Regional long term climate change ( ) in the midtropical Atlantic and its impacts on the hydrological cycle of Puerto Rico

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116,, doi: /2010jd015414, 2011 Regional long term climate change ( ) in the midtropical Atlantic and its impacts on the hydrological cycle of Puerto Rico Daniel E. Comarazamy 1 and Jorge E. González 2 Received 1 December 2010; revised 22 March 2011; accepted 31 March 2011; published 18 June [1] Large scale climate data for the north tropical Atlantic (NTA) region show that air temperatures have increased during the past 50 years ( to ) with moderate warming near the Caribbean islands to considerable heating in the northern region. This pattern may be driven by sea surface temperature anomalies in the same region of study that follow relatively small changes in the Caribbean basin to stronger anomalies in the northeast. These changes might be associated with changes in the long term pattern of the NTA high pressure system that drives climate in the region. A series of mesoscale numerical experiments were designed to study the regional impacts these large scale changes have on the hydrological cycle of the island of Puerto Rico. Results indicate that increased easterly surface winds for the time frame disrupts a pattern of inland moisture advection and convergence zone, increasing cloud base heights and reducing the total column liquid water content over high elevations. This combination of factors produces a reduction in precipitation over the central and eastern mountains of Puerto Rico. Citation: Comarazamy, D. E., and J. E. González (2011), Regional long term climate change ( ) in the midtropical Atlantic and its impacts on the hydrological cycle of Puerto Rico, J. Geophys. Res., 116,, doi: /2010jd Introduction [2] Human activity has profound climate and environmental impacts. Most of these impacts are considered to have negative effects on living conditions on planet Earth, and therefore require special attention for humans to produce adequate adaptation and mitigation policies (depending on the scale and severity of such impacts). The scales of these impacts range from global, to regional, to local and include, but are not limited to, changes in the atmospheric composition and radiative forcing, global warming, effects on the water cycle, ocean changes and rising sea levels, ecosystem modifications, among many others. [3] At the global scale, the link between human activities and climate change is well documented and widely studied, with the most prominent issue being a marked increase in global average temperatures, i.e., global warming, mainly due to an increase of anthropogenic emissions of atmospheric greenhouse gases [Intergovernmental Panel on Climate Change, 2007; Trenberth et al., 2007]. However, our understanding of how these changes reflect at regional scales and what impact they have to the local population is incomplete at specific regions. This leads to the major fundamental question addressed by this research: What is the regional climate impact of global anthropogenic induced 1 NOAA CREST Center, City College of New York, New York, New York, USA. 2 Department of Mechanical Engineering, City College of New York, New York, New York, USA. Copyright 2011 by the American Geophysical Union /11/2010JD atmospheric and oceanic changes, e.g., global warming and increased sea surface temperatures (SST), on the Caribbean Basin? The northeastern region of Puerto Rico is used as the test case to study the climate impacts of global climate change on the region s hydrological cycle. This region offers a unique opportunity for this type of study because of the close proximity of the San Juan Metropolitan Area (SJMA), Luquillo Experimental Forest (LEF, hereafter referred to as El Yunque), and Central Mountain range, and because of evidence of global and local effects on regional climate. [4] The Caribbean rainfall pattern, which spans from April to November, is characterized by a bimodal behavior. The two modes represent the early rainfall season (ERS, April June) and the late rainfall season (LRS, August November), separated by what is referred to as the midsummer drought (MSD) [Magaña et al., 1999]. This pattern is clearly observed on the island of Puerto Rico (Figure 1) and the rest of the Caribbean islands, Central America, and southern Mexico with slight variations at each location [Chen and Taylor, 2002; Curtis and Gamble, 2007; Gamble et al., 2008; Angeles et al., 2010]. [5] In Puerto Rico, a precipitation peak in May defines the ERS. Differential heating and local transport of moisture uphill the Central Mountains, which provide the lifting mechanism, produce most of the precipitation in Puerto Rico [Comarazamy and González, 2008]. The late season begins after the summer months and peaks in October. Included in the LRS is the hurricane season, occurring between June and November. The precipitation maximum in October occurs mostly because of enhanced low level convergence to the east of the Lesser Antilles islands, low 1of18

