Assessment of Remotely-Sensed Precipitation Products Across the Saudi Arabia Region

Transcription

1 16-17 December, 2014, Riyadh, Saudi Arabia Assessment of Remotely-Sensed Precipitation Products Across e Saudi Arabia Region Marwan M. Kheimi and Saud Gutub Department of Civil Engineering, Faculty of Engineering, Water Resources, King Abdulaziz University Abstract: Precipitation events have a huge impact on e economy, e environment and e society, especially in e largely arid countries. Recently, wi e leap into satellite-retrieved precipitation products wi high special resolution and global coverage which resulted in a new source of sustainable precipitation estimates. However, e incorporation between satellite- retrieved estimates and e operational decision making are not well recognized due to lack of information towards uncertainties and consistency. In is study, e primary goal was to evaluate e performance of satellite products rainfall estimator (TRMM 3B42), (CMORPH), (GSMaP_MVK) and (PERSIANN) around Saudi Arabia. While Taking into consideration for all products e period from Jan Dec and for GSMaP_MVK were collected from e period of January 2003 to November Independent rain gauge data were collected from over 29 local precipitation gauge stations from all irteen provinces located in Saudi Arabia. After aggregation and interpolation, is data was specifically used to diagnose systematic differences between in-situ based rainfall and satellite derived rainfall using an extensive selection of validation metrics. The results show according to e probability of detecting rainfall amounts and volume of correctly identified precipitation, TRMM 3B42 offers e best possibility for accurate estimation and variability of precipitation of is high spatial resolution. In fact, e validation results show at all e products can predict rainfall in e study area reasonably well but overestimates rainfall in e regions. However, is bias is comparatively less in e semi-arid part of e country where most of e rain falls. Key words: TRMM 3B42 PERSIANN GSMaP_MVK CMORPH Remote Sensing Water Resources INTRODUCTION pesticide and gasoline can be released into e rivers, lakes, bays and ocean, killing maritime life. Therefore, Water resource is an indispensable commodity for rainfall measurements are important meteorological data. every human being and every living creature in e Rainfall rate and quantity interact wi many oer factors ecosystem. The water environment characterized in e to influence erosion, vegetative cover, groundwater hydrological cycle including flooding, drought and all of recharge, stream water chemistry and runoff of non-point its crucial and beneficial forms are playing a significant source pollution into streams. Weaer and atmospheric role in economy, heal, urbanization and environment. researches rely on e conventional rainfall measuring Scarcity of water supplies, rainfall, surface runoff and instruments at observe and monitor precipitation and its aquifer recharge will seriously influence e social oer forms. The reading is susceptible to natural and survival and e welfare of communities. On e oer non-natural influences at may lead to errors in e hand, floods make an enormous impact on e reading. The most significant influences on e accuracy environment and society. Floods destroy drainage of precipitation measurement are e environment and systems in cities, causing raw sewage to spill out into wind at e installation site raer an e performance of bodies of water. In addition, in cases of severe floods, e instrument itself. [1] The environment of e buildings can be significantly damaged and even instrument's location significantly influences observation destroyed. This can lead to catastrophic effects on e of precipitation and erefore, e surroundings of e environment as many toxic materials such as paint, observation site must be considered before final Corresponding Auhtor: Marwan M. Kheimi, Department of Civil Engineering, Faculty of Engineering, King Abdulaziz University-Rabigh branch. mmkheimi@kau.edu.sa 315

