Assessment of global precipitation measurement satellite products over Saudi Arabia

Journal of Hydrology 559 (2018) 1–12

Contents lists available at ScienceDirect

Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol

Research papers

Assessment of global precipitation measurement satellite products over Saudi Arabia Mohammed T. Mahmoud a, Muhammad A. Al-Zahrani b,⇑, Hatim O. Sharif c a

Department of Civil and Environmental Engineering, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia Department of Civil and Environmental Engineering, King Fahd University of Petroleum and Minerals, Water Research Group, Dhahran 31261, Saudi Arabia c Hatim O. Sharif, Department of Civil and Environmental Engineering, University of Texas at San Antonio, USA b

a r t i c l e

i n f o

Article history: Received 2 July 2017 Received in revised form 6 February 2018 Accepted 8 February 2018 Available online 13 February 2018 This manuscript was handled by Marco Borga, Editor-in-Chief, with the assistance of Francesco Marra, Associate Editor Keywords: Global Precipitation Measurement (GPM) IMERG Rain gauge Precipitation Saudi Arabia

a b s t r a c t Most hydrological analysis and modeling studies require reliable and accurate precipitation data for successful simulations. However, precipitation measurements should be more representative of the true precipitation distribution. Many approaches and techniques are used to collect precipitation data. Recently, hydrometeorological and climatological applications of satellite precipitation products have experienced a significant improvement with the emergence of the latest satellite products, namely, the Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (GPM) mission (IMERG) products, which can be utilized to estimate and analyze precipitation data. This study focuses on the validation of the IMERG early, late and final run rainfall products using ground-based rain gauge observations throughout Saudi Arabia for the period from October 2015 to April 2016. The accuracy of each IMERG product is assessed using six statistical performance measures to conduct three main evaluations, namely, regional, event-based and station-based evaluations. The results indicate that the early run product performed well in the middle and eastern parts as well as some of the western parts of the country; meanwhile, the satellite estimates for the other parts fluctuated between an overestimation and an underestimation. The late run product showed an improved accuracy over the southern and western parts; however, over the northern and middle parts, it showed relatively high errors. The final run product revealed significantly improved precipitation estimations and successfully obtained higher accuracies over most parts of the country. This study provides an early assessment of the performance of the GPM satellite products over the Middle East. The study findings can be used as a beneficial reference for the future development of the IMERG algorithms. Ó 2018 Elsevier B.V. All rights reserved.

1. Introduction Precipitation, which is a major component of the hydrological cycle, falls in many different forms according to the meteorological conditions. Moreover, precipitation is one of the most important components of the global energy cycle (Kidd and Huffman, 2011; Ebert et al., 2007). Measurements of precipitation provide one of the primary inputs for hydrological, meteorological and climate models, which are used to predict different natural hazards such as landslides, floods, and droughts (Li et al., 2013; Wu et al., 2012). The responses of the hydrological and energy cycles depend upon not only the precipitation amount but also other characteristics, including the spatial pattern, intensity, and duration of precip-

⇑ Corresponding author. E-mail addresses: [email protected] (M.T. Mahmoud), mzahrani@ kfupm.edu.sa (M.A. Al-Zahrani), [email protected] (H.O. Sharif). https://doi.org/10.1016/j.jhydrol.2018.02.015 0022-1694/Ó 2018 Elsevier B.V. All rights reserved.

itation (Heistermann and Kneis, 2011; Sorooshian et al., 2011). Therefore, precipitation estimates with a high spatiotemporal resolution are always needed for various applications. The acquisition of precipitation measurements and the quality control of precipitation products are the most important steps prior to performing any analysis or constructing any hydrologic model (Li and Shao, 2010). The most widely used techniques to estimate precipitation are point measurements (i.e., rain-gauges), commercial microwave links, satellite-based sensors, and ground-based weather radar (Li et al., 2013; Raich et al., 2017). Rain gauges can provide direct measurements of precipitation, and they are considered to provide ground truth for precipitation observations due to their accuracy when compared with other sensors. There are many types of recording rain gauges. However, only three types, namely, the tipping bucket, the universal weighingtype gauge, and the float-type gauge, are commonly used by hydrometeorological agencies to measure precipitation. All three

2

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

types of gauges have numerous measurement problems that can be summarized as follows: (1) the underestimation of heavy precipitation due to splashing; (2) instrumental problems; (3) external factors such as wind and evaporation of precipitation; and (4) observer errors (Tapiador et al., 2012). In addition, gauge measurements represent point values and not aerial measurements, and thus, they cannot describe spatial variations in precipitation. Radar instruments measure precipitation indirectly by making use of the backscattering of electromagnetic waves via hydrometeors (i.e., water drops). The main benefit of radar is its ability to monitor large areas with a high, real-time resolution (Germann et al., 2006). However, radar measurements also have many error characteristics, such as range-dependent systematic errors, mean-field systematic errors, random errors, and obstruction by topography. In addition, radar networks do not cover all parts of the world (Tang et al., 2016). During the past thirty years, the utilization of multiple satellite sensors to measure the global precipitation has increased significantly (Ebert et al., 2007; Tang et al., 2016). The first devoted satellite that was used to measure precipitation was the Tropical Rainfall Measuring Mission (TRMM), which became operational in 1997 (Li and Shao, 2010; Tian et al., 2007; Prakash et al., 2016; Ning et al., 2016). The TRMM satellite was intended to measure moderate to heavy rainfall to provide a better understanding of the precipitation distribution around the globe and generate near-real-time precipitation products (Li and Shao, 2010; Tian et al., 2007; Prakash et al., 2016; Zhou et al., 2008). Several free-access satellite precipitation products have been extensively studied, and they have been verified both globally and regionally and subsequently released for public use (Tang et al., 2016). Examples of these products include the Precipitation Estimation from Remote Sensed Information using Artificial Neural Networks (PERSIANN) technique (Sorooshian et al., 2000), the Climate Prediction Center (CPC) MORPHing technique (CMORPH) (Joyce et al., 2004), the PERSIANN-Cloud Classification System (PERSIANN-CCS) (Hong et al., 2004), the TRMM Multi-satellite Precipitation Analysis (TMPA) (Huffman et al., 2007), and the Global Satellite Mapping of Precipitation (GSMaP) project (Kubota et al., 2007). Using these products, many studies demonstrated that Earth-observing satellites can reasonably estimate precipitation rates and that they are also able to represent the spatiotemporal variations in the precipitation over most parts of the world with a high resolution (Wang et al., 2017; Tang et al., 2016; Tian et al., 2007; Dixon and Wilby, 2015). The newly introduced Global Precipitation Measurement (GPM) mission is an international constellation of satellites grouped to provide next-generation measurements of global rain and snow at high spatial (0.1°
0.1°) and temporal (half-hourly) resolutions. The constellation consists of one main observatory satellite surrounded by ten partner satellites (NASA, 2016a). The GPM Core

