Multiscale assessment of TRMM (3B42 V7) and GPM

Abstract. The performance of Tropical Precipitation Measurement Mission (TRMM) and its successor, Global Precipitation Measurement (GPM), has provided hydrologists with a source of critical precipitation data for hydrological applications in basins where ground-based observations of precipitation are sparse, or spatially undistributed. The very high temporal and spatial resolution satellite precipitation products have therefore become a reliable alternative that researchers are increasingly using in various hydro-meteorological and hydro-climatological applications. This study aims to evaluate statistically and hydrologically the TRMM (3B42 V7) and GPM (IMERG V5) satellite precipitations products (SPPs), at multiple temporal scales from 2010 to 2017, in a mountainous watershed characterized by the Mediterranean climate. The results show that TRMM (3B42 V7) and GPM (IMERG V5) satellite precipitation products have a significant capacity for detecting precipitation at different time steps. However, the statistical analysis of SPPs against ground observation shows good results for both statistical metrics and contingency statistics with notable values (CC > 0.8), and representative values relatively close to 0 for the probability of detection (POD), critical success index (CSI), and false alarm ratio (FAR). Moreover, the sorting of the events implemented on the hydrological model was performed seasonally, at daily time steps. The calibrated episodes showed very good results with Nash values ranging from 53.2 % to 95.5 %. Nevertheless, the (IMERG V5) product detects more efficiently precipitation events at short time steps (daily), while (3B42 V7) has a strong ability to detect precipitation events at large time steps (monthly and yearly). Furthermore, the modeling results illustrate that both satellite precipitation products tend to underestimate precipitation during wet seasons and overestimate them during dry seasons, while they have a better spatial distribution of precipitation measurements performance, which shows the importance of their use for basin modeling and potentially for flood forecasting in Mediterranean catchment areas.



40
Precipitation is a major force in global climate change and plays an important role in hydrological and 41 meteorological applications (Yuan et al., 2017). As a significant phenomenon in nature; precipitation has complex 42 characteristics of spatiotemporal variations. It is one of the critical components of the global exchange of the 43 surface material, the hydrological cycle, and disaster prevention (Bollasina et al., 2011;Zhu et al., 2012).

44
The variability of precipitation in mountainous areas directly affects local agriculture and ecological environment 45 (Xia et al., 2015;Jiang et al., 2017). Moreover, the heavy precipitation events that occurred in mountainous areas 46 frequently generate flash floods (Borga et al., 2010). Therefore, the acquisition of reliable and accurate precipitation information in mountainous areas is of great significance to social and economic development and 48 related scientific researches (Germann et al., 2006). Rain gauge observation could provide a moderately accurate 49 method for point-based precipitation measurement. However, rain gauges in mountainous regions are often scarce, 50 irregular, and sometimes unavailable (Xia et al., 2015;Hrachowitz et al., 2011). Thus, in the applications that need 51 high spatiotemporal resolution precipitation data, such as flood disaster forecasts, gauge data are regularly 52 insufficient (Mei et al., 2014;Yi et al., 2018). Contrary to rain gauge precipitations, satellite remote sensing has 53 the advantages of completely scanning the entire study region and convenient access to the data, providing an 54 alternate way to monitor precipitation at regional and global scales (Chen et al., 2018).  (Huffman et al., 2007), and Integrated Multi-satellitE Retrievals for (GPM) mission (IMERG) (Hou et al., 2014).

62
Compared to these satellite precipitation products, the TRMM 3B42V7 precipitation product performance is higher 63 than other products, especially in estimating extreme precipitation events in several areas around the world (Tong

68
Given the excellent successes of the TRMM, the GPM Core Observatory satellite was set in motion by NASA and 69 JXAX as a successor of TRMM in February 2014. Compared with TRMM, the potential of GPM to detect liquid 70 and solid precipitation is improved by carrying space-borne dual-frequency precipitation radar (Chandrasekar et 71 al., 2015). Additionally, the GPM Core Observatory carrying a conical scanning multichannel microwave imager 72 offers a wider measurement range (Hou et al., 2014). The lately released IMERG further expands quasi-global 73 coverage from (50°N-50°S) to (60°N-60°S) and provides precipitation estimates with a finer spatial resolution of 74 (0.1° X 0.1°) and temporal resolution of 30 minutes (Liu et al., 2017).

75
Since the deliverance of IMERG products, Many studies have been conducted to evaluate and compare the 76 performance of TMPA and IMERG products regarding rain gauges observations in many regions, such USA

101
This sub-basin is characterized by Mediterranean climate strongly influenced by altitude. Taferiat hydrometric 102 station controls the discharge of the Zat Basin, and also serves as rain gauge. It receives an annual rainfall average 103 ranges from 133 mm /year to 913 mm /year; precipitation is mainly concentrated during the rainy period from

104
October to April and a hot and dry period from May to September. Therefore, this study region is subject to 105 frequent flash floods and droughts.

122
The TRMM (3B42 V7) precipitation products were generated by using the TRMM 3B42 Version 7 algorithm 123 (Huffman et al., 2007). It was designed to combine various microwaves MW, and infrared IR satellite-based 124 precipitation estimates with gauge adjustments observations to provide 3-hourly quasi-global quantitative 125 precipitation estimates (Hou et al., 2014). The 3B42 V7 product is derived by bias-adjusting the near-real-time 126 product with the GPCC monthly gauge-analysis precipitation data set, and it has two-month latency (Yuan et al., 127 2017). The product can produce rational precipitation estimates in a 0.25° spatial resolution with a quasi-global 128 coverage (50°S-50°N). In this study, the TRMM 3B42 V7 daily precipitation product was acquired from the     ). Therefore, a direct comparison is used in this study. To evaluate these two SPPs, we only considered the 153 grids that cover the data of the single gauging station present in the Zat basin. Therefore, the grids not covering 154 the gauge station were excluded from the assessment.

