fields using the TRMM data, and the UHI simulated by the TEB/urban model that created the realistic gradients that accurately maintained the convergence zone over Mumbai. In summary, the SSTs and landscape pattern helped transport large amount of moisture over the Western Ghats region, while the mesoscale boundaries created by the urban land surface and sea surface temperature gradients helped the positioning by the convergence zone and heavy precipitation focus over Mumbai.

38 Figure 2.1 Simulation model domains and idealized urban region over Mumbai city (inner most domain with TEB). Dashed line in third domain indicates the cross-section (at 19.15N) that is analyzed in this study. 25

39 26 (a) (b) Figure 2.2 Comparison of SST initial conditions: (a). Default / climatological SST ( C) between May and August from 1950 to 1980; (b). TRMM satelliteobserved weekly averaged SST ( C) July, 2005.

40 Figure 2.3 Comparison of NCEP reanalysis and model simulated wind circulation (m s -1 ). 27

41 Figure 2.4 NCEP reanalysis data for zonal divergence (10-5 s -1 ) and zonal winds (m s -1 ) between 72.5E and 73.5E 28

42 Figure 2.5 Precipitation time series for over Santacruz Airport (19.08N, 72.83E): Circles = observed rain gauge rainfall (mm), solid line with open squares=teb Run (mm), green dashed line with dots=control run (mm). 29

43 30 a) TMI precipitation (mm) b) CMORPH (mm) c) Gauge data (cm) d) control (mm) e) TEB (mm) f) TEB - control (mm) Figure 2.6 Spatial precipitation (mm) captured by TRMM, CMORPH, gauge data, and model.

44 31 a) Vertical sounding at 00Z July 26 th GMT b) Vertical sounding at 12Z July 26 th GMT Figure 2.7 Comparison of observed sounding data and model at the Santacruz airport. Solid line with open circles=observation; green dashed line with dots=teb; yellow dashed line with open circles=control.

45 32 a) 1000 hpa air temperature (K) at 00Z 26 July b) 1000 hpa air temperature (K) at 06Z 26 July c) 1000 hpa air temperature (K) at 12Z 26 July

46 33 d) 1000 hpa air temperature (K) at 18Z 26 July e) 1000 hpa air temperature (K) time series over urban location: (19.15N, 72.9E); rural location: (19.15N, 73.2E); sea surface location: (19.15N, 72E). Black line with open squares: control run; green line with closed squares: TEB run. Figure hpa air temperature changes due to urbanization (using TEB).

47 34 (a) (b) Figure 2.9 Cross-section of (a). Vertical air temperature (K) and (b). virtual potential temperature (K) profiles over the Mumbai urban region.

48 35 a) 00Z 26 July 2005 b) 06Z 26 July 2005

49 36 c) 12Z 26 July 2005 d) 18Z 26 July 2005

50 37 e) Sensible heat flux (W m -2 ) time series over urban location: (19.15N, 72.9E); rural location: (19.15N, 73.2E); sea surface location: (19.15N, 72E). Black line with open squares: control run; green line with closed squares: TEB run. Figure July sensible heat fluxes (W m -2 ) and differences in both simulations (TEB control) Figure 2.11 Same as Figure 2.10(e) but for latent heat flux (W m -2 ).

51 38 a) b) c) Figure 2.12 Cross-section of 12Z 26 July vertical wind and its change (m s -1 ) after using urban model.

52 39 a) Cross-section wind (m s -1 ) b) Cross-section convergence (10-5 s -1 ) Figure 2.13 Cross-section of 12Z 26 July u and w winds (m s -1 ), and their convergence (10-5 s -1 ).

53 Figure 2.14 Cross-section of 12Z 26 July rainwater mixing ratio (g kg -1 ) at 19.15N from 71E to 75E. 40

54 41 a) b) c) Figure 2.15 Cross-section of 12Z 26 July cloud water mixing ratio (g kg -1 ) arrows show the location of updrafts and downdrafts.

55 42 (a) (b)

56 43 (c) Figure 2.16 Simulated convergences (10-5 s -1 ) at 9Z Jul.26th: (a). various model simulations: control, TRMM SST, urban model, TRMM SST and urban model. (b) Analysis where the panels in Figure 2.15 a were used to study the effects of each model parameterization. Positive values (solid contours) represent areas of convergence increase, and negative values (dashed contours) represent areas of divergence increase. (c). The same with (b) but for cross section at N.

57 Figure 2.17 Factor separation analysis for accumulated precipitation (mm). 44

58 Figure 2.18 Time series of accumulated precipitation (mm) over Santacruz airport: control, the effect of TRMM SST, the effect of urban model and the effect of their interactions. 45

59 46 Table 2.1 Model Configuration and setup. Category Governing equations Vertical coordinate Grid stagger and configuration Time differentiation Turbulence closure Lower boundary and urban surface Upper boundary Options 3D, non hydrostatic, compressible Terrain-followingδ z Arakawa-C grid; three fixed, nested grids Leapfrog and forward time difference K from deformation scaled by stability LEAF-2 with TEB (For Domain 4) Rigid lid with modified Rayleigh friction layer Lateral boundaries Klemp and Wilhelmson (1978) Microphysics Single-moment bulk microphysics Cumulus parameterization Modified Kuo Tremback (1990) Radiation Chen and Cotton (1983)

60 47 Table 2.2 Key input variables for TEB. Key variables value Average urban Building height (m) 8 Building width (m) 3 Z 0 (m) 0.8 Traffic sensible heat release (Wm 2 ) 12 (average) Traffic latent heat source (Wm 2 ) 0 Industrial sensible heat source (Wm 2 ) 20 (average) Industrial latent heat source (Wm 2 ) 10 (average) Constant temperature inside building (K) 300 Table 2.3 Temporal rainfall variations at Santacruz on 26, 27 July Date Hour IST Rainfall (mm) Accumulated Rainfall (mm) 26 July (27 th ) July Total 24 hrs

61 48 CHAPTER 3. THE EFFECT OF THE INDIANAPOLIS URBAN AREA ON SUMMER THUNDERSTORM BEHAVIOR Abstract We examined radar and observation-based storm climatology for the time periods for the Indianapolis, Indiana urban area. The hypothesis of this research is that the Indianapolis urban landscape alters the intensity and composition/structure of approaching thunderstorms due to the heterogeneity of the land surface characteristics present between the urbanrural interface, especially the urban heat island and roughness. This research focuses primarily on the spring and summer seasons from May through August. A total of 81 unique thunderstorm cases in and around the Indianapolis region over a 61-day period are examined. Results indicate that 62% of the analyzed thunderstorms changed their morphology according to radar base reflectivity images, with the noticeable feature that thunderstorms break up over the Indianapolis urban area, although other alterations also exist. A climatology table was created to assess changes in storm composition, as well as orientation and propagation, synoptic setting, and the time of each event. A case study under storm event on 22 June 2006 is examined using the Regional Atmospheric Modeling System (RAMS 4.3) coupled with the explicit urban model Town Energy Budget (TEB) simulations. Data from NARR reanalysis, CoCoRaHS rain gauge observational data, AmeriFlux and regional climate center is also used. Results confirm many of the findings seen in the earlier studies that an explicit urban energy balance model was not used. These simulations show good representations of atmospheric conditions by comparing temperature, ground flux and precipitation

62 49 distribution with storm observations. The simulation results show increased precipitation over the downwind area of Indianapolis, diverse changes over the urban area, and limited changes over areas upwind of the Indianapolis area. Sensitivity experiments mainly indicate that the city s roughness length in simulation can increase low level convergence, the urban heat island could decrease the rainwater mixing ratio over Indianapolis, and urban sprawl might decrease the precipitation over the city, but increase the precipitation over downwind Indianapolis Introduction Background Population growth has led to significantly increased urbanization since the Industrial revolution, with the resulting land cover changes caused significant local-scale anthropogenic forcing of weather and climate is expected (Cotton and Pielke, 2007). During in USA, 13.9 million acres of farmland were converted to urban use, including about 5.4 million acres of prime farmland (ERS-USDA, based on National Resources Inventory data). These local-scale changes act as both external and internal forcings: external forcing of surface climate through alterations of local surface sensible heat, latent heat, and momentum fluxes, and internal forcing through emissions of short-lived or highly-concentrated chemicals or aerosols. Such forcing can produce large changes in the thermodynamic, kinematic, and microphysical properties of the atmosphere, resulting in modifications of a full range of meteorological phenomena including thunderstorms, tornados and even climate events (Shepherd, 2005; Diem and Mote, 2005; Dixon and Mote, 2003; Zhong and Doran, 1995). Urban areas can influence thunderstorm behavior through their unique energy balance and roughness profile. Buildings and concrete areas in urban

63 50 areas store more heat than the surrounding rural terrain, which leads to a continuously varying diurnal temperature gradient across the urban rural interface known as the urban heat island (UHI) (Oke, 1988). To date, a majority of the investigations of urban impact have focused on UHI-induced temperature changes and have relied heavily on the analysis of short-term field observations or long-term climate records (Kalnay and Cai, 2003; Zhou et al., 2004). Though thunderstorm behavior has been widely studied, thunderstorm behavior changes when passing through urban areas are not thoroughly understood (Shepherd, 2005). Studies such as Shepherd (2003) found that radar-derived urban morphological parameters (e.g. roughness length) significantly impact the precipitation variability in numerical simulations of coastal Houston. Moreover, each city is unique, so different city landscapes might cause different effects, and the forecasting as well as regional climate implication need to be studied. The complex interactions between urban-scale anthropogenic forcing, the associated local or urban-scale phenomena, and the regional or largescale atmospheric processes are poorly understood (Shepherd et al., 2002). Moreover, mesoscale weather prediction and regional climate models (e.g. the community-based NCAR-Penn State Fifth Generation Mesoscale Model (MM5) (Grell et al., 1995) or the Weather Research Forecast (WRF) model) often time fail to successfully simulate these complex interactions because they lack detailed information on urban land use, as well as a land surface model (LSM) that adequately treats urban landscapes and their interactions with the atmosphere above them. By improving the treatment of the urban energy balance in weather forecasting models, it is expected that the explicit urban model will help to quantify the effect of urban land cover on atmospheric dynamics, and thus improve the prediction of thunderstorm location, intensity, and duration (Niyogi et al., 2006). Enhancement of the atmospheric patterns in the urban environment using the urban canopy model (UCM) can clarify current difficulties in urban forecasting-related issues.

64 51 In his study of thunderstorm radar images from over Central Indiana, Pyle (2007) used MM5 and found that non-urban landscape could not initiate a 13 June thunderstorm, but urban landscape could initiate it. Recent developments in parameterization of the urban land surface within land surface / mesoscale models provide us an opportunity to answer the following questions: What is the impact of utilizing a more detailed urban energy balance model as compared to a simplified scheme used in previous coupled mesoscale modeling studies of urban thunderstorm behaviors? What is the ability of coupled urban models to simulate the thunderstorm structure over a city such as Indianapolis urban area? What is the role of urban area on the thunderstorm ad rainfall over urban region? We addressed these questions by building on Pyle (2006) and Pyle et al. (2008) study of Indianapolis thunderstorms. His radar climatology from 2000 to 2005 summer storm was extended until 2007 and simulation by RAMS with TEB setup was added to analyze a thunderstorm event as discussed ahead Indianapolis urban area Indianapolis is the capital city of Indiana and is approximately km 2 in size, being the 12 th largest city in United States. It is located in central Indiana (39.79N, 89.15W) and the U.S. Ohio Valley. Indianapolis population has almost doubled since 1950 and increased approximately 40,000 since 1990 from 741,952 to 781,870. (U.S. Census Bureau, 2000). Similar to the urbanization occurring all over the world, central Indiana/Indianapolis continues to urbanize through the conversion of agricultural lands to built-up urban landscapes. The Indianapolis landscape is a valuable urban-rural study area because it has a sharp contrast in land use/land cover from row crops to

65 52 urban land within a relatively short distance and almost without topography changes (Fig.3.1) Precipitation climatology for Indiana Average annual precipitation ranges from 940 mm in northern Indiana to 1194 mm in the south. Table 3.1 shows the 30 year normal (1971 to 2000). The period from April to August has more rainfall than other months. May is the wettest month of the year with an average rainfall between 4 and 5 inches across the state. The average rainfall decreases slightly as summer progresses. Autumn and winter months are drier with 76 mm of rainfall typical in each month. February is the driest month of the year statewide, when precipitation increases in March and April as the spring soil moisture recharge season begins. Then, on average, precipitation occurs every third day. Fig.3.2 shows a 30-year normal precipitation distribution, covering all seasons, for the Indianapolis urban region. Fig.3.2 a is from 1961 to 1990 and Fig.3.2 b is from 1971 to A noticeable feature in Fig.3.2 a-b is the low rainfall near the Indianapolis region compared to its adjacent area. Because no significant topographical or other natural geographic heterogeneity in this area except the Indianapolis urban region, a reasonable explanation for this feature is the effect of the urban rural heterogeneity (Pyle et al., 2008) Methodology A storm climatology assessment was performed for the Indianapolis urban region valid during the time period of May April 2007, focusing on the summer months of May through August. Pyle (2007) completed his analysis from 2000 to 2005, and the analysis was extended until The methodology is similar to that described in Pyle (2007). Each event was found through the Storm Prediction Center (SPC) data archive of severe weather

66 53 events ( A total of 81 summer thunderstorm cases covering 61 days from 2000 to 2007 were studied. Each case was examined using observed base reflectivity radar plots for storm composition change as the storm entered and passed through the urban region. Storm composition change is assumed to occur when the storm was initiated, broken up, intensified, dissipated, or converged in or around the urban or rural region. It is important to note that base reflectivity analysis might contain certain biases that may alter observation results when considering storm intensity. 3-D version radar images from GR2Analyst software were also used to analyze the vertical structure of the storms, and the intensity and structure changes. Pyle (2007) also studied storm event using coupled model and was able to identify important land-atmosphere feedbacks. Pyle used the MM5 model with a slab land use model to simulate the ground surface conditions. In this study, an explicit 3-D land surface model: the town energy budget (TEB) were employed coupled to RAMS 4.3. NARR data, AmeriFlux observations and CoCoRaHS data were used to evaluate the simulation. In order to better understand the mechanism of the thunderstorm-urban landscape interactions on precipitation over Indianapolis, RAMS was run with and without the explicit urban model TEB to study the effect of the urban representation on precipitation over Indianapolis. Sensitivity experiments were also undertaken to study various urban area factors such as albedo, roughness length, and urban size. Our main hypotheses include: (1). Scattered thunderstorms or squall lines will become organized and strengthened when approaching an urban area and this is due to an enhanced convergence near urban boundary. (2). Thunderstorms will decrease in intensity when passing through the urban area due to an increased saturation water vapor mixing ratio caused by