Observatory satellite was launched through the cooperation between the National Aeronautics and Space Administration (NASA) and the Japanese Aerospace Exploration Agency (JAXA) on February 27, 2014. The GPM adopts modern technology and instruments, thereby raising the standard of precipitation measurements. The GPM mission aims to improve the existing knowledge about the global water and energy cycles, enhance the ability to predict extreme events, and provide precipitation products that can be applied directly to all scientific fields (NASA, 2016b). The GPM has many products, which are classified into four categories according to NASA. The GPM categories (i.e., levels) are Level 0, Level 1, Level 2 and Level 3 as shown in Table 1. The recommended product for use by researchers is Level 3. This product is provided by the Integrated Multi-satellitE Retrievals for GPM (IMERG) algorithm, which is designed to incorporate, merge, and inter-calibrate all precipitation microwave (MW) estimates along with infrared (IR) satellite estimates, ground precipitation gauges, and all other precipitation estimators involved in the era of Tropical Rainfall Measuring Mission (TRMM) satellites (Huffman et al., 2015; NASA, 2016c). NASA provides three main IMERG products: ‘‘Early” near-real-time run products, ‘‘Late” run products, and ‘‘Final” run products. Some studies investigated the accuracy of the GPM IMERG products using earlier satellite products, such as the Tropical Rainfall Measurement Mission (TRMM) Multi-satellite Precipitation Analysis (TMPA) 3B42-V7 product, and demonstrated their potential for hydrological applications. Recent studies reported a significant improvement in the detection and measurement of rainfall intensities using the GPM satellites compared with earlier satellites (e.g., Khodadoust Siuki et al., 2017; Wang et al., 2017). Other studies evaluated the GPM products against ground measurements and reported a good agreement between the satellite estimates and the ground observations (e.g., Asong et al., 2017; Sungmin et al., 2017). The TRMM rainfall products for Saudi Arabia were evaluated by (Almazroui, 2011) for the period from 1998 to 2009. As the results showed varying degrees of accuracy for the TRMM across different events and regions, the study recommended using TRMM rainfall products only for ungauged regions and for those regions with extremely sparse rain gauge networks to supplement the rainfall data in the country. Another recent study (Tekeli and Fouli, 2016) recommended using TRMM rainfall products for flood warning purposes in urban areas and concluded that TRMM rainfall products could provide limited input information for flood warning systems, and thus, they cautioned against relying solely upon those product during large rainfall events. Both a comparison and an evaluation of the GPM IMERG products against ground observation gauges are highly important for different regions across the world because those products are being continuously refined; moreover, feedback on the perfor-

Table 1 Definition of GPM data product levels. Product

Input Data

Distributed to Users

Description

Level 0 Level 1A

 Level 0

No No

Level 1B and 1C

Level 1A

No

Level 2

Level 1A, 1B, and 1C

Yes

Level 3

Level 1 or Level 2

Yes

Depacketized data by an Application Process Identifier (APID). These data are managed as the master data in the mission operation system. Main parameters: sensor output value, satellite altitude and location information, sensor conditions, conversion parameters. Products created via geometric collection and processing. Main parameters: received power, brightness temperature. Products containing various physical quantities related to precipitation. Main parameters: radar cross section of the Earth’s surface, precipitation type, bright band altitude, attenuation-compensated radar reflectivity factor and precipitation intensity, spectral latent heating. Products created via spatiotemporal statistical processing. Main parameter: Precipitation rate.

3

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

mance of those products in different regions will be invaluable for the development of subsequent product versions. This study evaluates the IMERG over an arid region with a very sparse weather station network. This study is conducted to evaluate the accuracy of the GPM IMERG early, late and final run precipitation products using all available daily ground-based observational data over Saudi Arabia as a reference during the period from October 1, 2015, to April 30, 2016. The results will shed light on the accuracy of native satellite products and the effectiveness of improvements to the late products in the study region. In addition, the results will provide guidance on whether satellite products that are becoming available in semi-real-time with high temporal and spatial resolutions can be adequate for different hydrometeorological applications in the region.

Table 2 Saudi Arabia sub-regions and their numbers of rain gauge stations. No.

Region

Region ID

Number of Stations1

1 2 3 4 5 6 7 8 9 10 11 12 13

Riyadh Makkah Madinah Qassim Eastern Region Asir Tabuk Hail The Northern Border Jazan Al-Baha Al-Jouf Najran

RD MK MD QS ER AS TK HL NB JZ BH JF NJ Total

38 14 21 21 16 27 2 12 7 13 8 10 None2 189

1

The number of stations is limited to the availability of the data. Data were not available for the Najran region during the period of October 1 – April 30.

2. Study area

2

This study encompasses the whole Kingdom of Saudi Arabia, which covers the area located between 34°320 0000 55°400 0000 E and 32°150 0000 16°220 0000 N as shown in Fig. 1. This location is characterized by tropical and subtropical desert regions (Hag-elsafi and El-Tayib, 2016; Al-Zahrani and Husain, 1998). It has a dry desert climate with generally light winds and high temperatures in most regions. Saudi Arabia hosts thirteen administrative regions: Riyadh, Makkah, Madinah, Qassim, Asir, Tabuk, Hail, Jazan, Najran, Al-Baha, Al-Jouf, the Eastern Region and the Northern Borders Region. Each region contains a number of rain gauge stations. Table 2 summarizes the sub-regions and the number of rain gauges within each region. Fig. 1 shows the distribution of the rain gauge stations over Saudi Arabia. The rain gauge data were downloaded from the website of the Ministry of Water and Electricity (MOWE) – Saudi Arabia for the period from October 1, 2015, to April 30, 2016 (website: http://app.mowe.gov.sa/DailyRainsNews/Rain_Dams.aspx, accessed May – Nov 2016). In Saudi Arabia, each season has distinctly different weather features. Generally, summer rains are caused by the Intertropical Convergence Zone and the northward advance of the southwesterly monsoon (Sen and Al-Subai, 2002), the effects of which decline

from north to south, not including the highlands where the uplift factor is dominant (Hag-elsafi and El-Tayib, 2016; Abdullah and Al-Mazroui, 1998). Meanwhile, winter rains are produced by westerly waves in the upper atmosphere and disturbances from the Mediterranean Sea and the Sudan Trough (Hag-elsafi and ElTayib, 2016). The spatial and temporal variations in the precipitation amounts are a major consideration of water projects and other activities. The regional and local climates are both influenced by largescale atmospheric movements along with surficial features. In addition, atmospheric movement changes constitute important aspects of the regional climate (Hasanean and Almazroui, 2015). The rainfall features vary both locally and regionally based on atmospheric circulation patterns resulting from the El Niño Southern Oscillation (ENSO), the North Atlantic Oscillation (NAO), and other variability patterns (Hag-elsafi and El-Tayib, 2016; Hasanean and Almazroui, 2015; Almazroui, 2011). Generally, the rainy season in Saudi Arabia extends from October to April (Hasanean and Almazroui, 2015; Almazroui, 2011; The

Fig. 1. Distribution of rain gauge stations.

