RAINFALL ESTIMATION OVER BANGLADESH USING REMOTE SENSING DATA

Transcription

1 RAINFALL ESTIMATION OVER BANGLADESH USING REMOTE SENSING DATA Final Report June 2006 A. K. M. Saiful Islam M. Nazrul Islam INSTITUTE OF WATER AND FLOOD MANAGEMENT & DEPARTMENT OF PHYSICS BANGLADESH UNIVERSITY OF ENGINEERING AND TECHNOLOGY 1

2 Contents TABLE OF CONTENTS Page no. TABLE OF CONTENTS... i LIST OF FIGURES... ii LIST OF TABLES... v ACKNOWLEDGEMENTS... vi ABSTRACT... vii Chapter 1: Introduction Background and Present State of the Problem: Objective with Specific Aims and Possible Outcome: Outline of Methodology / Experimental Design:... 2 Chapter 2: Use of TRMM in determining the climatic characteristics of rainfall over Bangladesh Introduction Data and Method of Analysis Results Point-to-point comparison of TRMM 3B42 and RNG rates Day-to-day comparison of TRMM 3B42 and RNG rates Rain climatology obtained from TRMM 3B42 and RNG Rain climatology at five selected stations over Bangladesh Rain climatology in different rainy periods in Bangladesh Long-term rain climatology derived from TRMM in Bangladesh Accuracy of TRMM estimates in Bangladesh Discussion Chapter 3: Diurnal variation of rainfall in and around Bangladesh Introduction Data and methods Wind condition in Bangladesh Results and discussion A. Diurnal variation of precipitation in and around Bangladesh Vertical structure of precipitation in and around Bangladesh Chapter 4: Statistical Analysis of TRMM and Rain-gauge rainfall Introduction Data and assumptions Statistical Methods Results of statistical analysis Spatial distribution of the statistical indicators Time series comparison Comparison of mean monthly rainfall over five years period Chapter 5: Conclusions and Recommendations Summary of rainfall characteristics Summary of diurnal variation of rainfall Summary of Statistical Analysis Recommendations REFRENCES i

3 LIST OF FIGURES Figure Page no. Figure 2.1 Daily rainfall (mm) determined by RNG (left of plus mark) and TRMM (right of plus mark) at different rain-gauge sites (plus mark) throughout Bangladesh. Data averages for 1 March to 30 November from The shaded arrow represents the route of monsoon progression Figure 2.2 Daily rainfall (DRF) measured by RNG and TRMM at five selected stations (a) Teknaf, (b) Dhaka, (c) Sylhet, (d) Rangpur, and (e) Khulna in Vertical lines have been used to divide the entire rainy season into three periods Figure 2.3 Same as Fig. 2.2 but at Teknaf in (a) 1999 and (b) Figure 2.4 Rainfall (mm/day) measured from TRMM 3B42 and rain-gauge (RNG) data for a 5-year period at each selected station Figure 2.5 Same as Fig. 2.2 except that rainfall is averaged from 31 stations throughout Bangladesh in (a) 1998 and (b) Figure 2.6 Comparison of the rainfall estimated by TRMM and RNG in different years averaged for 31 stations throughout the country. Rainy days (in %) determined from RNG and TRMM in different years and averaged for (98-02). A rainy day is defined as any day that has a measurable amount of rainfall, and match days is any day detected as rainy by both TRMM and RNG Figure 2.7 Same as Fig. 2.2 except that the averages are for 5 years ( ) and at (a) Khulna and (b) Teknaf Figure 2.8 Daily rainfall determined from RNG (left of plus mark) and TRMM (right of plus mark) at different stations (plus mark). Data averaged for 1 June to 30 September from The shadowed region represents the region where the rainfall was underestimated by TRMM Figure 2.9 Daily rainfall measured by rain-gauge (RNG) and TRMM 3B42. Averages from for (a) 5 stations and (b) 31 stations throughout Bangladesh Figure 2.10 Anomaly of the average daily humidity (in %) over Bangladesh. Averages from 1 March to 30 November in each year for Positive and negative values represent surpluses and deficits of humidity, respectively. Positive and negative values also represent wet and dry regions, respectively ii

4 Figure 2.11 Rainfall (mm/day) averaged for the wet region (high humidity) and the dry region (low humidity). Data averages for 17 stations in the wet region and 14 stations in the dry region throughout Bangladesh from Figure 2.12 Typical structure of the precipitation field of (a) pre-monsoon: 23 May 2002 at 19:14 LST and (b) monsoon: 31 July 2001 at 08:44 LST. The available data from the PPI scan of BMD at 18:04 LST and the vertical profile of the precipitation field along line AB are also shown in (a). The available data from the PPI scan of BMD at 08:46 LST and the vertical cross section of the precipitation field along line AB are shown in (b). The dashed lines show the TRMM pass Figure 3.1 Diurnal variation of precipitation determined by TRMM-3B42RT during monsoon period (JJAS) Figure 3.2 Diurnal variation of precipitation determined by TRMM-3B42RT during premonsoon period (MAM) Figure 3.3 Diurnal variation of rainfall determined by TRMM-3B42RT and RNG Figure 3.4 Same as Fig. 5 except for averaged for entire rainy season Figure 3.5 Vertical extension of precipitation field obtained by TRMM-2A25 data over Bangladesh a) 27 April, over Bay of Bengal b) 23 July and c) 30 October Figure 3.6 Vertical extension of precipitation field obtained by TRMM-2A Figure 3.7 Echo base-and top height determined by both TRMM-2A25 in a) April, b) July and c) October Figure 3.8 Echo base-and top height and maximum rain rate determined by TRMM-2A Figure 4.2 Scatter plot of Daily Rainfall (mm/day) estimated from TRMM 3B42 data and rain-gauge over Bangladesh during Figure 4.4 Contour maps of (a) slope (monthly data) and (b) slope (daily data) of linear regression line, (c) RMSD (monthly data) and (d) RMSD (daily data) over Bangladesh during Figure 4.5 Time series plots of Mean Daily Rainfall (mm/day) obtained from the average of all 33 rain-gauge stations and TRMM 3B42 over Bangladesh during Figure 4.6 Time series of Monthly Rainfall (mm/month) estimated from TRMM (3B42) and rain-gauge over Bangladesh during iii

5 Figure 4.7 Histograms of Monthly Average Rainfall (mm/month) estimated from TRMM 3B42 data and rain-gauge over Bangladesh during iv

6 LIST OF TABLES Table Page no. Table 1.1. Sourcest, availability, collection frequency and resolution of data... 2 Table 2.1. Rainfall and rainfall biases calculated from TRMM and RNG rainfall averaged from 1 March to 30 November in each year at different stations over Bangladesh Table 2.2. Average daily rainfall calculated from TRMM and RNG data in the premonsoon, monsoon, and post-monsoon period and the entire rainy period (March- November). Averages from 5 well-separated stations (5st5Yr) and 31 stations (31st5Yr) over Bangladesh Table 3.1. Rainfall (mm/day) determined by rain-gauge (RNG) and TRMM-3B42RT in different periods and averaged from 2002 to Table 3.2. Echo Parameters determined by TRMM-2A Table 4.1. Statistical Analysis of between rain-gauge and TRMM (3B42) dataset of monthly rainfall accumulation over Bangladesh during v

7 ACKNOWLEDGEMENTS We hereby express our sincerest appreciation for the encouragement and support that we received from our Institute in carrying out the study. We have been immensely benefited by many informal discussions that we held with our learned colleagues at different stages of the study. We are thankful to the BMD staff for supplying us rainfall and radar data. Mr. Yamamoto and Shingo Shimizu of HyARC, Nagoya University, Japan, are gratefully acknowledged for their assistance with copying TRMM 2A25 data. The TRMM 2A25 data were provided by JAXA from HyARC, Nagoya University, Japan. The TRMM 3B42 data were acquired from their website. TRMM is an international project jointly sponsored by the Japan National Space Development Agency (NASDA) and the U. S. National Aeronautics Space Administration (NASA) Office of Earth Science. vi

