Modeling run-off flow hydrographs using remote sensing data: an application to the Bashar basin, Iran

Precipitation is a complex parameter in the hydrologic cycle that influences the environmental characteristics of a geographical area (Hassan et al. 2020). Iran, a semi-arid country with low annual precipitation, experiences large short-time high-intensity rainfalls, often resulting in significant financial and human losses. The spatial variability of these events is likely influenced by the diverse topography of the country, necessitating the placement of dense rain gauge stations. However, Iran lacks adequate rain gauge station coverage (Arab Amiri & Mesgari 2019). This limitation leads to challenges in precipitation-runoff simulations, as the base rain gauge station may be far from the catchment control point or not representative of the entire watershed. Additionally, point-wise rain gauge measurements encounter instrumental and interpretation problems (Tapiador et al. 2012). The spatiotemporal resolutions and low quality of precipitation data pose major challenges in many regions worldwide, including Iran (Pai et al. 2014, Salio et al. 2015; Salmon et al. 2015; Hussein et al. 2020).

Advanced remote sensing technology has emerged as a viable alternative for areas with limited or irregularly distributed rain gauge stations. Satellite remote sensing offers temporal and spatial coverage that can bridge the gaps in ground measurements and mitigate issues caused by station heterogeneity. It provides a better understanding of precipitation as a crucial parameter across large areas. Several satellite rainfall sources have been used for precipitation estimation, including the Tropical Rainfall Measuring Mission (TRMM) satellite, specifically the TRMM Multi-satellite Precipitation Analysis (TMPA) (Hong et al. 2012), the Climate Prediction Center MORPHing (CMORPH) technique data (Joyce et al. 2004) and the half-hourly Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (GPM) (IMERG) rainfall data (Huffman & Bolvin 2013). While many studies have focused on rainfall estimation using satellite data, few have evaluated its application in rainfall-runoff modeling. For instance, Kawo et al. (2021) evaluated GPM-IMERG v6 over the Lake Hawassa catchment using a linear scaling bias correction approach and comparing the satellite rainfall products with ground observations. They found a good correlation between IMERG's estimated rainfall and observed rainfall after bias correction, with variable performance across seasons. Prakash et al. (2015) evaluated rainfall in India, a region prone to heavy rainfall and floods, and found that GPM-IMERG data perform better than TMPA data in heavy rainfall areas. GPM satellite-based estimates offer advantages over rain gauge measurements, such as high temporal resolution and near-global spatial coverage (Sharifi et al. 2016). Many scientists have investigated the performance of IMERG GPM rainfall estimates in hydrological applications (Sungmin et al. 2017; Gosset et al. 2018; Mohsan et al. 2018). Satellite-based estimates, including IMERG data, have been identified as suitable substitutes for ground-based precipitation data, as they cover areas that are difficult to access with other instruments (Alsumaiti et al. 2020). Tang et al. (2016) evaluated the IMERG, 3B42V7 and 3B42RT precipitation data from GPM and TRMM satellites, identifying IMERG data as a good substitute for TRMM satellite precipitation data. Wang et al. (2017) evaluated the performance and hydrological tools of the new generation of GPM and TRMM satellites, specifically IMERG and 3B42-V7, respectively, using ground station data in the Beijing River basin of China. The results demonstrated satisfactory accuracy for IMERG satellite data, with a high correlation coefficient (CC) and relative deviation ratio. Mahmoud et al. (2018) assessed the accuracy of GPM-IMERG products in Saudi Arabia by comparing them with daily rainfall. Their study indicated the potential of using the IMERG Final run product to complement or replace ground precipitation observations in poorly gauged and ungauged regions.

Jawad (2021) assessed the accuracy of different GPM era satellite precipitation products (SPPs) by using them to develop a hydrological model of the sparsely gauged and trans-boundary Brahmaputra catchment. Flow data for model calibration and validation were available at the catchment outlet, hindering the calibration process. IMERG-Late performed better than other two precipitation products for hydrological modeling. However, for flood inundation mapping, any of the three selected products can be used with good results. Al-Areeq et al. (2021) developed a physically based, fully distributed hydrologic model using three IMERG high-resolution satellite rainfall products and a limited number of ground observations to simulate two flood events for model calibrations and validation. The model simulations underestimated rainfall and resulted in an amplified underestimation of the runoff for peak discharge events.