155
Furthermore, to evaluate the ability of the SPPs to reproduce rainfall events, it was decided to use them as input  (Table 2).

167
Four statistical measures were selected, including correlation coefficient (CC), root mean square error (RMSE), 168 relative bias (RB), and bias (bias), which were calculated to statistically evaluate the two PPS products.

169
The Pearson Correlation Coefficient (CC) measures the agreement between the PPS products and the gauge data.

170
The (RMSE) was used to represent the mean magnitude of the error. The (RB) and (bias) was applied to evaluate 171 the systematic bias between the SPPs and gauge data in percent and amount of precipitation, respectively. The   divided by the total number of successes, false alarms, and failures. Table 3 shows the formulas for these metrics.   The rainfall time series of the two selected satellite products and the rain gauge at different timescales in the Zat 207 basin are presented in (Figure 3). In general, the 3B42 V7 and IMERG V5 products present similar chronological 208 precipitation patterns to those of the gauge. However, it can be seen that the product 3B42 V7 slightly 209 overestimated the daily precipitation, while the product IMERG V5 showed good performance on the daily 210 timescale ( Figure 3A). Regarding the monthly precipitation series, the product 3B42 V7 underestimated the 211 monthly precipitation, while IMERG V5 clearly showed good initial agreement with the observed precipitation,

212
although from 2015 IMERG V5 slightly overestimated the monthly precipitation ( Figure 3B). As for the annual

240
The SPPs were statically compared against the ground observations to evaluate their accuracy and reliability.    Figures (4 and 5 B), Represents scatterplots and boxplot of precipitation from 3B42 V7 and IMERG V5 at 263 monthly scale. Compared with gauge data, it can be seen that both products slightly underestimate the precipitation.

282
The categorical statistical metrics of 3B42 V7 and IMERG V5 at different time scales are shown in (Table 5).

304
The HEC-HMS model was used to calibrate the daily rainfall events from (1/09/2010) to (31/08/2017) according 305 to the different seasons, at the level of the Zat basin, using the rainfall and Runoff data from the Taferiat gauge 306 station, and satellite precipitation products, the four episodes that we chose to present are the three most 307 representative of the data series. The hydrological simulations were carried out according to two different 308 scenarios:

309
Scenario 1: Simulation and calibration by implementing the model with observed rainfall and discharge data.

310
Scenario 2: Simulation and calibration using rainfall from both satellite products with observed flows.   Hydrographs were well reproduced for both the rise and the recession; peak flows were achieved and evaluation 321 criteria are very satisfactory with RMSE of 0.5 and 0.4 and Nash of 77% and 81.11% respectively.

322
The simulation results of scenario 2 are better than those of scenario 1. This is explained by the fact that the

335
The hydrograph of scenario 1 represents a simulated flow curve quite illustrative, the rising curve and the recession 336 were well reproduced, contrary to the peak flow which has not been reached, the evaluation criteria are moderately 337 good representing values of RMSE = 0.6 and Nash = 63.1%, this is induced by the fact that the snowy fraction has 338 not been taken into account due to the irregularity of the precipitation measuring stations.

339
On the other hand, the hyetogram of scenario 2 illustrates a good spatial distribution of precipitation, although the 340 curves of the simulated flows are quite well reproduced, and the peak flows were not reached, the evaluation 341 criteria are acceptable with RMSE of 0.7 for IMERG V5 and 3B42 V7 and Nash of 57.8% and 53.2% respectively.

342
Given this, an intense underestimation of the precipitation was noticed while obtaining the calibration results,

375
It is important to note that the basin response is relatively slow due to the initial soil conditions, additionally to the 376 overestimation of precipitation from the satellite products during the high-temperature seasons.

392
To evaluate the accuracy of 3B42 V7 and IMERG V5 satellite precipitation products, several quantitative, 393 categorical, and graphical statistical measurements were used, (R, RMSE, MAE, R Bias, Bias) are used to 394 quantitatively analyze the accuracy of satellite precipitation products, and (POD, CSI, FAR, and FBI) were used 395 to evaluate the precipitation detection capability of satellite precipitation products, and to simulate satisfactorily 396 the flooding events in hydrological model.

397
The conclusions resulting from this study are summarized as follows:

398
(1) IMERG V5 and 3B42 V7 products performed well in estimating daily, monthly, and annual precipitation 399 compared to observed data from the Taferiat station. SPPs products slightly underestimated the daily and annual 400 precipitation, while 3B42 V7 slightly underestimated the monthly precipitation, especially during winter periods.

401
(2) Compared to the ground applications, 3B42V7 and IMERG V5 showed good correlation results at the daily very good, ranging from 53.2% to 95.5% for the 3B42 V7, IMERG V5, and observed precipitation respectively.

426
The main point to remember is that both satellite precipitation products tend to underestimate precipitation during 427 wet seasons and overestimate them during dry seasons. The proposed method is an interesting approach to apply 428 for solving the problem of insufficient observed data in the Mediterranean regions.

429
Therefore, the results of this study are of great importance for analyzing the prospect's application of SPPs at