8 ABSTRACT In recent years, rainfall estimation from remotely sensed data has been considered a viable alternative over the traditional rain-gauge measurements. In this report, daily rainfall data with one degree resolution measured by Tropical Rainfall Measuring Mission (TRMM) satellite with the 3B42 processing algorithm has been used to quantify the precipitation over Bangladesh. TRMM 3B42 data has been compared to understand climatic characteristics of rainfall and to established correlation with traditional raingauge data collected by the Bangladesh Meteorological Department (BMD) from 33 raingauge stations located all over the country during the period Daily rainfall measured by TRMM 3B42 was compared to that of rain-gauge values from pre-monsoon to post-monsoon months (March-November). The time sequence patterns of the rainfall determined by TRMM and those from rain-gauges were remarkably similar. The spatial and temporal averages of rainfall revealed good estimations of rainfall: during March to November, the TRMM 3B42- and rain gauge-estimated daily rainfall was 8.12 and 8.34 mm, respectively. The average percentage of rainy days determined by TRMM data with respect to the rain-gauge value was 96%. TRMM 3B42 is useful for estimating the average values of rainfall in Bangladesh. The prominent difference between rainfall estimated by rain-gauge and TRMM 3B42 was found to be period- and locationdependent. TRMM 3B42 overestimated the rainfall during the pre-monsoon period and in dry regions but underestimated it during the monsoon period and in wet regions. The reason for the differences according to season and locations is considered to be the vertical cross section of convection obtained by TRMM-PR 2A25 data. During the monsoon, in the southeast of Bangladesh, convection at a moderate height prevailed, making the 3B42 estimation of rainfall smaller than that from a rain-gauge. In this manner, the merit of using TRMM data for climatological studies of rainfall over Bangladesh is shown. This research also studied the diurnal variation of precipitation using TRMM-3B42RT data of from pre-monsoon to post-monsoon periods in and around Bangladesh. It is found that maximum rainfall peak over Bangladesh and northeast of the vii

9 Bay of Bengal were appeared at 06 LT (local time) with the minimum at 21 LT during monsoon period. The maximum rainfall peak over India was found at 18 LT with the minimum at 09 LT. The morning maximum rainfall at 06 LT over Bangladesh determined by TRMM-3B42RT is confirmed after compared with the same obtained from ground-based rain-gauge (RNG) data. Another secondary peak maximum rainfall is observed at 15 LT from both TRMM and RNG data. Post-monsoon rainfall followed the similar patterns of diurnal variations as monsoon whereas pre-monsoon maximum peak is observed from late-night to mid-night. This study also represents the vertical structure of precipitation using TRMM-2A25 data of TRMM-2A25 data analysis reveals that pre-monsoon, monsoon and post-monsoon precipitations are strong, moderate and less intensified, respectively. Strong rain rates are found at higher altitudes for premonsoon and relatively at lower altitudes for later periods. Averages from 4 years data, it is found that maximum rain rate in April, July and October is , and mm/h, respectively. In general, pre-monsoon echoes are tall compared to monsoon and post-monsoon periods. However, the maximum echo top height of about 18.25, 18.8 and km is found during the pre-monsoon, monsoon and post-monsoon periods, respectively. This analysis revealed that in south Asia, the overestimation and underestimation of rainfall by TRMM-3B42RT depends on different vertical structures of precipitations in corresponding periods. Statistical analysis reveals that high correlation (Pearson s correlation coefficient is 0.97) between these satellite and ground based rain-gauge datasets has found for monthly accumulated rainfall. However, daily data shows moderate correlation with a correlation coefficient is The pattern observed from the time series plot of the mean value of rainfall from these two datasets is matched quite well. Though TRMM 3B42 data shows very close estimation with rain-gauge, it underestimates in calculation of rainfall in places where mean daily rainfall is above 6.0 mm/day. In heavy rainfall zones located very close to the coastal regions like Teknaf or Coxsbazar, this underestimation is nearly 50% which shows the huge mismatch of the TRMM 3B42 with rain-gauge data. Hence, it can be concluded that TRMM 3B42 underestimated rainfall in the coastal and heavy rainfall regions of the country. viii

10 Chapter 1: Introduction 1.1. Background and Present State of the Problem: Rainfall data is considered an important parameter for studying water resources related problems like flood which is a common phenomenon in Bangladesh. However, developing a rainfall forecasting model is a challenging task considering the technical and political circumstances. Technical problems arise from the lack of ground-based rainfall measuring equipments such as radar network or automated rain-gauges. The solution to overcome these problems lies on the use of satellite based remote sensing devices. In and around Bangladesh, the rainy season is divided into three periods (Das, 1995): (a) pre-monsoon (March-May), (b) monsoon (June-September), and (c) postmonsoon (October-November). In Bangladesh, about 20%, 62.5%, 15.5%, and 2% of the annual rainfall (~2700 mm) occurs during pre-monsoon, monsoon, post-monsoon, and winter periods, respectively (Islam & Uyeda, 2005). Research conducted on the estimation of rainfall in Bangladesh using remote sensing data remains inadequate. Recently, Ohsawa et al. (2001) studied the relationship of T BB from GMS (Geostationary Meteorological Satellite) data to RNG rainfall in Bangladesh. Islam et al. (2002; 2003) shows an early attempt to estimate rainfall in Bangladesh from GMS-5 satellite data using Convective Stratiform Technique (CST) and these estimated rainfall were calibrated with rain-gauge rainfall for the Bangladesh Meteorological Department (BMD) rain-gauge stations located at different parts of the country. Although preliminary results can estimate rainfall successfully in some areas, the study suggested long time verification of the results and using of radar data. In that context, we have planned to use three dimensional reflectivity data of Tropical Rainfall Measuring Mission (TRMM) to find the cloud characteristics and precipitation system developed around Bangladesh. Since, TRMM is the first satellite earth observation mission to monitor tropical rainfall, which closely influences to global climate and environment change (NASDA, 2001), this study would helpful for the disaster prevention as well as water management of Bangladesh. 1

11 1.2. Objective with Specific Aims and Possible Outcome: The specific objectives of this study are: 1) to investigate cloud characteristics of precipitation systems developed over Bangladesh. 2) to develop a linear relationship between TRMM-PR satellite data and actual rainfall over Bangladesh Outline of Methodology / Experimental Design: This study will identify the cloud characteristics and pattern of precipitation systems developed over Bangladesh. Three different types of data will be used in this project: 1) daily average precipitation data all over the Bangladesh, 2) precipitation radar (PR) data (2A25) from TRMM satellite in and around Bangladesh, and 3) Dhaka radar data. Detailed descriptions of data products including data types, availability, collection frequency and resolution are given in Table 1. Table 1.1 : Sourcest, availability, collection frequency and resolution of data Data Type Available Collection Frequency Resolution Rain-gauge Rainfall Daily precipitation data from rain-gauge TRMM-PR data TRMM satellite data 3B42 products one image per day Dhaka Radar PPI Collected at Dhaka scans data for 1 hour with 2 hours leisure. 33 rain gauge stations located all over the Bangladesh 4km radius 2.5 km x 2.5 km Major steps of data analysis can be summarizes as follows: (1) collection of three different types of data, (2) processing of TRMM-PR 2A25 data for Bangladesh grid (0.1 km by 0.1km grid), (3) processing of Dhaka radar data for same Bangladesh grid, (4) 2

12 statistical analysis of these processed data to find correlation and linear regression coefficients, (5) identify the cloud characteristics based on 3D cloud structures for a given latitude or a longitude. We have been some initial data acquisition and processing of TRMM-PR data using open source algorithms and C++ programs. Further analysis and processing of yearly data requires powerful computer and data storage facilities. Methods of analyses are explained in detail in each chapter. 3