2 Figure 1. Thirty year ( ) monthly precipitation climatology for 15 COOP stations distributed throughout Puerto Rico (adapted from the work of Comarazamy and González [2008]). vertical wind shear, high sea surface temperatures, greater amounts of deep layer moisture across the tropical North Atlantic, and tropical systems approaching the region (i.e., tropical waves and tropical cyclones) [Taylor et al., 2002]. [6] There are numerous studies that have addressed the relationship between warm El Niño Southern Oscillation (ENSO) events and dry LRS in the Caribbean [Hastenrath, 1976; Aceituno, 1988; Ropelewski and Halpert, 1996]. The focus on the LRS is due to its abundance of rain and its coincidence with the hurricane season. It is precisely the relatively smaller amounts of rainfall during the ERS what makes it important, because even small variations in precipitation can lead to seasonal droughts or flooding. Recent studies have reported a strong positive relation between the early season Caribbean rainfall and warm ENSO events [Enfield and Alfaro, 1999; Giannini et al., 2000; Giannini et al., 2001a; Taylor et al., 2002; Chen and Taylor, 2002]. A small research project performed by the NWS and Weather Forecast Office (WFO) in San Juan showed that ENSO events enhance precipitation during normally wet seasons [Wang et al., 2006]. The correlation is such that warm winter anomalies in the ENSO index are related to positive ERS departures in the Caribbean basin one season after, the El Niño + 1 years. This teleconnection accounts for almost half of the season s precipitation variability. The propagation of the warm ENSO event into the Caribbean ERS is attributed to positive spring north tropical Atlantic (NTA) sea surface temperature anomalies (SSTA) that develop in response to the wintertime equatorial Pacific anomalies [Aceituno, 1988; Giannini et al., 2000; Chen and Taylor, 2002]. The ERS, in addition to being correlated with winter Pacific SSTAs, also shows a robust correlation with concurrent warm Caribbean SSTAs [Enfield and Alfaro, 1999; Taylor et al., 2002]. Numerical experiments performed with global circulation models (GCMs) show that increased early season rainfall is characterized by lower than normal surface pressures, a warmer and convergent lower atmosphere, a divergent upper atmosphere, and a low shear environment. These atmospheric conditions are consistent with a region biased toward increased precipitation [Taylor et al., 2002; Angeles et al., 2010]. [7] The relationship between the North Atlantic Oscillation (NAO) and the climate record across the Caribbean basin was determined using the statistical Burnaby test, which gives rise to an E value [Burnaby, 1953]. A significant positive E value indicates a direct relationship between the series, whereas a negative value indicates an inverse relationship. Anomalies in Caribbean annual precipitation have been linked to the variability of the NAO index, where fluctuations in annual rainfall amounts are synchronous with variations in the wintertime NAO and are not controlled by ENSO. The relation by Malmgren et al. [1998] was calculated using the winter NAO index and the standardized mean annual precipitation for five stations in Puerto Rico, and reported to be: E NAO Pcp = During years of a high winter NAO index, when the axis of moisture transport in the North Atlantic changes to a southwesterly northeasterly orientation, annual precipitation in Puerto Rico is lower than average. As the precipitation in the Caribbean is believed to be strongly influenced by the North Atlantic high, which is closely connected to the Icelandic low, the relationship between the NAO and precipitation patterns in the Caribbean may result from variations in displacement and strength of the high and the associated trade winds [Jury and Winter, 2010; Giannini et al., 2001b; Malmgrem et al., 1998]. [8] It has also been reported that local land cover and land use (LCLU) changes have an effect on local precipitation and cloud cover patterns. It was found that lowland deforestation is leading to increases in cloud base heights and thinner clouds in Central American rain forests [Lawton 2of18

3 Figure 2. Three point running mean of the bimonthly El Niño Southern Oscillation index (ENSO_I) for 1950 to 2006, standardized departures from the mean. Warm/cold periods, denominated as El Niño/La Niña, respectively, occur when a +/ 0.5 C threshold is met for five consecutive overlapping periods as defined by the Climate Prediction Center (CPC) and are represented in red/black horizontal lines. Black dashed boxes represent the two time frames selected for the study. et al., 2001; Nair et al., 2003; Ray et al., 2006], which is resulting in increased severity of regional droughts. Regional model simulations of pasture and forested scenarios show that lowland and premontane deforestation increases air temperatures and sensible heat fluxes and decreases dew point temperatures and latent heat fluxes of air masses that when lifted eventually lead to the formation of orographic cloud banks at higher elevations. However, a similar numerical experiment conducted in Puerto Rico reported a similar increase in cloud base heights, although contradictory results were produced when a forested island was used [van der Molen, 2002; van der Molen et al., 2006]. These studies concluded that under clear and calm conditions the development of the sea breeze circulation dominates atmospheric flow over the island. Differential heating during forested runs produced a stronger sea breeze with stronger updrafts and a more defined convergence front. As a result the enhanced updrafts transport moisture to higher levels and thus increase cloud base heights when compared to pasture simulations. [9] Given that there could be a combination of global and local factors affecting precipitation in northeastern Puerto Rico, any comprehensive study focused on analyzing such impacts should include the possible contribution of numerous factors related to the total climate change observed locally. In section 2, large scale climate changes are analyzed at two different time frames in the Caribbean region to reduce the global oscillation indices influence on the region s ERS climate. Section 3 focuses on the results of a series of numerical mesoscale atmospheric model climate simulations that are designed to study the regional and local effects these large scale atmospheric and oceanic changes, possibly in combination with LCLU changes, have on the hydrological cycle of Puerto Rico. Section 4 summarizes the major findings of this research, with recommendations for future work, and possible implications in related fields. 2. Large Scale Climate Changes Reflected in the Caribbean Region 2.1. Time Frame Selection [10] As discussed in section 1, the Caribbean early season climate is greatly influenced by variations of several important large scale oscillations, most notably the ENSO and NAO indices [Malmgren et al., 1998; Enfield and Alfaro, 1999; Malmgren and Winter, 1999; Chen and Taylor, 2002; Taylor et al., 2002]. In the case of the ENSO index, the correlation is such that warm winter ENSO anomalies are related to positive ERS departures in the Caribbean basin one season after the observed Equatorial Pacific anomaly, noted as El Niño + 1 years [Chen and Taylor, 2002]. The relationship between the NAO index and the climate record was determined as a high inverse correlation between the winter NAO index and the standardized mean annual precipitation for five stations in Puerto Rico [Malmgren et al., 1998]. Using this information, the long term record ( ) of ENSO and NAO indices (Figures 2 and 3) as archived by the Climate Prediction Center (CPC), was analyzed to select two 5 year periods of comparable low warm ENSO values and low NAO variability that are at widely separated time periods to reflect a significant change in relevant climate variables. [11] The 5 year periods from 1955 to 1959 and from 2000 to 2004 are the best available in the long term record, in 3of18

4 Figure 3. Three point running mean of the daily North Atlantic Oscillation Index (NAO_I) for (top) , (middle) , and (bottom) , standardized departures from the mean. Solid horizontal lines represent the standard deviation of the data, ±s (±0.6213) and ±2s (±1.2427). Black dashed boxes represent the two time frames selected for the study. 4of18