2 Fig. 1: Location map of e rainfall gauge recording stations. installation of e gauge. e point measurement have and different spectra of e wavelengs to produce high accuracy, ground-based precipitation network precipitation data. Furer, accuracy of ese estimates like United States is not common in most parts of e also varies over different regions of e world. globe. Therefore, it limits e development and use of Consequently, an in dep validation of a satellite product hydrologic models to monitor and warn for flood or is necessary before using e data wi confidence in droughts for decision-making systems [2] in ose decision support systems or in hydrologic models. This regions. Consequently, satellite precipitation is study focuses on assessing several satellite products increasingly in demand to provide rainfall information at against ground observation over e Saudi Arabia region a spatial scale of interest. Some of e satellite data are to find out which product best describes e regional now available and accessible in near-real time wi almost climate dynamics and later can be used in hydrologic global coverage over e oceans and parts of e world models or climate models. However, e western part of where conventional ground-based observations (rain e country, which has seen several devastating flash, gauges and radars) are very sparse or nonexistent [3]. flood events (e.g. on November 25, 2009 at Jeddah and Furer, wi continuous improvement in sensor May 5, 2010 at Riyadh, among many oers) in recent technology and new meods in merging various data years, have of less an 5 rain gauges [5]. Consequently, sources, e satellite precipitation data are now available magnitude of e rainfalls in e impacted area could not at high measurement accuracy at sub-daily temporal scale be assessed wi desired accuracy. Thus, for Saudi [4]. In arid and extremely arid regions, e magnitude and Arabia, satellite derived precipitation data can be a distribution of ese parameters vary spatially and suitable alternative to rain gauges. The study area (Saudi temporally affecting e hydrological cycle of e area. Arabia) is a special case due to its geographic location Discrepancy and prediction of e rainfall variability in and arid, dry atmosphere. The remote-sensing space and/or in time are fundamental requirements for a precipitation data wi high agreement wi observation wide variety of human activities and water project will play an important role in preparing flood map or future designs. In Saudi Arabia, ere are only 29 conventional flood forecasting using hydrologic models or climate rain gauges installed across e country (Figure 1). forecasting using climate models. However, e satellite- retrieved estimates can be an These products can have as short as 3-hourly to as overestimation or underestimation of e actual rainfall long as daily temporal resolution. Four satellite amount, which could lead to uncertainty in e operational precipitation products has been chosen, Precipitation decision-making. The difference in data products is mainly Estimation from Remotely Sensed Information using because each satellite product uses different algorim Artificial Neural Networks (PERSIANN), Precipitation 316

3 Estimation from Global Satellite Mapping of Precipitation (GSMaP) project, product called as Global Satellite Mapping of Precipitation Microwave-IR Combined Product (GSMaP_MVK V ), Tropical Rainfall Measurement Monitoring (TRMM) and National Oceanic and Atmospheric Administration, Climate Prediction Center (NOAA,CPC) Morphing Technique (CMORPH) precipitation products has been selected because ey have finer spatial resolution. Course resolution product cannot capture all e local dynamics and so a finer resolution product is always highly regarded as best alternative for conventional rain gauge. In a study prepared by Dinku et al. [3] PERSIANN and CMORPH products have been investigated its accuracy over e region of Sou America, Colombia. The validation was done for whole as well as different parts of e country. The validation results of in comparison between e two products indicates at PERSIANN has a serious overestimation while CMORPH exhibited e best performance among e products investigated. Moreover, Novella and Thiaw [6] precipitation data was better an PERSIANN and TRMM products to detect rain versus non-rain events against rain gauges. In anoer study performed by AghaKouchak et al. [2] across e central United States to evaluate e satellite-retrieved extreme precipitation it found at CMORPH and PERSIANN products data sets lead to better estimates an TAMPA-RT and TAMPA-V6. In addition, Sohn et al. [7] arrived to e same conclusion regarding e better agreement wi observation in e Korean Peninsula region. Thiemig et al. [4] concluded at CMORPH showed specific streng in rainfall estimating over mountainous area under sparse conditions and TRMM 3b42 succeeded in detecting seasonal variability and timing of rainfall events in e African region while Feidas et al. [8] in Greece e TRMM 3b42 showed reasonable skill wi detecting rainfall. Therefore, e selected satellite products will be evaluated to identify specific weaknesses and strengs of respective products using traditional statistical meod of analysis. Furermore, ere has been very limited studies on assessing satellite products over Saudi Arabia and us raises e need for e validation of ese products as an alternative for conventional rain gauge. Study Area and Datasets: Satellite data were validated against rain gauges over e entire country of Saudi Arabia, which is located in e Souwest of Asia. It covers one ird of e Arabian Peninsula and it links Asia wi Africa [9]. It is located in e sub-tropical belt and is bounded by latitudes 12 N and 35 N and longitudes 30 W and 57 W (Figure. 