13 Chapter 2: Use of TRMM in determining the climatic characteristics of rainfall over Bangladesh 2.1. Introduction The Tropical Rainfall Measuring Mission (TRMM), cosponsored by NASA of the U. S. A. and JAXA of Japan, has collected data since November 1997 (Kummerow et al., 2000). Tropical rainfall, which falls between 35 N to 35 S, comprises more than twothirds of global rainfall. TRMM is a long-term research program designed to study the Earth's land, oceans, air, ice, and life as a total system. Previous estimates of tropical precipitation were usually made on the basis of weather models and the occasional inclusion of very sparse surface rain-gauges (RNGs) and/or relatively few measurements from satellite sensors. The TRMM satellite allows these measurements to be made in a focused manner. TRMM is NASA's first mission dedicated to observing and understanding tropical rainfall and how it affects the global climate (Wolff et al., 2005). The TRMM spacecraft fills an enormous void in the ability to calculate worldwide precipitation because ground-based radars that measure precipitation cover a very small part of the planet. Ground-based radars cover only 2 percent of the area covered by TRMM, and RNGs are limited to specific geographic points. The TRMM Ground Validation (GV) Program began in the late 1980s and has yielded a wealth of data and resources for validating TRMM satellite estimates by providing rainfall products for four sites: Darwin, Australia (DARW); Houston, Texas (HSTN); Kwajalein, Republic of the Marshall Islands (KWAJ); and Melbourne, Florida (MELB). Wolff et al. (2005) provide extensive details on the TRMM GV program, site descriptions, algorithms, and data processing. For years, other groups studied various locations to validate TRMM data, such as Ikai et al. (2003), who calculated rain rates over the Ocean, Nicholson et al. (2003), who validated TRMM rainfall for West Africa, and Barros et al. (2000), who studied a monsoon case in Nepal. As the TRMM mission winds down, with the eventual atmospheric re-entry of the satellite, focus has begun to plan for the Global Precipitation Measurement (GPM) GV program. The GPM satellite will commence operations in The main goal of the TRMM GV program is to provide rain estimates at various sites throughout the globe in order to compare with and hopefully help to improve GPM 4

14 satellite retrievals (Bidwell et al., 2002). Unfortunately, so far, there has been no research done to validate TRMM data over Bangladesh. This paper describes how accurately the TRMM satellite can determine surface rainfall in Bangladesh. In Bangladesh, RNG data are the only representation of precipitation throughout the country. Inadequate RNG networks throughout the country sometimes provide inadequate information on the distribution of rainfall throughout the country. The use of remote sensing data in estimating rainfall in Bangladesh, thus, becomes a major task. Neither RNGs nor satellite-based estimates are perfect indicators of rainfall (Nicholson et al., 2003). Morrissey and Greene (1993) and Xie and Arkin (1995; 1996) show in their examinations that all satellite estimates have non-negligible biases when compared with concurrent in situ observations. With RNGs, biases are introduced by gauge type, maintenance, and placement (Legates & Willmott, 1990; Sevruk 1982) as well as by spatial sampling (Huffman, 1997; Huffman, Adler, Schneider, & Keehn, 1995; Morrissey & Greene, 1993; Rudolf, Hauschild, Rüth, & Schneider, 1994). Xie and Arkin (1995) concluded that these are small when compared with the bias in satellite estimates. Even when these biases are included in the measurements, RNG rainfall is used as a ground-based measurement and is more reliable than satellite-based rainfall measurements. As mentioned above, very little validation work has been conducted on the rainfall estimated by TRMM over Bangladesh, and the GPM GV is an on-going project. A number of researchers are working on the development of instrumentation and algorithms for GPM. Information about the distribution of rainfall and the structure of precipitation systems from a heavy rainfall region, such as Bangladesh, is important for these developers. In this work, attempts have been made to compare the rainfall determined by TRMM 3B42 products, which is the combination of TRMM Precipitation Radar (PR) and TRMM Microwave Imager (TMI), with the values of ground-based RNGs throughout Bangladesh. Pre-monsoon, monsoon, and post-monsoon rain rates have distinctive features in Bangladesh as well as in parts of Asia that experience monsoons. These are also obtained from TRMM 3B42, TRMM-PR 2A25, and RNG data. Radar PPI scans data of the Bangladesh Meteorological Department (BMD) are used to support the TRMM-PR horizontal images. 5

15 2.2. Data and Method of Analysis Rainfall data for 3-hour periods from 1998 to 2002, measured and collected by RNGs placed at 31 locations throughout Bangladesh by the BMD, were used to prepare the RNG rainfall data. TRMM produced a daily 1 1 microwave-calibrated IR rain estimate (TRMM Science Data and Information System (TSDIS) 3B42). The daily data from TRMM 3B42 products estimated gridded rainfall of 1 1 resolution for the same analysis period. The RNG values were not uniformly gridded because the RNGs of BMD are not positioned at uniform grid distances. Therefore, the comparison between rainfall estimated by TRMM and that measured by RNG was performed on a point-to-point basis, i.e., one RNG must be in a TRMM data grid box. The analysis period of was chosen because it was the only one with a full complement of data for the TRMM and the collected RNGs rainfall in Bangladesh. In the analysis, the entire rainy period of March- November has been accounted for, which is nearly the complete rain period of Bangladesh, with 98% of the total annual precipitation. Except for 2A25, TRMM 3B42 products are available for each day of the 5-year analysis period. Four types of analyses were carried out. First, the mean rainfall field for March to November (MAMJJASON) was prepared for the TRMM 3B42 dataset. Second, day-to-day comparisons of rainfall for five well-separated and selected stations throughout Bangladesh were conducted. The zero-mm rainfall within 3-hour periods has been accounted to calculate the daily rainfall amount. Third, point-to-point comparisons of rain climates were obtained from the TRMM 3B42 and RNG datasets for the three rainy periods. Fourth, rainy days for each month, each year, and the overall 5-year period were detected with the RNG dataset as a reference. Any amount of rainfall recorded by an RNG within 3 hours was defined as a rainy day, whereas zero rainfall within a 24-hour period was defined as a rain-free day. Rainy days simultaneously detected by both TRMM and RNG were declared as match days. The dry and wet regions were defined as areas in which the relative humidity was below and above the averaged relative humidity value of the country, respectively. The monsoon onset date in a particular region was determined by a change of the surface wind field from mainly NW to S/SE and recorded precipitation during at least 2-3 consecutive days. Anomalies were defined by deviations in a country-averaged value from individual ones. 6

16 2.3. Results This section will show several comparisons of RNG rain intensities and satellite retrievals from the TRMM Microwave Imager (TMI) and Precipitation Radar (PR) algorithms. The TRMM data used for these comparisons was obtained from the 3B42 gridded rainfall product developed by TSDIS Point-to-point comparison of TRMM 3B42 and RNG rates Daily rainfall (mm) estimated from TRMM 3B42 (right of plus mark, Fig. 2.1) and RNG data (left of plus mark, Fig. 1) averaged for the period of 1 March to 30 November at the location of each RNG (plus mark, Fig. 2.1) throughout Bangladesh are shown in Fig. 1. TRMM underestimated the rain rate in the northeastern and southeastern parts of Bangladesh, which are the heavy-rainfall regions of the country (Islam, Terao, Uyeda, Hayashi, & Kikuchi, 2005). The determination of the rain rate in the remaining parts of the country by both techniques is comparable Day-to-day comparison of TRMM 3B42 and RNG rates For the day-to-day comparison, five stations, Teknaf (southeast), Dhaka (center), Sylhet (northeast), Rangpur (northwest), and Khulna (southwest), were selected, as shown in Fig A comparison of the daily rainfall measured by TRMM 3B42 and RNG at the five selected stations in 2001 is shown in Fig Vertical lines in Fig. 2.2 were used to divide the entire rainy season into pre-monsoon, monsoon, and post-monsoon periods. To represent the date of the monsoon onset in a particular region, a down arrow is used. Teknaf is located in the southeastern coastal region of the country (see Fig. 2.1), and the pre-monsoon system does not generally reach that region. Therefore, the lowest rates were recorded in this region during the first two months (March and April) of the premonsoon period. At the end of the first week or in the beginning of the second week of May, rainfall was first observed in this region when the monsoon circulation was established (not shown) in the South Bay and adjoining areas (south of 13(N)). Generally, the monsoon circulation begins to have an impact in the last week of May. Historically, 7