5 Table 1. Simulation Matrix Run Land Cover and Land Use Driving Conditions PRESENT Present atmospheric and oceanic conditions PRESENT Past atmospheric and oceanic conditions PAST (Preurban) Present atmospheric and oceanic conditions PAST (Preurban) Past atmospheric and oceanic conditions terms of the ENSO and NAO indices, to perform the simulations proposed in the run matrix in Table 1. The period of the late 1950s presents an El Niño event toward the end of the decade, making the years 1959/1960 potential El Niño + 1 years; regardless, the two periods selected present a similar ENSO climatology (Figure 2). In terms of the NAO index, the period of the early 2000s shows less variability of the index within the ±s and ±2s range as compared to other periods in the record. The 1950s decade shows stronger individual warm and cold events, but has in general less variability than the rest of the data, as it stays between the ±2s bands (Figure 3). The two NAO time frames do not show a clear indication of periodicity, but slight indications of periodicity are present in both periods. An analysis of the global oscillations and the Caribbean climate for the two periods selected yielded similar teleconnection results as those found in the literature Large Scale Climate Changes [12] Current ( ) and preurban ( ) conditions of atmospheric 4 D fields of air temperature, horizontal wind components, and relative humidity are provided by the National Centers for Environmental Prediction (NCEP) Reanalysis 2.5 large scaledata[kalnay et al., 1996]. Past and present specifications of SSTs are derived from the Smith and Reynolds Extended Reconstruction Sea Surface Temperature (ERSST v3b) [Smith and Reynolds, 2003; Smith et al., 2008]. The horizontal distribution of differences in atmospheric temperatures and SSTs in Figures 4 and 7 are shown in the geographic area covered by the modeling grid 1 (Figure 8). Figure 4 shows the large scale temperature changes at the times of local overnight low and daytime high temperatures over the Caribbean region, where moderate increases in temperature are observed on the order of 1.0 to 1.8 C, respectively; these values increase toward the North Atlantic. These changes are found to penetrate vertically from the surface to about the 700 mbar pressure level over the island of Puerto Rico, located at approximately 66 west longitude (Figure 5). Figure 6 shows the SST time series, values are averaged over grid 1. Although there is a general increasing trend of SST in the Caribbean, associated with differences between Caribbean and North Atlantic near surface pressures [Jury and Winter, 2010; Angeles et al., 2007], there are periods where the increase is more pronounced. These periods are identified as and from the early mid 1970s to 2005; the later period is of great importance for the research presented in this paper because is within the two time frames selected to perform the numerical mesoscale simulations. Changes in pressure difference between the Caribbean and the North Atlantic are evident in the SST changes, as reflected in the Caribbean basin (Figures 7). Changes in SST for the ERS peak are only about C around the island of Puerto Rico, but increase to the northeast of the island reaching values closer to 1 C in the northeast region of modeling grid 1. The difference in warming rates between near surface air temperatures and SSTs shown in Figures 4 and 7, respectively, has been previously associated with an accelerated Hadley circulation, with sinking motions over the Caribbean corresponding with increasing rising motion over the Amazon [Jury and Winter, 2010]. The sinking motions induce a faster rate of warming and drying at 850 mb than at other levels and much of this trend is attributable to physical mechanisms. 3. Numerical Experiments 3.1. General Model Configuration [13] To examine how these large scale changes affect climate in the northeastern region of Puerto Rico, an ensemble of numerical atmospheric model simulations that allows the separation of the LCLU changes and global warming signals was performed. The model chosen for the study is the Regional Atmospheric Modeling System (RAMS), a highly versatile numerical code developed at Colorado State University to simulate and forecast meteorological phenomena [Pielke et al., 1992; Cotton et al., 2003]. The version of RAMS used in this investigation, v.4.3, contains an upgraded cloud microphysics module described by Saleeby and Cotton [2004], as opposed to the original package available in the current model release [Meyers et al., 1997]. [14] The simulations focus on the northeastern coast of Puerto Rico, with special interest in the SJMA and El Yunque, and were conducted with three grids making use of the grid nesting capabilities of the model. Grid 1 covers great part of the Caribbean basin with a horizontal resolution of 25 km. Grid 2, which is nested within grid 1, covers the island of Puerto Rico at a resolution of 5 km. Grid 3, nested within grid 2, is centered in the city of San Juan, and covers the SJMA, El Yunque, relatively nondeveloped regions west and south of the city, and ocean areas to the north (Figure 8). The vertical grid structure consists of a grid spacing Dz of 30 m near the surface and stretched at a constant ratio until a Dz of 1000 m is reached. The depth of the model reaches approximately 25 km. [15] A 5 year climatology is constructed from model results for numerous climate variables in order to clearly identify trends in accumulated precipitation. All runs performed will span the 3 month period of April June, i.e., the Puerto Rican ERS (Figure 1). This part of the year is the most convenient to conduct LCLU changes studies in Puerto Rico [Velazquez Lozada et al., 2006], and represents the end of the dry season and the onset of the midsummer drought [Magaña et al., 1999], a critical period in the annual hydrological cycle of the island, during which the atmospheric model has performed satisfactorily [Comarazamy 5of18

6 Figure 4. Large scale temperature differences ( C) between the and time frames in the Caribbean basin calculated from the National Centers for Environmental Prediction (NCEP) Reanalysis 2.5 resolution data averaged at (top) 0200 and (bottom) 1400 local standard time (LT), the two closest times in the four hourly data to the local overnight low and daytime high temperatures, respectively, during the Caribbean 3 month early rainfall season (ERS). The thick solid vertical line in both panels represents the location of the north south vertical cross section in Figure 5. and González, 2008]. A spin up time of one week is specified at the start of each 3 month simulation to allow for numerical stabilization of the atmospheric model and the numerous submodels and parameterizations available in the modeling system. [16] To answer the proposed research question, a set of numerical simulations is configured combining two largescale atmospheric conditions (representing the atmospheric and oceanic climate change presented in section 2.2 and their corresponding levels of green house gases [GHGs]) with two LCLU scenarios (representing current and preurban conditions), resulting in the run matrix in Table 1. The atmospheric mesoscale modeling system and the modeling methodology used in this study has been extensively validated against surface and upper air observational data that guarantees model performance for the prediction of 6of18