67 54 the UHI. Test of this hypothesis could be also an explanation of climatologic precipitation over Indianapolis urban area. (3). Thunderstorms will be intensified downwind of an urban area. These hypotheses will be tested by analyzing the radar datasets and by performing a case study using RAMS mesoscale model coupled to the TEB explicit urban energy balance scheme (described in Chapter 2) Model configuration RAMS was set up with three telescopic nested grids that were two-way interactive and had 80, 20, and 5 km grid spacing with center point (39.77N, 86.16W) as shown in Fig.3.3. Grid 1 had 64 x 48 horizontal grid points covering the entire USA and a time step of 90 s. The second domain simulated mesoscale condition with 82 x 74 horizontal grid points and a time step of 30 s. The third grid has 94 x 94 horizontal grid points, and a time step of 10 s. The grid mostly covered the state of Indiana. The model was configured with 36 vertical layers with spacing from 0.05 to 1.27 km in the first 8.5 km of the atmosphere; and the vertical spacing was maintained at 1.27 km for heights above 8.5 km. The total depth of the model atmosphere was set to 21 km. The urban model (TEB) scheme was coupled within the surface energy balance scheme only over the inner-most region that covered the Indianapolis urban area Model initializations Model initialization and boundary conditions were developed as horizontally inhomogeneous fields following with forcing data from the National Centers for Environmental Prediction (NCEP) and the Global Data Assimilation System (GDAS) (Kalnay et al., 1996). All simulations were performed for 72 h from 21 June to 23 June Soil moisture was

68 55 initialized at 35% of soil moisture saturation and followed the GDAS fields. Soil temperature is also prescribed as default GDAS value. Soil texture variations followed the USGS land topography datasets. The urban land use was developed by reviewing the high resolution Google Earth photos, street maps, and was loosely based on the coefficients available for Washington DC area based on the work of T.Nobis (Personal communication, 2006). The variables defined the urban morphology and are shown in table 3.2. Kuo s Cumulus parameterization scheme was used for domain 1 and 2, while the inner most domain had explicit convection. All other model parameter choices were either set to default or as described in Chapter Results RADAR level II base reflectivity images were collected and analyzed from May 2000 through September A total of 81 thunderstorm cases covering 61 days were studied, and summarized in Table 3.3. This table includes the cases Pyle (2007) analyzed from 2000 to Intensity and distribution of thunderstorms are analyzed. In the first 5 years of the 61 cases investigated, 41 occurred during the day and 20 were at night. Of the 41 daytime events, 29 or (71%) of the storms changed composition and 6 or (30%) of the nighttime events changed composition. In the last 2 years from 20 cases, 13 were for the daytime conditions and 7 occurred at night. Of the 13 daytime events, 9 or (69%) of the storms changed composition and 6 or (85%) of the nighttime events changed composition. The two separate analyses are consistent in statistics. (This increase for the night is partially because of the small amount of nighttime cases from 2005 to 2007). Overall, Table 3.4 shows the statistics of the storm urban area interactions: Fifty of 81 (62%) storms showed obvious morphology alteration when passing over the Indianapolis urban region. Of the 50 storms which changed morphology, half (25) broke up when over the urban area. 8 out of

69 56 25 broken storms became intensified just after passing over the urban region, and 14 more storms became intensified without breaking up following a track over the Indianapolis region. Five storms appear to have been initiated due to the urban rural boundary, and 6 storms changed their distribution or dissipated when passing over the Indianapolis urban area. These results point to the previously posed hypothesis: thunderstorms break up or split when passing over the Indianapolis urban region. This phenomenon decreased the precipitation over metropolitan Indianapolis, and potentially increased the rainfall downwind of the urban area. This could potentially explain the low rainfall value over Indianapolis and higher values in vicinity as seen in the climatological analysis (Fig.3.2). Fig.3.4 shows Level II radar images of 4 representative thunderstorm cases. The data used in generating this figure was obtained from National Climatic Data Center (NCDC) and plotted using the GrLevel2 Radar data Software Package (available online at and respectively). These storms occurred on 14 June 2005, 22 June 2006 (two thunderstorms) and 18 July Fig.3.4 a shows a single cell moving from west to east Indianapolis, and splitting into two parts over the urban area. The two parts combined again after passing the Indianapolis area. Fig.3.4 b is a time sequence valid for Z 22 June. The 3-D structure of the thunderstorm at 1849 Z (Fig.3.4 c) shows clearly that almost no base reflectivity occurred over the city, indicating an absence of liquid water. Fig.3.4 d shows another thunderstorm passing over Indianapolis on 22 June. This storm will be discussed in detail ahead. Fig.3.4 e shows the thunderstorm on 18 July 2006 and indicates that a storm was initiated downwind of Indianapolis, and then intensified over the southern part of the city. The radar analysis shows clear indications regarding changes in thunderstorm morphology as they interact with urban landscapes.

70 Case study 22 June 2006 Thunderstorm Case: Synoptic conditions: Fig.3.5 shows the surface map for 22 and 23 June 2006 and a 500 hpa map for 22 June. A cold front over the Great Lakes and a castoff of a synoptic low pressure system in Canada brought showers and light rain to northern Indiana. As the cold front progressed to the east, a trailing stationary front entered Indiana, resulting in heavier rains throughout the day over a large regions including Indianapolis. The region saw small drops in temperatures as a result of the passing disturbance. Air pressure remained relatively unchanged as the stationary front passed, mostly due to high pressure following the system. From radar reflectivity images, two thunderstorms were seen passing Indianapolis. Of these two storms, one was from 18Z to 19Z on 22 June (Fig.3.4 b) and the other one is from 00Z to 01Z on 23 June (Fig.3.4 d), which means they both occurred on 22 June from noon to the afternoon local time. The Indianapolis airport recorded mostly westerly and southwesterly winds on 22 June, and radar images also show the thunderstorms moving from west to the east. At 1818Z, some reflectivity is seen to the west of Indianapolis, but it is not strong. Then after 16 minutes at 1834Z, a well-organized southwest to northeast squall line appears to the west of Indianapolis and close to the urban area. At 1847Z when the thunderstorm is just passing over the urban area, a very limited liquid water zone clearly shows over downtown Indianapolis, while the surrounding area is covered with a highly concentrated cloud layer (high base reflectivity area). When the thunderstorms leave Indianapolis at 1909Z, and become intensified and even spread out over the downwind area of the city. The second thunderstorm shows similar evolution, the difference being that it moved over the city from a southwesterly direction. At 2333Z of 22 June, both thunderstorms were well-organized and picked up strength before approaching Indianapolis. Interestingly when the storm begins to pass over

71 58 Indianapolis, the system more or less disappears, and then reappears, when it has passed the Indianapolis urban region (2342Z). Just after passing the Indianapolis urban area at 0015 Z on 23 June, the thunderstorm intensified again north and northeast of Indianapolis. The northern Indianapolis city area was expected to get higher precipitation. These storms are representative of typical thunderstorm case days for our study (shown in Table 3.2), and this is why we selected this event for simulation. Observations and validation of simulation for this event: The June precipitation map (Fig.3.6) obtained from the Midwestern Regional Climate Center/Indiana State Climate Office shows widespread rainfall over Indiana on that day. The southern part of Indiana received more precipitation than northern part of the state: the precipitation amount shows at about a half-inch in the northern region, while it is more than 1 inch in the south. Due to the coarse resolution, the color representing 0.8 to 1 inch covered Marion County. From this map we can validate the simulation of the precipitation distribution. The relatively finer precipitation distribution map was obtained from the CoCoRaHS data available from Indiana state climate office (Fig.3.7). (CoCoRaHS stands for the Community Collaborative Rain, Hail, and Snow Network). Fig.3.7 shows CoCoRaHS precipitation reports on 23 June 2006 over Marion County. Most observations ranged from 0.78 to 1.07 inches. Two of the reports, however, show a lower amount of precipitation, and one of these is the closest (of the ten reports) to the Indianapolis downtown area. This observation along with radar reflectivity further corroborates the hypothesis that urban regions experience less rain than the adjacent rural regions. A sensible heat flux and latent heat flux validation was performed by comparing the simulations with AmeriFlux Network observations. The AmeriFlux network was established in The network provides continuous observations of ecosystem level exchanges of CO 2, water, energy and

72 59 momentum spanning diurnal, synoptic, seasonal, and inter-annual time scales, and is currently composed of sites from North America, Central America, and South America ( Data from Bondville, Illinois, about 180 km to the west of the Indianapolis urban region are used. Fig.3.8 a shows the observed and simulated sensible heat flux; they matched well, both in variations and absolute values: before and after sunrise/sunset they both showed negative values and reached maximum value in one day, at approximately 3 pm. The variations of latent heat flux (Fig.3.8 b) were similar, while the simulated daytime latent heat flux was significantly higher than the observations. The difference was about 100 W m -2. Overall, simulations of surface fluxes over our domain were reasonable, except the latent heat flux value was overestimated by the model at the Bondville site. There is another AmeriFlux site in southern Indiana. Results (not shown) show larger difference between observations and simulations mainly because of errors in rainfall locations. While no sounding data is available over Indiana, the nearest sounding site is located in Wilmington, Ohio. We chose the standard observations at 12 Z 22 June The observation spawned from 981 to meters at 86 levels, and the simulations had 35 layers from 24 to meters high. We compared air temperature in Fig.3.9, and both air temperatures match well. The observational dew point temperature is close to the air temperature line at heights of approximately 2000 and 4000 meters, which indicates of clouds. Ground surface air temperature was validated by comparing it with NARR reanalysis data. Fig.3.10 a-b shows the NARR and RAMS simulated temperature field on 22 June Both plot temperature gradients ranging from 298 K at Lake Michigan to 293 K southeast of Indiana. The air temperature is higher in the NARR data, and the average air temperature is about 1 K higher than our simulation. Near the Indianapolis area, the surface air temperature is about 1 K higher in the NARR fields than in the control simulation. Thus overall, the model results are broadly realistic of the regional conditions as revealed by point observation.

73 Effect of urban environment on thunderstorm evolution In this chapter, we use RAMS coupled with the explicit urban model TEB to further study the effect of the Indianapolis urban area on precipitation Temperature In Fig.3.11, we can see the averaged ground surface temperature with and without the urban model on 22 June and 23 June Over the Indianapolis urban area, a warm region is visible due to the use of the urban landscape. Near downtown Indianapolis (39.7N, 86.18W), the urban heat island intensity is highest, which is about 1 K higher using TEB than without using TEB. The entire urban area shows warming, and the warming effect decreases from the city center at 0.8 K to surrounding regions at 0.2 K; 30 km from downtown Indianapolis, the urban heat island effect is negligible. The vertical effect of UHI affect is seen up to 400 m above the ground. This value is reasonable in a mid-latitude area compared to the 600 meter measure in tropical regions (Freitas, 2006). The cross-section shown in Fig.3.12 shows the distance from the surface to the 50 meters height, which is about the order of magnitude for the atmospheric surface layer height, the UHI is seen with about 1 K air temperature higher. Then from 50 meters to 400 meters, the warming effect varies from 0.6 K to zero. The upper atmosphere as expected does not show an obvious temperature change resultant from application of TEB. Ground surface air temperature also has a reasonable diurnal variation. Fig.3.13 indicates that the warming effect is most obvious from 18Z, which is in the afternoon. In the afternoon air temperature is 1.5 K warmer due to urbanization mostly because the albedo is lower in urban areas than in rural areas, and more radiation is absorbed by the urban area. Then the higher heat capacity can hold the heat until nighttime. The warming effect is

74 61 relatively small in the early morning. Presence of clouds and the rain also impact the night time temperature and hence the diurnal variation in the urban heat island Surface fluxes Within the Indianapolis urban region (Fig.3.14), the average sensible heat flux on 21 and 22 June increased from 20 W m -2 to 80 W m -2, especially over downtown where TEB was used. The average sensible heat flux increase due to TEB was about 40 W m -2 over Indianapolis metropolitan. The fluxes over surrounding suburban area experienced limited change. The diurnal variation during the period reached its maximum value around 18Z for both days which is 12 PM for local time (Fig.3.15). But the sensible heat flux change due to the presence of the urban area reached its peak point at 0Z on 22 June (Fig.3.15). Results from the rapid decrease in the control run s sensible heat flux in the afternoon. At 0Z, the control run had a slightly negative sensible heat flux possibly due to cloudiness and rainfall, while the urban area maintained a positive sensible heat flux. This situation persisted until 12Z when the control run became positive again, and their difference reached its minimum within one day. Similar to sensible heat flux, the latent heat flux changed significantly over the Indianapolis urban area, with the difference in the latent heat flux being a decrease that was simulated in the city (not shown). The average simulated latent heat flux for two days was 240 W m -2 over the urban area, which decreased about W m -2 after coupling with TEB to 190 W m -2. The diurnal variation is shown in Fig.3.16 and indicates that the latent heat flux has its maximum value at 18Z, and is steady at 0 during the nighttime. Since much less evapotranspiration (ET) occurs in the city, latent heat flux is proportionately much less. Our simulation results showed that the latent heat flux decreased with more than 200 W m -2 when using the urban model, as

75 62 compared with the normal simulated value of 600 W m -2. As discussed in the previous section, the model maybe overestimates latent heat flux. However, since the variation of latent heat flux was well simulated, we believe that the decrease in urban area latent heat flux is realistic. The horizontal soil moisture and flux gradients are critical for conditioning the pre-storm environment. Existing research (Zhong et al. 1996) indicates that the soil moisture gradient can significantly enhance the low level jet in a convection system. Due to the predominantly westerly wind direction, the convective system moved from west to east. The sensible heat flux and latent heat flux gradients averaged between 39.7N and 39.9N from west to east were analyzed for 3 hours before both squall lines had passed the urban area. Fig.3.17 shows that the second storm had higher gradients in both sensible heat flux and latent heat flux. The sensible heat flux gradient was 1.5 W m -2 km -1 near the urban area, and in the second thunderstorm, the latent heat flux was about 7 W m -2 km -1 while compared to the first thunderstorm, sensible heat flux and latent heat flux were 0.5 W m -2 km -1 and 1 W m -2 km -1. Obviously, environmental conditions during the second thunderstorm were more suitable for convergence Moisture and rainwater mixing ratio The rainwater mixing ratio is an important variable which shows the potential available water that can be converted to rainfall. Because the air temperature increased significantly after coupling with TEB, an increased saturation mixing ratio over the Indianapolis urban area was expected. Fig.3.18 shows the diurnal rain water mixing ratio during the simulation period. The 3-D radar base reflectivity map indicates that most high storms area extend from the surface to 4000 m (Fig.3.18). The model results showed a maximum available rain water mixing ratio at 09Z on 21 June and 03Z on 23 June, with the water mixing value about 0.2 g kg -1 (This