17 the monsoon onset is 31 May in this region (Das, 1995). The mean monsoon onset dates in Bangladesh have been clarified by Ahmed and Karmaker (1993). However, the rainfall in the last week of May is sometimes indistinguishable from the monsoon onset. The setup of surface wind fields from northeasterly to southwesterly with a consecutive rainfall for about 2 to 3 or more days determined the monsoon onset in this region, and BMD used this criterion for the definition of the onset. BMD concluded that the monsoon onset in the southern parts of Bangladesh (Teknaf) had occurred on 9 June in 1998, 27 May in 1999, 3 June in 2000, 3 June in 2001, and 06 June in 2002 (the onset dates for and 2002 are not shown). However, in this region, TRMM underestimated the rainfall, especially during monsoon months. Dhaka is located in the center of the country. Some rainfall events occurred during the pre-monsoon months during the observed period, and rainfall measured by TRMM compared well to that of the RNG values. Sylhet is located in the northeastern part of the country, and heavy rainfall was recorded in this region from the pre-monsoon to monsoon periods. The pre-monsoon rainfall measured by TRMM is almost comparable with that of the RNG values of Sylhet, but, during the monsoon period, TRMM underestimated the rainfall. At Rangpur, the TRMM calculation of rainfall was quite comparable with the RNG values, whereas, in the monsoon period, the TRMM estimation did not compare well with the point values. In some cases of 2001 (also 1999 and 2002, not shown), the estimation patterns of TRMM showed good similarity with the RNG values, but the amount of deviation was very high, which may be due to the short life of the cloud cells and the single pass of TRMM during the 24-hour period. At Khulna, the average rainfall detected by both TRMM and RNG in 2001 was quite comparable. In some cases in 1999 and 2000 and many cases in 2002, the TRMM estimation was very low in comparison to the RNG values (not shown), supporting the conclusion that the TRMM estimation is very low for heavy rainfall events. Daily rainfall (DRF) measured by TRMM and RNG at Teknaf in 1999 and 2002 is shown in Fig. 2.3 as an example. Rainfall and rainfall biases (bias = TRMM - RNG) in all analyzed years ( ) and at all 5 stations are summarized in Table 2.1. As in Fig. 3, the TRMM and RNG patterns are quite similar, but rainfall measured by TRMM differs substantially from the RNG values. TRMM fails to detect rain intensities with biases of about -6.5 and -4.8 mm/day in 1999 and 2002, respectively (Table 2.1). The 8

18 biases at Teknaf are -6.8, -8.7, and -5.3 mm/day in 1998, 2000, and 2001, respectively. On average, for the five-year analysis, the bias was about 1.7 mm/day at Dhaka, 1.9 mm/day at Khulna, -1.7 mm/day at Rangpur, -3.2 mm/day at Sylhet, and -6.4 mm/day at Teknaf. Therefore, of the 5 locations, TRMM fails to determine rainfall at 3 (negative bias) and succeeded or overrated at 2 (positive bias). In 1999, daily rainfall measured by TRMM and RNG was 10.4 and 16.9 mm/day, respectively, at Teknaf, whereas the values were 9.2 and 14 mm/day in Therefore, in 1999 and 2002, TRMM underestimated 6.5 and 4.8 mm/day, respectively. Hence, it is obvious that the magnitude of the errors in rainfall measurement from TRMM is largely dependent on the calculation site and period. Table 2.1 Rainfall and rainfall biases calculated from TRMM and RNG rainfall averaged from 1 March to 30 November in each year at different stations over Bangladesh. Rainfall and rainfall biases (mm/day) Dhaka Khulna Rangpur Sylhet Teknaf RN TRM Bias RNG TRM Bias RNG TRM Bias RNG TRM Bias RNG TRM Bias G M M M M M Ave The comparative bar diagram of Fig. 2.4 shows the daily rainfall measured from TRMM and RNG data at five selected stations for five years. The daily rainfall from TRMM data is almost always higher than the RNG value at Dhaka and Khulna. In the case of the three other sites, the opposite was observed. Therefore, TRMM underestimated the rainfall at locations that lie within the eastern and northern parts of the country and overestimated the rainfall at locations that lie within the central and southwestern parts of the country. Figure 2.5 shows the comparison of daily average rainfall throughout the country in 1998 and 2001 determined from TRMM and RNG. The down arrow indicates the date of the monsoon onset throughout the country as determined by the BMD. The date of the monsoon onset was delayed in 1998 and normal in Pre-monsoon rainfall prevailed in 1998, while rainfall was deficient during the peak monsoon months, particularly, in 9

19 July and August, The DRF patterns between the RNG and TRMM values have good similarity during the observed period, but there is a slight deviation in calculating the average rainfall; particularly, during the pre-monsoon and post-monsoon months, the TRMM estimates are higher than those of RNG, whereas, during the monsoon months, the TRMM estimates are lower. Hence, it is again noticeable that the magnitude of the errors in rainfall determination from TRMM is largely dependent on the calculation period or on the season of the occurrence of precipitation. Figure 2.6 shows the comparison of rainfall (mm/day) estimated by TRMM and RNG and the percentage of rainy days (in %) and match days (in %) in different years averaged for the 31 stations throughout the country. The TRMM estimation is very close to the RNG values for country-scale averaged daily rainfall, and the variation in the rainfall lies between 0.08 to 0.51 mm. In 2000, measurements between these two systems are very close, but, in 1998, 1999, and 2002, TRMM slightly underestimates the rainfall, while, in 2001, it slightly overestimates it. As explained in Fig. 2.5, 1998 was a normal monsoon year, and, in 2001, there was lack of rainfall during the peak monsoon months. TRMM underestimates the rainfall in the normal monsoon year and overestimates it in an unusual monsoon year during which there is a lack of rainfall during the monsoon months. The rainfall anomaly (individual annual average rainfall minus the annual average value) lies between and 0.52 mm/day for TRMM values and between and 0.52 mm/day for RNG values. On average, from the 5-year analysis, the daily rainfall calculated by TRMM and RNG is 8.12 and 8.34 mm, respectively. Therefore, the TRMM estimation is very close to the RNG values for daily rainfall. The number of rainy days determined by TRMM in 1998 and 2002 was higher than that for RNG and lower in 1999 and However, the number of rainy days determined by RNG and TRMM obtained on average for five years ( ) was almost the same as and very close to that for match days, respectively. There are always fewer match days than rainy days in individual years. The average percentage of rainy days determined by TRMM data with respect to the RNG value is 96%. Hence, TRMM succeeds in detecting rainy days very well. Therefore, TRMM can be used as a tool for estimating rainfall in the parts of Asia, such as Bangladesh, that experience monsoons. 10

20 2.4. Rain climatology obtained from TRMM 3B42 and RNG Rain climatology at five selected stations over Bangladesh Average daily rainfalls (DRFs) for the 5-year period ( ) at Khulna and Teknaf are shown in Fig From the analysis of the average daily rainfall, the estimated value by TRMM is comparable to the RNG value at Khulna (Fig. 2.7 (a)) and Dhaka (not shown) except on some days in which TRMM underestimated it. However, at Teknaf (Fig. 2.7 (b)), TRMM underestimated the rainfall during the monsoon period. The same situation was observed at Sylhet and Rangpur (not shown). Thus, the calculation site and period are shown again to be important factors in estimating rainfall from TRMM data Rain climatology in different rainy periods in Bangladesh The daily average rainfall determined by TRMM and RNG for 31 locations during the three observed periods of the rainy season (pre-monsoon, monsoon, and post-monsoon seasons) is described below: Pre-monsoon: Daily rainfall measured by TRMM together with that of the RNG value at 31 locations during the pre-monsoon period showed that TRMM overestimated the rainfall at 24 locations, had the same measurement at 5 locations, and underestimated it at 1 location (not shown). In this season, the rainfall determined by both processes at different locations throughout the country had a deviation within 1 to 3 mm in some isolated places. Monsoon: Daily average monsoon rainfall values measured by TRMM and RNG at 31 locations throughout Bangladesh are shown in Fig During this period, TRMM- and RNG-estimated rainfall values in the central and western parts of the country are almost the same, with only a small variation. However, TRMM underestimated the rainfall in the eastern, southern, and northern parts, with a variation of up to 12 mm (the shadowed region), and these are the regions through which monsoon progression occurred as indicated in Fig In this period, TRMM overestimated the rainfall at 14 locations, had the same measurement at 3 locations, and underestimated it at 14 locations. Post-monsoon: During the post-monsoon period, daily average rainfall measured by TRMM and RNG throughout the country showed that both techniques determined 11