7 Figure 5. Vertical cross sections of temperature differences ( C) between the and time frames through the north south line in Figure 4 averaged at (top) 0200 and (bottom) 1400 LT, the two closest times in the four hourly data to the local overnight low and daytime high temperatures, respectively, during the Caribbean 3 month ERS. temperature and precipitation, making it an appropriate tool to conduct climate impact studies across Puerto Rico [Velazquez Lozada et al., 2006; Comarazamy et al., 2006; Comarazamy and González, 2008; Comarazamy et al., 2010; Comarazamy, 2010] Analysis Methods [17] To study the impact of global warming in specific geographical regions, the northeastern region of the Caribbean island of Puerto Rico was used for the test case. Climate impacts on accumulated precipitation are analyzed by calculating the difference between the possible combinations of simulations in Table 1 that have a physical meaning. These combinations explain the climate impacts due to global warming with present LCLU (PRESENT1 PRESENT2), due to global warming with past LCLU (PAST1 PAST2), and what is determined to be Total Change due to LCLU changes and global warming (PRESENT1 PAST2). Since there is the possibility that the climate impacts due to LCLU changes under present atmospheric conditions (PRESENT1 7of18

8 Figure 6. Twelve point running mean of the monthly Smith and Reynolds Extended Reconstructed Sea Surface Temperature (ERSST) data set for 1900 to Sea surface temperature (SST) values are domain averaged over the Caribbean region covered by modeling grid 1. PAST1), and under past conditions (PRESENT2 PAST2) are important, these combinations are also analyzed. [18] Also incorporated into the analysis is the factor separation method described by Stein and Alpert [1993], which calculates the individual contribution of the two factors in question, determined to be two of the calculations presented above and the contribution to the Total Change due to the nonlinear interaction between the two factors. The results presented in the next sections refer to the 5 year climatology of each of the simulation scenarios described in Table Impacts on the Hydrological Cycle of Northeastern Puerto Rico [19] The influence of topography in the precipitation pattern of Puerto Rico is well documented [Daly et al., 2003; Comarazamy and González, 2008; Jury et al., 2009]. Precipitation climatologically occurs when the easterly trade winds transport available moisture from ocean evaporation up the mountain slopes, which acts as a lifting mechanism, forming a convergence zone atop the mountain ridge. This in turn leads to cloud formation, rain development, and consequently surface accumulated precipitation [Comarazamy and González, 2008]. Any disruption to the wind pattern that sets the mechanism in motion is likely to have a profound effect on the precipitation pattern, or at least in the amounts of precipitation and fresh water available at the surface. [20] Figure 9 shows how that wind pattern has changed from the PAST2 simulations, which consists of past LCLU specifications with past atmospheric conditions and SSTs, to the PRESENT1 simulations, i.e., present LCLU with present atmospheric and SST conditions; this is referred to as the Total Change. Figure 9 shows the daily evolution of the two wind vector fields to better visualize the differences in wind patterns and identify the axis of convergence produced by the past and present simulated climatologies. In Figure 9 it is seen how the two vector fields have a similar magnitude and that they follow a similar directional pattern from midnight to the early morning hours, from 0000 to 0800 LT (Figures 9a, 9b, and 9c). At 1200 LT (local standard time), the position of the convergence zone along the Central Mountains in the past climatology (red vectors) and along the northern region just off the foothills of the mountain range in the present climatology (black vectors) is evident (Figure 9d). This pattern is still in full effect at 1600 LT (Figure 9e). The convergence in the present climatology is at its strongest point at 1200 and 1600 LT (Figures 9d and 9e, respectively) in the narrow valley where Caguas is located, identified by the thick black contour south of the SJMA, between the Central Mountains and El Yunque, identified by the two thick green contours in Figure 9. This strong convergence zone localized in such a narrow geographical area could be due to the downwind low pressure generated by a free flow around an obstacle, in this case El Yunque; during the past climatology, the flow is slower and approaches the coastline at a different angle. These results are in agreement with observed changes in trade winds magnitude for the Caribbean basin. Figure 10 shows the regional scale domain wide averaged trade winds speed 8of18

9 Figure 7. Regional scale sea surface temperature differences ( C) between the and time frames in the Caribbean basin calculated from Smith and Reynolds ERSST data set for April. change between the two climate periods analyzed from the NCEP Reanalysis data at times closest to local overnight low and daytime high temperatures. The observed increases in trade winds magnitude have been attributed to an increase in the North Atlantic high [Jury and Winter, 2010; Angeles et al., 2007]. [21] The convergence zone in the PRESENT1 simulations generates vertical motions that increase the liquid water mixing ratio. The subsidence that the circulation cells generate on the mountaintops acts to reduce the liquid water content in the atmosphere above the ridge. The liquid water mixing ratio (g kg 1 ) contains the amount, in total mass, of cloud droplets and raindrops per kilogram of dry air, and therefore accounts for the two completely nonfrozen cloud microphysics hydrometeors in the atmospheric regional model cloud module that indicate the cloud formation and rain development process in the tropics. It has been reported that changes in wind patterns and vertical wind motions should also produce changes in cloud base heights, an important parameter in tropical montane cloud forests [Hamilton et al., 1993; Bruijnzeel and Proctor, 1993; Still et al., 1999; Nair et al., 2003; Ray et al., 2006; Lawton et al., 2011]. A montane cloud forest is defined as a tropical or subtropical moist forest in elevated terrain in which the cloud base is persistently or seasonally at or below the vegetation canopy level. The cloud base height, in turn, is defined as the lowest atmospheric level at which the air contains perceptible amounts of cloud particles. In this sense, the liquid water mixing ratio is the ideal model produced variable to analyze these parameters. [22] The cloud base height is determined by identifying the first level in the model vertical coordinate, from the surface up, where measurable amounts of liquid water (i.e., 0.01 g kg 1 ) are present, and then the climate impact calculations are performed and presented in Figure 11. Figure 11a shows the results for cloud base height difference due to LCLU change, while driving the model with present atmospheric and SST conditions, Figure 11b also shows the LCLU signal but by driving the model with the past climate conditions. Although the two LCLU signal cases did not produce a pattern of cloud base height differences that clearly follows the regions of larger LCLU changes, the case with past climate (Figure 11b) shows large increases in cloud bases along the north coast, less so along the Central Mountains, and decreased heights along the eastern and southeastern coast. For the cases that represent the impact due to global climate change, both with present and past LCLU specifications (Figures 11d and 11e, respectively), cloud base heights appear higher at 500, 1000, and 1500 m over the higher elevations of the Central Mountains, and by an average of 500 m over El Yunque, while decreasing along the coastal plains. A pattern reproduced almost exactly for the Total 9of18