1). Geography of Saudi Arabia is dominated by e Arabian Desert and its largest desert, known as e Empty Quarter (locally called "Rub' al Khali"), occupies 647,000 km2 of e Souern part of e country[10]. There are no rivers or lakes in e country, but wadis are numerous [10]. Nouh [11] defined "wadis" as streams, which run full for short time after a heavy rainfall. Souwest of e country is more mountainous and e two mountain ranges, e Hijaz to e nor and e Asir farer sou, lies along e western coast [12]. Topography also plays an important role in e country's climate. Most part of e study area is arid and have desert climate [13, 14] but souwest is semi-arid. This exception in e climatology of e souwest can be explained by e role of mountains and air masses at proceeds from different directions over e country during e year. During winter (December-February), e Mediterranean cyclones migrate from west to east in association wi upper troughs and active phases of subtropical and polar jets [14]. This front furer picks up more moisture from e Red sea [9]. However, its potential decreases from nor to sou except for e mountainous sou-west region, where e topographic effects of Hijaz escarpments modify air mass. Orographic effect is e main cause of rainfall during is period. During e summer season (June-August) e circulation pattern is altered [14]. Monsoonal air mass from Indian Ocean is predominant, creating understorms along e escarpment and e souern part of e Red Sea coast [15]. Nevereless, norern part of e country remains dry because cold air mass adverts from e Atlantic Ocean [14]. Moist soueasterly monsoon air causes rain during spring (March-May) as inter-tropical front move norwards. This rainfall is mainly along e leeward side of e mountains and e Red Sea coast [15]. This soueasterly air weakens because of increasing norern westerly air front during Autumn (September-November). Wi a strong convergence of two fronts, tropical phenomenon rises and widespread rainfall occurs along e mountains of e souwest and e Red Sea coast. Thus, souwest of e country represents a unique climate. It receives more rain an any oer part of e country and characterized by precipitation events roughout e year [5]. Rainfall in e rest of e country at generally falls from October rough April is scarce, irregular and unreliable [16]. Figure. 3 shows e temporal and spatial distribution of rainfall using rain data from 1985 to 2005 in different parts of Saudi Arabia. Annual average rainfall in e souwest is about 200 mm (Figure. 2) while most of e rain falls 317

4 Fig. 2: The average monly precipitation events from Fig. 4: Average annual precipitation represented in over e major cities in Saudi Arabia. isohyet contour lines for rain gauges over Saudi Arabia. elevations. Station "Gizan" along e Coast of Read Sea is at 7.2 mm while station "Abha" has e highest elevation (2,093.4 m) and at e sou-west of e country. The rainfall data from e ground observation stations was used to validate e satellite data. The annual average precipitation recorded in e rain gages from 1985 to 2005 is shown in Figure. 4 and Figure. 5 shows temporal distribution of rainfall at e observation sites. Rain gauge "Abha" close to e Asir Mountain records annual average of about 300 mm, e highest amount of precipitation (Figure. 4). As mentioned Fig. 3: The spatial distribution of rain gauges over e earlier, is region is subjected to monsoons from Indian provinces of Saudi Arabia. Ocean, usually occurring between October and March and approximately 60 percent of e annual total falls during spring season. The duration of rainfall is usually during is period (Figure. 5). short but can consist of one or two high-intensity Generally, e highest amount of rainfall over understorms [5]. Duration, intensity and return periods kingdom occurs in e spring season and lowest amount of rainfalls affects wadis which bring in surface runoff in e summer. The lowest amount of rainfall occurs over from high elevated lands to low level coastal areas [17]. e nor and norwest areas. The sou east, a deserted Nevereless, e desert soil of Saudi Arabia does not area, does not have any meteorological station and no soak water easily and us, even a small storm can cause information on rainfall is known. flash flood. Satellite Data: There are number of satellite-derived Station Data: Meteorological data were collected from precipitation products were