4

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

3.2. Satellite-based precipitation data set

Ministry of Water, 1984). During the other months, almost no precipitation can be observed throughout Saudi Arabia except for the southwestern region (Almazroui, 2011). The rainfall pattern in the southwestern region has a high spatial variation due to the complex topography of the area, which affects the satellite accuracy. The topographically driven convective rain in the southwestern region manifests as multi-rain events in this specific region of Saudi Arabia (Al-Zahrani and Husain, 1998; Al-Mazroui, 1998; Subyani, 2004). Overall, in the desert areas, the mean annual precipitation is less than 100 mm, while it ranges between 250 and 300 mm in the mountainous areas (Hag-elsafi and El-Tayib, 2016; Al-Zahrani and Husain, 1998; Hasanean and Almazroui, 2015).

The Integrated Multi-satellitE Retrievals for GPM (IMERG) algorithm was used to estimate the precipitation measurements from the constellation of satellites. This algorithm has been sustainably used since the launch of the GPM Core Observatory satellite to estimate the amount of precipitation. The IMERG algorithm produces precipitation estimates at a high spatial and temporal resolution of 0.1°
0.1° (longitude
latitude) at every half-hour interval. IMERG provides three precipitation products, namely, early run, late run, and final run products, based on the latency and the need of the product. The final run product is a post-real-time product provided with a latency of four months, while the early and late run products are near-real-time products with latencies of 6 h and 18 h, respectively. The GPM IMERG products were downloaded from the NASA Precipitation Processing System (PPS) FTP servers [Early run product (NASA, 2016d), Late run product (NASA, 2016e), and Final run product (NASA, 2016f)].

3. Precipitation data sets 3.1. Ground-based precipitation data set Rain gauges are employed globally as references for precipitation data, as they provide a direct physical record of the precipitation at a given point. In this study, daily rainfall data were collected from all regions of the Kingdom of Saudi Arabia via the online database service provided by the Ministry of Water and Electricity (MOWE) – Saudi Arabia. The day in Saudi Arabia begins 3 h ahead of GMT (GMT + 3). The GPM products were aggregated to the local day. The MOWE gauges used in this study are not included among the rain gauges used by the Global Precipitation Climatology Center (GPCC) to perform adjustments to the final run IMERG product. This allows us to perform a completely independent evaluation of the final run product, which is highly dependent on gauge adjustments. The MOWE database covers the 13 administrative regions. Each region has many sub-regional stations (see Table 2). The data collection was determined according to the availability of rainfall records during the study period ranging from October 1, 2015, to April 30, 2016 (i.e., the rainy season). A total of 1455 observations were recorded from 189 stations. The minimum amount of rainfall detected during the study period was 0.1 mm/day. The rain gauge network in Saudi Arabia is unevenly distributed, as can be observed from Fig. 1, which shows that the stations are relatively densely spaced in the western region (Madinah) and in the southwestern regions (Jazan and Asir), where the climate is wet, as well as in the middle region (Riyadh), where there is a dense population. However, the eastern and northern regions have relatively sparse rain gauge station densities. The variations in the data provided by the MOWE were evaluated for each region through a descriptive statistical analysis, which is presented in Table 3. The statistics shown in Table 3 were computed for the rain records observed during the study period from October 2015 to April 2016.

4. Methodology To evaluate the IMERG products, several steps were carried out (Fig. 2). The process starts with the preparation of the ground measurement data in addition to the GPM data processing and preparation. The subsequent step consists of data analysis, which includes three major steps: an event determination using a script developed by the authors based on visual basic (VB.NET), an analysis of GPM files (the downloaded GPM files are converted into an ASCII format), and a coordinate matching process (CMP) using a VB.NET script. The last step is a comparison of the observed satellite (GPM) data with the reference data (rain gauge data) using widely used statistical performance measures. 4.1. Data processing 4.1.1. Ground measurements and data preparation The rainfall ground measurement data were downloaded from the MOWE website, and the coordinates of each station were obtained. The GEOPLANER website was used to determine the elevation (altitude) of each rain gauge station (http://www.geoplaner.com/). The coordinates provided were in decimal degrees. The UTM CONVERTER application was used to convert the coordinates from decimal degrees to Universal Transverse Mercator (UTM) coordinates for each rain gauge station (http://www.utmconverter.com/ ). Finally, a station ID was given to each station consisting of two characters (used to identify the region) and one number referring to the sequence in the database list.

Table 3 Descriptive statistical analysis of the rainfall ground observations (all units in mm/day).

*

Region

Mean

Standard Error

Median

Mode

Standard Deviation

Variance

Kurtosis

Skewness

Range

Min*

Max.

Rainfall Sum

Rain Records

Riyadh Makkah Madinah Qassim Eastern Region Asir Tabuk Hail Northern Borders Jazan Al-Baha Al-Jouf

9.3 16.2 11.7 7.5 6.8 19.9 23.0 4.7 10.5

0.7 2.5 1.2 0.6 0.6 1.4 0.6 0.4 1.0

5 10.8 9 5.1 3.5 13 23 4 7

2 10 6 1 1 8 22 4 6

11.2 13.5 11.2 8.3 8.6 19.1 1.2 3.7 9.5

126.0 180.9 125.4 68.6 74.7 363.4 1.3 13.6 89.9

4.8 1.0 4.7 24.8 12.7 4.9 6.0 9.3 1.8

2.23 1.41 2.06 3.83 3.09 2.06 0.00 2.59 1.50

56.9 50 53.6 76.8 62.9 109.5 2 21 39.3

0.1 3 0.4 0.2 0.1 1.5 22 1 0.5

57 53 54 77 63 111 24 22 39.8

2556 485.9 1073.8 1534.3 1362.5 3831.6 92 395.5 1011.3

276 30 92 206 201 193 4 84 96

16.5 17.5 4.5

1.5 1.6 0.5

10.3 13 4.3

2 3 8

17.2 18.3 2.5

294.9 334.3 6.5

4.3 8.4 1.1

1.81 2.43 0.19

95 118 8

1 1 1

96 119 9

2089.8 2189.5 99

127 125 22

Note: The minimum amount of rainfall measured during the study period was 0.1 mm/day.

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

5

Fig. 2. Flowchart of the evaluation process.

4.1.2. GPM data preparation The GPM data were available in an RT-H5 file format (RT refers to real-time and H5 denotes an HDF5 file). The precipitation products cover most of the globe. The data were extracted and converted into an ASCII format using an algorithm written in R language. In addition, the data were adjusted to match a Saudi Arabian day, which begins 3 h ahead of GMT (GMT + 3). 4.2. Data analysis 4.2.1. Events determination The data collected from the MOWE contained more than 1450 rain gauge observations covering all of Saudi Arabia. To determine the rainfall events that occurred during the study period, a code was developed using VB.Net to analyze all of the data and to specify the storm events according to a minimum number of occurrences (i.e., the number of rain gauge observations), which is set to 30 stations in this study. The reason for the selection of 30 stations was to provide rain records from at least 16% of the total available number of rain gauges from all over Saudi Arabia; this 16% was proven to cover more than 40% of the land of Saudi Arabia. Fig. 3. Coordinate matching process.