10 Figure 8. Model grids, with topography information, used in this research. (top left) Grid 1 shows the Caribbean basin and islands (contour interval 750 m); (top right) grid 2 covers the island of Puerto Rico and adjoining Vieques, Culebra, and Virgin Islands to the east of the domain (contour interval 150 m); and (bottom) grid 3 shows the northeastern region of Puerto Rico, the main area of interest in this research (contour interval 150 m). Change (Figure 11c). For the case of cloud base height differences presented in Figure 11, Figures 11a and 11b show different patterns, indicating that cloud base height differences due to LCLU changes are not independent of the large scale climate conditions and that their difference, the nonlinear interaction between the two factors (Figure 11f), cancels the LCLU signal along the coast from the Total Change. This produces a Total Change pattern (Figure 11c) that closely resembles the global climate change signal in Figures 11d and 11e. [23] The decreasing cloud base heights over lowland coastal plains might be due to the slight coastal cooling observed in the easternmost region of the island [Comarazamy, 2010]. These results are in agreement with the increases in cloud base heights observed over the Monteverde region of Costa Rica due to localized deforestation [Lawton et al.,2001;nair et al., 2003; Ray et al., 2006], although these previous studies did not include the effects of global climate change in the configuration of their modeling experiments and conclusions. [24] To complement the cloud base heights information, the total column liquid water content (g kg 1 ) was also analyzed. This gives an indication of how thick or dense the cloud is, and is independent of where in the column the liquid water appears, effectively eliminating some of the problems encountered in the cloud base height differences due to trace amounts of liquid water at high atmospheric levels. Results for total column liquid water content change, Figure 12, show that there is little impact on the differences observed due to LCLU changes when driving the model with present climate conditions (Figure 12a) and due to LCLU changes under past atmospheric and SST conditions (Figure 12b). The impact on total liquid water content are mainly due to the global warming signal, whether with present or past LCLU (Figures 12d and 12e, respectively), 10 of 18

11 Figure 9. Daily cycle of the wind vector fields produced by the PRESENT1 simulations (black vectors) and the PAST2 simulations (red vectors). The thick black line delineates the urban area, and the thick green line delineates the 350 m topography contour. and this is reflected on the Total Change pattern. Results for the Total Change in total atmospheric column liquid water content in Figure 12c shows that liquid water decreases in high amounts all along the Central Mountains ( 0.03 to 0.15 g kg 1 ) and in the El Yunque area and close surroundings ( 0.03 to 0.12 g kg 1 ); both areas are identified in Figure 12 by the thick green contours. These changes in total atmospheric liquid water content are associated with the higher cloud bases observed over the Central Mountain ridge and El Yunque (Figure 11c), that in turn occur because of the northward shift of the convergence zone that produced convective activity over these regions (Figures 9d and 9e). The Total Change in atmospheric column liquid water content also shows increasing values (0.03 to 0.12 g kg 1 )in the valley between the Central Mountains and El Yunque, south of the SJMA where the residential area of Caguas is located; both of these urban areas are identified in Figure 12 by the thick black contour, and where the enhanced vertical motions are found. Although this valley did not experience a significant change in cloud base height, the increased convective activity produced a deeper or denser cloud layer above the region. [25] The changes in wind magnitude and direction, cloud base heights, and total column liquid water content result in dramatic changes of total surface accumulated precipitation during the ERS in Puerto Rico, especially over regions where precipitation and low cloud cover occurrence is of extreme importance. Figure 13 shows the change in the 3 month surface accumulated precipitation. Figure 14 presents these differences in terms of percent of change relative to the simulation with past LCLU (PAST1 and PAST2) for the LCLU signal cases, to the simulation with past climate conditions (PRESENT2 and PAST2) for the global warming signal cases, and to the simulation with past LCLU and past atmospheric and SST conditions (PAST2) for the Total Change case (Table 1). [26] As with the previous variables discussed in this section, the LCLU change signal appears to have a limited contribution to the Total Change of the 3 month accumulated surface precipitation during the five year periods, as seen in the results for the impact due to LCLU changes with present climate (Figure 13a) and with past climate conditions (Figure 13b). The precipitation differences that these two scenarios produced do not account for a significant percent of change (Figures 14a and 14b). Figures 14d and 14e represent the global climate change signal for accumulated precipitation change (Figure 13) and percent change (Figure 14), for the simulation with present and past LCLU specifications, respectively. The similarity between Figures 13a and 13b and 14a and 14b indicates that changes due to LCLU changes are independent of global climate conditions, and vice versa for the similarity between Figures 13d and 13e and Figure 14d and 14d. This also indicates, combined with the low impact the LCLU signal has on precipitation changes, that the global warming signal has a much stronger contribution to the Total Change and that it will closely follow the pattern observed in Figures 13d and 13e and 14d and 14e (see Figures 13c and 14c). The Total Change (Figures 13c and 14c) presents precipitation difference values that range from 45 to 150 mm throughout the Central Mountain range and El Yunque, regions identified by the thick green contours, with a maximum negative precipitation difference of up to 150 mm at the higher elevations in the domain in the Central Mountains. 11 of 18