available at very high spatial Presidency of Meteorology and Environment (PME) of (0.25 latitude 0.25 longitude grid size) and temporal Saudi Arabia. There are only 29 gauging stations across (ree hourly) resolution. A four satellite products were e country. The stations have continuous daily data on evaluated: TMPA product [18] 3B42, which are produced rainfall, temperature, relative humidity, wind speed and by e TRMM project at e National Aeronautics and some of em stated collected weaer data as early as Space Administration; PERSIANN [19] from e Automated sensors are used to collect e weaer University of California, Irvine; CMORPH [20], which is data. The data was manually checked for consistency and produced by NOAA/CPC; and GSMaP from Osaka accuracy. Spatial distribution of e observation stations Prefecture University in Japan [21]. GSMaP products, are shown in Figure. 3. The gauges are installed at various GSMaP moving vector wi Kalman filter (GSMaP-MVK). 318

5 Fig. 5: Average annual precipitation isohyetal contour lines over Saudi Arabia in (a) e summer, (b) e autumn, (c) e winter, and (d) spring season. Table 1: Summary of High-Resolution Precipitation Products Selected in This Study The Tropical Rainfall Measuring Mission (TRMM) is a joint U.S.-Japan satellite mission to monitor tropical and subtropical rainfall. The primarily rainfall sensors on board e TRMM spacecraft are e Precipitation Radar (PR), e TRMM Microwave Imager (TMI) and e Visible and Infrared Scanner (VIRS). The TRMM standard products are classified into ree levels: level one product is e calibrated and geolocated raw data. Level two products are derived geophysical parameters at e same resolution and location as ose of e level one source data. Level 3 product, known as climate rainfall products, is e time-averaged parameters mapped onto a uniform space-time grid [22]. For is study, TRMM climate rainfall product 3B42 is used which is derived from TRMM Multi-Satellite Precipitation Analysis (TMPA) algorim. The Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) algorim uses a ree-layer feed forward artificial neural network (ANN) technique to estimate rainfall rates from IR images [19] of e global geosynchronous satellites provided by e Climate Prediction Center (CPC), NOAA [23], as e main source of information. ANN, an adaptive training technique, uses infrared brightness temperature (Tb) of e pixel, mean Tb of e 3 3 and 5 5 pixel windows around e pixels of interest and standard deviations of Tbs in ese windows. Initially e ANN was trained using radar data and e input was limited to IR data.the recent version also uses daytime visible imagery [24] and e TRMM microwave Imager rainfall estimates (2A12) to update e ANN parameters [25]. The rainfall data is available since 2000 at daily timescale and at 0.25-degree-by-0.25-degree spatial resolution and is used in is study. Figure. 6 shows e rain estimation by PERSIANN algorim over e Saudi Arabia region on 319

6 Fig. 6: Precipitation estimated by CMORPH TRMM 3b42 (a), PERSIANN (b), GSMaP_MVK v (c), and (d) satellites over e study region on Nov 10, Nov 25, CMORPH (CPC MORPHing technique) rainfall product is e produced at NOAA's Climate Prediction Center (CPC). The product produces global precipitation analyses at very high spatial (grid size of 8 km at e equator) and temporal resolution (half hourly) by incorporating various passive microwave rain estimates derived from e following passive microwaves sensors. 320

7 The Global Satellite Mapping of Precipitation moving vector wi Kalman filter (GSMaP-MVK) project was initiated to produce a high precision, high-resolution global precipitation map using satellite data and is sponsored by Core Research for Evolutional Science and Technology (CREST) of e Japan Science and Technology Agency (JST). Similar to e algorim of CMORPH, e GSMap-MVK algorim combines PM rainfall estimates from TRMM TMI, Aqua AMSR-E, DMSP SSM/I and Special Sensor Microwave Imager/Sounder (SSMIS), NOAA AMSU-A/-B and Microwave Humidity Sounder (MHS), MetOp MSU-A and MHS of European Organization for e Exploitation of Meteorological Satellites (EUMETSAT) wi IR rain data from A Geostationary Operational Environmental Satellite 8 and 10, Meteosat 5 and 7 and Geostationary Meteorological Satellite (GMS). But e difference between CMORPH and GSMap-MVK is at CMORPH uses e ermal IR observations only to extract information about e time evolution of e PM rain rates, while GSMaP-MVK not only uses IR information for time evolution but also uses e IR rainfall estimates at times e PM estimates are not present along wi e propagated PM estimates wiin a Kalman filter framework [21]. Data is available from 2003 to 2009 at 0.1 spatial and hourly temporal resolution. Figure 6 and