21 rainfall with almost the same accuracy. In this period, TRMM overestimated the rainfall at 10 locations, had the same measurement at 9 locations, and underestimated it at 10 locations (not shown). The deviation of the rainfall measurement was within 1 to 2 mm in some isolated places. The rain climatology obtained from TRMM 3B42 and RNG rainfall averaged for 5 wellseparated stations and for 31 stations throughout the country for the years is tabulated in Table 2.2. Table 2.2 Average daily rainfall calculated from TRMM and RNG data in the premonsoon, monsoon, and post-monsoon period and the entire rainy period (March-November). Averages from 5 well-separated stations (5st5Yr) and 31 stations (31st5Yr) over Bangladesh. Average daily rainfall (MM/DAY) for RNG- TRMM- 5st5Yr 5st5Yr RNG-31st5Yr TRMM-31st5Yr Pre-monsoon Monsoon Post-monsoon March-November The rain climatology in different rainy periods shows that the daily average rainfall calculated by using the TRMM and RNG data has good similarity in the pre-monsoon and post-monsoon period and the entire rainy period (March-November) for 5 wellseparated stations and 31 stations of Bangladesh. The overall estimation by TRMM is about 97.36% of the RNG rainfall. Hence, TRMM can be used as a successful tool for rainfall estimation in Bangladesh Long-term rain climatology derived from TRMM in Bangladesh The overall periodic analysis of the average DRF for five stations (Fig. 2.9 (a)) shows that there is a good comparison between the TRMM values and the RNG values, with the exception of the mid-june to mid-august period, when TRMM fails to obtain precise estimates. Therefore, TRMM data can be used for long periodic rainfall estimation as well as to prepare rain climatology. The average daily rainfall determined by the TRMM 12

22 and RNG dataset at 31 locations during the observed period ( ) is shown in Fig. 2.9 (b). The patterns of DRF determined by both processes depict good agreement in the case of rainfall amounts as well as the variation of average rainfall. Hence, the comparison is much better for the average from a large number of sites and over the long term Accuracy of TRMM estimates in Bangladesh There are several potential problems that may cause the TRMM satellite to fail, many of which are related to precipitation characteristics in a particular region and a particular period, i.e., the measurement depends on the site and the period. There are many possible reasons for the discrepancies between the RNG and TRMM measurements, such as the fact that TRMM-PR measures the exact rain rate, while measurements from the TMI depend on the rain characteristics in pre-monsoon, monsoon, and post-monsoon periods, and the calculation sites, such as whether they are in heavy- or low-rainfall regions. TRMM overestimated 20.73% and underestimated 11.3% and 0.54% of the surface rain during the pre-monsoon, monsoon, and post-monsoon periods, respectively. The percentages of TRMM to RNG rainfall are 98.27%, 97.54%, 100.4%, %, and 94.59% in 1998, 1999, 2000, 2001, and 2002, respectively, as explained in Fig. 6. It is reasonable to say that, on an annual scale, the RNG and TRMM estimates agree to within % with an average of 98.34% for the 5-year period. In known heavy rainfall regions, such as the northeastern and southeastern parts of Bangladesh, for example, TRMM data results in large differences, and TRMM fails to accurately estimate the rainfall. In the central and western parts of the country, TRMM overestimates the rainfall. The drop size distributions may play a large role in these regional/temporal differences (Berg et al., 2004). Additional studies will be necessary to refine these results using radar rain rates and detailed characteristics of precipitation structures using TRMM-PR 2A25 records in Bangladesh Discussion Daily average rain rates estimated by TRMM data are lower than the RNG values in the 13

23 regions of northeastern and southeastern Bangladesh (shown in Fig. 2.1), which are the regions with the heaviest rainfall (Islam et al., 2005). An analysis of humidity anomalies (individual station average minus the average from all stations), as shown in Fig. 2.10, shows that the eastern, southern, and northern parts of the country are more humid than the central and western parts. The less humid area lies on the southeastern boundary and is hilly. Based on excess (positive anomaly) and deficit (negative anomaly) average humidity, Bangladesh can be divided into two regions: wet and dry. The sites located in the regions of positive and negative anomalies can be applied to determine the wet and dry regions, respectively. The wet region is also in the path of the monsoon progression over the country. The average DRF measured by both TRMM and RNG techniques in the wet and dryregion is shown in Fig Averages are shown from 17 sites located in excessively humid regions in the wet region and from 14 sites located in deficiently humid regions in the dry region during TRMM underestimated 0.97 mm/day in the wet region but overestimated 0.67 mm/day in the dry region. In other words, TRMM underestimated 10.08% of the surface rain in the wet region and overestimated 9.82% in the dry region. On average, TRMM can determine 97.36% of the surface rain in Bangladesh. The structure of the precipitation fields and the strength of the rain rate in Bangladesh have been observed using TRMM 2A25 data records. The typical structure of premonsoon and monsoon period precipitation fields has been analyzed and is shown in Fig The precipitation fields had intensities of up to 65 mm/h in the southeastern part of Bangladesh on 23 May 2002 at 19:14 LST in the pre-monsoon season. The available data from the Plan Position Indicator (PPI) scans of BMD radar at 18:04 LST indicated the development of convection on the same day at the same location. The very southern part of the domain is outside of the radar coverage. The vertical profile of the convective cloud along line AB had vertical extensions of about 17.5 km and a maximum intensity of 45 mm/h. An example of a similar case of a cloud that developed with vertical extensions of 19 km and rain rates beyond 30 mm/h was observed in May 1998, as reported in the TRMM report (2002). During the monsoon period, the rain rate had a 14

24 lower value of about 13 mm/h on 31 July 2001 at 08:44 LST. The vertical extension was about 11.5 km. The BMD radar PPI scan at 08:46 LST on the same date confirmed the development of this event. Hence, it is clear that tall and intense convective clouds developed during the pre-monsoon season whereas, in the monsoon period, less intense short convective clouds developed in Bangladesh. In fact, TRMM 3B42 could successfully detect the development of tall convection during the pre-monsoon period, whereas it failed to detect the development of low-level short convection during the monsoon period. It is reasonable to say that, during the monsoon period, the low-level monsoon flow carrying water vapor from the Bay of Bengal assisted in the development of low-level short convections over the monsoon progression region of the country. Detailed research will be necessary on the precipitation structure in different periods. 15

25 Figure 2.1 Daily rainfall (mm) determined by RNG (left of plus mark) and TRMM (right of plus mark) at different rain-gauge sites (plus mark) throughout Bangladesh. Data averages for 1 March to 30 November from The shaded arrow represents the route of monsoon progression. 16

26 Figure 2.2 Daily rainfall (DRF) measured by RNG and TRMM at five selected stations (a) Teknaf, (b) Dhaka, (c) Sylhet, (d) Rangpur, and (e) Khulna in Vertical lines have been used to divide the entire rainy season into three periods. 17

27 Figure 2.3 Same as Fig. 2.2 but at Teknaf in (a) 1999 and (b)

28 Figure 2.4 Rainfall (mm/day) measured from TRMM 3B42 and rain-gauge (RNG) data for a 5-year period at each selected station. 19

29 Figure 2.5 Same as Fig. 2.2 except that rainfall is averaged from 31 stations throughout Bangladesh in (a) 1998 and (b)