12 Figure 10. Large scale basin averaged horizontal wind magnitude (m s 1 ) between the and time frames and between pressure levels 1000 and 700 mbar, calculated from the NCEP Reanalysis 2.5 resolution data averaged at (top) 0200 and (bottom) 1400 LT, the two closest times in the four hourly data to the local overnight low and daytime high temperatures, respectively, during the Caribbean 3 month ERS. These precipitation reductions are equivalent to a range of 6 to 13% change based on total rainfall totals of mm for this region based on the PAST2 simulations. In the valley between the Central Mountains and El Yunque, the increased convective activity that lead to increased liquid water content also produced increased surface accumulated precipitation in the order of mm, which is a 4% change from the 700 mm of surface accumulated precipitation over that region. [27] These results for accumulated precipitation are of particular importance for small island states, mainly because tropical rain and montane cloud forests provide much of the natural resources, including fresh water, consumed in these countries [Hamilton et al., 1993; Bruijnzeel and Proctor, 1993; Franco et al., 1997; Pounds et al., 1999; Still et al., 1999; Lawton et al., 2011]. Therefore, it is equally important that the results for accumulated precipitation differences be compared with other data sets. Although a reliable longterm record of station data for precipitation, dating back to at least the early 1950s, is not available for the past climate in Puerto Rico the results presented in this study show that changes in precipitation appear to be mainly a large scale global climate change impact, as opposed to be due to LCLU changes. Since most of the warming for the second half of the twentieth century occurred since the middle to late 1970s, the same trend should be evident in precipitation records from 1975 to 2004, a period for which reliable data exist for El Yunque. The analysis for precipitation change 12 of 18

13 Figure 11. Differences in the cloud base height (m), at the 0.01 g kg 1 liquid water mixing ratio threshold, (a) due to land cover and land use (LCLU) changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) due to LCLU changes while driving the model with past atmospheric conditions, (c) total change due to LCLU changes and global warming, (d) due to global warming with present LCLU specifications, (e) due to global warming with past LCLU specifications, and (f) the contribution due to the nonlinear interaction among the LCLU and global warming factors. The thick black line delineates the urban areas and the thick green line delineates the 350 m topography contour. for the El Yunque cooperative station show a linear trend of precipitation decrease during the ERS at 20.7 mm per decade, during the five decades that span the current study this results in a total precipitation reduction from 1955 to 1959 to of mm, which is lower than the 150 mm shown by the model results, but as mentioned before the station results exclude the trend from 1955 to A large scale global data set with a 5 5 horizontal resolution reports a reduction in precipitation for the Caribbean basin of between 20 to 40% per century during the period [Trenberth et al., 2007]. This percentage is based on the means for the 1961 to 1990 period, and represents changes in year round precipitation. 4. Summary and Conclusions [28] The work presented here is an investigation of the climate impacts of global climate change (global warming, GW) using the Regional Atmospheric Modeling System (RAMS) as the main research tool, using the northeastern region of the Caribbean island of Puerto Rico as the test case. To achieve this goal, large scale climate data was analyzed, and a numerical modeling matrix was designed to combine two large scale atmospheric conditions to drive the model. Two LCLU scenarios for the corresponding atmospheric time frames provided the surface characteristics for the simulations. The two 5 year periods from 1955 to 1959, for the past conditions, and from 2000 to 2004, for the present conditions, were selected under a series of criteria that minimizes the influences of global scale oscillations (e.g., ENSO and NAO) in the Caribbean ERS climate, and represents a significant climate change in the region of interest. [29] The large scale temperature changes at the times of local overnight low and daytime high temperatures over the 13 of 18

14 Figure 12. Differences in the total column liquid water content (g kg 1 ) (a) due to LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) due to LCLU changes while driving the model with past atmospheric conditions, (c) total change due to LCLU changes and global warming, (d) due to global warming with present LCLU specifications, (e) due to global warming with past LCLU specifications, and (f) the contribution due to the nonlinear interaction among the LCLU and global warming factors. Caribbean region show moderate increases in temperature on the order of 1.0 to 1.8 C, respectively; these values increase toward the North Atlantic. These changes penetrate vertically from the surface to about the 700 mbar pressure level over the island of Puerto Rico. Domain wide trade winds speed changes between the two climate periods show an increase in wind magnitudes that have been attributed to an increase in the North Atlantic high pressure system. This change in the pressure difference between the Caribbean and the North Atlantic can also be evident in the SST change, as reflected in changes in sea surface temperature for the ERS peak only at about C around the island of Puerto Rico, but increase to the northeast of the island reach values closer to 1 C in the region covered by modeling grid 1. [30] This global climate change signal was found have an important role in producing changes in the region s wind pattern by increasing the magnitude of the approaching trade winds and by changing its direction to an easterly or slightly southeasterly flow. These changes cause a shift in the location of a convergence zone at the time of highest temperature difference values in the daily evolution of the Total Change case. The convergence zone had its axis along the ridge of the Central Mountain range in the simulations driven by past climate conditions, and this axis changed to a northern position just off the range s foothills in the simulations driven by the present climatology. Without the added orographic lifting mechanism that enhances convection above the mountaintops and accounts for cloud formation and rain development in those areas, the base of the clouds are higher and the total column liquid water content is decreased along the Central Mountain ridge and El Yunque. This combination of elements translates into a dramatic decrease in surface accumulated precipitation during the ERS in the region of interest, with strong precipitation 14 of 18

15 Figure 13. Differences in the total 3 month accumulated surface precipitation (mm) (a) due to LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) due to LCLU changes while driving the model with past atmospheric conditions, (c) total change due to LCLU changes and global warming, (d) due to global warming with present LCLU specifications, (e) due to global warming with past LCLU specifications, and (f) the contribution due to the nonlinear interaction among the LCLU and global warming factors. 15 of 18