76 63 represents two thunderstorms passing through, which will be discussed further ahead). They both decreased 0.05 g kg -1 after coupling with TEB, and the decrease ratio is approximately 25%. This decrease is believed to be due to the urban heat island-induced air temperature increase. The average relative humidity from the surface to 4000 meters altitude of simulation also decreased due to coupling with TEB throughout the simulation period (Fig.3.19). Meanwhile, because thunderstorms had moved in from other places, and the urban area does not release significant water vapor, the total water vapor amounts over Indianapolis changed due to different land cover conditions. Simulation results also support this observation (not shown) which measured a slight decrease in the total water vapor amount (less than 1 mm out of 50 mm) due to a decrease in water vapor flux. The following process partially explains the decrease in the rain water mixing ratio: air temperature increase saturation water mixing ratio increases, although the actual water mixing ratio stays the same more water stays in a vapor condition instead of a liquid condition rain water mixing ratio decreases less precipitation Precipitation Fig.3.20 a indicates the first thunderstorm s decrease in accumulated precipitation due to the urban area, an amount of approximately 3 mm, the decreased portion being about 7% of the precipitation amount in control run. However, the situation is different in the second thunderstorm simulation, where the precipitation increased due to urban representation in the model. The increased amount was about 7 mm, and the increased portion is 10%. The first thunderstorm is consistent with the rain water mixing ratio condition, though simulation did not capture the thunderstorm event time well, and most of the rainfall occurred at 12Z 12 June. The second thunderstorm was wellsimulated in 0-3Z 23 June, but the reason the precipitation was intensified

77 64 over the urban area remains a problem which will be further discussed in a later section. Our work related to Mumbai precipitation discussed in Chapter.2 which suggested urban area can increase the precipitation upwind and downwind urban region. Fig.3.20 b shows the precipitation over the upwind direction of the Indianapolis urban area. The precipitation difference between the with and without urban representation is small, though using TEB slightly increased the precipitation amount (1 mm) over the upwind direction of the urban area. As previous studies (e.g. Shepherd et al., 2002) demonstrated, downwind precipitation over a large urban area is higher than upwind precipitation. In our results (Fig.3.20 c), urban area precipitation does not change significantly during the first thunderstorm, and increases by about 1-2 mm. For the second thunderstorm, downwind precipitation increased around 5 mm. Spatially, precipitation has the same distribution as our previous analysis. In the first thunderstorm (Fig.3.21 a), precipitation decreased by approximately 3 mm over the Indianapolis downtown area, and also decreased towards the southeast direction of the urban region which experienced a precipitation decrease of almost 4 mm. In the north-east direction of urban Indianapolis (the downwind direction) precipitation increased from 2 mm to 4 mm. In the second thunderstorm (Fig.3.21 b), precipitation increased both over the urban area and in the downwind direction of the Indianapolis urban area; the amount of the increase was as much as 20 mm, about 25% of the total precipitation amount. Over the upwind direction over Indianapolis, the precipitation amount shows a 10 mm decrease.

78 Vertical velocity, convergence, and CAPE Vertical velocity and convergence are two important variables to the study of dynamic processes in the convection system. For consistency, we analyzed cross-section vertical velocity averages between 39.7N and 39.9N, and between 87.5W and 85W. Fig.3.22 shows updrafts for the first thunderstorm with a downdraft over the Indianapolis area in both simulations (with and without TEB). The downward movement (due to TEB) increased from 0.5 m s -1 to 0.7 m s -1. Indianapolis and its surrounding areas nearby showed an upward movement at 0.1 to 0.2 m s -1. In the downtown area, the vertical wind was increased with TEB at 0.15 m s -1 at a height of 1000 meters. Considering that the updraft and downdraft persisted over a small area, convergence and divergence should occur. Conditions during the second thunderstorm were different due to a high horizontal flux gradient; over the urban area and downwind, the vertical movement was upward at 0.02 m s -1 with a downward movement in an upwind direction at 0.04 m s -1. Urban representation caused an increase in updrafts over the urban area and downwind urban area by 0.04 m s -1 near the ground. The enhanced updrafts over urban area and downdrafts over the upwind urban area caused localized circulation. Convergence can enhance convection, and convergence also impacts the vertical velocity (Fig.3.23). For the first thunderstorm, a divergence occurred over the Indianapolis urban area and convergence downwind Indianapolis. After coupling with TEB, the divergence over the urban area increased by 10-5 s -1 and convergence also increase by 10-5 s -1. This change appears to decrease the precipitation over the urban area and increased the precipitation downwind urban area. In the simulation of the second thunderstorm, the convergence conditions in the control run were negative (divergence) near Indianapolis and its downwind area. Coupling with TEB caused the convergence to increase to 0.3 x 10-5 s -1 both horizontally and vertically.

79 66 The CAPE values near Indianapolis do not show significant change due to TEB. During most of the simulation period, the urban area decreases CAPE compared to the simulation without the urban model Sensitivity experiments Due to the complex mechanisms of urban atmosphere interactions, the identification of the detailed effects of urban energy balance on micro-scale physics is difficult. In an effort to explore these physical processes and further understand urban thunderstorm interactions, we designed process based experiments to understand the effect of roughness length, albedo, and urban area size on the thunderstorm simulation Effect of Roughness change over the urban area Due to urban scheme in downtown Indianapolis is higher when considering the TEB run, changes in roughness length can change the near surface wind and hence change the urban canopy exchange. A sensitivity experiment was undertaken by reducing the roughness of the explicit urban representation from approximately 1.5 m to 0.1 m near the downtown Indianapolis area. As shown in Fig.3.24, following this increase of the roughness setup over the Indianapolis urban area from a small (0.1 m) to a normal (1.5 m) value, precipitation caused by the first thunderstorm showed an increase over the urban region and downwind of Indianapolis (northeast of Indianapolis), and decrease over upwind of urban area (southwest of Indianapolis). Precipitation caused by the second thunderstorm showed a similar change of precipitation, but the urban area presented different feedback due to increased roughness length; the upwind urban area received less rainfall than the downwind urban area. The difference between the first and the second thunderstorm was that

80 67 the positive rainfall change area moved in the downwind direction for about half a degree. The difference in the rainfall amount was about 2 mm in the first thunderstorm and 15 mm rainfall in the second thunderstorm, which was much smaller than the difference with and with out TEB. Therefore, roughness does not seem to be the main reason to change rainfall significantly by urban representation. After analysis, we found that roughness changed low level convergence by decreasing the low level horizontal wind. From the ground surface to 4 meters (Fig.3.25), wind speed was decreased by about 1 m s -1 due to roughness, while vertical wind speed was increased by roughness. Both thunderstorms showed a significantly increasing low level convergence over the urban area and downwind of Indianapolis. The first storm showed a 0.8 x 10-5 s -1 increase and the second storm showed a 3 x 10-5 s -1 increase of convergence Effect of Albedo UHI is the main feature of the urban area, and is generally represented by the low albedo in slab land surface models. This experiment was performed to check the potential effect of the albedo setup in TEB as well as UHI. We increased and decreased the albedo over the wall, roof, and road by 20% so that we could study the albedo effects. Air temperature over the Indianapolis urban area increased about 0.3 K after the albedo decreased from above control conditions to below normal conditions by 20%; this amount is smaller than the temperature change due to explicit urban representation. Based on the previous discussion of the temperature-dependent saturation water mixing ratio change, temperature increase may cause precipitation decrease. Also, because the convergence did not change significantly over the urban area purely due to the decrease in

81 68 albedo, we speculate that the precipitation decrease was due to a decrease in the rain water mixing ratio caused by an increase in air temperature. Fig.3.26 shows the spatial distribution of precipitation change caused by albedo change. We can see that both the upwind and downwind precipitation is increased by an albedo decrease, but the precipitation over the urban area of Indianapolis decreased by more than 10 mm. This precipitation decrease could help explain the climatological low precipitation over Indianapolis urban areas Effect of urban expansion The size of urban areas is an important issue for urbanization impact and can be used to assess future weather conditions. We designed an experiment that expanded the Indianapolis urban area to twice its current size. The expanded areas are adopted residence areas with a building height setup at 8 m, as expected with sprawl. The expanded area changed the precipitation distribution near Indianapolis: the large urban area received less rain than is received over the current Indianapolis city center. Instead southern Indianapolis may receive more rainfall after the city expansion. Fig.3.27 a-b shows that both thunderstorms caused decreased precipitation after the urban area expansion. The first thunderstorm precipitation decreased by about 2 mm and increased by 2 mm in the southern part of the Indianapolis area. The second thunderstorm precipitation changed more significantly; it decreased by 10 mm over the city and more than 15 mm over the downwind of urban area (north east of Indianapolis). Precipitation also increased over southern Indianapolis by more than 15 mm. Fig.3.28 shows the time series of simulated precipitation, and it indicates that the change in precipitation is consistent throughout the simulation period.

82 69 The sensible heat flux and latent heat flux patterns are similar to the current urban size condition and have a wider range between high sensible heat flux and low latent heat flux. The convergence values also support the precipitation change. The expansion of the urban area may cause decrease in the low level (from ground surface to 5000 m) convergence condition over the urban area, while it increased over the south Indianapolis area (not shown) Conclusions After analyzing the radar base reflectivity, the vertical 3-D radar image statistics and simulations by RAMS coupled with TEB, we draw the following conclusions: About 62% of the thunderstorms between 2000 and 2007 showed modified behavior when passing over the Indianapolis urban area. The effect of Indianapolis urban area on thunderstorm morphology near the urban area is complex, but statistics showed that the thunderstorms breaking up over the city or intensity decrease is important features among other effects. This change is one explanation of the climatologically negative precipitation over Indianapolis compared to its adjacent areas. RAMS can reasonably simulate the thunderstorms over Indianapolis urban region when setup with proper schemes. When coupled with the explicit urban model TEB, mesoscale model performance was better. Simulation results show that the 22 June 2006 thunderstorm case was well simulated by RAMS; RAMS main bias lies in the overestimation of the precipitation amount. The urban area increases the downwind precipitation for the thunderstorms cases, and it shows variable effects on precipitation over the Indianapolis urban area. Upwind precipitation does not change significantly due to urbanization, though the convection is better organized and may impact the downwind precipitation.

83 70 The identification of different urban factors which play a role in changing precipitation is difficult, but sensitivity experiments suggest that roughness can increase thunderstorm convergence by decreasing low level wind velocities; the UHI can decrease the rainwater amount, and the size of the urban area might increase the precipitation over downwind region.

84 Figure 3.1 Land use land cover map showing Indianapolis urban region and central Indiana. 71

85 72 Average annual precipitation, Average annual precipitation, a) b) Figure year normal average annual precipitation: a). From 1961 to 1990; b). From 1971 to 2000.

86 Figure 3.3 Model simulation domains, TEB is adopted over the inner most domain. 73

87 Z 0024Z 0029Z 0038Z 0042Z 0051Z a) Radar base reflectivity images on Z 1845Z 1849Z 1853Z 1902Z 1931Z b) Radar base reflectivity images on

88 75 c) 3-D radar base reflectivity image for 1 st thunderstorm on Z 2342Z 2350Z 2354Z 0002Z 0015Z d) Radar base reflectivity images on

89 Z 2244Z 2257Z 2314Z 2339Z 2352Z e). Radar base reflectivity images on Figure 3.4 Sample radar base reflectivity images for 4 thunderstorm cases.

90 77 a) b) c) Figure 3.5 a) Surface weather map for 22 June 2006; b) surface weather map for 23 June 2006; c) 500 hpa map for 22 June 2006.

91 Figure 3.6 Accumulated precipitation (inch) map over Indiana from 21 to 23 June

92 Figure 3.7 CoCoRaHS recorded precipitation (inch) on 23 June

93 80 a) Sensible heat flux over Bond Ville on June , x-axis is time with 24 hour format (from 0 to 2400) and y-axis is sensible heat flux (W m -2 ). b) The same with last figure but for latent heat flux (W m -2 ). Figure 3.8 Observed surface fluxes in Bondville, Illinois, AmeriFlux site: a). sensible heat flux (W m-2); b). latent heat flux on June (W m-2).

94 sounding simulation dew point Figure 3.9 Observed and simulated vertical temperature ( C).

95 82 NARR temperature Simulated temperature Figure hpa air temperature (K) from North American Regional Reanalysis (NARR) data and model.

96 83 Figure 3.11 Horizontal surface air temperature (K) averaged from 22 to 23 June Figure 3.12 Vertical cross-section of air temperature (K) averaged from 22 to 23 June 2006.

97 84 Figure 3.13 Diurnal surface air temperature (K) over Indianapolis from 0Z 21 June to 0Z 24 June. Figure 3.14 Average sensible heat flux (W m -2 ): LEAF2, TEB and their difference (TEB control).

98 85 Figure 3.15 Diurnal sensible heat flux (W m -2 ) variation over Indianapolis (short dashed line: control; long dashed line: TEB run; solid line: TEB control) Figure 3.16 Same as Figure 3.15 but for latent heat flux (W m -2 ).

99 86 a) b) c) d) Figure 3.17 Longitudinal heat flux gradient (W m -2 ) averaged from 39.7 N to 39.9 N: a) Sensible heat flux gradient in first thunderstorm; b) Sensible heat flux gradient in second thunderstorm; c) latent heat flux gradient in first thunderstorm; d) latent heat flux gradient in second thunderstorm (open circle= control run with LEAF2, close circles= TEB run).