4.2.2. Analysis of GPM files The GPM data processing results were analyzed to ensure that the extracted data were located within Saudi Arabia and that they precisely covered the analysis period. Additionally, headers were added to the processed files (ASCII files) to facilitate the searching process conducted within the developed code. 4.2.3. Coordinate matching process The coordinate matching process was programmed using VB. Net. This program was aimed at achieving the following goals: (1) search for every rain gauge record during each event using the record date and UTM zone of the station to select the appropriate GPM file from the processed GPM data; (2) read the selected GPM file and explore the UTM coordinates for each GPM intersection point to compare each intersection point with the location coordinates of the rain gauge station; (3) pick up the closest intersection point when that intersection point lies atop the latitude-

longitude 0.1°
0.1° quadrangle (approximately 11
11 km), as indicated in Fig. 3; and (4) compare the rainfall measurements from the specified rain gauge station and the selected GPM intersection.

4.3. Performance measures With the intent of assessing the performances of the GPM satellite precipitation products, six commonly used statistical performance measures were selected. These statistical measures are generally grouped into three main classes on the basis of their applications (Tang et al., 2016; Ning et al., 2016; Yong et al., 2010). The first group is used to describe the biases and errors within the satellite data in comparison with the rain gauge data; it consists of the mean absolute error (MAE), the root mean

6

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

Table 4 Statistical performance measures. Statistical Measures

Equation

Mean Absolute Error (MAE) Root Mean Squared Error (RMSE) Relative Bias (RB) Correlation Coefficient (CC) Probability of Detection (POD) Critical Success Index (CSI)

1 n

Pn

i¼1 jX i

Application  Yij

ffi qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Pn 2 1 i¼1 ðX i  Y i Þ n Pn ðX i Y i Þ i¼1 P
100% n Y i¼1 i Pn 1  

Describe the bias and error in satellite data compared with rain gauge data

1 n

n

i¼1

ðX i XÞðY i YÞ

rx ry

PSG PSG þPG

Describe the agreement between rain gauge data and satellite estimates Describe the contingency of satellite estimates

PSG PSG þPS þPG

Table 5 Contingency table for evaluating the satellite detection by measuring the POD and CSI. Rain-gauge Observation

GPM product

Yes No

Yes

No

PG

PS 

squared error (RMSE), and the relative bias (RB). The second group includes only one statistical measurement, namely, the correlation coefficient (CC), which describes the consistency between rain gauge data and satellite estimates. The third group describes the contingency of the satellite estimates through two statistical parameters, namely, the probability of detection (POD) and the critical success index (CSI). Tables 4 and 5 summarize these statistical performance measures and their expressions used for the comparison between the rain gauge data and satellite data. 4.4. Comparison process Three main comparison or evaluation processes were conducted, namely, region-based, station-based, and event-based evaluations. The region-based comparison was conducted to evaluate the accuracy of the rainfall detected by the GPM satellite for each sub-region versus that of the rainfall measured by the rain gauges within the same sub-region. The principle is to evaluate the overall accuracy of the measurements for the 13 administrative regions in Saudi Arabia. The station-based evaluation was performed to investigate the spatial distribution of the accuracy of each rain gauge station and determine their performance. The event-based evaluation was conducted based on the selection of the largest events that occurred during the study period ranging from October 2015 to April 2016. To ensure that the appropriate record was available to conduct the analysis, a minimum threshold of 30 rain records observed by different rain gauges every day was adopted in this study. Accordingly, 11 events resulted from the available records. 5. Results and discussion 5.1. Regional evaluation In this study, the IMERG products were evaluated both regionally and nationally using six statistical performance measures over all of Saudi Arabia during the study period. The POD and CSI mea-

sures are shown in Fig. 4a–f. The IMERG early run product performed well in detecting the precipitation throughout most of the regions in Saudi Arabia except for Makkah, which exhibited a relatively low detection (less than0.6). The IMERG late run product showed a better performance in detecting the precipitation over all of the regions of Saudi Arabia, including Makkah (greater than0.7). Moreover, the final run product showed a robust detection rate of almost 70% throughout Saudi Arabia with POD and CSI measures exceeding 0.9. The RB was tested in all regions over the study period (Fig. 4g– i). The results for the IMERG early run product indicate that the Eastern Region, Qassim, Riyadh, and Madinah showed a low RB of ±1%, which is considered a reliable performance in estimating the precipitation. The GPM satellite overestimated the precipitation over the Northern Borders region, Tabuk, Al-Jouf, and Hail with RB values of more than 20%. Meanwhile, over the remaining regions, namely, Jazan, Asir, Al-Baha, and Makkah, the GPM satellite underestimated the precipitation by more than 5%. The IMERG late run product showed the same performance for most of the regions except for Asir, which had a significantly lower RB of about ±1%. The IMERG final run product showed a substantial improvement over all of the Saudi Arabian regions. This can be noticed clearly in Fig. 4g–i, where the final run product has the lowest relative bias (±1%) compared with the other products. Overall, the final run product showed adequate results over 70% of the study area. The MAE results for the IMERG early run product showed low errors over Riyadh and the Eastern Region of less than 10 mm (see Fig. 5a–c), while Tabuk, Hail, and Al-Jouf exhibited the highest errors (more than 25 mm). This reveals that the satellite performed poorly in these regions. The rest of the regions showed a moderate performance in terms of the MAE with values ranging from 15 mm to 25 mm. Although the IMERG late run product revealed a relatively lower MAE than the IMERG early run product, that improvement was not significant in most of the regions. However, the IMERG final run product showed a significant improvement in the accuracy of the precipitation estimates (see Fig. 5a–c). All of the regions displayed an MAE of less than 10 mm, except for Hail, Madinah, and Asir, for which the MAE ranged between 10 mm and 15 mm. The RMSE results for the IMERG early run product showed a relatively low error for Riyadh and the Eastern Region of less than 10 mm, indicating that the satellite performed well over these regions (see Fig. 5d–f); meanwhile, the Northern Borders region and Makkah exhibited a moderate RMSE of 20 mm. The rest of the regions, namely, Tabuk, Hail, Al-Jouf, Madinah, Qassim, and Al-Baha, had the highest RMSE values (more than 30 mm), revealing that the satellite showed poor performance over these regions. Moreover, the IMERG late run product demonstrated considerable improvement through a reduction in the RMSE from 40 mm to 20 mm for Asir and Jazan. In contrast, the IMERG final run product revealed a substantial improvement in the accuracy of the precipitation estimates (see Fig. 5d–f). All of the regions exhibited an RMSE of less than 15 mm, except for Hail, Madinah, and Asir, in which the RMSE ranged between 15 mm and 30 mm. Riyadh, Tabuk, and Jazan had a high CC (more than 0.5, pvalues less than 0.001) for the IMERG early run product. Meanwhile, Al-Jouf and Hail displayed a moderate CC (from 0.4 to 0.5, p-value less than 0.05). In addition to Asir and Qassim, these regions exhibited a higher CC for the late run product, as shown in Fig. 5g–i. However, the Northern Borders region, the Eastern Region, Makkah, Al-Baha, and Madinah had a relatively low CC (less than 0.2, p-value greater than 0.05) for both products. In contrast, the final run product proved its accuracy with a significant increase in the CC values for almost all of the regions throughout Saudi Arabia (more than 0.5, p-value less than 0.05) with the

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

7

Fig. 4. Spatial distribution of the regional statistical performance measures for the GPM-IMERG products. (a), (b), and (c) Probability of detection (POD) measures for the IMERG early run, late run, and final run products, respectively. (d), (e), and (f) Critical success index (CSI) measures for the IMERG early run, late run, and final run products, respectively. (g), (h), and (i) Relative bias estimations for the IMERG early run, late run, and final run products, respectively.

exception of Al-Baha, Makkah and Madinah, which had CC values that ranged from 0.1 to 0.2 (p-value greater than 0.1).