16 Figure 14. Percent change in the total 3 month accumulated surface precipitation (%) (a) due to LCLU changes while driving the atmospheric mesoscale model with present atmospheric conditions, (b) due to LCLU changes while driving the model with past atmospheric conditions, (c) total change due to LCLU changes and global warming, (d) due to global warming with present LCLU specifications, (e) due to global warming with past LCLU specifications, and (f) the contribution due to the nonlinear interaction among the LCLU and global warming factors. reductions in the Central Mountain range and El Yunque. The shift in the convergence zone leads to increased vertical motions over its present location that do not cause a lifting of the cloud base heights along its axis in the present location, but increased amounts of total column liquid water content are present in the region between El Yunque and the Central Mountains. This suggests the presence of either a thicker cloud layer in that column, or a cloud with increased concentration of cloud droplets and/or raindrops over Caguas that turn leads to an increase of surface accumulated precipitation in that zone. [31] To improve the results presented in this study, an investigation of remotely sensed parameters should be performed in order to update the atmospheric model surface characteristics. Parameters that are used in the regional atmospheric model hydrological cycle parameterizations and schemes include volumetric soil moisture content, which could be incorporated into the measurements obtained by the next generation of remote sensors. In addition to an improvement of input parameters, in terms of analysis of the model results, a better representation of the cloud base height differences should be explored. Projections for future climate changes in tropical coastal regions could be performed using both different GHG emissions scenarios [Meehl et al., 2007] and statistical data for future population growth as a proxy for urban development [Velazquez Lozada et al., 2006]. Downscaled and detailed climate change projections need to be a priority in order for critical mitigation strategies and adaptation policies to be established. [32] Acknowledgments. This research was partially funded by the NASA EPSCoR program of the University of Puerto Rico and by NOAA CREST grant NA06OAR The atmospheric model simulations were performed at the High Performance Computing Facilities of the University of Puerto Rico at Río Piedras. 16 of 18