100 Figure 3.18 Rain water mixing ratio (g kg -1 ) over the Indianapolis urban area from 0Z 21 June to 0Z 24 June. (Short dashed line= control; long dashed line= TEB run; solid line = TEB control) 87

101 88 Figure 3.19 Same as Figure 3.18 but for relative humidity (%) over Indianapolis (short dashed line= control; long dashed line = TEB run; solid line = TEB control)

102 89 a) b)

103 90 c) Figure 3.20 Precipitation time series (mm) throughout the simulation period: a). Over the Indianapolis area; b) upwind of Indianapolis; c) downwind of Indianapolis. (Short dashed line= control; long dashed line=teb run; solid line= TEB control)

104 91 a) b) Figure 3.21 Spatial accumulated precipitation (mm): a) First thunderstorm; b) Second thunderstorm.

105 92 a) b) Figure 3.22 Vertical velocities (m s -1 ): a). First thunderstorm; b) Second thunderstorm

106 93 a) b) Figure 3.23 Same as Figure 3.22 but for convergence (10-5 s -1 ).

107 94 a) b) Figure 3.24 Precipitation (mm) for low roughness length conditions and explicit urban representations: a). First thunderstorm; b). Second thunderstorm.

108 95 a) b) Figure 3.25 Vertical cross section for convergence (10-5 s -1 ) averaged from 39.7N to 39.9N for low and normal roughness length condition: a). First thunderstorm; b). Second thunderstorm

109 Figure 3.26 Accumulated precipitation (mm) in the simulation with high albedo condition, low albedo condition and their difference. 96

110 97 a) Figure 3.27 Accumulated precipitation (mm) for the two thunderstorms before and after the synthetic expansion of Indianapolis urban area: a). First thunderstorm; b). Second thunderstorm. b)

111 98 a) Figure 3.28 Precipitation time series averaged over: a). Indianapolis urban area and b). southern part of Indianapolis (long dashed line with closed circles= current explicit urban representations; short dashed line with open circles= expanded urban area; solid line with open circles= differences) (mm). b)

112 99 Table 3.1 Indiana climate normals ( ). Indiana Climate Normals ( ) Month Mean Max. Min. Precipitation Temperature Temperature Temperature Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

113 100 Table 3.2 key variables to describe Indianapolis urban morphology Key variables Value Average downdown urban 15 Building height (m) Building width (m) 3 Z 0 (m) 1.5 Traffic sensible heat release 8 (average) (Wm 2 ) Traffic latent heat source (Wm 2 ) 0 Industrial sensible heat source 15 (average) (Wm 2 ) Industrial latent heat source 10 (average) (Wm 2 ) Constant temperature inside 295 building (K)

114 101 Table 3.2 Analysis of thunderstorms from May 2000 to July Date Storm Orientation and Propagation Synoptics Time of Event Urban Notes 9 May 2000 NE/SW Line Moving ENE CF + S/W Day None 12 May June 2000 NE/SW Line Moving E 1) Scattered convection moving NE 2)NE/SW Line Moving NE 1) Warm sector convection 2) CF CF Night None Both Day 1)Some break up of line and convergence after Indy 2)Line converges after passing Indy 20 June 2000 Two events: NE/SW Line(s) Moving ENE 1) Warm sector line 2) Pre-frontal line 1) Day 2) Night 1) Skirts around the city 2) None 2 August 2000 NE/SW Moving ESE CF Day Line breaks before and over Indy 6 August August August July 2001 E/W line moving ESE Two events 1) MCV moving SSE 2) E-W moving SSE forms near and after (south) of INDY Two events 1) Strong E-W cluster connects with second event 2) Bow like feature Both events moving ESE Two events 1) NW flow strong bow echo cell 2) MCS Pre-frontal trough (MCS outflow) Night None 1) MCV 2) CF WF 1) Night 2) Day Both Day 1) Stationary surface boundary 2) MCS Both Day 1) None 2) Tail end of line intensifies in and east of Indy 1) None 2) Minor splitting 1) Bowing of cell 2) None 18 August May May 2002 NE/SW Line Moving ESE CF + S/W Day Break and discontinuity Two events 1) Multicell cluster moving ESE 2) Scattered strong cells moving ENE NE/SW Line Moving E WF CF 1) Day 2) Night Day 1) None 2) Skirts downtown Indy Intensification near/over Indy followed by break up after 31 May June 2002 ENE/WSW line Moving SE CF Day Break up of line NE/SW Small line moving E CF (warm sector) Night Some dissipation of line

115 August 2002 Multi-cell cluster Moving E CF (warm sector) Day Weak intensification 1 May 2003 N/S Bow Line CF (pre-frontal convergence zone) Day Break up and intensification 6 May 2003 Moderate precipitation with embedded strong cells SF Night Break up and intensification 10 May ) MCV moving E 2) NE/SW line moving NE 1) MCV 2) CF 1) Day 2) Night 1) Intensification 2) None 14 May May 2003 MCS moving ESE MCS Night None Multi-cell cluster moving SE S/W trough Day Break up and intensification 4 July ) NE/SW Line moving SE 2) E/W Line moving SSE S/W 1) Day 2) Night 1) None 2) Dissipation 5 July 2003 E/W Multi-cell cluster moving ESE S/W Day Break up 8 July ) Scattered cells moving E 2) NE/SW Multi-cell cluster moving E 3) NE/SW Line moving ESE Pre-existing E/W surface boundary with S/W All 3 Day 1) Intensification 2) Break up 3) Dissipation 9 July ) Multi-cell cluster moving E 2) E/W line moving S SF Both Day 1) None 2) None 11 July 2003 NE/SW line moving SE Vort Max with small low level convergence Day Break up 20 July 2003 MCV moving SE MCV (need synoptic) Night None 18 May ) Single cell moving ENE 2) Two cells moving ENE 3) Multi-cell cluster moving E S/W All 3 Day 1) Dissipate 2) Break up 3) Intensification 23 May May 2004 Multi-cell cluster moving E CF Night None N/S bowed line moving ENE WF Night None 30 May ) Supercell moving NE 2) NE/SW line moving ENE (-) tiltled trough with 989mb surface low in MN (warm sector) Both Day 1) None 2) None 10 June 2004 Multi-cell moving E SF north of region Day Intensification 13 June 2004 NE/SW line moving ENE NE/SW surface boundaries with low pressure north Day Break up and intensification 3 July 2004 NW/SE line moving NE CF (warm sector) Day None

116 July 2004 MCS moving SE MCS Night Break of NE side of MCS 21 July ) NE/SW line moving SE 2) Multi-cell line moving SE Surface boundary from old MCS Both Night 1) None 2) None 11 May 2005 E/W line moving ESE SF Day Break up 13 May ) Multi-cell cluster moving NE 2) NE/SW line moving E S/W (prefrontal trough) Day, Night 1) Intensification 2) None 13 June 05 Single isolated cell Moving NE CF (warm sector) Day Intensification and break up 28 June ) Single cell possibly moving E 2) NE/SW oriented cluster moving ESE S/W (upper level low) Both Day 1) Possible urban initiation 2) None 30 June 2005 Single cell moving east CF (warm sector) Day Possible urban initiation 20 July 2005 NE/SW oriented cluster moving SE Pre-existing outflow boundary + WF Night Break up 21 July 2005 NE/SW oriented MCS MCS Night None 26 July ) Scattered convection moving NE 2) NE/SW oriented line moving ESE 1) CF (warm sector) 2) CF Day, Night 1). None 2). None 19 Sep 2005 Scattered convection moving E CF (warm sector) MCS Day, Night 1).Break up 2). Initiation 06 Nov 2005 NE/SW moving E CF (warm sector) Night Break up at the beginning 31 Mar ) NE/SW moving SE 2) Scattered convection CF (warm sector) MCS Day Intensification and possibly initiation 03 Apr 2006 NW/SE Moving NE CF (warm sector) Prefrontal trough Night Intensification 30 May 2006 Scattered convection, Moving NE CF Night Urban initiation 1 Jun 2006 Scattered convection, Moving NE CF Day Possible urban initiation and intensification

117 Jun 2006 NE/SW moving south CF Day Possible intensification 22 Jun 2006 NE/SW moving east Occluded front Day Break up and intensification 18 Jul 2006 Scattered convection Moving South CF Day Possible urban initiation 1 Dec 2006 Scatted convection Moving SE CF MCV Day None 13 Feb 2007 Wide spread snow MCS Day None 15 Mar 2007 Scattered convection moving east Two CF Day None 28 Mar ). Scattered convection 2). Moving to NE SF Night Possible intensification 1 Apr 2007 N/S thunderstorm moving east CF (warm sector) Night Break up and intensification 26 Apr 2007 N/S thunderstorm moving east MCS, WF Day Break up 3 Jun 2007 Scattered convection Moving east Cold front Day, night 1). None 2). None 21 Aug 2007 NE/SW Line Moving EAST CF (warm sector) MCS Day Intensification 9 Sep 2007 Scattered convection moving north east CF (warm sector) Day Break up

118 105 Table 3.3 Summary of thunderstorm cases. Number of thunderstorms examined 81 Number of thunderstorms showing obvious 50 urban effect when passing over urban area Number of thunderstorms showing no 31 obvious change when passing over urban area Number of thunderstorms breaking up 25 Number of thunderstorms intensifying after breaking up Number of thunderstorms intensifying without breaking up Number of thunderstorms initiating over urban area Number of thunderstorms changing or dissipating over urban area

119 106 CHAPTER 4. CONCLUSIONS, LIMITATIONS AND FUTURE WORK 4.1. Summary and conclusions The effect of land surface processes, especially urbanization, on precipitation patterns in vicinity of urban regions was investigated using observational data and mesoscale model simulations. High resolution RAMS coupled with TEB could reasonably simulate the precipitation events in both tropical and mid-latitude areas. Model runs with and without TEB experiments as well as sensitivity experiments with TEB variables helped us better understand the atmospheric feedback near the urban region. Due to different moisture sources and precipitation types, the mid-latitude urban area and tropical urban area each have varying effects on the atmosphere: the midlatitude urban area might change the precipitation distribution while the tropical urban area might change the precipitation as well as the total amount. These variations could also be related to the coastal versus inland urban feedbacks and need to be further investigated. Over the coastal city of Mumbai, the urban area interacted primarily with sea surface temperature in creating temperature gradients that impacted the convergence and precipitation patterns shown in our study. In the case study of 26 July 2005, coupling with explicit urban model can increase the rainfall amount upwind of Mumbai city. For Mumbai region, the Arabian Sea provided the moisture source and the urban area provided mesoscale convection rigger that can significantly increase precipitation by enhancing the land sea breeze convergences. Over the Indianapolis region, our results indicated that the precipitation increase downwind of Indianapolis was largely due to urban feedbacks, as

120 107 confirmed by both observations and simulations. This phenomenon is linked to the intensification of thunderstorms when they moved over the urban area. Prevailing wind in the summer over Indianapolis is most westerly, so rainfall amounts east of Indianapolis can be expected to be greater than that west of Indianapolis. Though climatological precipitation data indicate decreasing precipitation over the city due to the urban area, some of the mesoscale complex interactions processes were not well simulated in the model. UHI can increase the water mixing ratio in saturation condition, thus decrease the rain water mixing ratio. High horizontal flux and temperature gradient can led to stronger convergence and vertical movement, but the location of this process might also be affected by the urban area. Both in tropical and mid-latitude areas, urbanization might cause more extreme weather, such as heavy precipitation and severe thunderstorms Limitations and future work plan Due to the rapid development of urban areas and the limitations of remote sensing data, urban morphology datasets used in the models are not adequately updated. Further, some data have been created based on our subjective interpretation of various data resources. For both Mumbai and the Indianapolis region, one case study alone cannot completely explain the general results. More case studies perhaps with simpler models may provide additional insights. More reliable radar and satellite-derived data need to be assimilated within modeling systems to develop ensemble simulations. Future assessments will need to consider more detailed rural-urban heterogeneity and more detailed model processes such as photosynthesis and hydrological feedbacks (Niyogi et al., 2006).

121 108 LIST OF REFERENCES Arnfield, A.J. 2003: Two Decades of Urban Climate Research: A Review of Turbulence, Exchanges of Energy and Water, and the Urban Heat Island. Int. J. Climatol., 23, Avissar, R. and R.A. Pielke, 1989: A parameterization of heterogeneous land surfaces for atmospheric numerical models and its impact on regional meteorology. Mon. Wea. Rev., 117, Bohra, A.K., S. Basu, E.N. Rajagopal., G.R. Iyengar, M.D. Gupta, R. Ashrit and B. Athiyaman, 2006: Heavy rainfall episode over Mumbai on 26 July 2005-Assessment of NWP guidance. Current Science, 90, Bounoua, L., R. Defries, G. J. Collatz, P. Sellers, and H. Khan, 2002: Effects of land cover conversion on surface climate. Clim. Change, 52, Braham, R.R., and D. Wilson, 1978: Effects of St. Louis on Convective Cloud Heights. J. Appl. Meteorol., 17, Brown, S., A.J.R. Gillespie, and A.E. Lugo, 1991: Biomass of tropical forests of south and southeast Asia. Can. J. For. Sci., 21, Chang, H., A. Kumar, D. Niyogi, U.C. Mohanty, F. Chen, and J. Dudhia, 2008: The role of land surface processes on the mesoscale simulation of the July 26, 2005 heavy rain event over Mumbai, India. Glob. Plan. Change, In press. Changnon, S.A., F.A. Huff, and R.G. Semonin, 1971: METROMEX: an Investigation of Inadvertent Weather Modification. Bull. Amer. Meteor. Soc., 52, Chase, T. N., R. A. Pielke, Sr., T. G. F. Kittel, R. R. Nemani, and S. W. Running, 2000: Simulated impacts of historical land cover changes on global climate in Northern Winter. Clim. Dyn., 16, Chin, H.N.S., M.J. Leach, G.A. Sugiyama, J.M. Leone, H. Walker, J.S. Nasstrom, and M.J. Brown, 2005: Evaluation of an Urban Canopy Parameterization in a Mesoscale Model Using VTMX and URBAN 2000 Data. Mon. Wea. Rev., 133,