30 Figure 2.6 Comparison of the rainfall estimated by TRMM and RNG in different years averaged for 31 stations throughout the country. Rainy days (in %) determined from RNG and TRMM in different years and averaged for (98-02). A rainy day is defined as any day that has a measurable amount of rainfall, and match days is any day detected as rainy by both TRMM and RNG. 21

31 Figure 2.7 Same as Fig. 2.2 except that the averages are for 5 years ( ) and at (a) Khulna and (b) Teknaf. 22

32 Figure 2.8 Daily rainfall determined from RNG (left of plus mark) and TRMM (right of plus mark) at different stations (plus mark). Data averaged for 1 June to 30 September from The shadowed region represents the region where the rainfall was underestimated by TRMM. 23

33 Figure 2.9 Daily rainfall measured by rain-gauge (RNG) and TRMM 3B42. Averages from for (a) 5 stations and (b) 31 stations throughout Bangladesh. 24

34 Figure 2.10 Anomaly of the average daily humidity (in %) over Bangladesh. Averages from 1 March to 30 November in each year for Positive and negative values represent surpluses and deficits of humidity, respectively. Positive and negative values also represent wet and dry regions, respectively. 25

35 Figure 2.11 Rainfall (mm/day) averaged for the wet region (high humidity) and the dry region (low humidity). Data averages for 17 stations in the wet region and 14 stations in the dry region throughout Bangladesh from

36 Figure 2.12 Typical structure of the precipitation field of (a) pre-monsoon: 23 May 2002 at 19:14 LST and (b) monsoon: 31 July 2001 at 08:44 LST. The available data from the PPI scan of BMD at 18:04 LST and the vertical profile of the precipitation field along line AB are also shown in (a). The available data from the PPI scan of BMD at 08:46 LST and the vertical cross section of the precipitation field along line AB are shown in (b). The dashed lines show the TRMM pass. 27

37 Chapter 3: Diurnal variation of rainfall in and around Bangladesh 3.1. Introduction Bangladesh is one of the summer monsoon active regions in the world. In this region premonsoon convections are characterized by short lifetime and high intensity rainfall whereas monsoon rainfalls are more continuous and less intensified. There is little research work on the diurnal variations of precipitation in Bangladesh using remote sensing data. Some preliminary works are reported by Islam et al. (2004; 2005). There is almost no work on the vertical structure of precipitation over Bangladesh. So, attempts have been taken to determine the vertical structure and diurnal variation of precipitation by TRMM 2A25 and 3B42RT datasets Data and methods TRMM-3B42RT 3-hourly data are analyzed from pre-monsoon to monsoon and postmonsoon periods for the years in a large domain of E and 06-30N. TRMM-2A25 data are analyzed in a relatively smaller domain of 86-94E and 18-28N including Bangladesh to obtain vertical structure in different seasons of the years from 2000 to In this connection, precipitation of 80 vertical layers, each 250 m depth are placed one on top of the other. This provides the vertical extension of precipitation field up to 20 km from the surface. The month of April, July and October are selected for detail analysis as the peak month of pre-monsoon, monsoon and post-monsoon periods, respectively from each year. Rain-gauge data collected at 33 stations by the Bangladesh Meteorological Department (BMD) are also utilized Wind condition in Bangladesh In Bangladesh, low-level wind blows from the northeast and northwest in dry and premonsoon seasons, respectively. Usually, thunderstorms occur from the beginning of premonsoon season. With the advancement of the month, wind direction turns from the 28

38 northwest to the southwest that begins the monsoon (Islam, 2004). The monsoon wind plays an important role in carrying water vapor from the sea to the land Results and discussion A. Diurnal variation of precipitation in and around Bangladesh In and around Bangladesh, precipitation substantially varies with space and time as shown in Fig Daily rainfall (mm/d) determined by TRMM-3B42RT averaged for 1 June to 31 September from 2002 to 2004 at 00, 03, 06, 09, 12, 15, 18 and 21 LT (times are mentioned on each panel) is shown in Fig Over the northeast of the Bay of Bengal, rainfall increases from 00 LT and shows maximum at 06 LT. After this time it decreases and shows minimum at 21 LT. It is also found that heavy rain falls in the northern and the southern parts of Bangladesh, which are the heavy-rainfall regions in the country (Islam et al., 2005), from 03 to 06 LT where 06 LT is the time of heaviest rainfall. The time of heaviest rainfall over Bangladesh at 06 LT is confirmed using raingauge rainfall data and discussed later. The time of minimum rainfall over Bangladesh is 21 LT. Inspecting all figures during monsoon period it is found that 18 LT is the time for maximum rainfall over India with minimum at 12 LT. Another interesting point of the contrast of the variation of rainfall over land, considering huge Indian landmass, and over ocean is disclosed in this research. Over ocean rainfall peak is obtained at about noon whereas over land it is at about evening. These results are consistent with the reported result of Islam et al. (2004). 29

39 Figure 3.1 Diurnal variation of precipitation determined by TRMM-3B42RT during monsoon period (JJAS). 30

40 Figure 3.1 (Continued) The characteristics of rainfall depend on place to place and time to time. As mentioned above, the time of peak rainfall over Bangladesh and India are different even the event are occurred over land. On the other hand, for pre-monsoon or post monsoon the situations are not same as for monsoon. As example, the time of maximum rainfall in premonsoon over Bangladesh is at 00 LT (Fig. 3.2) that differs from 06 LT in monsoon period. The time of minimum rainfall over Bangladesh is at 09 that also unalike from 21 LT in monsoon. 31

41 Figure 3.2 Diurnal variation of precipitation determined by TRMM-3B42RT during premonsoon period (MAM). Figure 3.3 shows the diurnal variation of rainfall (mm/day) determined by TRMM- 3B42RT and RNG in pre-monsoon (Fig. 3.3 a), monsoon (Fig. 3.3 b) and post-monsoon (Fig. 3.3 c) periods. Rainfalls are obtained by averaging from 33 stations and from 2002 to 2004 in corresponding periods. In pre-monsoon, rainfall dominates from evening to early morning whereas monsoon rainfall dominates from morning to afternoon. In the case of post-monsoon, rainfall dominates during day time. The patterns of the variation of rainfall determined by both TRMM and RNG are almost similar. The rainfall peak at 00 LT in pre-monsoon is consistent with the explanation of Fig Actually this peak comes from the mid night heavy rainfall in the northeastern part of the country. 32

42 12 25 Rainfall (mm/day) a) Pre-monsoon RG-MAM_ B42RT-MAM_ LT Rainfall (mm/day) b) Monsoon RG-JJAS_ B42RT-JJAS_ LT Rainfall (mm/day) c) Post-monsoon RG-ON_ B42RT-ON_ LT Figure 3.3 Diurnal variation of rainfall determined by TRMM-3B42RT and RNG. For the entire rainy season (March to November), diurnal variation of rainfall (mm) determined by TRMM-3B42RT and RNG averaged for 33 stations and from 2002 to 2004 (Fig. 3.4), shows the maximum rainfall in Bangladesh occurred at 06 LT and the secondary maximum rainfall peak appeared at 15 LT. These findings are consistent with the result of Islam et al. (2004). 33

43 12 10 March-November Rainfall (mm/day) RG-MAMJJASON_ B42RT-MAMJJASON_ LT Figure 3.4 Same as Fig. 5 except for averaged for entire rainy season. Daily rainfall at different hours in different seasons obtained by TRMM-3B42RT and RNG averaged for 33 locations over Bangladesh and from 2002 to 2004 is tabulated in Table 3.1. Daily rainfall determined by TRMM is higher than the same determined by RNG in pre-monsoon. In monsoon, TRMM underestimates rainfall while in postmonsoon TRMM measures almost similar to RNG. These deviations come from different intensities of precipitation at various altitudes in corresponding periods in and around Bangladesh as discussed in next sub-section. Table 3.1 Rainfall (mm/day) determined by rain-gauge (RNG) and TRMM-3B42RT in different periods and averaged from 2002 to LT RNG MAM 3B42RT MAM RNG JJAS 3B42RT JJAS RNG ON 3B42RT ON Daily