Khodadoust Siuki et al. (2016) evaluated 3-h 3B42V7 and half-hour IMERG rainfall data using rain gauges in the Razavi Khorasan area. They assessed the accuracy of the data using statistical indicators (Critical Success Index (CSI), mean absolute error (MAE), false alarm ratio (FAR), probability of detection (POD), root-mean-square error (RMSE)) and linear correlation. The results showed a higher correlation between the rain gauge data and estimates from IMERG compared to 3B42-V7. Shirmohammadi-Aliakbarkhani & Akbari (2020) assessed the daily rainfall data from the TRMM-3B42-V7 and GPM products with rain gauge data over a semi-arid climate in northeast Iran using 13 stations for a period of 4 years. The comparison was performed on daily, monthly and seasonal timescales. All SPPs correlated well with measurements from rain gauge measurements on the monthly timescale but moderately on the daily timescale. TRMM data performed slightly better than GPM products. Parisuoj et al. (2018) evaluated two rainfall sources, TRMM and Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), using the Hydrologic Engineering Center-Hydrologic Modeling System (HEC-HMS) rainfall-runoff model to simulate runoff. The results indicated higher simulation accuracies of TRMM precipitation compared to those of PERSIANN on a daily scale. Eini et al. (2018) used the Soil and Water Assessment Tool model to simulate runoff on a monthly scale and investigated the efficiency of two rainfall sources: Climatic Research Unit (CRU) and National Centers for Environmental Prediction-Climate Forecast System Reanalysis (NCEP-CFSR) in the Maharloo watershed from 1980 to 2013. They concluded that the average Nash–Sutcliffe index for observational and computational data was about 0.91. For surface runoff estimation, the CRU source performed better than the NCEP-CFSR source.

Few studies have evaluated the effectiveness of choosing alternative precipitation data types for runoff simulation, especially in Iran, a country frequently experiencing severe floods and subsequent damages. This study aims to evaluate and compare the accuracy of the HEC-HMS rainfall-runoff simulation model using two different types of precipitation data (station and satellite).

A work-flow showing the study process is shown in Figure 2. To achieve the objectives of the study, rain gauge station data were incorporated into the HEC-HMS software to calibrate the model parameters. This involved comparing three output hydrographs with observed data from the hydrometric stations and adjusting the model parameters to align the predicted flow hydrographs with the observed ones. Subsequently, the calibrated model underwent validation using two additional hydrographs to ensure its suitability under various hydrologic conditions. Simultaneously, satellite precipitation data were integrated into the calibrated model. The model outputs were validated against observed data, specifically two recorded hydrographs corresponding to rain events. Finally, the accuracy of the two different sources of precipitation (rain gauge station and satellite) was evaluated by comparing their results to the observed data.

The HEC-HMS, developed as part of the Research and Development Program by the US Army Corps of Engineers (USACE), was initially introduced by the HEC in 1992. This computer application enables the simulation of rainfall–runoff-routing processes under both typical and constrained conditions. According to the USACE-HEC (2000), the HEC-HMS serves as a replacement and substitute for the Flood Hydrograph Package HEC-1 and its various versions. The initial version of the HEC-HMS, referred to as version 1.0, encompassed all the functionalities of HEC-1 with some enhancements. Subsequently, version 2.0 was released, incorporating the soil moisture accounting method, which expanded the program's capabilities from event-based to both continuous simulation and event-based simulation. The third significant release, version 3.0, introduced a new graphical interface, expanded the meteorological model to include potential evapotranspiration and snowmelt methods, and introduced new infiltration representation methods in the basin model. The most recent significant release is version 4.0, which added sediment transport and surface erosion features to the computation. At present, the latest version of the HEC-HMS is 4.2.1. This program offers various exceptional features for precipitation–runoff-routing simulation, including options for precipitation specification, loss models for runoff volume assessment, a transform model for excess rainfall transformation, hydraulic routing models for accounting storage and energy flux, a baseflow model and models for water control measures such as diversion and storage.

The rainfall recorded at rain gauge stations and rainfall estimated from satellite data were utilized as inputs for the model. The computational methods employed in the model were the SCS-CN method for converting excess precipitation into runoff and the Massickingam method for river flood routing. The station weighting method was used in the meteorological model. Precipitation-runoff simulation scenarios in the HEC-HMS model included the following scenarios.

First scenario: the model being calibrated using the rainfall data recorded at rain gauge stations for the three selected events and validation was carried out using the data from two other events by comparing the simulated floods characteristics with those recorded at the Pataveh hydrometric station.