5.2. Station-based evaluation The station-based evaluation was performed by testing each station individually. Then, the resultant measures were interpolated using the Geographic Information Systems (GIS) inverse distance weighting (IDW) tool to investigate the variation in each statistical measure over all of Saudi Arabia beginning with the POD and CSI, which are presented in Fig. 6a–f. The evaluation of the satellite estimates using these statistical measures for the IMERG early run product showed a substantial performance in detecting the precipitation over most parts of Saudi Arabia (more than 0.8) with the exception of the southeastern part of the country, which had values ranging from 0.5 to 0.7. The southeastern part of Saudi Arabia contains the desert of Rub’ al-Khali, and the low POD and CSI values reflect the inferior performance of the satellite over deserts. Additionally, a small portion of the western part showed a low detection of precipitation events. However, these results were improved for the IMERG late run product with POD and CSI values reaching more than 0.75 in areas of lower detection. The final run product revealed a substantial detection of precipitation throughout Saudi Arabia with values greater than 0.85 for the POD and CSI. Once again, the IMERG final run product proved that it contains the most accurate precipitation estimates compared with the other IMERG products.

The MAE, RB and RMSE were used to evaluate the accuracy of the different IMERG satellite rainfall products over the region. The MAE results showed that the eastern and western parts of Saudi Arabia have lower mean absolute errors of less than 10 mm for both the early run and the late run products (see Fig. 7a– c). Meanwhile, the MAE results were very high (more than 30 mm), indicating an inferior performance of the satellite in this region. In contrast, the IMERG final run product revealed a significant improvement in the accuracy of the precipitation estimates (see Fig. 7a–c). All of the regions had an MAE of less than 10 mm with the exception of the southwestern part, in which the MAE was between 15 mm and 25 mm. The RB and RMSE values supported the MAE results, which can be observed clearly in Fig. 7d– i. These figure panels present the same trend as the MAE results; the eastern and western parts demonstrate superiority with remarkable satellite estimations compared with the ground observations for the early run and late run products. Therefore, the IMERG final run product proved its superiority in estimating the precipitation with very low errors and biases compared with the other IMERG products. The CC distribution for the early run product (Fig. 6g–i) suggested that the satellite estimates over the southern, northern, and northeastern parts of Saudi Arabia had a good correlation with the ground observations. These results improved significantly for the late run and final run products in the southern and eastern parts of the country. However, in the northern part, the CC was lower in the late run product due to a high bias in the estimation of the precipitation, as was mentioned in the first group evaluation.

8

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

Fig. 5. Spatial distribution of the regional statistical performance measures for the GPM-IMERG products. (a), (b), and (c) Mean absolute error (MAE) measures for the IMERG early run, late run, and final run products, respectively. (d), (e), and (f) Root mean squared error estimations for the IMERG early run, late run, and final run products, respectively. (g), (h), and (i) Correlation coefficient estimations for the IMERG early run, late run, and final run products, respectively.

Nevertheless, the CC results for the northern part showed some improvement in the final run product. The rest of the country showed high variations in the CC due to the heterogeneity of the rainfall distribution and the variation in the topography of the middle and western parts of Saudi Arabia. 5.3. Event-based evaluation The IMERG products were also validated by exploring the accuracy of each product in the detection of major rain events/storms that occurred during the study period over all of Saudi Arabia. The events were considered major or large only when the number of stations that observed precipitation exceeded 30 (i.e., if the rainstorm covered more than 40% of the total land area of Saudi Arabia) on the same date. The event determination process revealed that eleven major or large events occurred during the study period as shown in Table 6. The performance measures from the eventbased evaluation are summarized in Table 7. The POD and CSI showed similar values for each event; the reason for this is that the precipitation measurements acquired by the satellite but not observed by a rain gauge station were not included within the analyzed data. Overall, for the IMERG early run product, the POD and CSI varied from 0.68 up to 1. Meanwhile, the IMERG late run product revealed a major improvement in the detection of precipitation by increasing the POD and CSI to a range from 0.71 to 1. The final run product showed a high level of perfection in the detection of precipitation events with very high POD and CSI values compared with the other IMERG products. In general, the accuracy of detection for precipitation events by the GPM satellite could be

considered substantial due to the high POD and CSI values for the three products (Table 7). Moreover, the CC values for each event for the IMERG early run product varied from 0.02 (significance level p greater than 0.1) to 0.33 (significance level p less than 0.01), except for the events observed in April 2016. These events had CC values that varied from 0.5 (significance level p less than 0.005) to 0.75 (significance level p less than 0.0005), and they were more correlated than the 2015 events. However, the CC values for the IMERG late run product improved significantly (150% on average) for the same events. The IMERG final run product showed a significant improvement compared with the other products; for instance, the CC values for the early, late and final run products for the event on December 30, 2015, were 0.005 (p greater than 0.5), 0.11 (p greater than 0.1) and 0.34 (p less than 0.05), respectively. The MAE and RMSE results indicated that the errors in the satellite estimates for all of the events with the exception of those on November 17 and December 2 were relatively low for the early run and late run products, respectively. However, the values of the two parameters decreased significantly when compared with those for the final run product. The high errors associated with these two events were caused by a storm extending over the northern, middle and western parts of Saudi Arabia, all of which have a complex topography, which increases the variability in the rainfall intensity as well as the error in the satellite measurements. The MAE for the IMERG early run product ranged from 5 mm to 36 mm; meanwhile, it decreased for the IMERG late and the IMERG final run products by approximately 20% and 45% on average, respectively. The RB and RMSE also exhibited the same trend of

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

9

Fig. 6. Spatial distribution of the statistical matrices of the GPM-IMERG products used in the station-based evaluation. (a), (b), and (c) Probability of detection (POD) measures for the IMERG early run, late run, and final run products, respectively. (d), (e), and (f) and Critical success index (CSI) measures for the IMERG early run, late run, and final run products, respectively. (g), (h), and (i) Correlation coefficient (CC) estimations for the IMERG early run, late run and final run products, respectively.

the mean absolute error (Table 7). Overall, the MAE, RSME, and RB showed significant improvements among the GPM products. The MAE, RSME, and RB values decreased from the early run to the late run products and from the late run to the final run products; for example, the MAE, RMSE, and RM values decreased from 6 mm to 3 mm, from 8.6 mm to 6.24 mm, and from 0.75% to 0.22%, respectively, for the event on December 30, 2015.