122 109 Craig, K. J., and R. D. Bornstein, 2002: MM5 simulations of urban induced convective precipitation over Atlanta. Preprints, Fourth Symp. on the Urban Environment, Norfolk, VA, Amer. Meteor. Soc., 5 6. Cotton, W.R. and Pielke R.A., 2007: Human Impacts on Weather and Climate. Cambridge University Press, New York, 90pp. Cotton, W.R., R.A. Pielke Sr., R.L. Walko, G.E. Liston, C.J. Tremback, H. Jiang, R.L. McAnelly, J.Y. Harrington, M.E. Nicholls, G.G. Carrio and J.P. McFadden, 2003: RAMS 2001: Current status and future directions, Meteorol. Atmos. Phys., 82, Deardorff, J., 1978: Efficient prediction of ground surface temperature and moisture with inclusion of a layer of vegetation. J. Geophys. Res., 20, Dickinson, R.E., 1984: Modelling evapotranspiration for three-dimensional global climate models. Climate Processes and Climate Sensitivity, Geophys. Monogr., 29, Diem, J.E. and T.L. Mote, 2005: Interepochal Changes in Summer Precipitation in the Southeastern United States: Evidence of Possible Urban Effects near Atlanta, Georgia. J. Appl. Meteorol., 44, Dixon, P.G. and T.L. Mote, 2003: Patterns and causes of Atlanta s urban heat island-initiated precipitation. J. Appl. Meteorol., 42, Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land-surface model advances in NCEP operational mesoscale Eta model. J. Geophys. Res., 108, D22, 8851, doi: /2002JD Francis, P.A., and S. Gadgil, 2006: Intense rainfall events over the west coast of India. Meteorol. Atmos. Phys., 94, Freitas, E.D., C.M. Rozoff, W.R. Cotton and P.L.S. Dias, 2006: Interactions of an urban heat island and sea-breeze circulations during winter over the metropolitan area of São Paulo, Brazil. Boundary-Layer Meteorol., 122:43 65 DOI /s Feddema, J., K. Oleson, G. Bonan, L. Mearns, L.E. Buja, G. Meehl and W.M. Washington, 2005: The importance of land-cover change in simulating future climates. Science, 310,

123 110 Flint, E.P. and J.F. Richards 1991: Historical analysis of changes in land use and carbon stock of vegetation in south and southeast Asia. Can. J. For. Res., 21, Gero, A.F. and A.J. Pitman, 2006: The Impact of Land Cover Change on a Simulated Storm Event in the Sydney Basin. J. Appl. Meteorol. Climatol., 45, Goswami, B.N., V. D. Sengupta, M.S. Madhusoodanan and P.K. Xavier, 2006: Increasing Trend of Extreme Rain Events Over India in a Warming Environment. Science, 314, Grell, G. A., J. Dudhia, and D. R. Stauffer, 1995: A description of the fifth generation Penn State/NCAR mesoscale model (MM5). NCAR Tech Note NCAR/TN-3981STR, 138 pp. Grimmond, C.S.B. and T.R. Oke, 1999a: Heat storage in urban areas: observations and evaluation of a simple model. J. Appl. Meteorol., 38, Grimmond, C.S.B. and T.R. Oke, 1999b: Aerodynamic properties of urban areas derived from analysis of surface form. J. Appl. Meteorol., 38, Hafner, J. and S.Q. Kidder, 1999: Urban heat island modeling in conjunction with satellite-derived surface/soil parameters. J. Appl. Meteorol., 38, Hjemfelt, M. R., 1982: Numerical simulation of the effects of St. Louis on mesoscale boundary-layer airflow and vertical motion: Simulations of urban vs. non-urban effects. J. Appl. Meteorol, 21, Houghton, R. A., 1990: The global effects of tropical deforestation. Env. Sci. Tech., 24, Holt, T., D. Niyogi, F. Chen, M. A. LeMone, K. Manning, A. L. Qureshi, 2006: Effect of Land - Atmosphere Interactions on the IHOP May 2002 Convection Case. Mon. Wea. Rev., 134, Howard, L., 1833: The climate of London, deduced from meteorological observations. London: Harvey and Darton. Huff, F.A. and S.A. Changnon Jr., 1973: Precipitation modification by major urban areas. Bull. Am. Meteorol. Soc. 54,

124 111 Huff, F. A., 1986: Urban hydrological review. Bull. Amer. Meteorol. Soc., 67, Jenamani, R.K., S.C. Bhan and S.R. Kalsi, 2006: Observational/forecasting aspects of the meteorological event that caused a record highest rainfall in Mumbai. Current Science, 90, Kalnay, E, M. Kanamitsu, R. Kistler, W. Collins, D. Deaven, L. Gandin, M. Iredell, S. Saha, G. White, J. Woolen, Y. Zhu, M. Chelliah, W. Ebisuzaki, W. Higgins, J. Janowiak, K.C. Mo, C. Ropelewski, J. Wang, A. Leetma, R. Reynolds, R. Jenne and D. Joseph, 1996: The NCEP/NCAR 40-year reanalysis project. Bull. Am. Meteorol. Soc., 77, Kalnay, E. and M. Cai, 2003: Impact of urbanization and land use change. Nature, 423, Kim, H.H., 1992: Urban heat island. Inter. J. of Remote Sensing, 13, Kishtawal, C., D. Niyogi, M. Tewari, R. A. Pielke Sr., M. Shepherd, 2007: Heavy Rainfall During Indian Summer Monsoon: A Clear Urban Bias. Geophys. Res. Lett., in revision. Koster, R.D., P.A. Dirmeyer, Z. Guo, G. Bonan, E. Chan, P. Cox, C.T. Gordon, S. Kanae, E. Kowalczyk, D. Lawrence, P. Liu, C. Lu, S. Malyshev, B. McAvaney, K. Mitchell, D. Mocko, T. Oki, K. Oleson, A. Pitman, Y.C. Sud, C.M. Taylor, D. Verseghy, R. Vasic, Y. Xue, and T. Yamada, 2004: Regions of strong coupling between soil moisture and precipitation. Science, 305, Koster, R.D., and M.J. Suarez, 1996: Energy and water balance calculations in the Mosaic LSM. NASA Tech. Memo., ,. 9, 59 pp. Kusaka, H., H. Kondo, Y. Kikegawa and F. Kimura, 2001: A Simple Single- Layer Urban Canopy Model for Atmospheric Models: Comparison with Multi- Layer and Slab Models. Boundary-Layer Meteorol, 101, Kumar, V., D.K. Bhagat, U.R. Kumar and M.S. Ganesh, 2006: Impact of low level jet on heavy rainfall events over Mumbai. Mausam, 58, Lee, T. J. 1992: The impact of vegetation on the atmospheric boundary layer and convective storms. Atmospheric Science Paper No. 509, Colorado State University, Fort Collins, CO, 137 pp. Loose, T., and R.D. Bornstein, 1977: Observations of Mesoscale Effects on Frontal Movement Through an Urban Area. Mon. Wea. Rev., 105,

125 112 Masson, V. 2000: A physically-based scheme for the urban energy budget in atmospheric models. Boundary Layer Meteorol., 94, McPherson, R. A., D. J. Stensrud, K. C. Crawford, 2004: The impact of Oklahoma s wheat belt on the mesoscale environment. Mon. Wea. Rev., 132, Mote, T.L., M.C. Lacke, and J.M. Shepherd, 2007: Radar signatures of the urban effect on precipitation distribution: A case study for Atlanta, Georgia. Geophys. Res. Let., 34, L20710, doi: /2007gl Mukhopadhyay, P., J. Sanjay, W.R. Cotton and S.S. Singh, 2005: Impact of surface meteorological observations on RAMS forecast of monsoon weather systems over the Indian region. Meteorol. Atmos. Phys., 90, Nair, U.S., M.R. Hjelmfelt and R.A. Pielke, 1997: Numerical Simulation of the 9 10 June Black Hills storm using CSU RAMS. Mon. Wea. Rev., 125, NCDC: National Climatic Data Center- Climate of 2005 accessed at NTS.HTML#GLOBE on 10 March Niyogi, D., T. Holt, S. Zhong, P.C. Pyle, and J. Basara 2006: Urban and land surface effects on the 30 July 2003 mesoscale convective system event observed in the Southern Great Plains. J. Geophys. Res., 111, D19107, doi: /2005JD Oke, T.R. 1988: The Urban Energy Balance, Prog. Phys. Geogr., 12, Peterson, T. C., 2003: Assessment of urban versus rural in situ surface temperatures in the contiguous United States: No difference found. J. Clim, 16, Pielke, R.A., W.R. Cotton, R.L. Walko, C.J. Tremback, W.A. Lyons, L.D. Grasso, M.E. Nicholls, M.D. Moran, D.A. Wesley, T.J. Lee and J.H. Copeland 1992: Comprehensive meteorological modeling system RAMS. Meteor. Atmos. Phys., 49, Pielke, R.A., and D. Niyogi, 2008: The role of landscape processes within the climate system. In: Otto, J.C. and R. Dikaum, Eds., Landform - Structure, Evolution, Process Control: Proceedings of the International Symposium on

126 113 Landforms organised by the Research Training Group 437. Lecture Notes in Earth Sciences, Springer, Vol. 115, in press. Pielke, R.A., T.J. Lee, J.H. Copeland, J.L. Eastman, C.L. Ziegler and C.A. Finley 2001: Use of USGS-provided data to improve weather and climate simulations. Ecol. Appl.,7, Pielke, R.A., G. Marland, R.A. Betts, T.N. Chase, J.L. Eastman, J.O. Niles, D. Niyogi, and S.W. Running, 2002: The influence of land-use change and landscape dynamics on the climate system: relevance to climate-change policy beyond the radiative effect of greenhouse gases, Phil. Trans. Royal Soc. of London, Ser. A., 360, Pielke, R. A. S., J. Adegoke, A. Beltran-Przekurat, C.A. Hiemstra, J. Lin, U.S. Nair, D. Niyogi, and T.E. Nobis, 2007: An overview of regional land use and land cover impacts on rainfall. Tellus B, 59B, Pyle, P., D. Niyogi, S.P. Arya, M. Shepherd, F. Chen, and B. Wolfe, 2008: An Observational and Modeling-based Storm Climatology Assessment for the Indianapolis, urban region. J. Appl. Meteorol. Clim., in revision. Roy, S. B., G.C. Hurtt, C.P. Weaver, and S.W. Pacala, 2003: Impact of historical land cover change on the July climate of the United States. J. Geophys. Res., 108, doi: /2003jd Rozoff, C.M., W.R. Cotton, and J.O. Adegoke, 2003: Simulation of St. Louis, Missouri, Land Use Impacts on Thunderstorms. J. Appl. Meteorl., 42, Shepherd, J.M., H. Pierce, and A.J. Negri, 2002: Rainfall Modification by Major Urban Areas: Observations from Space borne Rain Radar on the TRMM Satellite. J. Appl. Meteorl., 41, Shepherd, J.M. and S.J. Burian, 2003: Detection of Urban-Induced Rainfall Anomalies in a Major Coastal City. Earth Interactions, 7, Shepherd, J.M., 2005: A Review of Current Investigations of Urban-Induced Rainfall and Recommendations for the Future. Earth Interaction Shyamala, B. and C.V.V. Bhadram, 2006: Impact of mesoscale synoptic scale interactions on the Mumbai historical rain event during 26th 27th July Current Science, 91, 12, Stein, U. and P. Alpert, 1993: Factor separation in numerical simulations. J. Atmos. Sci., 50,

127 114 Stensrud, D.J., 1996: Importance of Low-Level Jets to Climate: A Review. J. Clim, 9, Vaidya, S.S. and D.R. Kulkarni, 2006: Simulation of heavy precipitation over Santacruz, Mumbai on 26th July 2005, using Mesoscale model. Meteol. Atmos. Phys., doi: /s Walko, R.L., L.E. Band, J. Baron, T.G.F. Kittel, R. Lammers, T.J. Lee, D. Ojima, R.A. Pielke Sr, C. Taylor, C. Tague, C.J. Tremback and P.J. Vidale, 2000: Coupled atmosphere biophysics hydrology models for environmental modeling. J. Appl. Meteorol., 39, Zhong, S., J.D. Fast, and X. Bian, 1996: A Case Study of the Great Plains Low-Level Jet Using Wind Profiler Network Data and a High-Resolution Mesoscale Model. Mon. Wea. Rev., 124, Zhong, S., and J. Doran, 1995: A Modeling Study of the Effects of Inhomogeneous Surface Fluxes on Boundary-Layer Properties. J. Atmos. Sci., 52, Zhou, L., R. E. Dickinson, Y. Tian, J. Fang, Q. Li, R. K. Kaufmann,C. J. Tucker, and R. B. Myneni, 2004: Evidence for asignificant urbanization effect on climate in China, Proc. Nat. Acad. Sci. 101,

128 115 PUBLICATION(S) 1. M. Lei*, D. Niyogi, C. Kishtawal, R. Pielke Sr., A. Beltrán-Przekurat, T. Nobis, and S. Vaidya, 2008, Effect of explicit urban land surface representation on the simulation of the 26 July 2005 heavy rain event over Mumbai, India, Atmos. Chem. Phys., submitted. 2. Roy, S. S., R. Mahmood, D. Niyogi, M. Lei, S. A. Foster, K. G. Hubbard, E. Douglas, and R. Pielke, Sr., 2007: Impacts of the agricultural Green Revolution-induced land use changes on air temperatures in India, J. Geophys. Res., 112, D21108, doi: /2007jd

129 116 Impacts of the Agricultural Green Revolution Induced Land Use Changes on Air Temperatures in India Shouraseni Sen Roy 1 Department of Geography and Regional Studies, University of Miami Coral Gables, FL Rezaul Mahmood Department of Geography and Geology, Western Kentucky University Bowling Green, KY Dev Niyogi, Ming Lei Departments of Agronomy and Department of Earth and Atmospheric Sciences Purdue University, West Lafayette, IN Stuart A. Foster Department of Geography and Geology, Western Kentucky University Bowling Green, KY Kenneth G. Hubbard School of Natural Resources, University of Nebraska-Lincoln Lincoln, NE Ellen Douglas Environmental, Earth and Ocean Sciences, University of Massachusetts, Boston Boston, MA Roger Pielke Sr. CIRES, University of Colorado, Boulder, CO 80309

130 117 Abstract: India has one of the most intensive and spatially extensive irrigation systems in the world developed during the1960s under the agricultural Green Revolution (GR). Irrigated landscapes can alter the regional surface energy balance and its associated temperature, humidity, and climate features. The main objective of this study is to determine the impacts of increased irrigation on long-term temperature trends. An analysis of the monthly climatological surface datasets at the regional level over India showed that agriculture and irrigation can substantially reduce the air temperature over different regions during the growing season. The processes associated with agriculture and irrigation-induced feedback are further diagnosed using a column radiation-boundary layer model coupled to a detailed land surface/ hydrology scheme, and 3D simulations using a Regional Atmospheric Modeling System. Both the modeling and observational analysis provide evidence that during the growing season, irrigation and agricultural activity are significantly modulating the surface temperatures over the Indian subcontinent. Therefore irrigation and agricultural impacts, along with land use change, and aerosol feedbacks need to be considered in regional and global modeling studies for climate change assessments. Keywords: Climate Change, Land Use Change, Surface Temperature, India, Agriculture Impacts, Irrigation.