Figure 7 shows e precipitation estimated by e four satellites over e study region on November 25, 2009 when a major flood event occurred on e western side of e country. On is day, TRMM 3B42 and PERSIANN show more an 40 mm rainfall on e souwest of e study area. Bo CMORPH and GSMaP_MVK have identified rain over e region but wi 49 mm. MATERIALS AND METHODS Analyses of monly estimate of all above-mentioned precipitation remote sensing products. The analysis meods used to validated ese products are Probability of Detection (POD), False Alarm Ratio (FAR), Frequency of bias (FBS), Probability of False Detection (POFD), Bias, Mean Error (ME), Mean Absolute Error (MAE), Efficiency (Eff), Correlation Coefficient (CC), Root Mean Square Error (RMSE), using all of ose statistical meods of validation to assess e accuracy of each product against e rain gauge ground observations and compare eir percent of agreement against each oer and e precipitation ground observing network. According to e Presidency of Metrological and Environment (PME), rain gauges network includes 29 ground Table 2: The Contingency Table Shows e Binary and Corresponding Observations Event Observed at Rain Gauge Event forecast/predicted bysatellite YES NO YES A B NO C D Note: The rainfall reshold is >= 1.0 mm. gauges distributed vastly across e region, which e study is focusing on from 2003 to 2011 period. The location lacks sufficient precipitation observing networks required for water resources controlling, atmospheric analyses and natural dangers mitigation. This is true in particularly wi complex terrain regions where urban areas and infrastructures are sparse as in our study area. The region needs e categorical statistics to evaluate e binary hits/ misses estimates of type of statement an event will or will not occur. In Table 2, it shows e contingency corresponding to observation of hit or miss depending on e observation coming from rain gage or satellite. Probability of Detection (POD): The measure at examines e event by measuring e proportion of observed events at actually occurred and detected by e rain gauge is Probability of Detection: POD hits = hits + misses Range of POD is zero to one, as one is e perfect score. It is also known as e Hit Rate. The POD gives e Hit rate, which gives e relative number of real rainfall events. POD is sensitive to hits but takes no account for false alarms. It is our intention to get e maximum number of hits and minimize e number of false alarms and misses to get adequate estimates for e tested satellite product. Frequency Bias (FBS): The amount of estimation errors is e frequency of binary precipitation events compares e frequency of precipitation estimates to e frequency of e actual occurrence and is represents by ratio. FBS ranges between zero to infinity, an unbiased score= 1. Wi FBS >1 (<1), e precipitation estimate system exhibits overcasting (under-forecasting) of events. hits + false alarm FBS = (2) hits + misses False Alarm Ratio (FAR): This meod detects e false alarm events from e data sets of e population investigated: (1) 321

8 FAR = false alarm hits + false alarm (3) (7) Range of FAR is zero to one, as zero is a perfect The MAE range is from zero to infinity and, as wi score. Contrary to POD, FAR is sensitive to false alarms ME, a perfect score= 0. The MAE measures e average but takes no account of misses. FAR has a negative magnitude of estimated errors in a given dataset and orientation. Also, FAR is very sensitive to climatological erefore is a scale of estimated accuracy. frequency of e event. Efficiency: Evaluate e performance of e satellite Probability of False Detection (POFD): It is e opposite products in estimating e amount of satellite e rainfall. of False Alarm Ration purpose (8) (4) Efficiencies can range from (-8 to 1). An efficiency of The result magnitude of POFD is again one to zero 1 (Eff = 1) corresponds to a perfect match of modeled wi a perfect score = 0. POFD is generally associated discharge to e observed data. An efficiency of 0 wi e evaluation of probabilistic of estimate e (Eff = 0) indicates at e model predictions are as evaluation of probabilistic of estimate by combining it accurate as e mean of e observed data. wi POD. G = gauge rainfall measurements, = average of e Correlation Coefficient (CC): Person's Correlation gauge measurements, S = satellite rainfall estimate and Coefficient measures of e streng and direction of e N = number of data pairs. linear relationship between two variables at is defined in terms of e (sample) covariance of e variables Bias: The bias of gauge vs. satellite product verified divided by eir (sample) standard deviations. compares e frequency of estimated to e frequency of e actual occurrence and represented by e ratio: (9) (5) The correlation coefficient ranges from -1 to 1. A Range of bias is zero to infinity, an unbiased score= value of 1 implies at a linear equation describes e 1. If Bias >1 e estimated system exhibits overestimation relationship between rain gauge data and satellite data and if Bias <1 en it exhibits an underestimation. perfectly. Mean Error (ME): This step is to compute e simple Root Mean Square Error (RMSE): Root mean square is average difference between e estimated precipitation common accuracy measure. It is a statistical measure of amount and e observation from e rain gauge, e e magnitude of a varying quantity. It is especially useful Mean Error: when it used to assess e accuracy of e satellite data versus raingear data. (6) (10) The mean error is easiest and most familiar scores, it can provide useful information on e local behavior of a Root mean square error has e same unit as e given weaer parameter such as precipitation. The ME estimate satellite precipitation parameter. The RMSE ranges from infinity and minus infinity and e perfect ranges from zero to infinity wi a perfect score = 0. The score is = 0. However, it is possible to reach a perfect RMSE is e squared difference between e estimate score for dataset wi large errors, if ere was a negative satellite precipitation and observation. magnitude, which place errors. The ME is not an accuracy measure, as it does not give information of e amount of RESULTS estimation errors. To measure e accuracy of satellite rainfall Mean Absolute Error (MAE): It is a simple test to estimates from TRMM 3B42, PERSIANN and compensate for e potential positive and negative errors CMORPH rainfall data were collected for e of e ME, which is represented in is equation: study period of January 2003 to December 2011 and 322

9 Fig. 7: Monly validation statistics of (a) Bias, (b) Efficiency, (c) Correlation Coefficient, and (d) Mean Absolute Error of e four satellite products over e region. Table 3: Validation Statistics Comparing e Performance of Daily Satellite Rainfall Estimate TRMM 3b42 PERSIANN GSMa PMVK CMORPH FBS POD POFD FAR Bias CC ME MAE RMSE precipitation data from GSMaP_MVK (V5.222) were collected from e period of January 2003 to November The satellite precipitation data is evaluated at daily, 10-daily and monly time scale using different conventional statistical meods at are described above. Rainfall estimates from each satellite pixel at overlapped at least one rain gauge location is considered for validation study. Different validation statistics were also assessed at monly scale. Bias, efficiency, means absolute error and correlation coefficient at different mons is compared in Figure 7. TRMM shows overall very little bias. However, in during e summer mon TRMM show increased bias. Similarly, PERISANN, CMORPH and GSMap overestimate rainfall in e summer as well as in autumn and underestimates in e winter mons. Among e four products, PERSIANN shows highest bias in summer and autumn. Results of efficiency also show at PERSIANN has no forecast skill in summer and autumn mons. TRMM has overall better forecast efficiency. In Table 3, e four products show varying correlation coefficients (CC) roughout e year. All e products show increased correlation coefficients in e mons of February and August. TRMM shows least correlation coefficient in March and PERSIANN has least CC value in March. TRMM, PERSIANN and CMORPH have positive mean absolute error (MAE) roughout e year while PERSIANN show highest variance in MAE value. Rain estimates from GSMap has negative MAE values in winter and positive MAE rest of e year. Scatter plot of monly-accumulated rainfall from gauge and satellite observation is shown in Figure 8. At monly scale, all e four satellite products presents greater agreement wi gauge observation but TRMM 3b42 has more symmetrical scatter. probability and TRMM has e least false alarm ratio. Statistics calculated to measure 323

10 Fig. 8: Scatter plot rainfall estimates over Saudi Arabia at monly time scale and 0.25 long/ lat special resolution of four different satellite products. accuracy of estimates show TRMM is a better rainfall TRMM, PERSIANN, CMORPH rainfall data have estimator wi least ME, MAE, RMSE values. Alough coarser (0.25 ) gird resolution an GSMaP_MVK (0.1 ). all e products overestimate rainfall, TRMM has e least Before applying e satellite precipitation data in e bias. model, accuracy of e products was assessed by Comparison wi monly rainfall accumulation show robust inter-comparison between rain-gauges seasonal variation of validation matrices. Figure 8. Scatter observations and satellite-retrieved estimates of plot rainfall estimates over Saudi Arabia at monly time monly rainfall data. Study period for e TRMM, scale and 0.25 