6. Conclusions This research focuses on a validation of the Integrated MultisatellitE Retrievals for Global Precipitation Measurement (GPM) mission (IMERG) satellite rainfall products (i.e., the early, late, and final run products) using ground-based rain gauge observations as a reference over the Kingdom of Saudi Arabia for the period from October 2015 to April 2016. The evaluation was carried out using six commonly used statistical performance measures: the MAE, RMSE, RB, CC, POD, and CSI. Three main comparisons or evaluations were conducted, namely, event-based, region-based, and station-based evaluations. The region-based evaluation showed low errors and biases over the Eastern Region, Riyadh, Madinah, and Qassim for the IMERG early run, late run and final run products, while Tabuk, Al-Jouf, the Northern Borders region, and Hail showed relatively high errors and biases for the early run and late run products. However, for the final run product, the errors and biases decreased significantly, leaving only Tabuk and Al-Jouf with relatively high values. A comparison between the satellite estimates and ground observa-

tions revealed that Riyadh, Tabuk, Jazan, Al-Jouf, and Hail had relatively high CC values compared with the other regions for all IMERG products. The late run product showed a high correlation in Asir and Qassim. Furthermore, the correlation results showed a remarkable improvement within the final run product with a CC of 0.5 over more than 70% of the total land area of Saudi Arabia. Moreover, in terms of the satellite detection accuracy, all of the IMERG products tended to be consistent in detecting precipitation events, although the final run product exhibited a higher detection accuracy than the other two products. The station-based evaluation for the IMERG early run product revealed high biases and errors for the stations located in the northern and northwestern parts of Saudi Arabia, while the middle and southern parts showed moderate errors and biases. Conversely, the eastern part and portions of the western part exhibited very low biases and estimation errors. Meanwhile, the stationbased evaluation revealed a decrease in the biases and estimation errors within the late run product over the southern and western parts of the country. However, the northern and middle parts of Saudi Arabia still suffered from high estimation errors and biases. In contrast, the final run product showed a significant improvement over more than 80% of the country. In terms of the correlation between the satellite estimates and ground-based observations, the southern, northern, and northeastern parts of Saudi Arabia were shown to have high correlations for the early run product. The results for the late run and final run products were also significantly enhanced in the southern and eastern parts of the country. However, in the northern part, the correlation was lower for the late run product and higher for the final run product.

10

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

Fig. 7. Spatial distribution of the statistical matrices of the GPM-IMERG products used in the station-based evaluation. (a), (b), and (c) Mean absolute error (MAE) measures for the IMERG early run, late run, and final run products, respectively. (d), (e), and (f) Relative bias (RB) estimations for the IMERG early run, late run, and final run products, respectively. (g), (h), and (i) Root mean squared error (RMSE) estimations for the IMERG early run, late run, and final run products, respectively.

Table 6 Descriptive statistical analysis of major rainfall events (all units in mm/day). Event Date

Mean

Standard Deviation

Minimum

Maximum

Total Rainfall

No. of Stations

28-10-2015 16-11-2015 17-11-2015 23-11-2015 24-11-2015 2-12-2015 23-12-2015 30-12-2015 4-4-2016 12-4-2016 13-04-2016

8.5 15.7 7.4 11.5 23.9 11.7 8.4 5.1 13.2 11.2 30.8

16.2 14.4 6.8 8.3 18.2 8.4 8.6 3.5 14.7 10.4 25

0.5 1 1 0.5 1 1 0.5 0.5 1.3 0.5 1

86 53 30 39 77 33.5 42 15 84 64 96

288.3 439.2 297.5 656.2 1287.7 397.8 461.4 223.2 658.4 750 955.7

34 28 40 57 54 34 55 44 50 67 31

The rest of the country showed high variations in the correlation due to the heterogeneous rainfall distribution and variations in the topography of the middle and western parts of Saudi Arabia. Furthermore, the satellite detection of rainfall events was precise for all of the IMERG products, although the final run product showed the best detection among all of the products. For the event-based evaluation, the IMERG final run product showed a much higher accuracy than the other IMERG products. For almost all of the events, the final run product had a higher correlation between the satellite estimations and ground observations compared with the other IMERG products. Additionally, the mean absolute error, relative bias, and root mean squared error were relatively low for the final run product. Moreover, the probability of detection and critical success index parameters were remarkably

higher for the final run product. In particular, higher errors and biases were detected for two events, namely, those on November 17 and December 2, using the early run and late run products; meanwhile, for the final run product, the errors and biases for the same events were lower by at least 50%. In general, the poor results observed over the northern and southern parts of Saudi Arabia can be interpreted as a result of the high heterogeneity in the rainfall distribution in conjunction with the high variability in the terrestrial topography in these regions. In addition, many researchers indicated that the satellite detection accuracy for precipitation is quite low in high-latitude regions, reinforcing the poor results in the southern region. Moreover, the northern region contains a sparse distribution of rain gauge stations relative to the middle and western parts of the

11

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12 Table 7 Event-based comparison of all GPM satellite products (Early, Late and Final run). Event Date

28-10-2015 16-11-2015 17-11-2015 16+17-11-2015 23-11-2015 24-11-2015 23+24-11-2015 2-12-2015 23-12-2015 30-12-2015 4-04-2016 12-04-2016 13-04-2016 12+13-04-2016 *

Correlation Coefficient (CC)

Mean Absolute Error (MAE)*

Root Mean Squared Error (RMSE)*

Relative Bias (BIAS)

Probability of Detection (POD)

Critical Success Index (CSI)