131 118 Introduction The complex interrelationships between land cover and land use change in the context of variable climatic conditions have been widely investigated (e. g., Chase et al. 2000; Bounoua et al. 2002; Roy et al., 2003; McPherson et al, 2004; Feddema et al., 2005a). According to Houghton (1990), changes in land uses have contributed to about 25% enhanced levels of greenhouse gases in terms of anthropogenic activities. (Pielke Sr. 2002) Pielke et al. (2002) concluded that land use change is an important climate policy consideration beyond the radiative effects of green houses. The majority of studies in the last century have focused on the effects of deforestation and denudation of natural landscapes and associated impacts on the surrounding environment (Brown et al. 1991; Flint and Richards, 1991). Some of the findings from these studies are related to greater temperature variations, decreased proportions of soil retention of carbon, and increased levels of pollution and changes in precipitation pattern (Houghton, 1994; Kauppi et al. 1992; Pielke et al. 2007(Pielke Sr. 2002; Pielke 2007)). However, in recent years awareness about the impact of changes in agricultural land use in terms of cropping practices and irrigation on the resulting local weather conditions has substantially increased (Foley et al., 2005; Douglas et al.,2006, Douglas et al. in revision; Pielke R. A. Sr. et al. 2007b). One such study by Ramankutty and Foley (1999) systematically focused on changes in land use patterns over the last three centuries from 1700 to They used a combination of historical land use data and satellite imagery to monitor the changes in cropland acreage over different time periods. This study found, globally, significant conversion of forest lands to croplands since the year A primary example of this trend is the northwestern (NW) Indo-Gangetic Plain of the Indian subcontinent which extends eastwards along the foothills of the Himalaya in north-central (NC) India; here the gradual intensification of cropland has been replacing

132 119 forests/woodlands since This regional expansion of agricultural cropland has been even greater since 1947 due to independent India s rising population and its subsequent increased demand for agricultural products. Another significant development in India s land use history, is the so called Green Revolution (GR) that has substantially changed the NW and NC Indian landscape. The GR started in 1965 with the introduction of improved varieties of seeds, fertilizers, and an intricate canal/irrigation network throughout NW India which made the region less dependent on precipitation. The benefits of the GR were concentrated initially in NW and NC India where the Indo- Gangetic river system ensured adequate water supply. In 1960, a total of approximately 1.9 million hectares with high yields of wheat, rice, and other grains in several varieties rapidly increased after the introduction of irrigation to 15.4 million hectares by 1970 and 43.1 million hectares in Thus, the region saw a 20-fold expansion of irrigated land use over a 20-year period. Since 1980, the land use change has largely stabilized in these regions (Green Revolution in India, 2006). The main agricultural seasons in India are kharif (June to September), rabi (November to May), and zaid (March to June). Moreover, the peak rabi and zaid growing seasons are from February through April, and March through May, respectively. These periods coincide with the maximum vegetative growth and associated higher crop water requirements. During the kharif season, the summer monsoon rainfall meets most of the irrigation needs.. However, during the rabi and zaid seasons, regional agriculture fully depends on irrigation. Wheat is one of the primary crops grown during the rabi season, and oilseeds and other types of cash crops are planted during the zaid season, the driest of the three growing periods. The relatively greater availability of adequate water resources during these drier periods has likely modulated the land-atmosphere interaction in the region.

133 120 The climatological impacts of the GR in NW and NC India have not been fully studied. Given the large scale changes in land use resultant from the GR agricultural intensification in NW and NC India, the key objective of this paper is to determine the impacts of the introduction of irrigation to the following three components of this regions climatology: 1) the observed longterm monthly maximum and minimum temperatures, 2) diurnal temperature range (DTR), and 3) average temperatures at the seasonal time-scale. The main causative process resulting from the introduction of extensive irrigation is the greater availability of water for evaporation/transpiration (ET) leading to increased partitioning of energy into latent heat (Mahmood et al., 2004; Douglas et al., 2007 in revision). As a result, we hypothesize that an overall cooling trend in the long-term surface temperatures might be occurring. In the present study we have focused on the impact of land use changes associated with the GR on the long-term temperatures in NW and NC India, where the impact is generally considered greatest (Figure 1). In the 1960s, the GR was viewed as a critical socio-economic force for transforming the lives of millions of people in India. Over the subsequent years of the introduction of irrigation to this region, this model of agrarian transformation has also been adopted in other parts of the world. In order to understand the processes contributing to irrigation impact on the surface temperatures over the NW and NC India, and inductively the implications of comprehensive irrigation and its impact in other areas where it is practiced, this study applied a coupled boundary layer land surface model with detailed hydrology and vegetation response (Gottschalk et al. 2000, and Niyogi et al. 2004). The model was applied to understand the processes contributing to irrigation impact on the surface temperatures over the NW and NC India. The present study provides a unique opportunity to assess the impacts of GR-driven extensive agricultural intensification/ land use change and irrigation on regional temperature within the monsoon domain based on observed data in the NW and NC India.

134 121 Background Studies: The Intergovernmental Panel on Climate Change (IPCC) Third Assessment Report (TAR), as well as the Fourth Assessment for Policy Makers, revealed the relatively poor understanding of the impact of land use changes on the long-term trends shown in environmental variables. This knowledge gap was re-emphasized by Pielke et al. (2002) and Pielke et al. (2007), who recommended an original approach to the quantification of changes in vegetation cover as this impacts energy partitioning at different spatial scales of the climate system. Their study also indicated a need for comparative climatological investigations and suggested contrasting the impacts of land use change on remote, local, and regional climate. Previous studies have examined the impact of irrigation on near surface air temperatures in order to better understand the role of soil moisture on the general temperatures at different spatial scales. One of the earlier studies showing differences in temperatures between irrigated and nonirrigated areas was conducted by Idso et al. (1981). They found a temperature difference of about 12 K between irrigated and non-irrigated alfalfa fields. Similarly, a temperature difference of nearly 10 K between irrigated and non irrigated land uses was reported by Segal et al. (1989) in eastern Colorado. Other studies showing a similar relationship between soil moisture and current temperatures, and following monthly temperatures, include Walsh et al. (1985) and Williams (1992). Recently, Mahmood et al. (2004) examined the long-term monthly maximum, minimum, and average temperatures for several irrigated and non- irrigated stations in Nebraska. The results of the study indicate a clearly decreasing trend in the mean maximum and average temperatures for the irrigated sites, with an increasing trend observed for the non-irrigated sites. The physical reasoning that

135 122 supports this difference is that irrigated areas are becoming cooler due to partitioning of the incoming solar radiative flux into increased latent energy flux as a result of increased levels of soil moisture. The role of soil moisture on daily temperatures has been analyzed under different geographical conditions. Hogg et al. (2000) found a cooling trend in deciduous forests of interior western Canada during summer as a result of increased latent heat flux. Fitzjarrald et al. (2001) identified a decreasing trend in the mean temperatures during the spring season in the eastern US. The main explanation for the aforementioned findings stems from a variety of energy balance studies that identify a feedback cycle in which changes in land surface such as increased soil moisture availability (as from excess rain or irrigation) can lead to changes in the surface Bowen ratio due to increased latent energy flux, evaporative cooling and generally lowered air temperatures (Stull 1988, Niyogi et al., 1999). The results of the above empirical studies have also been supported by modeling studies including Mahmood and Hubbard (2002 and 2003), who used a soil moisture-energy balance model for three types of land uses in Nebraska representing spatially wet- to-dry conditions. The results of these studies identified the modification in rates of evapotranspiration (ET) and near surface soil moisture content due to the changes in land use. Kueppers et al. (2007) reported the modeling results of irrigation s cooling effect on near surface air temperatures in California, referring to it as the Irrigation Cooling Effect (ICE). Chase et al. (2000) also found similar results from model simulation which indicated changes in latent and sensible heat flux at the local and regional scales, which might not be detectable at the global level. Recently, the modification of the summer monsoon has been attributed to land cover changes, modification of the surface energy balance, and the subsequent cooling over Asia (Feddema et al., 2005a). Overall, the role of the local geographical setting has been found to be critical to a better understanding of

136 123 the physical processes contributing to overall temperature trends (Niyogi et al., 2002; Mahmood et al., 2004; Feddema et al., 2005a, b). Although extensive literature exists for the North American continent, a relative absence of any detailed analysis investigating the impacts of land use changes on near surface climate in other parts of the world, including India, is available. One exception is a modeling study by Douglas et al. (2006) who reported a 7% increase in latent heat fluxes in the wet season (kharif) and a 55% increase in the dry season (rabi), from a pre-agricultural and contemporary land cover. Two-thirds of these increases were attributed to irrigation. A follow-up numerical study (Douglas et al., 2007 in revision) showed decreases in sensible heat flux of 100 W m -2 or more due to irrigation in northwestern India. Douglas et al. also found that intensive irrigation in the northwest and along the southeast coast has suppressed the air temperature by 1 to 2 K and has increased the water vapor content by more than 1g/kg in the lowest atmospheric layer (68 m). In this context, the present study investigates the impacts of the wide-spread adoption of irrigation on near surface air temperatures in NW and NC India. Since the land use change is associated with the GR, this assessment provides an opportunity to evaluate the GR s impacts, and will add to the general understanding of the impacts of land use change in other regions of the globe. Methodology: a. Data sources and analytical approach This study uses the dataset of the regional monthly maximum and minimum temperature time series compiled by the Indian Institute of Tropical Meteorology (IITM) Indian Monthly Surface Air Temperature (updated January 18 th, 2006). Created from a all-india network of 121 stations in seven homogenous land regions across the subcontinent. This dataset initially covered the time period from 1901 to 1990 and was later extended to 2003

137 124 with data obtained from Indian Daily Weather Reports. The demarcation of the seven homogenous regions was based on similarities in geographical, topographical, and climatological features. In order to develop a more realistic temperature dataset onto the selected network of 121 stations, the climatological normals of monthly mean maximum and minimum temperatures during for 388 geographically well-distributed stations were obtained from the India Meteorological Department. The available station data were converted into a monthly anomaly time series for the entire time period extending from 1901 to 2003, with respect to individual station normal values. Next, the station level monthly temperature anomaly values were interpolated into a 0.5 by 0.5 grid for the entire period. Then the climatological normals (1951 to 1980) of temperatures for the 388 stations were interpolated onto the same grid that resulted in a high resolution grid point temperature climatology for the entire subcontinent. Finally, the regional level temperature series were computed using the averages of the constituent grid point data in the different regions. Detailed methodology regarding the construction of this dataset is available in Kothawale and Rupa Kumar (2005). Other useful references for the dataset include Pant and Rupa Kumar (1997) and Rupa Kumar et al. (1994). Given the main objective of determining the impacts of increased irrigation on long-term temperature trends, the present analysis is limited to the last fifty years from 1947 to 2003 for the NW and NC region of the Indian subcontinent (Figure 1). The NW region includes the states of Punjab and Haryana, two states that benefited most from the GR, and also parts of the states of Uttaranchal and Rajasthan. The NC region consists of most of Uttar Pradesh and parts of Madhya Pradesh, Jharkhand, Bihar, and Orissa. In order to detect the differences in near surface air temperatures, this study completed an analysis of temperatures for individual months for the pre-gr and the post-gr periods. In this investigation, the pre-gr period spans from 1947 through 1964, while the post-gr period covers 1980 through The

138 125 transitional period from 1965 through 1979 was excluded from the analysis, in order to accurately describe the impact of GR on temperatures. The overall long-term trend assessment was also conducted using linear trend analysis for the rabi and the zaid seasons. As noted above, these seasons represent the periods of maximum irrigation application. A trend analysis was also conducted separately for the peak growing season in order to more exactly determine the atmospheric effects of irrigation. b. Land surface boundary layer model We perform two types of model analyses. The first involves sensitivity using a 1D coupled land-atmospheric boundary layer model. The second modeling analysis is performed using a 3D Regional Atmospheric Modeling System (Pielke et al. 1992). The 1D model consists of a subsurface, transition, surface, and a mixed layer continuum model. The model can develop clouds as a function of the initial profiles as well as the atmospheric thermodynamics and thus modify the radiation reaching the surface. Surface and the boundary layer responses are parameterized using the surface-layer similarity and the mixedlayer theory. It also accounts for mosaic vegetation cover, which was considered uniform for this case of crop cover. The soil moisture response is adopted by a detailed water balance, and the impact to the atmosphere is regulated by a surface moisture availability term which controls the plant response and humidity and temperature changes. The model follows detailed eddy diffusivity and boundary layer formulations using well-tested techniques, having been extensively tested for both mid-latitude and tropical conditions. The vegetation is represented following Taconet et al. (1986) and Carlson and Lynn (1991) for multiple vegetation and land surface interactions for the canopy, the bare ground, the non-leaf part of the canopy, and the interaction of the vegetation with the canopy boundary layer. The model is able to estimate partial canopy resistance as a function of leaf area index and can