long/ lat special resolution of four different PERSIANN and CMORPH is January 2003 to December satellite products. During summer, when e region has 2011 and for GSMaP_MVK is January 2003 to November less rain, all e satellite products show rainfall. Thus bias and MAE increases, efficiency and correlation coefficient The products have shown variation in accuracy at decreases. different mons. Relatively higher correlation between satellite estimates and gauge data during e mons of DISCUSSION AND CONCLUSION February, August and October and lower correlation in March, May, September and November for bo Four widely available remotely sensed precipitation products, PERSIANN and TRMM. CMORPH and satellite products were evaluated over e entire country GSMaP_MVK almost have e same pattern of of Saudi Arabia to study hydrologic response in a rainfall estimate during e year. Bo have a lower semi-arid watershed located at e sou west of e correlation in March, June and from September rough country. The four satellite products are TRMM, November a low correlation value is dominant. However, PERSIANN, CMORPH and GSMaP_MVK. All e it is recognizable at higher error is correlated wi higher products are available at daily temporal scale but bias values. From e evaluation of e four products for 324

11 e long-term average of precipitation suggest at TRMM has number of different products available TRMM 3B42 offers e best possibility for accurate and in is study 3B42 was used. This product is estimation and variability of precipitation of is high gauge adjusted and so more likely to have a better spatial resolution. performance an oers. The results also reflect at. The second product PERSIANN aggregated However, validation using TRMM real time (RT) spatially to (0.25 ) resolution, accumulated to monly product would have shown if ere was indeed any totals. PERSIANN monly events illustrated a poor improvement wi bias adjusted product. correlation for all mons except for e mon of Number of rain gauges is very sparse over e study February, which is in e winter season. Statistical meod region, which makes it difficult to conclude from is shows at PERSIANN performs poorly in accurately validation study which satellite product has e best estimating long-term averages. or worst performance. Furer, e country The evaluation of e ird product CMORPH experiences two different types of climate- semi arid revealed again e superiority of e TRMM 3B42 sou-west and arid climate in e rest of e region. over e comparison of daily rainfall estimate. A validation study over e sou-west, where most In monly results (Figures 6, a rough d), CMORPH of e rain falls, could offer more insight on e showed a dominant underestimation based on e Bias efficiency of e rainfall products. However, a results. regionalized validation is very limited in is case In terms of GSMaP_MVK, e assessment of is because of inadequate number of rain gauges over product yielded an underestimation roughout e period e souwest. Overall, a major conclusion at can of e study represented high values in POD, MAE, ME be drawn from is study is at some of e present which showed a good performance in detecting rainfall satellite daily rainfall products are potentially usable events comparing to TRMM 3B42. Eventually, in hydrologic response analysis over Saudi Arabia. GSMaP_MVK would not perform well in detecting rainfall However, ese results indicate ere is more room for as TRMM 3B42. improvement of ese products to remove errors and enhance rainfall estimates. Results from e Validation and Hydrologic Response However, e products have improved performance in Are Summarized Here: winter season when most of e rain falls. So, all e four satellite products were used for hydrologic The validation analysis of satellite rainfall evaluation. showed at all e products have good rainfall detection capability at daily time scale. Statistical Overall, a major conclusion at can be drawn from results for rainfall estimation accuracy show at all is study is at some of e present satellite daily rainfall e products overestimate rainfall in e region but products are potentially usable in hydrologic response TRMM 3B42 has less bias, MAE, RMSE an e analysis over Saudi Arabia. However, ese results oers. indicate ere is more room for improvement of ese Statistics at monly scale show at e products to remove errors and enhance rainfall estimates. accuracy of e products varies in different mons. All e products show increased bias ACKNOWLEDGEMENTS and MAE and decreased efficiency and correlation in e summer time when e region I would like to ank my advisor, Dr. Rebeka Sultana, has less rainfall. TRMM has least bias in e summer for her guidance and support roughout is entire while PERSIANN has e highest bias. TRMM also paper. Sincere anks goes to past and current members show higher performance in oer matrices an oer of e Presidency of Metrological and Environment (PME) products. for providing data, insight and valuable discussion related All e products were capable to capture e extreme to my research. In addition, I would like to ank Ph.D. events of e past decade. Therefore, a regional student. Muhammad Felemban at Purdue University, validation study would have provided furer insight West Lafayette and Master's student Fiasal Al-Dossari at satellite rainfall estimation accuracy. Rochester Institute of Technology, New York for eir 325