Second scenario: the model being calibrated using rainfall estimated from the GPM satellite for the same three selected events used and then validated using the data from the other two events.

In which R² is the correlation coefficient, RMSE and NRMSE are mean squared error and normalized mean squared error, respectively. Bias indicates the skewed statistics or distortion. NSE is the Nash–Sutcliffe efficiency index. RE_QP designates relative error in peak discharge calculation and MRE stands for relative mean error index. n is the number of precipitation data, x_i and y_i are ground station and satellite precipitation data, respectively. X_m and Y_m are the total means of x_i and y_i data, Q_O is the observed discharge at the hydrometric station, Q_PO is the peak discharge observed at the hydrometric station, V_O is the observed runoff volume, T_po is the time to reach the peak flow of observation data, Q_s is the calculated flow, Q_ps is the calculated maximum flow, V_S is the calculated model volume and T_ps is the time to reach the calculated peak flow.

According to Table 3, satellite precipitation data overestimate the amount of rainfall with a relative error of 19% at the basin level. About the overall agreement between the gauge observations and satellite precipitation data, the Pearson's CC is equal to 0.59, showing a good agreement. In terms of precipitation detection capability (POD equal to 0.62), GPM precipitation data present satisfactory performance and can be used to detect rain events over the basin with acceptable level of FAR (equal to 0.33). Several studies presented similar or comparable results (e.g., Hosseini-Moghari & Tang 2020). Overall, the results of this statistical evaluation indicated that the GPM achieved a reasonable performance at the basin scale.

To evaluate the accuracy of the calibrated HEC-HMS rainfall-runoff simulation model when using satellite rainfall statistics instead of station data, three important flood characteristics, namely peak discharge (Q_P), runoff volume (V) and time to reach peak discharge (T_P) obtained from the HEC-HMS, were compared to the corresponding measurements at the Patavah hydrometry station.

Comparing the calculated hydrographs for the two types of rainfall data and the observed hydrometric hydrograph of the 03/01/2018 flood (Figure 7(a)), it is evident that when using the gauge stationary rainfall instead of the satellite data, the estimated Q_P and V are closer to those obtained from the observed hydrometric hydrograph. However, more accurate T_P estimation is achieved when using the satellite rainfall data in the model, compared to the gauge station rainfall. The onset time of the flood hydrograph using satellite rainfall data occurs earlier compared to the hydrograph based on ground station data and the observed one at the hydrometric station. The comparison between the observed and computational flood hydrograph characteristics reveals that the model has underestimated the peak flow of the flood when using both types of rainfall data. The flood runoff volume is underestimated when using ground station rainfall data and overestimated when using satellite rainfall data. The model's estimated Q_P and T_P, using both types of rainfall data, are acceptable according to the applied statistical indices (Table 4). The values of RMSE, Std Dev, bias percentage, NSE and R² indicate a well-calibrated model for both types of rainfall data. According to Santhi et al. (2001), model performance can be said to be very good if NSE (0.75–1), R² (0.75–1) and BIAS (<10%); good if NSE (0.65–0.75), R² (0.65–0.75) and BIAS (10–15%); satisfactory if NSE (0.5–0.65), R² (0.5–0.65) and BIAS (15–25%) and unsatisfactory if NSE (<0.5), R² (<0.5) and BIAS (>25%).

Comparing the simulated hydrographs for the two types of rainfall data with the observed hydrograph on 04/09/2018 (Figure 7(b)), the model demonstrates very good performance in simulating the time of peak flow discharge and total flood volume, using both rainfall data types. The observed hydrograph shows an earlier onset time compared to the simulated hydrographs for both rainfall data types. A comparison between the observed and simulated hydrograph characteristics for the two rainfall data types reveals that the model underestimates Q_P for both rainfall data types. However, the flood volumes calculated by the model are very close to the volume of the observed hydrograph for both rainfall data sources. Although the simulated volumes are slightly smaller than the actual flood volume using either of the rainfall data types, the differences however are negligible. A good agreement is found between the predicted and observed times of peak flow discharge for both rainfall data types. As shown in Table 4, no significant differences were found between the statistical indices for each of the rainfall data sources for the flood event of 04/09/2018. The calculated values of RE_Qp, RE_V and RE_Tp are 17.5, 3.19 and 0% for rain gauge station rainfall and 16.3, 4.9 and 0% for satellite rainfall data, indicating a well-calibrated model for both rainfall data sources.