139 126 compute different roughness regimes for canopy separation (Niyogi and Raman 1997). The model utilizes a default wind profile for a single column representative of the region (Douglas et al. 2006). The observations focused on soil moistures measured at three different locations: Amritsar (31.64ºN, 74.87ºE) in northwest India, Meerut, (29.01ºN, 77.42ºE), and Kanpur (26.4ºN, 80.23ºE) in north central India. The model was run twice for each of the three locations, once with irrigation and once for water-stressed soils. These runs were completed assuming average March conditions for each location. The vegetation was represented by a leaf area index of 5, with a 95% fractional vegetation cover, which is typical of the fully-grown crop canopy. The soil type was assigned as sandy clay loam, and the crop was assumed at its peak greenness with a crop height of 1.5 m and a width of 0.15 m. The deep soil temperature was prescribed on the basis of soil climatology and was 24.6 ºC for Amritsar, 26 ºC for Meerut, and 28.6 ºC for Kanpur. All other values in the model were set to default values typical for the month of March over India (Douglas et al. 2006; Alapaty et al. 2001). The 1D model studies were further analyzed using a fully coupled 3D modeling study which used the Regional Atmospheric Modeling System (RAMS). We adopted a domain with 30 km grid spacing, with46 x 42 grids covering an area of 1380 km x 1260km with the domain center located at 28N, 78E (Figure 2). In the vertical, 35 stretched sigma levels were employed with a stretching factor of 1.15 up to a1500 m spacing level. All vertical grid spacing above this height were maintained at a constant 1500 m. The model s initial conditions were prescribed using the one-degree National Center for Environmental Prediction-Global Data Analysis System (NCEP GDAS). The lateral boundary conditions followed Klemp and Wilhelmson (1978) and were updated every 6 hours using analysis nudging. The radiation processes were parameterized following Chen (1983) with an update every 1200s. Surface energy balance was calculated in the LEAF2 model (Walko et al. 2000) and had nine soil layers. A similar modeling setup

140 127 has been applied in a number of irrigation-related studies (e.g. Mahmood et al. 2004, Adegoke et al. 2006). A one-week period from March 2006 was selected. This period generally coincided with the green-up phase of the growing season. The period was also selected as the region had little to no synoptic activity with relatively clear skies, and the surface-boundary layer feedback was expected to be a prominent forcing on the mesoscale processes. Two experiments were performed, a control run assuming no irrigation, and an irrigation run in which all the cropland within the domain was assumed to be irrigated. Note that, even though we chose a 7-day period due to computational limitations, irrigation was a daily event over the domain during the growing period, and the potential impacts could be extrapolated for the entire growing season. In the simulation we assumed the top 12 centimeters of soil (top 3 soil layers in the model), reached their field capacity due to irrigation, which was initiated once a day at 10 am LT. This idealized experiment was developed for further illustrating the significant influence irrigation can exert on regional surface temperatures. Results: a. NW India The present analysis is limited to the rabi (November to May) and zaid (March to June) growing seasons. Overall, November to June is the driest period of the year, when irrigation is most intense in order to meet crop growth requirements. The long-term trend in the seasonal mean maximum ( C per decade) and minimum temperatures (-0.05 C per decade) has been negative (cooling) for the zaid season (Table 1). Compared to seasonal maximum temperatures, minimum temperatures show a slightly greater decline (Table 1). During the rabi season, the trends were almost neutral to slightly positive, with almost no trend observed in the maximum temperatures. The linear trends were also calculated for the seasonal average temperatures

141 128 and diurnal temperature range (DTR). The DTR was reduced and average temperatures showed decreasing trends (cooling) during the zaid season, while for the rabi season the trends were nearly neutral to slightly positive. The decreasing trend during the zaid season was slightly greater for average temperatures (-0.05 ºC per decade) than that of the DTR (-0.02 ºC per decade). In order to specifically isolate the impact of irrigation on the partitioning of the energy budget, we further investigated the peak growing season temperatures for the two cropping seasons. The peak growing period for the rabi season was limited to February through April, while for the zaid, it was March through May, when crops experienced peak water requirements. The trends were negative (cooling) at ºC per decade for the peak zaid season mean maximum, mean minimum, mean, and mean DTR. In contrast, during the peak rabi season only the maximum temperatures showed a decline (-0.02 ºC per decade) (Table 1). The declining trends can be attributed to the increased availability of soil moisture from greater inputs of irrigation. The impact of soil moisture on the surface atmosphere energy budget was relatively greater during the zaid season. The reduction in DTR for both seasonal and peak seasonal periods can be attributed to evaporative cooling as suggested by Dai et al. (1999). In order to detect the differences in temperatures between the pre-gr and post-gr period, we divided the study period into two parts with the pregreen revolution period extending from 1947 to 1964, and the post-green revolution period extending from 1980 to The mean growing season and peak growing season maximum and minimum temperatures, DTR, and average temperatures were calculated for the pre-gr and post-gr periods. The results indicate a lower maximum and average temperature and DTR during both growing seasons and peak growing season months of the rabi and zaid seasons (Figure 3). For example, it was found that the mean peak growing season maximum temperature for the rabi crop was 31.7 ºC and 31.4

142 129 ºC during the pre- and post-gr periods, respectively. Moreover, the mean peak growing season DTR was 16 ºC and 15.8 ºC during the pre- and post- GR period, respectively. In other words, 0.34 ºC and 0.18 ºC cooling and decline had occurred for the mean growing season maximum temperature and DTR, respectively, during the post-gr period. For the zaid crop, the mean growing season maximum temperature was 36.2 and 35.9 ºC during the pre- and post-gr period, respectively. In addition, the mean growing season DTR was ºC and 15.5 ºC during the pre- and post-gr period, respectively. Hence, for the zaid season, 0.3 ºC and 0.06 ºC declines occurred for maximum temperature and DTR, respectively, during the post- GR period. These results are in agreement with findings of previous modeling and observed data-based studies indicating a cooling trend in daily maximum temperatures, leading to further reductions in DTRs over major agricultural areas of the USA (Bonan, 1997, Mahmood et al., 2004). The analysis suggests a 0.19 ºC and 0.27 ºC decrease of average temperatures for the rabi and zaid seasons, respectively, during the post-gr period. Declines in both the mean maximum and mean minimum temperatures in the post-gr period have resulted in this lowering of average temperatures. We further examined the monthly average temperatures before and after the GR in order to specifically identify the period of maximum impact. Figure 4 shows the monthly average maximum and minimum temperatures for the pre-gr and post-gr periods. The results evidence that, overall, the months of February, March, April, May, and June show lower average monthly maximum temperatures during the post-gr period. Further analysis reveals that the mean maximum growing season temperatures were 0.29, 0.51, 0.07, 0.38, and 0.28 ºC lower for the months of February, March, April, May, and June, respectively, during the post-gr period. Minimum temperatures for the entire period extending from March to June showed lower monthly averages for the post-gr period.

143 130 Statistical tests were performed to determine the significance of differences in means between the pre-and post-gr periods. The tests include a student s t-test, bootstrapping, and robust statistics (20%-trimmed mean approach). Even though the results are not statistically significant at a 05 confidence level, the lowering of temperatures are physically consistent with the theoretical understanding of the relationship between soil moisture, energy partitioning, and the Bowen ratio. Findings are also consistent with results from the Great Plains and other modeling studies (e.g., Mahmood et al., 2004; Kalnay and Cai, 2004; Adegoke et al., 2003; Roy et al., 2003; Zhao and Pitman, 2002; and Eastman et al., 2001). For additional verification, we also conducted a coupled land -atmosphere model simulation to determine the physical relationship between land use and temperature in NW India. The results are presented in a following section. b. NC India Although the impacts of the GR are most pronounced in NW India, its benefits have gradually spread over most of the Gangetic basin. As evident from Table 2, the long-term seasonal and peak growing season temperatures also experienced a cooling trend in the NC region. The rates of decline were greater during the zaid season, with the highest negative trend in the case of seasonal minimum temperature at ºC per decade, followed by per decade for seasonal maximum temperature (Table 2). The rates of cooling for maximum and minimum temperatures in general were also greater than those observed in the NW India temperature rates. However, in the case of the DTR, the trends were positive, as the ºC rate of decline in minimum temperatures was slightly greater than that of the ºC maximum temperature during the peak growing season. The temperatures calculated for the pre-gr and post-gr periods during the seasonal and peak growing months demonstrated changes similar to those found in the case of NW India (Figure 5). In most cases, the post-gr

144 131 temperatures were lower, except for the DTR which showed either no difference or a slight increase. This can be attributed to an equal, and in some cases, greater decline in minimum temperatures. The minimum temperatures at the seasonal level showed greater cooling (0.11 ºC), compared to the maximum temperatures (0.04 ºC), resulting in a higher DTR during the post GR period. The average DTR during the post-gr period was, however, slightly lower than the pre-gr period consistent with previous studies (Dai et al., 1999). Similar to NW India, NC India also showed greater cooling during the zaid season. For example, our study found a 0.55 ºC and 0.53 ºC cooling of minimum temperatures for the entire and peak growing season, respectively. The decline in maximum temperatures was lower compared to the minimum temperatures, which were 0.29 ºC and 0.22 ºC at the seasonal and the zaid peak growing season. The average temperatures were lower overall during both the rabi and zaid seasons. A maximum cooling of 0.42 ºC was observed in the zaid seasonal average temperatures, followed by a 0.3 ºC maximum cooling during the peak growing periods for both seasons. In addition, the temperatures during the pre-gr and post-gr periods were analyzed to further focus on the months of greatest cooling (Figure 6). As expected, the overall pattern showed irrigation-induced cooling during the first half of the year, which is also the driest period. The observed cooling was greatest during April, May, and June. For example, during pre-gr period for these months, the mean maximum temperature values were ºC, 39.5 ºC, and 37.0 ºC, respectively. These mean temperatures for the post- GR period were 36.9 ºC, 39.1 ºC, and 36.4 ºC, respectively. Hence, a 0.01 ºC, 0.4 ºC, and 0.53 ºC cooling has occurred for April, May, and June, respectively. However, as mentioned earlier, this cooling was greater in the case of the minimum temperatures for the months of April, May, and June with values of 0.57, 0.66, 0.67 ºC, respectively.

145 132 Model Sensitivity Results The above results were also verified by coupled land surface-boundary layer model runs for three sites, located over NW and NC India. These include Amritsar in the NW region, and Kanpur and Meerut in the NC region. The results of the model runs are in agreement with the observations and show a distinct cooling in the near surface air temperatures due to agricultural irrigation. The modeling results reveal the impact of irrigated crop land versus moisture stressed land surface conditions on the diurnal variation of the surface air temperature for the three locations. Typically, the irrigated crop landscape is about 2 ºC to 3 ºC cooler during the day (similar to the findings of Douglas et al., in review) and about 5 ºC cooler at night. The resulting temperature difference between the two model scenarios is due to the large difference in the evapotranspirative cooling produced by the irrigated crop. The results of model run for Meerut for March 15 th, 1999 is shown in Figure 7 with and without the irrigated crop consideration. Consistent with past studies for other regions, the irrigated crop surface partitions the available surface radiative flux leading to an enhanced moisture flux through a modified surface latent heat flux as shown in Figure 7a. The day time moisture flux is about 100% higher when the surface cropland is irrigated. The enhanced evaporation/transpiration leads to an altered humidity regime. It is evident in Figure 7b that the irrigated crop leads to about 3 g kg -1 or a 20% increase in the boundary layer- specific humidity. Similarly, the air temperatures at 2 m are cooler under the influence of irrigated crops (Figure 7c). The amount of cooling shown in Figure 7 is to the order of about 2 ºC both for the day and night time. The impact of the surface changes is dramatically depicted in the radiometric temperature data (such as those sensed by the satellites), and is shown in Figure 7d. For this variable, the difference between the irrigated crops versus default is of the order of 5 ºC during day and about 2 ºC during night.

146 133 Similar results are obtained in the 3D RAMS modeling study. Figure 8 shows the temperature time series for the week-long run. With irrigation, the average temperature is reduced by about 5 ºC. The regional distribution of the averaged temperature differences for the entire simulation, and for day and night, is shown in Figure 9 a-c. Interestingly, in a few locations, particularly in the northeast corner of the domain, the irrigation case shows a slight warming (by 0.5 to 1 ºC). This can be attributed to the regional feedbacks associated with changes in cloudiness and humidity in the irrigated regions. However, in the majority of the domain, the irrigation case shows an overwhelming cooling response by as much as 6 ºC. The cooling is more pronounced during the day time, but continues to be prominent even for the night time, typically of the order of 4ºC across the domain. Discussion and Concluding remarks: This study investigates the GR-induced large scale adoption of irrigation and its impacts on the long term temperatures in NW and NC India. The GR in India has brought about phenomenal growth in food production with the introduction of high-yielding varieties of seeds and the installation of a widespread irrigation network across this region. The results of the analysis broadly conform with the findings of previous studies for other parts of the world which examine the impact of irrigation on surface- atmosphere interactions which affect local level energy budgets (e. g., Chase et al., 2000; Adegoke et al., 2003; Mahmood and Hubbard, 2002; Mahmood et al., 2004; Feddema et al., 2005a, b; Mahmood et al., 2006a; Douglas et al. 2006, Douglas et al. in review). The main findings of the study are: The overall seasonal trends during the rabi season in NC India suggest a cooling, while NW India showed predominantly neutral to slightly positive trends (Tables 1 and 2). The trend in maximum temperatures

147 134 during the peak growing season was negative for both regions. These results are in agreement with previous findings by Mearns et al. (1995), Dai et al. (1999), Kalnay and Cai (2003) attributing them to increased surface ET rates. In the case of the zaid season, the overall trends were negative for both the seasonal and peak growing season months for both NW and NC India (Tables 1 and 2). This is the driest period of the year, when the impact of irrigation on temperature can be detected most clearly. The negative trends were relatively greater at the seasonal level than the peak season months. These findings are in agreement with the findings of an earlier study examining the trends in seasonal maximum and minimum temperatures, and DTR, by Sen Roy and Balling (2005), who also reported a declining trend over NW India. The comparative analysis of averages of DTR, maximum, minimum, and mean temperatures, revealed lower averages for the post-gr period for both the rabi and zaid seasons (Figures 2 and 4). The differences between the pre-gr and post-gr periods were generally similar for the entire growing season as well as the peak growing season. Lower averages during the post- GR period were also observed in the case of individual monthly mean temperatures (Figures 3 and 5). In the case of NC India, both the rabi and zaid season showed a stronger decline during the peak growing season months. However, due to an equal or greater decline in minimum temperatures, a slightly positive or neutral trend occurred in the DTRs long-term trends. The above results were also validated by the simulations from a coupled land - atmosphere model run carried out for three locations within the two study regions and a 3D modeling study using the Regional Atmospheric Modeling System. The irrigated areas showed a 3 ºC to 4 ºC daytime cooling (Figures 7, 8).