44 Vertical structure of precipitation in and around Bangladesh Vertical extension of precipitation field determined by TRMM-2A25 in and around Bangladesh in different periods of 2001 is shown in Fig Usually, intensified regions are embedded relatively higher altitudes for pre-monsoon echo (Fig. 3.5 a) compared to monsoon (Figs. 3.5 b) and post-monsoon (Fig. 3.5 c). On the other hand, less intensified regions are embedded at higher altitudes for monsoon and post-monsoon echoes compared to pre-monsoon. The echo top heights versus threshold rain rates for April (14 cases), July (29 cases) and October (12 cases) in 2000 are shown in Fig It is clear that 10 mm/h or higher rain rates are existed from about 2 to 14 km for April whereas the same rates are existed from about 2 to 7.5 km for July. In October, rain rates are comparatively lower than that of other months. These results are very consistent with the explanation of Fig For other analyzed years ( ) similar situations are obtained (not shown). 35

45 Figure 3.5 Vertical extension of precipitation field obtained by TRMM-2A25 data over Bangladesh a) 27 April, over Bay of Bengal b) 23 July and c) 30 October Echo top height (km) a) 14Cases APR Threshold rain rate (mm/h) Echo top height (km) b) 29Cases JUL Threshold rain rate (mm/h) Echo top height (km) c) 12Cases OCT Threshold rain rate (mm/h) Figure 3.6 Vertical extension of precipitation field obtained by TRMM-2A25. 36

46 Echo base- and top height and maximum rain rate in each case for April, July and October 2002 are shown in Fig Maximum rain rate is obtained in the domain of 86-94E and 18-28N while the echo base- and top heights are obtained along the maximum rainy areas for individual case. In each and every year, rainfall amounts are not same in a region but the characteristics of precipitation are almost similar. Comparing Figs. 3.7 (a), (b) and (c) it is clear that echo base height is higher in pre-monsoon and post-monsoon while rain rate is relatively low in monsoon and very low in post monsoon. The echo top height is about 16 km or higher for many cases in July and low rain rates are also found for some cases in April. The fact is that these depend on the development site-, time- and stage of the echo. The echo base- and top height, maximum rain rate are tabulated in Table 3.1. From single scan of TRMM-2A25 data in a day, it is difficult to obtain the stage of the echo whether it is in developing, mature or dissipating Height (km) a) Echo top Echo base Max rain rate Max rain rate (mm/h) Height (km) b) Echo top Echo base Max rain rate Max rain rate (mm/h) 0 03_ _ _ _ _ _ _1518 Date_Time (LT) April _ _ _ _ _ _ _ _ _ _2106 Date_Time (LT) July _ _ _ _ Height (km) c) Echo top Echo base Max rain rate (mm/h) 0 01_ _ _ _ _ _ Date_Time (LT) October 2002 Figure 3.7 Echo base-and top height determined by both TRMM-2A25 in a) April, b) July and c) October

47 Analyzing a number of cases as presented in Table 3.2, it is found that there are similarities in precipitation parameters in year to year and dissimilarities in month to month. Some exceptions are also found due to different stages, development sites and echo born-time. Averages from 4 years ( ) data, it is found that echo base height is 0.45, 0.34, 0.57, echo top height is 12.73, 12.02, and maximum rain rate is , 73.88, in April, July and October respectively. Table 3.2 Echo Parameters determined by TRMM-2A25 No Cases Echo base height (km) Echo top height (km) Max rain rate (mm/h) APR APR APR APR JUL JUL JUL JUL OCT OCT OCT OCT Hence, rain rate is higher (lower) in pre-monsoon (post-monsoon) month April (October) for all years. Echoes are developed at higher altitudes in post-monsoon period compared to other periods. These things are very clear from Fig

48 14 Echo base Echo top Max rain rate 120 Height (km) Max rain rate (mm/h) 0 APRIL JULY OCT 0 Figure 3.8 Echo base-and top height and maximum rain rate determined by TRMM-2A2. 39

49 Chapter 4: Statistical Analysis of TRMM and Rain-gauge rainfall 4.1. Introduction The estimation of rainfall over Bangladesh has currently done using ground based raingauge network of Bangladesh Meteorological Department (BMD). (Islam, 2004) At present, there are 33 rain-gauge stations used by BMD which are neither quite sufficient to cover the whole country nor dense enough to correctly estimates the true rain rates. On the other hand, rainfall estimation using remote sensing data can be a viable solution to this problem. The accuracy and availability of remotely sensed data has been increased over recent years that make it as an alternative to traditional rain-gauge. Therefore it is necessary to calibrate any measurement of rainfall with the ground based observational rain-gauges. As mentioned earlier, unfortunately a few research works have been carried out in this field in Bangladesh over the past. Researches of using GMS-5 satellite IR data and only one active radar of Bangladesh of 250 km effective range has conducted to analyzes the diurnal variation of precipitation in Bangladesh during 2000 monsoon seasons (Islam, 2004). They have shown that the diurnal cycle of precipitation in Bangladesh exhibits a morning peak with minimum at noon. Using radar and rain-gauge data, Islam et al. (2005) examine spatial and temporal variations of precipitation in and around Bangladesh. It has been found that the northwest part of the Bangladesh was largely affected by pre-monsoon convections, while the whole country was affected by the peak monsoon activities. After lunch of TRMM in 1997, TRMM satellite data products have been widely used in different parts of the globe to understand the cloud formation processes, rainfall process and quantification. Chokngam and Chiu (2004) compared daily rainfall obtained from rain-gauge with TRMM estimated rainfall in Thailand. They have pointed out that rain-gauge rainfall is lower than the 3B42 data except of the east part of the Thailand. However, TRMM 2A25 data underestimated the surface rainfall over western part of India (Suresh, 2002). Islam and Uyeda (2005) showed some early results by analyzing TRMM 3B42 and rain-gauge rainfall of some selected stations in Bangladesh. They found that 3B42 overestimates rainfall during pre-monsoon (March- May) and underestimates during monsoon (June-September) while alike during post- 40

50 monsoon (October-November) period. However, they have suggested for detailed study to be conducted to calibrate TRMM satellite data with rain-gauge in Bangladesh. Therefore, this study compares the rainfall obtained from TRMM 3B42 and rain-gauge over Bangladesh using statistical methods and develops a linear regression equation among these data sets Data and assumptions In this study, ground based rain-gauge rainfall and space based precipitation radar (PR) data of TRMM satellite are used. Rain-gauge rainfall has been collected by the BMD for 33 stations located in different parts of Bangladesh. Figure 4.1 shows the location of raingauge stations situated throughout Bangladesh. The rain-gauge data of those stations are available at an interval of 3 hours from 1998 to Daily rainfall is calculated by averaging eight of the 3 hourly values and finally multiplying with 24 to convert hourly rate (mm/hr) into daily rate (mm/day). TRMM 3B42 data provides daily precipitation estimated for each grid box. The data is available between latitude 40 0 N to 40 0 S and longitude W to E. TRMM 3B42 is a combined product from participation radar (PR) of TRMM satellite and geosynchronous Infra Red (IR) imagery of daily precipitation (NASDA, 2001). There have been a significant problem with satellite and rain-gauge data due to the difference of the characteristics of the measurement of satellite and ground based instrument (K. P. Bowman, 2005). Satellite data has very high spatial regulation with a very low temporal (only about once per day) while rain-gauge data is high temporal resolution (three hourly data) with a very low spatial resolution (point measurement in space). In order to compare rain-gauge data with TRMM 3B42 data, the following two assumptions are made: (a) the rain-gauge rainfall is a point measurement whereas TRMM 3B42 data are calculated for each 1 0 by 1 0 grid box. Therefore, we have considered raingauge rainfall as an equivalent representation of total rainfall of that grid box where it is situated, (b) the time lag between TRMM 3B42 and rain-gauge data is considered as zero although there is a small time different between these two types of measurements. 41