Early

Late

Final

Early

Late

Final

Early

Late

Final

Early

Late

Final

Early

Late

Final

Early

Late

Final

0.03 0.02 0.12 0.04 0.24 0.22 0.33 0.18 0.13 0.005 0.75 0.24 0.53 0.42

0.07 0.15 0.2 0.002 0.33 0.14 0.29 0.27 0.09 0.11 0.74 0.35 0.51 0.57

0.008 0.09 0.28 0.13 0.31 0.40 0.44 0.29 0.24 0.34 0.58 0.35 0.58 0.53

7.4 21.8 23.2 23.8 9.5 19.2 19.5 36.9 6.8 6 9.1 9.4 17.8 12.7

8.2 23.4 21.1 22.8 10 19.7 21 21.9 6.6 5.8 9 9.7 15.7 12.1

7.6 12.1 9.1 10.3 9.8 15 17.4 10.3 6.1 3.6 8.9 10.9 17.8 14.3

17.3 30.9 33.4 33.2 12.8 24.3 25 51.4 10.6 8.4 12.8 16.7 25.7 21.1

17.9 33.7 29.5 32 13.6 25.4 26.8 64.7 10.5 8.6 12.1 17.5 22.3 19.3

17.1 17.6 14.6 16.1 14.6 20.3 24.8 12.3 10.2 6.2 13.8 18 25.6 22

2.19 1.48 6.29 2.09 0.07 0.79 0.44 8.2 1.34 0.4 1.25 0.1 1.58 0.28

1.98 1.53 6.2 1.99 0.31 0.67 0.27 12.16 1.16 0.75 0.81 0.35 0.46 0.03

1.72 1.55 1.51 0.02 0.47 0.22 0.02 1.38 1.24 0.22 0.78 0.58 0.45 0.11

0.74 0.68 0.9 0.87 0.88 0.87 0.91 1 0.73 0.89 0.88 0.97 0.97 1

0.85 0.71 0.93 0.89 0.89 0.94 0.93 1 0.82 0.95 0.98 0.97 0.97 0.99

0.97 0.75 0.93 0.89 0.96 0.98 0.99 1 0.89 0.91 0.96 0.99 1 1

0.74 0.68 0.9 0.87 0.88 0.87 0.91 1 0.73 0.89 0.88 0.97 0.97 1

0.85 0.71 0.93 0.89 0.89 0.94 0.93 1 0.82 0.95 0.98 0.97 0.97 0.99

0.97 0.75 0.93 0.89 0.96 0.98 0.99 1 0.89 0.91 0.96 0.99 1 1

Units are in mm/day

country. This small number of stations increased the errors and biases in the estimation of the precipitation, as the scarcity of rain gauge stations in northern Saudi Arabia is inadequate to represent the variability in the precipitation distribution. Overall, the results revealed that the IMERG final run product may have the potential to play a significant role in complementing or replacing ground precipitation measurements, thereby offering a useful approach for ungauged or poorly gauged regions. In addition, the GPM products are available at relatively high spatial and temporal resolutions, making them useful for many hydrological applications in regions such as Saudi Arabia that have very sparse distributions of rain gauge stations, most of which report only at a daily interval. Moreover, the IMERG near-real-time products are available for most of the regions in Saudi Arabia, and thus, these products may be useful for early flood warning systems or other applications that require near-real-time rainfall data, although their accuracy must be continuously evaluated.

Acknowledgments The authors would like to acknowledge the support provided by the Deanship of Scientific Research (DSR) at King Fahd University of Petroleum and Minerals (KFUPM) for funding this work through project no. RG1305-1&2.

References Abdullah, M.A., Al-Mazroui, M.A., 1998. Climatological study of the southwestern region of Saudi Arabia. I. Rainfall analysis. Clim. Res. 9, 213–223. https://doi.org/ 10.3354/cr009213. Almazroui, M., 2011. Calibration of TRMM rainfall climatology over Saudi Arabia during 1998–2009. Atmos. Res. 99, 400–414. https://doi.org/10.1016/j. atmosres.2010.11.006. Al-Mazroui, M., Climatological study over the southwestern region of the Kingdom of Saudi Arabia with special reference to rainfall distribution. M.Sc. Thesis, King Abdulaziz University, Jeddah, Saudi Arabia, 1998. Al-Zahrani, M., Husain, T., 1998. An algorithm for designing a precipitation network in the south-western region of Saudi Arabia. J. Hydrol. 205, 205–216. https:// doi.org/10.1016/S0022-1694(97)00153-4. Asong, Z.E., Razavi, S., Wheater, H.S., Wong, J.S., 2017. Evaluation of Integrated Multisatellite Retrievals for GPM (IMERG) over Southern Canada against Ground Precipitation Observations: A Preliminary Assessment. J. Hydrometeorol. 18, 1033–1050. https://doi.org/10.1175/JHM-D-16-0187.1. Dixon, S.G., Wilby, R.L., 2015. Forecasting reservoir inflows using remotely sensed precipitation estimates: a pilot study for the River Naryn, Kyrgyzstan. Hydrol. Sci. J. 61. 150113093103002 10.1080/02626667.2015.1006227. Ebert, E.E., Janowiak, J.E., Kidd, C., 2007. Comparison of near-real-time precipitation estimates from satellite observations and numerical models. Bull. Am. Meteorol. Soc. 88, 47–64. https://doi.org/10.1175/BAMS-88-1-47.

Germann, U., Galli, G., Boscacci, M., Bolliger, M., 2006. Radar precipitation measurement in a mountainous region. Q. J. R. Meteorol. Soc. 132, 1669– 1692. https://doi.org/10.1256/qj.05.190. Hag-elsafi, S., El-Tayib, M., 2016. Spatial and statistical analysis of rainfall in the Kingdom of Saudi Arabia from 1979 to 2008. Weather 71, 262–266. https://doi. org/10.1002/wea.2783. Hasanean, H., Almazroui, M., 2015. Rainfall: Features and variations over Saudi Arabia, a review. Climate 3, 578–626. https://doi.org/10.3390/cli3030578. Heistermann, M., Kneis, D., 2011. Benchmarking quantitative precipitation estimation by conceptual rainfall-runoff modeling. Water Resour. Res. 47, 1– 23. https://doi.org/10.1029/2010WR009153. Hong, Y., Hsu, K.-L., Sorooshian, S., Gao, X., 2004. Precipitation estimation from remotely sensed imagery using an artificial neural network cloud classification system. J. Appl. Meteorol. 43, 1834–1853. https://doi.org/10.1175/JAM2173.1. Huffman, G.J., Bolvin, D.T., Nelkin, E.J., Wolff, D.B., Adler, R.F., Gu, G., Hong, Y., Bowman, K.P., Stocker, E.F., 2007. The TRMM multisatellite precipitation analysis (TMPA): Quasi-global, multiyear, combined-sensor precipitation estimates at fine scales. J. Hydrometeorol. 8, 38–55. https://doi.org/10.1175/JHM560.1. Huffman, G.J., Bolvin, D.T., Braithwaite, D., Hsu, K., Joyce, R., Kidd, C., Nelkin, E.J., Xie, P., 2015. NASA Global Precipitation Measurement (GPM) Integrated MultisatellitE Retrievals for GPM (IMERG). Natl. Aeronaut. Space Admin. Joyce, R.J., Janowiak, J.E., Arkin, P.A., Xie, P., 2004. CMORPH: a method that produces global precipitation estimates from passive microwave and infrared data at high spatial and temporal resolution. J. Hydrometeorol. 5, 487–503. https://doi. org/10.1175/1525-7541(2004) 005<0487:CAMTPG>2.0.CO;2. Khodadoust Siuki, S., Saghafian, B., Moazami, S., 2017. Comprehensive evaluation of 3-hourly TRMM and half-hourly GPM-IMERG satellite precipitation products. Int. J. Remote Sens. 38, 558–571. https://doi.org/10.1080/01431161.2016.1268735. Kidd, C., Huffman, G., 2011. Global precipitation measurement. Meteorol. Appl. 18, 334–353. https://doi.org/10.1002/met.284. Kubota, T., Shige, S., Hashizume, H., Aonashi, K., Takahashi, N., Seto, S., Hirose, M., Takayabu, Y.N., Ushio, T., Nakagawa, K., Iwanami, K., Kachi, M., Okamoto, K., 2007. Global precipitation map using satellite-borne microwave radiometers by the GSMaP project: Production and validation. IEEE Trans. Geosci. Remote Sens. 45, 2259–2275. https://doi.org/10.1109/TGRS.2007.895337. Li, M., Shao, Q., 2010. An improved statistical approach to merge satellite rainfall estimates and raingauge data. J. Hydrol. 385, 51–64. https://doi.org/10.1016/j. jhydrol.2010.01.023. Li, Z., Yang, D., Hong, Y., 2013. Multi-scale evaluation of high-resolution multisensor blended global precipitation products over the Yangtze River. J. Hydrol. 500, 157–169. https://doi.org/10.1016/j.jhydrol.2013.07.023. Ministry of Water and Electricity (MOWE), Rain Gauge Summary over the Kingdom of Saudi-Arabia. http://app.mowe.gov.sa/DailyRainsNews/Rain_Dams.aspx. NASA Constellation Partners | Precipitation Measurement Missions, http:// pmm.nasa.gov/GPM/constellation-partners (accessed on 20 Feb. 2016a). NASA Global Precipitation Measurement (GPM) Mission Overview | Precipitation Measurement Missions, http://pmm.nasa.gov/GPM (accessed on 20 Feb. 2016b). NASA GPM Data Downloads | Precipitation Measurement Missions, http:// pmm.nasa.gov/data-access/downloads/gpm (accessed on 20 Feb. 2016c). NASA GPM IMERG – Early Run Product, ftp://jsimpson.pps.eosdis.nasa.gov/NRTPUB/ imerg/early/, 2016d. NASA GPM IMERG – Late Run Product, ftp://jsimpson.pps.eosdis.nasa.gov/NRTPUB/ imerg/late/, 2016e. NASA GPM IMERG – Final Run Product, ftp://jsimpson.pps.eosdis.nasa.gov/NRTPU/, 2016f. Ning, S., Wang, J., Jin, J., Ishidaira, H., 2016. Assessment of the latest GPM-era highresolution satellite precipitation products by comparison with observation gauge data over the Chinese Mainland. Water 8, 481. https://doi.org/10.3390/ w8110481.