148 135 Lower average maximum temperatures and a reduction in the DTR during the post-gr period is possibly as a result of the greater availability of soil moisture as suggested from previous empirical and modeling studies conducted in other parts of the world (Cao et al., 1992; Mearns et al., 1995). Also, for the Indian region, the role of atmospheric aerosols leading to warming or cooling of the lower atmosphere needs to be considered (Ramanathan et al. 2002). We conducted one sensitivity experiment with the RAMS model setup with doubled optical depth representative of the aerosol rich air over NW India following Niyogi et al. (2007). Results indicate the aerosol induced cooling could be about half of that due to irrigation. We obtained cooling over the region which is of the order of less than 1 deg C (range: C; average: 0.9). While the effect of irrigation from our results is much larger i.e. order of 2 C (range: C; average: 2.3C). These values need to be treated with caution as the aerosol issue is complex and can cause both warming and cooling depending on the single scattering albedo and aerosol speciation (Menon et al. 2002). Niyogi et al. (2007) reviewed the Moderate Resolution Imaging Spectroradiometer (MODIS) aerosol optical depths and the Aerosol Robotics Network (AERONET) optical depth data over Kanpur, India. Their results indicate that the aerosol plumes are transitory with optical depths ranging around 0.5 for majority of the period and going in excess of 2 for few occasions. Thus the impact of aerosols can be an important, but variable, forcing that can increase or decrease the surface temperatures. On the other hand, irrigation activity is nearly permanent feature of the growing season. As a result of which, the irrigation effect on surface temperature can be consistently considered as one of cooling. The present study generally found cooling of growing season minimum temperatures. These results are somewhat analogous to the findings from various locations in the Great Plains (Mahmood et al., 2004, 2006a, b) which report both the cooling and warming of growing season mean minimum

149 136 temperatures at irrigated locations. Thus, we suggest that the cooling trends for mean minimum growing season temperatures in irrigated areas indicate that complex surface-atmosphere interactions are occurring in these regions. The lack of significant changes in monsoon activity during the two selected time periods (Pant and Rupa Kumar, 1997) also supports these results. On the basis of the above findings, it can be concluded that due to the introduction of widespread irrigation measures in NW and NC India, the potential exists for the cooling of maximum temperatures and a reduction in the DTR during the earlier part of the year, primarily from March to May. We suspect that the signal could have been improved if station level data were available rather than regional data. Also, inclusion of the arid Rajasthan desert in the regional data may have diluted the overall signal. Future studies that examine station level data may reveal much stronger irrigation-related signals in long-term temperatures. Thus, inclusion of land use/cover changes particularly those due to agricultural intensification can improve regional climate assessment for climate change conditions. Acknowledgements: This work has been partially supported by NASA Grant Nos. NAG and NNG04GL61G, NASA-THP NNG04GI84G (Dr. J. Entin), NASA-IDS NNG04GL61G (Drs. J. Entin and G. Gutman), NASA LULCC Program (Dr. G. Gutman), NSF-ATM (Dr. S. Nelson), and the Purdue Asian Initiative Grant. The data were obtained from the Indian Institute of Tropical Meteorology and are greatly appreciated. References: Adegoke, J. O., R. A. Pielke, Sr., J. Eastman, R. Mahmood, and K. G. Hubbard (2003), A regional atmospheric model study of the impact of irrigation on midsummer surface energy budget in the U. S. High Plains, Mon. Wea. Rev., 131,

150 137 Alapaty K., N. Seaman, D. Niyogi, A. Hanna (2001), Assimilating surface data to improve the accuracy of atmospheric boundary layer simulations, J. App. Met., 40, Bonan G. (1997), Effects of land use on the climate of the United States. Climate Change, 37, Bounoua, L., R. Defries, G. J. Collatz, P. Sellers, and H. Khan (2002), Effects of land cover conversion on surface climate, Clim. Change, 52, Brown, S., A.J.R. Gillespie, and A.E. Lugo (1991), Biomass of tropical forests of south and southeast Asia, Can. J. For. Sci., 21, Cao, H. X., J. F. B. Mitchell, and J. R. Lavery (1992), Simulated diurnal range and variability of surface temperature in a global climate model for present and doubles CO 2 climates., J. Climate, 5, Carlson T.N., and B. Lynn (1991), The effects of plant water storage on transpiration and radiometric surface temperature, Ag. For. Meteorol., 57, Chase, T. N., R. A. Pielke, Sr., T. G. F. Kittel, R. R. Nemani, and S. W. Running (2000), Simulated impacts of historical land cover changes on global climate in Northern Winter, Clim. Dyn., 16, Dai, A., K. E. Trenberth, and T. R. Karl (1999), Effects of clouds, soil moisture, precipitation and water vapor on diurnal temperature range, J. Climate, 12, Douglas, E.M., A. Beltran-Przekurat, D.S. Niyogi, R. A Pielke Sr., and C.J. Vörösmarty. (2007), Simulating changes in land-atmosphere interactions from expanding agriculture and irrigation in India and the potential impacts on the Indian monsoon, Global Planetary Change, (in revision). Douglas, E. M., D. Niyogi, S. Frolking, J. B. Yeluripati, R.A. Pielke Sr., N. Niyogi, C. J. Vörösmarty, and U. C. Mohanty (2006), Changes in moisture and energy fluxes due to agricultural land use and irrigation in the Indian Monsoon Belt, Geophys. Res. Lett., 33, L14403, doi: /2006gl

151 138 Eastman, J. L., M. B. Coughenour, and R. A. Pielke Sr. (2001), The regional effects of CO2 and landscape change using a coupled plant and meteorological model, Global Change Biol., 7, Feddema, J., K. Oleson, G. Bonan, L. Mearns, L.E. Buja, G. Meehl, W.M. Washington, (2005a), The importance of land-cover change in simulating future climates, Science, 310, Feddema, J., K. Oleson, G. Bonan, L. Mearns, W. Washington, G. Meehl, D. Nychka, (2005b), A comparison of a GCM response to historic anthropogenic land cover change and model sensitivity to uncertainty in present-day land cover representations, Clim. Dyn., 25, Fitzjarrald, D. R., O. C. Acevedo, and K. Moore (2001), Climatic consequences of leaf presence in the eastern United States., J. Climate, 14, Flint, E.P., and J.F. Richards (1991), Historical analysis of changes in land use and carbon stock of vegetation in south and southeast Asia., Can. J. For. Res., 21, Foley, J. A., R. DeFries, G. P. Asner, C. Barford, G. Bonan, S. R. Carpenter, F. S. Chapin, M. T. Coe, G. C. Daily, H. K. Gibbs, J. H. Helkowski, T. Holloway, E. A. Howard, C. J. Kucharik, C. Monfreda, J. A. Patz, I. C. Prentice, N. Ramankutty, P. K. Snyder, (2005), Global consequences of land use, Science, 309, Gottschalk, J.C., R.R. Gillies, and T.N. Carlson, (2000), The simulation of canopy transpiration under doubled CO2: The evidence of and impact of negative feedbacks on transpiration in two 1-D soil-vegetation-atmospheretransfer (SVAT) models, Ag. And Forest Meteo., 106, Green Revolution in India (2006), /WWW. INDIANCHILD. COM /GREEN_REVOLUTION_INDIA.HTM accessed on May 1st, Hogg, E. H., D.T. Price, and T. A. Black (2000), Postulated feedbacks of deciduous forest phenology on seasonal climate patterns in the western Canadian interior, J. Climate, 13,

152 139 Houghton, R. A. (1990), The global effects of tropical deforestation, Env. Sci. Tech., 24, Houghton, R.A. (1994), The worldwide extent of land-use change: In the last few centuries and particularly in the last several decades, effects of land use change have become global, Bioscience, 44(5), Idso, S.B., R.J. Reginato, D.C. Reicosky, and J.L. Hatfield (1981), Determining soil-induced plant water potential depressions in Alfalfa by means of infrared thermometry, Agron. J., 73, IMD (1999), Tables of Climatological Normals: , India Meteorological Department (IMD), Pune, 350 pp. Kalnay, E. and M Cai (2003), Impact of urbanization and land use change. Nature, 423, Kauppi, P.E., K. Mielikainen, and K. Kuusela (1992), Biomass and carbon budget of European forests, , Science, 256, Kothawale, D.R. and K. Rupa Kumar, (2005), On the recent changes in surface temperature trends over India, Geophys. Res. Lett., 32, L Kueppers, L.M., M.A. Snyder, and L.C. Sloan, (2007), Irrigation cooling effect: Regional climate forcing by land use change. Geophys. Res. Lett., 34, L Mahmood, R. and K. G. Hubbard (2002), Anthropogenic land use change in the North American tall grass-short grass transition and modification of near surface hydrologic cycle, Clim. Res., 21, Mahmood, R. and K. G. Hubbard (2003), Simulating sensitivity of soil moisture and evapotranspiration under heterogeneous soils and land uses, J. Hydrol., 280, Mahmood, R., K. G. Hubbard, and C. Carlson (2004), Modification of growingseason surface temperature records in the Northern Great Plains due to land use transformation: Verification of modeling results and implications for global climate change, Int. J. Climatol., 24,

153 140 Mahmood R., S. A. Foster, T. Keeling, K. G. Hubbard, C. Carlson, and R. Leeper (2006a), Impacts of irrigation on 20 th -century temperature in the northern great plains, Glob. Planet. Change, 54, Mahmood, R., T. Keeling, S. A. Foster, and K. G. Hubbard (2006b), Observed data-based evidence of impacts of irrigation on 20 th -century temperature in the Ogallala aquifer region, J. App. Met. Climat., (in review). McPherson, R. A., D. J. Stensrud, K. C. Crawford (2004), The impact of Oklahoma s wheat belt on the mesoscale environment, Mon. Wea. Rev., 132, Mearns, L.O., F. Giorgi, L. McDaniel, C. Shields (1995), Analysis of variability and diurnal range of daily temperature in a nested regional climate model: comparison with observations and doubled CO2 results, Clim. Dyn., 11, Menon S., J. Hansen, L. Nazarenko, Y. Luo (2002) Climate Effects of Black Carbon Aerosols in China and India, Science 297, DOI: /science Niyogi D., and S. Raman (1997), Comparison of four different stomatal resistance schemes using FIFE observations, J. Appl. Meteorol., 36, Niyogi D., H. Chang, L. Gu, S. Menon, R. A. Pielke Sr. (2007), Potential Impacts of Aerosol-Land-Atmosphere Interaction on the Indian Monsoonal Rainfall Characteristics, Natural Hazards - Monsoon Special Issue, DOI /s y. Niyogi D., S. Raman and K. Alapaty (1999), Uncertainty in specification of surface characteristics, Part 2: Hierarchy of interaction explicit statistical analysis, Boundary-Layer Meteorol., 91, Niyogi D., Y. Xue and S. Raman (2002), Hydrological land surface response in a tropical regime and midlatitudinal regime, J. Hydromet., 3, Niyogi D., K. Alapaty, S. Raman, and F. Chen (2007), A coupled gas exchange/photosynthesis based evapotranspiration model (GEM) for environmental applications, J. Appl. Meteorol.Clim., (in revision).

154 141 Pant, G.B. and K. Rupa Kumar (1997), Climates of South Asia. John Wiley & Sons, Chichester, 320 pp. Pielke, Sr, R.A., G. Marland, R.A. Betts, T.N. Chase, J.L. Eastman, J.O. Niles, D. Niyogi, and S.W. Running (2002), The influence of land-use change and landscape dynamics on the climate system: relevance to climate-change policy beyond the radiative effect of greenhouse gases, Phil. Trans. Royal Soc. of London, Ser. A., 360, Pielke R. A. Sr., J. O. A., T.N. Chase, C. H. Marshall, T. Matsui, D. Niyogi (2007a). A New Paradigm for Assessing the Role of Agriculture in the Climate System and in Climate Change. Agric. For. Meteorol. 132(Special Issue): Pielke, R. A. S., J. Adegoke, A. Beltran-Przekurat, C.A. Hiemstra, J. Lin, U.S. Nair, D. Niyogi, and T.E. Nobis (2007b). An overview of regional land use and land cover impacts on rainfall. Tellus B, 59B, Ramanathan, V., C. Chung, D. Kim, T. Bettge, L. Buja, J.T. Kiehl, W.M. Washington, Q. Fu, and D.R. Sikka (2005), Atmospheric brown clouds: impacts on south Asian climate hydrologic cycle, Proc. Nat. Acad. Sci., 102, Ramankutty, N. and J. Foley (1999), Estimating historical changes in global land cover: Croplands from 1700 to 1992, Global Biogeochem. Cycles, 13, Roy, S. B., G.C. Hurtt, C.P. Weaver, and S.W. Pacala (2003), Impact of historical land cover change on the July climate of the United States., J. Geophys. Res., 108, doi: /2003jd Segal, M., W. E. Schreiber, G. Kallos, G.R. Garratt, A. Rodi, J. Weaver, R.A. Pielke Sr. (1989), The impact of crop areas in northeastern Colorado on midsummer mesoscale thermal circulations, Mon. Wea. Rev., 117, Sen Roy, S., and R. C. Balling Jr. (2005) Analysis of trends in maximum and minimum temperature, diurnal temperature range, and cloud cover over India, Geophys. Res. Lett., 32, L12702.

155 142 Taconet, O., T.N. Carlson, R. Bernard, and D. Vidal-Madjar, (1986), Evaluation of a surface/vegetation model using satellite infrared surface temperatures. J. Clim. Appl. Meteor., 25, Walko RL, L.E. Band, J. Baron, T.G.F. Kittel, R. Lammers R, T.J. Lee, D. Ojima, R.A. Pielke Sr., C. Taylor, C. Tague, C. J. Tremback, and P.L. Vidale, (2000), Coupled atmosphere biophysics hydrology models for environmental modeling. J. Appl. Meteorol., 39, Walsh, J. E., J.E. Jasperson, and B. Ross (1985), Influence of snow cover and soil moisture on monthly air temperature, Mon. Wea. Rev., 113, Williams, K. R. S. (1992), Correlation between Palmer drought indices and various measures of air temperature in the climatic zones of the United States. Phys. Geog., 13, Zhao, M. and A.J. Pitman (2002), The impact of land cover change and increasing carbon dioxide on the extreme and frequency of maximum temperature and convective precipitation, Geophys. Res. Lett. 29, /2001GL

156 Figure 1. Shaded areas delineate the boundary of the NW (northwestern) and NC (north-central) regions of the Indian Regional Monthly Surface Air Temperature data set developed by IITM. The star symbols on the map represent locations for which land surface boundary layer model runs were conducted. 143