51 4.3. Statistical Methods TRMM 3B42 and rain-gauge data have been compared with some known statistical analysis. The statistical methods use in this study has been given along with their ranges below. (i) Bias is a term which refers to how far the average statistic lies from the parameter it is estimating, that is, the error which arises when estimating a quantity. Errors from chance will cancel each other out in the long run, those from bias will not. It has been calculated ( Rs Rg) from the following formula, e = where, Rs is a satellite rain estimate Rg is n rain-gauge rain estimate and n is the total number of data. (ii) Root Mean Square Difference (RMSD) between the satellite and rain-gauge is defined by, σ = [( Rs Rg)] estimated (TRMM) and actual (rain-gauge) values. n 2. RMSD represents the mean difference between (iii) Pearson s Correlation coefficient (r) is a dimensionless index that ranges from -1.0 to 1.0 is defined by, r = ( ( Rs Rs)( Rg Rg) Rs Rs) 2 ( Rg Rg) 2. The absolute value of the sample correlation must be less than or equal to 1 which represents absolute agreement between the estimated and actual values. (iv) Slope of regression line is a vertical distance divided by the horizontal distance between any two points on the line, which is the rate of change along the regression line is defined by, b = ( Rs Rs)( Rg Rg). The slope will be 1 or it has an angle 45 0 if 2 ( Rs Rs) the estimated and actual values match each other. (v) Correlation of the regression line (R 2 ) value is a number from 0 to 1 that reveals how closely the estimated values for the regression line correspond to the actual data. It is 42

52 2 n ( Rs Rs) 2 defined by, R = 1. A regression line is most reliable when its R n Rs ( Rs) value is at or near Results of statistical analysis Rainfall data from 33 rain-gauge stations has been compared with the TRMM 3B42 data using some common statistical methods. A summary of these analyses results has been presented in Table 4.1. It has been found from the table that the monthly rainfall shows much more agreement than the daily rainfall i.e., with the increase of time length the match between rainfall determined by rain-gauge and TRMM data is increased. The mean daily rainfall calculated from 3B42 has been overestimated (negative bias) from the rain-gauge rainfall for the stations where mean daily rainfall is less than 6.0 mm/day. On the other hand, TRMM 3B42 underestimated (positive bias) from the rain-gauge rainfall in the heavy rainfall areas. The highest bias 5.0 mm (40% of mean rainfall) for daily rainfall occurs at Cox s Bazar (coastal station) which is a heavy rainfall area of the country. The average daily and monthly biases for all the 33 stations of Bangladesh are 0.2 and 6 respectively. Another heavy rainfall area, Sylhet (inland station) has considerably lesser amount of bias 2.5 mm (25% of mean rainfall) for daily rainfall when we compare it with Cox s Bazar. The similar characteristics are found when we compare bias for monthly rainfall. The lowest monthly bias was found at Srimongal (inland station) 4.0 mm (2% of mean rainfall) and the highest is at Cox s Bazar mm (50% of mean rainfall). Another statistical indicator, root mean square difference (RMSD) was calculated to quantify the absolute disagreement between the TRMM 3B42 and raingauge data. Among all the stations, least disagreement for daily rainfall is 12 mm found at Satkhira, Ishurdi and Chuadanga (inland stations) whereas most disagreement is 30 mm found at Sandwip (a coastal station). RMSD of monthly data shows minimum value of 81 mm at Bhola (coastal station) and maximum value of 383 mm at Sandwip. The average daily and monthly RMSD from all the 33 stations of Bangladesh are 6 and 56 respectively. Hence it is clear that agreement and disagreement between rainfall determined by rain-gauge and TRMM depend on the local condition. Due to the lack of high spatial resolution observational data, at present it is difficult to obtain in detail the 43

53 characteristics of convective systems in this region. Using the climate model facilities such as mesoscale model MM5, in near future this problem may be solved and we will be able to understand the cause of local variation in rainfall (Kataoka & Satomura, 2005). The relationship between rain-gauge and TRMM 3B42 rainfall has been analyzed by fitting linear regression equation between these datasets. These equations are must have a zero intercept because both of these data should be zero when there is no rainfall. The scattered plot of daily rainfall data along with best fit line for rain-gauge at some selected stations are shown in Fig Plots are shown only for eight stations located different regions of the country and one additional plot also shown for the average of all the stations. The similar diagrams for monthly rainfall data are shown in Fig A slope of the best fit line with a value less than unity represents the trend of underestimation and the slope greater than unity represents the trend of overestimation of TRMM 3B42 rainfall data from the rain-gauge estimated data. The minimum slope of monthly daily rainfall value of 0.42 found at Coxs Bazar and the maximum slope of monthly value of 1.45 found at Jessor. The minimum slope of daily rainfall value of 0.21 found at Hatia and the maximum slope of monthly value of 0.58 found at Satkhira. Both maximum and minimum slope for daily data is less than unity represents that TRMM 3B42 underestimates the rain-gauge daily rainfall all over the country. The best match of trend of monthly rainfall data towards unity found at Srimongal of a value of The average daily and monthly slopes of the best fit line for all the 33 stations throughout Bangladesh are 0.82 and 0.92 respectively. Also, the correlation of the regression line (R 2 ) for monthly rainfall data shows a strong agreement for all the rain-gauge stations over Bangladesh (Figure 4.3). The strongest correlation has found at Bhola is about 0.85 and lowest value of correlation is 0.26 at Chittagong. However, when we take the average of all 33 stations the correlation value stands on This represents that regression model between monthly rainfall of rain-gauge stations and that from TRMM 3B42 has fitted very well. Unfortunately, the regression model for the daily rainfall data shows not promising for all the stations. The highest R 2 of the regression line for daily rainfall has found at Teknaf of about 0.38 and the lowest value is 0.01 found at Jessor. However, the average of all 33 rain-gauges stations shows moderate value of R 2 which is about

54 Table 4.1 Statistical Analysis of between rain-gauge and TRMM (3B42) dataset of monthly rainfall accumulation over Bangladesh during Stations Mean (Gauge) Mean (TRMM) Slope **r Bias *RMSD Monthly Daily Monthly Daily Monthly Daily Monthly Daily Monthly Daily Monthly Daily Barisal Bhola Bogra Chandpur Chittagong Chuadanga Comilla Coxsbazar Dhaka Dinajpur Faridpur Feni Hatia Ishurdi Jessore Khepupara Khulna Kutubdia Madaripur Maijdeecourt Mongla Mymensingh Rajshahi Rangamati Rangpur Sandwip Satkhira Sayedpur Sitakunda Srimangal Sylhet Tangail Teknaf Average over Bangladesh *RMSD = Root Mean Square Difference and **r = Pearson s Correlation Coefficient 45

55 Figure-4.1: Shows the location of rain-gauge stations situated at different parts of Bangladesh. 46

56 Figure 4.2 Scatter plot of Daily Rainfall (mm/day) estimated from TRMM 3B42 data and rain-gauge over Bangladesh during

57 Figure 4.2: (Continued) 48

58 Fig. 4.3 Scatter plot of Monthly Rainfall (mm/month) estimated from TRMM 3B42 data and rain-gauge over Bangladesh during

59 Figure 4.3: (continued) 50

60 (a) (b) (c) Figure 4.4 Contour maps of (a) slope (monthly data) and (b) slope (daily data) of linear regression line, (c) RMSD (monthly data) and (d) RMSD (daily data) over Bangladesh during (d) 51

61 Figure 4.5 Time series plots of Mean Daily Rainfall (mm/day) obtained from the average of all 33 rain-gauge stations and TRMM 3B42 over Bangladesh during

62 Figure 4.6 Time series of Monthly Rainfall (mm/month) estimated from TRMM (3B42) and rain-gauge over Bangladesh during

63 Figure 4.6: (Continued) 54

64 Figure 4.7 Histograms of Monthly Average Rainfall (mm/month) estimated from TRMM 3B42 data and rain-gauge over Bangladesh during

65 Figure 4.7: (Continued) 56