12

M.T. Mahmoud et al. / Journal of Hydrology 559 (2018) 1–12

Sungmin, O., Foelsche, U., Kirchengast, G., Fuchsberger, J., Tan, J., Petersen, W.A., 2017. Evaluation of GPM IMERG Early, Late, and Final rainfall estimates with WegenerNet gauge data in southeast Austria. Hydrol. Earth Syst. Sci. Discuss. 1– 21 (2017). https://doi.org/10.5194/hess-2017-256. Prakash, S., Mitra, A.K., AghaKouchak, A., Liu, Z., Norouzi, H., Pai, D.S., 2016. A preliminary assessment of GPM-based multi-satellite precipitation estimates over a monsoon dominated region. J. Hydrol. https://doi.org/10.1016/j. jhydrol.2016.01.029. Raıch, R., Alpert, P., Messer, H., 2017. Vertical Precipitation Estimation Using Microwave Links in conjunction with Weather Radar. In: 15th International Conference on Environmental Science and Technology. Sßen, Z., Al-Suba’i, K., 2002. Hydrological considerations for dam siting in arid regions: a Saudi Arabian study. Hydrol. Sci. J. – Journal des Sciences Hydrologiques, 47, 173–186. Sorooshian, S., Hsu, K.-L., Gao, X., Gupta, H.V., Imam, B., Braithwaite, D., 2000. Evaluation of PERSIANN system satellite-based estimates of tropical rainfall. Bull. Am. Meteorol. Soc. 81, 2035–2046. https://doi.org/10.1175/1520-0477 (2000) 081<2035:EOPSSE>2.3.CO;2. Sorooshian, S., AghaKouchak, A., Arkin, P., Eylander, J., Foufoula-Georgiou, E., Harmon, R., Hendrickx, J.M.H., Imam, B., Kuligowski, R., Skahill, B., SkofronickJackson, G., 2011. Advanced concepts on remote sensing of precipitation at multiple scales. Bull. Am. Meteorol. Soc. 92, 1353–1357. https://doi.org/ 10.1175/2011BAMS3158.1. Subyani, A.M., 2004. Geostatistical study of annual and seasonal mean rainfall patterns in southwest Saudi Arabia. Hydrol. Sci. J. – Journal des Sciences Hydrologiques. 49, 803–817. doi: 10.1623/hysj.49.5.803.55137. Tang, G., Ma, Y., Long, D., Zhong, L., Hong, Y., 2016. Evaluation of GPM Day-1 IMERG and TMPA Version-7 legacy products over Mainland China at multiple

spatiotemporal scales. J. Hydrol. 533, 152–167. https://doi.org/10.1016/j. jhydrol.2015.12.008. Tapiador, F.J., Turk, F.J., Petersen, W., Hou, A.Y., García-Ortega, E., Machado, L.A.T., Angelis, C.F., Salio, P., Kidd, C., Huffman, G.J., de Castro, M., 2012. Global precipitation measurement: Methods, datasets and applications. Atmos. Res. 104–105, 70–97. https://doi.org/10.1016/j.atmosres.2011.10.021. Tekeli, A.E., Fouli, H., 2016. Evaluation of TRMM satellite-based precipitation indexes for flood forecasting over Riyadh City, Saudi Arabia. J. Hydrol. 541, 471– 479. https://doi.org/10.1016/j.jhydrol.2016.01.014. The Ministry of Water, 1984. Water Atlas of Saudi Arabia. Riyadh, Saudi Arabia. Tian, Y., Peters-Lidard, C.D., Choudhury, B.J., Garcia, M., 2007. Multitemporal analysis of TRMM-based satellite precipitation products for land data assimilation applications. J. Hydrometeorol. 8, 1165–1183. https://doi.org/ 10.1175/2007JHM859.1. Wang, Z., Zhong, R., Lai, C., Chen, J., 2017. Evaluation of the GPM IMERG satellitebased precipitation products and the hydrological utility. Atmos. Res. 196, 151– 163. https://doi.org/10.1016/j.atmosres.2017.06.020. Wu, H., Adler, R.F., Hong, Y., Tian, Y., Policelli, F., 2012. Evaluation of global flood detection using satellite-based rainfall and a hydrologic model. J. Hydrometeorol. 13, 1268–1284. https://doi.org/10.1175/JHM-D-11-087.1. Yong, B., Ren, L.-L., Hong, Y., Wang, J.-H., Gourley, J.J., Jiang, S.-H., Chen, X., Wang, W., 2010. Hydrologic evaluation of multisatellite precipitation analysis standard precipitation products in basins beyond its inclined latitude band: a case study in Laohahe basin, China. Water Resour. Res. 46, 1–20. https://doi.org/10.1029/ 2009WR008965. Zhou, T., Yu, R., Chen, H., Dai, A., Pan, Y., 2008. Summer precipitation frequency, intensity, and diurnal cycle over China: A comparison of satellite data with rain gauge observations. J. Clim. 21, 3997–4010. https://doi.org/10.1175/ 2008JCLI2028.1.