17 References Aceituno, P. (1988), On the functioning of the Southern Oscillation in the South American sector. Part I: Surface climate, Mon. Weather Rev., 116, , doi: / (1988)116<0505:otfots>2.0.co;2. Angeles, M. E., J. E. Gonzalez, D. J. Erickson III, and J. L. Hernández (2007), Predictions of future climate change in the Caribbean region using global circulation models, Int. J. Climatol., 27, , doi: /joc Angeles, M. E., J. E. González, N. D. Ramírez Beltrán, C. A. Tepley, and D. E. Comarazamy (2010), Origins of the Caribbean rainfall bimodal behavior, J. Geophys. Res., 115, D11106, doi: /2009jd Bruijnzeel, L. A., and J. Proctor (1993), Hydrology and biogeochemistry of tropical montane cloud forests: What do we really know?, in Tropical Montane Cloud Forests: Proceedings of an International Symposium, edited by L. S. Hamilton, J. O. Juvik, and F. N. Scatena, pp , Springer, New York. Burnaby, T. P. (1953), A suggested alternative to the correlation coefficient for testing the significance of agreement between pairs of time series, and its application to geologic data, Nature, 172, , doi: / b0. Chen, A. A., and M. A. Taylor (2002), Investigating the link between early season Caribbean rainfall and the El Niño + 1 year, Int. J. Climatol., 22, , doi: /joc.711. Comarazamy, D. E. (2010), Climate impacts of land cover and land use changes in coastal tropical regions under conditions of global climate change, Ph.D. thesis, Dep. of Mech. Eng., Santa Clara Univ., Santa Clara, Calif. Comarazamy, D. E., and J. E. González (2008), On the validation of the simulation of early season precipitation on the island of Puerto Rico using a mesoscale atmospheric model, J. Hydrometeorol., 9, , doi: /2007jhm Comarazamy, D. E., J. E. Gonzalez, C. A. Tepley, S. Raizada, and R. V. R. Pandya (2006), Effects of atmospheric particle concentration on cloud microphysics over Arecibo, J. Geophys. Res., 111, D09205, doi: /2005jd Comarazamy, D. E., J. E. González, J. C. Luvall, D. L. Rickman, and P. J. Mulero (2010), A land atmospheric interaction study in the coastal tropical city of San Juan, Puerto Rico, Earth Interact., 14, 1 24, doi: /2010ei Cotton, W. R., et al. (2003), RAMS 2001: Current status and future directions, Meteorol. Atmos. Phys., 82, 5 29, doi: /s Curtis, S., and D. W. Gamble (2007), Regional variations of the Caribbean mid summer drought, Theor. Appl. Climatol., 94, 25 34, doi: / s Daly, C., E. H. Helmer, and M. Quiñones (2003), Mapping the climate of Puerto Rico, Vieques and Culebra, Int. J. Climatol., 23, , doi: /joc.937. Enfield, D. B., and E. J. Alfaro (1999), The dependence of Caribbean rainfall on the interaction of the tropical Atlantic and Pacific oceans, J. Clim., 12, , doi: / (1999)012<2093:tdocro>2.0. CO;2. Franco, P. A., P. L. Weaver, and S. Eggen McIntosh (1997), Forest resources of Puerto Rico, 1990, Resour. Bull. SRS 22, USDA For. Serv., South. Res. Stn., Ashville, N. C. Gamble, D. W., D. B. Parnell, and S. Curtis (2008), Spatial variability of the Caribbean mid summer drought and relation to north Atlantic high circulation, Int. J. Climatol., 28, , doi: /joc Giannini, A., Y. Kushnir, and M. A. Cane (2000), Interannual variability of Caribbean rainfall, ENSO and the Atlantic Ocean, J. Clim., 13, , doi: / (2000)013<0297:ivocre>2.0.co;2. Giannini, A., Y. Kushnir, and M. A. Cane (2001a), Seasonality in the impact of ENS0 and the North Atlantic high on Caribbean rainfall, Phys. Chem. Earth, 26(2), Giannini, A., M. A. Cane, and Y. Kushnir (2001b), Interdecadal changes in the ENSO teleconnection to the Caribbean region and the North Atlantic Oscillation, J. Clim., 14, , doi: / (2001) 014<2867:ICITET>2.0.CO;2. Hamilton, L. S., J. O. Juvik, and F. N. Scatena (1993), The Puerto Rico tropical cloud forest symposium: Introduction and workshop synthesis, in Tropical Montane Cloud Forests, edited by L. S. Hamilton, J. O. Juvik, and F. N. Scatena, pp. 1 16, Springer, New York. Hastenrath,S.(1976),Variations in the low latitude circulation and extreme climatic events in the tropical Americas, J. Atmos. Sci., 33, , doi: / (1976)033<0202:villca>2.0.co;2. Intergovernmental Panel on Climate Change (2007), Summary for policymakers, in Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., pp. 1 18, Cambridge Univ. Press, Cambridge, U. K. Jury, M. R., and A. Winter (2010), Warming of an elevated layer over the Caribbean, Clim. Change, 99, , doi: /s Jury, M. R., S. Chiao, and E. W. Harmsen (2009), Mesoscale structure of trade wind convection over Puerto Rico: Composite observations and numerical simulation, Boundary Layer Meteorol., 132, , doi: /s Kalnay, E., et al. (1996), The NCEP/NCAR 40 year reanalysis project, Bull. Am. Meteorol. Soc., 77, , doi: / (1996) 077<0437:TNYRP>2.0.CO;2. Lawton, R. O., U. S. Nair, R. A. Pielke Sr., and R. M. Welch (2001), Climatic impact of tropical lowland deforestation on nearby montane cloud forests, Science, 294, Lawton, R. O., U. S. Nair, D. K. Ray, A. Regmi, A. Pounds, and R. M. Welch (2011), Quantitative measures of immersion in cloud and the biogeography of cloud forest, in Mountains in the Mist: Science for Conserving and Managing Tropical Montane Cloud Forest, edited by L. A. Bruijnzeel, F. N. Scatena, and J. O. Juvik, Univ. of Hawaii Press, Honolulu, in press. Magaña, V., J. A. Amador, and S. Medina (1999), The mid summer drought over Mexico and Central America, J. Clim., 12, , doi: / (1999)012<1577:tmdoma>2.0.co;2. Malmgren, B. A., and A. Winter (1999), Climate zonation in Puerto Rico based on principal components analysis and an artificial neural network, J. Clim., 12, , doi: / (1999)012<0977: CZIPRB>2.0.CO;2. Malmgren, B. A., A. Winter, and D. Chen (1998), El Niño Southern Oscillation and North Atlantic Oscillation control of climate in Puerto Rico, J. Clim., 11, , doi: / (1998)011<2713: ENOSOA>2.0.CO;2. Meehl, G. A., et al. (2007), Global climate projections, in Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., pp. SM10 1 SM10 8, Cambridge Univ. Press, Cambridge, U. K. Meyers, M. P., R. L. Walko, J. Y. Harrington, and W. R. Cotton (1997), New RAMS cloud microphysics parameterization. Part II: The twomoment scheme, Atmos. Res., 45, 3 39, doi: /s (97) Nair,U.S.,R.O.Lawton,R.M.Welch,andR.A.PielkeSr.(2003), Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of cumulus cloud field characteristics to lowland deforestation, J. Geophys. Res., 108(D7), 4206, doi: /2001jd Pielke, R. A., et al. (1992), A comprehensive meteorological modeling system RAMS, Meteorol. Atmos. Phys., 49, 69 91, doi: / BF Pounds, A. J., M. P. L. Fogden, and J. H. Campbell (1999), Biological response to climate change on tropical mountain, Nature, 398, , doi: / Ray, D. K., U. S. Nair, R. O. Lawton, R. M. Welch, and R. A. Pielke Sr. (2006), Impact of land use on Costa Rican tropical montane cloud forests: Sensitivity of orographic cloud formation to deforestation in the plains, J. Geophys. Res., 111, D02108, doi: /2005jd Ropelewski, C. F., and M. S. Halpert (1996), Quantifying southern oscillation precipitation relationships, J. Clim., 9, , doi: / (1996)009<1043:QSOPR>2.0.CO;2. Saleeby, M. S., and W. R. Cotton (2004), A large droplet mode and prognostic number concentration of cloud droplets in the Colorado State University Regional Atmospheric Modeling System (RAMS) Part I: Module descriptions and supercell test simulations, J. Appl. Meteorol., 43, , doi: / (2004)043<0182:almapn>2.0. CO;2. Smith, T. M., and R. W. Reynolds (2003), Extended reconstruction of global sea surface temperatures based on COADS data ( ), J. Clim., 16, , doi: / Smith, T. M., R. W. Reynolds, T. C. Peterson, and J. Lawrimore (2008), Improvements to NOAA s historical merged land ocean surface temperature analysis ( ), J. Clim., 21, , doi: / 2007JCLI Stein, U., and P. Alpert (1993), Factor separation in numerical simulations, J. Atmos. Sci., 50, , doi: / (1993)050<2107: FSINS>2.0.CO;2. Still, C. J., P. N. Foster, and S. H. Schneider (1999), Simulating the effects of climate change on tropical montane cloud forests, Nature, 389, Taylor, M. A., D. B. Enfield, and A. A. Chen, (2002), Influence of the tropical Atlantic versus the tropical Pacific on Caribbean rainfall, J. Geophys. Res., 107(C9), 3127, doi: /2001jc Trenberth, K. E., et al. (2007), Observations: Surface and atmospheric climate change, in Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of 17 of 18