Comparing the simulated hydrographs for the two types of rainfall data with the observed hydrograph on 12/08/2019 (Figure 7(c)), it is observed that using rainfall data from rain gauge stations in simulating the estimation of runoff volume and time to peak discharge yields better results than using satellite rainfall. Conversely, when using satellite rainfall in the model, the simulation of the flood peak discharge is improved. The flood start time when using ground station data occurs earlier compared to both the observed flood start time and the start time when using satellite rainfall data. Evaluating the observed and simulated flood characteristics for both rainfall data types shows that the model has estimated the peak discharge well when using satellite rainfall, with slight over-estimation that can be considered negligible. Similarly, when using rainfall from ground stations, the model performs well in the simulation of flood peak discharge, albeit with underestimation. The model performs well in estimating the flood volume under both rainfall data, with slight over-estimation, and performs better in estimating the runoff volume when using ground station rainfall. Additionally, the model performs well in simulating the time to reach the flood peak discharge when using both rainfall data types. The values of RMSE, Std Dev, bias percentage, NSE and R² indicators for the event of 12/08/2019 are 0.4, 2.26, 0.824 and 0.86, respectively, when using the rainfall of ground stations. Similarly, the values for the event on 12/08/2019 when using satellite precipitation are 0.4, 13.22, 0.803 and 0.84, respectively. These values indicate that the model is well-calibrated for both rainfall data types.

The parameters obtained from the calibration of the three flood events are shown in Table 5. The optimized parameters include the CN, initial losses (I_a), lag time (T_lag), the impact weight of the incoming flow on the amount of storage (X) and the parameter indicating the navigation time in the main river (K).

The comparison of the calculated hydrographs and observed data on the flood occurrence date (02/04/2019) is shown in Figure 8(a). According to this figure, the model performs acceptably in simulating the peak discharge and runoff volume while using both types of rainfall data. The flood start time occurs earlier when ground station data are implemented compared to satellite rainfall data. The table of observed and estimated flood characteristics produced by two types of rainfall data reveals that the model performs well in simulating the flood peak discharge regardless of input rainfall data type. However, a small over-estimation of peak discharge can be detected when using ground station data. Conversely, using satellite precipitation data resulted in slightly lower peak discharge value. Nevertheless, the model performs well in calculating the runoff volume regardless of the input rainfall data origin. Despite a minor over-estimation, the time to peak parameter is estimated more accurately when using satellite data. The values of RMSE, Std. Dev., bias percentage, NSE and R² indices are 0.6, 11.23, 0.646 and 0.67, respectively when ground stations data are implemented on the 11/15/2018 event (Table 6). The same order of indices is 0.3, −8.30, 0.919 and 0.92, respectively, when satellite data are considered on the 15/11/2018 event. These values prove that the model is well calibrated whether the ground or satellite precipitation data are used as inputs.

The comparison of the calculated and observed hydrographs simulated by the two different sources of rainfall data on the 02/25/2020 event (Figure 8(b)) reveals that ground station data provide more accurate outputs in estimating the peak flow discharge, runoff volume and the time to peak. The flood start time occurs earlier when using ground station data compared to the satellite rainfall data. A comparison between the observed and predicted flood characteristics in Table 6 for both types of rainfall data shows that despite negligible overestimations, the model was effective in simulating the peak discharge, using ground station data. The use of satellite rainfall led to an acceptable accuracy in predicting peak flow discharge with a relative error of 11.67%. Using satellite data also resulted in a properly estimated runoff volume in this event. The time to peak flow was adequately assessed by both data types; however, the ground station rainfall data resulted in a more accurate estimation compared to that obtained by satellite data, with relative errors of 10 and 20.9%, respectively. The values of RMSE, Std. Dev., bias percentage, NSE and R² indicators were 0.4, 18.64, 0.853 and 0.87, respectively, when using ground stations data on the 02/25/2020 event. The same values were 0.4, 22.83, 0.803 and 0.83, respectively, when satellite data were implemented at the same event. Such values indicate satisfactorily accurate model simulations for both satellite and ground station data.

Estimations of Q_P and V closely matched observed values when using gauge station rainfall, whereas T_P estimation showed greater accuracy when utilizing satellite rainfall data, leading to an earlier flood onset. Nevertheless, the model consistently underestimated peak flow and runoff volume, despite achieving acceptable Q_P and T_P estimations according to statistical indices. It demonstrated strong performance in simulating peak flow discharge and total flood volume for both types of rainfall data, effectively predicting the time to peak flow discharge.