12 assistance wi computer programming and oer 10. Wikipedia "Saudi Arabia: Geography." computer-related issues. I also, appreciate e huge Wikipedia < /Saudi_ endorsement and e financial aid from e government of Arabia # Geography> (August 2013). Saudi Arabia represented by e Saudi Arabia Cultural of 11. Nouh, M., 20.11, Wadi flow in e Mission (SACM). Arabian Gulf states. Hydrological processes, pp: REFERENCES 12. Wikipedia, 2013, 6. Geography of Saudi Arabia. Retrieved , from wiki/geography_of_saudi_arabia. 1. Japan Meteorological Agency, Koppen, W. and R. Geiger, Klima der Erde, "Cp6-precipitation measurements." Japan Walter de Gruyter, Berlin. Meteorological Agency, < jp/jma/ 14. Abdullah, M.A. and M.A. Al-Mazroui, jma-eng/jma-center /ric/material/1 _ lecture _ notes/ "Climatological study of e souwestern region of Cp6-Precipitation.Pdf.> (Nov. 2012). Saudi Arabia." I. Rainfall Anal. Climate Res., 9: 2. AghaKouchak, A., A. Behrangi, S. Sorooshian, K. Hsu and E. Amitai, "Evaluation of 15. Subyani, A.M., Geostatistical study of annual satellite-retrieved extreme precipitation rates across and seasonal mean rainfall patterns in souwest e central United States." J. Geophys. Res. 115(D2), Saudi Arabia. Hydrol Sci., 49: doi: /2010JD Atlas, Water atlas of Saudi Arabia. Riyadh: 3. Dinku, T., F. Ruiz, S.J. Connor and P. Ceccato, Water resource department, Ministry of Agriculture "Validation and intercomparison of satellite rainfall and Water. estimates over Colombia." J. Appl. Meteorol. 17. Subyani, A.M.D., Multivariate geostatistical Climatol., 49: meods of mean annual and seasonal rainfall in 4. Thiemig, V., R. Rojas, M. Zambrano-Bigiarini, V. souwest Saudi Arabia. Arabian Journal of Levizzani and A. De Roo, "Validation of Geosciences, 2.1: satellite-based precipitation products over sparsely 18. Huffman, G.J., R.F. Adler, D.T. Bolvin, G. Gu, gauged African river basins." J. Hydrometeor., E.J. Nelkin, K.P. Bowman, Y. Hong, F. Erich, 13: E.F. Stocker and D.B. Wolff, "The TRMM 5. Almazroui, M., "Sensitivity of a regional climate Multisatellite Precipitation Analysis (TMPA): model on e simulation of high intensity rainfall Quasi-Global, multiyear, combined-sensor events over e Arabian peninsula and around precipitation estimates at fine scales." J. Jeddah (Saudi Arabia)." Theoret. Appl. Climatol., Hydrometeor. 8(1): (1): Hsu, K.X., Precipitation estimation from 6. Novella, N. and W. Thiaw, Validation of remotely sensed information using artificial neural satellite-derived rainfall products over e Sahel, networks. J. Appl. Meteorol., 36: Wyle Information Systems, NationalOceanic and 20. Joyce, R.J., J.E. Janowiak, P.A. Arkin and P. Xie, Atmospheric Administration, Climate Prediction "CMORPH: A meod at produces global Center, Camp Springs, MD, pp: precipitation estimates from passive microwave and 7. Sohn, B.J.J.K., Validation of Satellite-Based infrared data at high spatial and temporal resolution." High-Resolution Rainfall Products over e Korean J. Hydrometeor. 5: Peninsula Using Data from a Dense Rain Gauge 21. Okamoto, K., T. Iguchi, N. Takahashi, T. Ushio, J. Network. J. Appl. Meteor. Climatol., 49: Awaka, S. Shige and T. Kubota, "High precision 8. Feidas, H., Validation of satellite rainfall and high resolution global precipitation map from products over Greece. Theoretical and Applied satellite data." Proc. Intl. Symp. Antennas and climatology, 99(1-2): Propagation, ISAP, pp: Subyani, A.M.M.A., Topographic, Seasonal 22. Feidas, H., Validation of satellite rainfall and Aridity Influences On. Journal of Environmental products over Greece. Theoretical and Applied Hydrology, 18(2). climatology, 99(1-2):

13 23. Janowiak, J.E., A real-time global half-hourly 25. Sorooshian, S.E., Evaluation of PERSIANN pixel-resolution IR dataset and its applications. Bull. system satellite-based estimates of tropical rainfall. Amer. Meteor. Soc., 82: Bulletin of e American Meteorological Society, 24. Hsu, K.l.e Estimation of physical variables from 81.9: multichannel remotely sensed imagery using a neural network: Application to rainfall estimation. Water Resources Research, 35(5):