According to Moriasi et al. (2007), CC between 0.8–1.0 and 0.75–0.80 indicate very high and high correlations between variables, respectively. For R² values lower than 0.6, no satisfactory correlation can be recognized between the recorded and predicted data. High R² values obtained in the calibration and validation processes indicate good and very high correlations between the model-simulated and the observed hydrograph characteristics for both rainfall datasets. As expressed by Foglia et al. (2009), the NSE index higher than 0.6 indicates the good performance of a model in simulating the flood hydrograph.

According to Table 4, for the event on 03/01/2018, flood peak discharge and volume are predicted more accurately when ground station data are considered. The estimation of the time to peak parameter in the model is more accurate when satellite data are implemented. In the event on 04/09/2018, the model performs better in runoff volume estimation when rain gauge stations data are considered. However, satellite data provide more reliable peak discharge outputs. Moreover, the model performs well in estimating the time to peak regardless of precipitation data origin. According to the simulations for the flood event on 12/08/2019, the model had a better performance in estimating the peak flow discharge when satellite rainfall data were used. However, runoff volume and time to peak parameters were estimated more accurately when ground station data were considered. Additionally, all events used in the calibration stage reveal high CCs. Based on the results obtained from the evaluation of the data in the calibration step, it was found that the HEC-HMS model performs well regardless of the rainfall data origin.

In the flood event of 02/04/2019, the model performed better in estimating peak discharge when using ground station rainfall (Table 6). However, more accurate estimations were observed for runoff volume and time to peak when satellite data were used. The R² index also indicated that the model was able to simulate the recorded flood well with satellite data. In this event, using satellite rainfall data resulted in more precise simulations of the flood hydrograph compared to using rain gauge stations. According to Table 6, in the event of 02/25/2006, better performance in calculating the peak flood discharge, estimating the runoff volume and the time to reach the peak flood discharge were observed when using rainfall from ground stations, as implied by all the indicators used. The model showed high CC compared to the conditions of using satellite precipitation. Two recorded events on 02/04/2019 and 02/25/2006 were used to validate the model. In the first event, a CC of 0.92 and NSE index of 0.919 were obtained. In the latter event, the corresponding values were 0.83 and 0.803, respectively, which can be categorized as excellent and very good according to the criteria of Moriasi et al. (2007) and Foglia et al. (2009). The results of the model validation demonstrated the high precision of the HEC-HMS model in simulating the Bashar basin floods, whether GPM satellite or ground station data were used as precipitation data.

The model accurately captured the daily time series of stream flow as well as the trend during calibration and validation periods. As indicated in Tables 3 and Table 7, ‘very good’ performance in terms of capturing the observed stream flow parameters during the calibration period and ‘good’ performance during the validation period. Both calibration and validation periods and on the standards set out by Moriasi et al. (2007), for evaluating hydrological model performance, rate the HEC-HMS model as ‘good’ for the Bashar watershed. Furthermore, a comparison of the statistics for the calibration and validation periods suggests better performance of the hydrological model during the calibration period compared to the validation period. This aligns with the findings of Moriasi et al. (2007), in which model performance during the calibration period outperforms the validation period. However, the model's performance is still acceptable during the validation period, indicating that the HEC-HMS model can be applied to study rainfall-runoff relations outside of the calibration period. Similar results were reported by Majidi & Shahedi (2012) in their study on the Abnama basin in Iran, where they utilized the HEC-HMS model for rainfall-runoff simulation during five storm events. They utilized SCS unit hydrograph and Muskingum methods for transforming excess rainfall into direct runoff and for flood routing, respectively. They found a strong correlation (0.89) between the observed and simulated discharge, confirming the model's ability for runoff simulation in the Abnana basin. Higher values of NSE and R², along with smaller bias values, were obtained when using satellite data instead of in situ rain gauges, indicating better performances of the HEC-HMS with satellite-derived precipitation in the event of 2/4/2019. However, relatively similar performances were found using the two different datasets when modeling the flood event on 2/5/2020.

In general, results differed among various flood events, showing enhanced peak discharge estimation using satellite rainfall data in one event and improved runoff volume and time to peak estimation with ground station data in another. Strong CCs during calibration indicated consistent model performance irrespective of the rainfall data source. The HEC-HMS model exhibited superior performance during calibration, consistent with prior research, while maintaining acceptable performance during validation, suggesting wider utility.

The efficiency of the numerical model in predicting flood properties, based on two types of rainfall data, was evaluated using the average relative error index. Table 8 presents the overall RE values for each parameter. The results indicate that the model provides more accurate estimates of the time to peak flow when satellite precipitation is considered. However, the model's performance in predicting peak discharge remains acceptable, regardless of the rainfall data source. Similar findings were observed for flood volume predictions.

Calibration and validation phases revealed that, irrespective of the rainfall data type (ground station/satellite), the HEC-HMS model consistently estimated hydrograph peak discharge values that were relatively smaller than those observed in the hydrometric stations. This discrepancy may be attributed to the unsatisfactory accuracy of GPM rainfalls at an hourly time scale, making it challenging to directly utilize satellite data for flood forecasting due to error amplification in rainfall-runoff calculations. To enhance the precision of rainfall monitoring and potential hydrological applications in medium and small catchments, Min et al. (2020) suggested merging satellite and gauged rainfall data. Meanwhile, the time to peak parameter was estimated with high accuracy in all events, with no significant advancement or delay in the hydrographs' time of peak flow. Additionally, the flood volume relative error analysis demonstrated that all flood events had absolute RE_v values below 10%, potentially influenced by the sub-humid characteristics. Tassew et al. (2019) reported that flood runoff is primarily controlled by infiltration excess during short but intense rainfall events, or a combination of infiltration and saturation excess during long-lasting rainfall events of varying intensities.

It is illustrated that, despite inherent errors affecting flow predictions, satellite-derived precipitation estimates show significant promise for hydrologic forecasting and water management. This potential is expected to expand with the introduction of new satellites. As a next step in this analysis, it is recommended to select diverse hydroclimatic regions across the country, encompassing varying basin sizes and rain gauge network densities, to thoroughly evaluate the potential benefits of using satellite-derived precipitation estimates for flow prediction. Further studies involving a larger number of flood events and/or multiple satellite-derived precipitation products will improve their suitability for basin-scale hydrologic applications.

When comparing rainfall data from the station with gridded products derived from GPM satellite precipitation data, it was observed that the satellite data overestimates the amount of rainfall by 19% at the basin level. The Pearson's CC revealed a strong agreement between gauge observations and satellite precipitation data. Additionally, analysis of POD values demonstrated that GPM precipitation data exhibited satisfactory performance in detecting rain events over the basin, with an acceptable level of FARs.

The aim of this study is to assess the accuracy of the HEC-HMS rainfall-runoff simulation model when using GPM satellite rainfall data instead of in situ rain gauge stations. The comparison is conducted by analyzing the flood hydrographs generated by the model using both precipitation sources and comparing them to observed data from the downstream hydrometric station, located in the Bashar watershed. The HEC-HMS model was calibrated using three major rainfall events and validated during two additional events, employing evaluative statistical indicators, RMSE Std Dev, Nash-Sutcliffe, Bias and R². The hydrograph time series and their three key parameters (Q_P, T_P and V) were the aforementioned comparisons.

The results revealed varying reliability between model simulations using satellite rainfall data and ground gauge rainfall data across different events. For instance, the accuracy of V and T_P estimated from rain gauge station data on 12/08/2019 was higher than those obtained from satellite data, while the accuracy of Q_P values on the same date showed the opposite trend. Similarly, on 02/04/2019, Q_P values estimated from ground rain gauge data exhibited higher relative accuracy compared to satellite-based estimation, whereas the use of satellite data led to smaller values of relative error in the assessment of V and T_P. The time to peak estimations were similar to the observed values in all events regardless of precipitation data type or origin.

The model demonstrated good performance in accurately capturing observed stream flow parameters during both calibration and validation phases. However, the peak discharge values modeled by the HEC-HMS were consistently lower than those observed at hydrometric stations, likely due to the unsatisfactory accuracy of GPM rainfalls at an hourly time scale. Consequently, direct utilization of satellite data presents challenges in meeting the requirements of flood forecasting, primarily due to the amplification effect of errors in rainfall-runoff calculations. To improve rainfall monitoring accuracy and enable hydrological applications in medium and small catchments, integrating satellite rainfall and ground gauge data is suggested, as it is expected to yield significant improvements. Furthermore, it is important to consider longer statistical periods for analysis to obtain more precise and robust results. Additionally, developing methods to integrate multiple sources of rainfall data is crucial for enhancing hydrological research utilizing satellite data.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.