Evaluation of the Extreme Precipitation and Runoff Flow Characteristics in a Semiarid Sub-Basin Based on Three Satellite Precipitation Products

Abstract

In this study, we analyzed the suitability of using the CHIRPS, CMORPH and TRMM platforms in monitoring extreme precipitation events, precipitation–runoff relationships, and seasonal/year-to-year variability in the Saltito semiarid sub-basin in the Mexican state of Durango. Satellite precipitation products (SPP) in 16 sites were contrasted point to point with data from rainfall gauge stations and with a daily temporal resolution for the period of four years (2015–2019). Using this information, we constructed Rx1d, Rx2d, R25mm, and RR95 extreme rainfall indices. For the precipitation–runoff relationships, a runoff model based on the Storm Water Management Model (SWMM) was calibrated and validated with gauge data, and we obtained the Qx1d, Qx2d, and Qx3d runoff indices. We used the bias volume (%), MSE, correlation coefficient, and median bias to evaluate the ability of satellite products to detect and analyze extreme precipitation and run flow events. Although these sensors tend to overestimate both precipitation levels and the occurrence of extreme precipitation events, their high spatial and temporal resolutions make them a reliable tool for the analysis of trends in climate change indices. As a result, they serve as a useful resource in evaluating the intensity of climate change in the region, particularly in terms of precipitation patterns. They also allow hydrological modeling and the observation of precipitation–runoff relationships. This is relevant in the absence of precipitation and hydrometric information, which is usually common in vast regions of the developing world.

1. Introduction

Climate change has intensified the frequency and magnitude of extreme precipitation events, particularly in semiarid regions, where water resources are already scarce and highly variable. The accurate monitoring and analysis of these events is crucial for effective water resource management, flood forecasting, and climate change adaptation [1,2]. However, in many developing regions, including the semiarid sub-basins in Mexico, the lack of dense ground-based precipitation and hydrometric networks poses significant challenges for hydrological studies [3,4]. Satellite-derived precipitation products (SPP) have become a viable solution, providing extensive global coverage and high-frequency data, making them especially beneficial for areas with limited access to ground-based measurements [5,6].

Precipitation, a key element of the Earth’s water cycle, links atmospheric and surface processes, playing a critical role in hydrology, meteorology, and climatology [7,8]. Under the influence of global warming, with the enhancement of evapotranspiration at the surface, the atmospheric moisture content in the air increases, water cycle movement intensifies, and floods, storms, snowstorms, typhoons, and other extreme precipitation events have significantly increased in frequency and intensity [9,10]. Floods are caused by extreme precipitation, droughts, snowstorms, and other extreme climate hazards and disasters, with a significant impact on society’s economic development, ecosystems, and human activities [11,12].

Precipitation is the main component for the estimation of the runoff in a region. Various authors [13,14] warn that a single weather station can lead to false predictions if it is not representative of the hydrographic surface, due to the importance of precipitation as the primary input for runoff modeling. Researchers have used radar to obtain precipitation data to perform runoff simulations and build flood forecast computational models. They have successfully reflected the hydrological situation of the study area [9,13]. However, achieving sufficient spatial coverage using radar precipitation technology in large basins is challenging, and it still has the disadvantage of spatial distribution limitations.

Satellite-based precipitation data have gained increased popularity in hydrological and climatological research because of their capacity to deliver spatially comprehensive and temporally uniform information, particularly in areas where ground-based monitoring is sparse [7,15]. Several studies have demonstrated the utility of these products in monitoring extreme precipitation events and their impacts on runoff and water resources [16,17]. CHIRPS combines satellite imagery with in situ station data to provide high-resolution precipitation estimates. It has been widely used in regions with sparse ground-based networks, such as Africa and Latin America [18,19]. Studies have shown that CHIRPS performs well in capturing seasonal precipitation patterns and extreme events, particularly in semiarid and arid regions [9,20]. For example, in East Africa, CHIRPS was found to accurately represent the spatial and temporal distribution of rainfall, making it a valuable tool for drought monitoring and early warning systems [9,21].

CMORPH uses a morphing technique to blend satellite-derived precipitation estimates, providing a high temporal resolution (30 min to 3 h) and global coverage [22,23]. Studies have emphasized CMORPH’s effectiveness in detecting the intensity and spatial patterns of extreme precipitation events. However, it has been observed to overestimate light rainfall and underestimate heavy rainfall in certain areas. In semiarid regions, CMORPH has been utilized to analyze the effects of extreme rainfall on flash flooding and soil erosion [14,24]. TRMM, although no longer operational, has been a cornerstone in satellite precipitation research. Its successor, the Global Precipitation Measurement (GPM) mission, continues this legacy [25,26]. TRMM has been extensively used to study extreme precipitation events and their relationships with runoff in various climatic regions. For example, in the Amazon Basin, TRMM data were used to analyze the impact of extreme rainfall on river discharge, demonstrating its utility in large-scale hydrological modeling [24,27]. In semiarid regions, TRMM has been used to assess the variability of monsoon rainfall and its impact on water resources [28,29].

Satellite data have the advantages of a continuous spatial distribution, large coverage area, and uniform distribution. Therefore, studying extreme climate characteristics in small local watersheds with dense ground station data can achieve good results [30,31]. However, the accuracy is poor in large watersheds with a sparse site distribution. To compensate for the lack of traditional climate stations in some basins, some researchers use spatial interpolation methods to lattice the data of ground stations in order to achieve the effect of a uniform distribution, but the accuracy of the results needs to be further verified [32,33].

Currently, research on extreme climates focuses on the simple analysis of precipitation and temperature data based on data collected at meteorological sites, while studies on re-analysis fusion data are rare. With the continuous development of satellite remote sensing, data fusion, and other advanced technologies, atmospheric data fusion products have been widely used in areas with no or scarce data [28,34]. The approach in the initial evaluation of uncorrected satellite estimates is critical to understand systematic errors before applying adjustment techniques [35]. High-quality gauge data, required for reliable bias correction, may be sparse or unavailable in a study region, and correction methods assume stationarity, which may not hold for extreme events or long-term trends.

In this study, we evaluate the suitability of three widely used satellite precipitation platforms, namely the Climate Hazards Group Infrared Precipitation with Station (CHIRPS) data [18,19], the CPC MORPHing (CMORPH) technique [14,36], and the Tropical Rainfall Measuring Mission (TRMM) [37,38], for the monitoring of extreme precipitation events, the analysis of precipitation–runoff relationships, and the assessment of seasonal and year-to-year variability in the Saltito semiarid sub-basin, located in the Mexican state of Durango. This region is characterized by low annual rainfall, high climatic variability, and frequent extreme weather events, making it an ideal case study for the evaluation of the performance of satellite products in semiarid environments [28,31].

2. Materials and Methods

2.1. Study Area

The polygon of the study area was obtained from the hydrological network available at INEGI (https://www.inegi.org.mx/temas/hidrologia) (accessed on 11 August 2024). El Saltito (saS) (Figure 1) is located in the Mexican state of Durango, in the northern part of the country. It lies between the parallels corresponding to 22°40′ and 26°50′ north latitude and between the meridians 102°25′55′′ and 107°08′50′′ west latitude about the Greenwich Meridian [39]. The study area corresponds to an approximate drained area of 11,942 km². It is within the Mezquital River basin, with the main river stream being El Rio Durango. The hydrological network begins in the Sierra de La Cacaria stream, formed by the Mimbres Stream and the Sauceda River. The hydrometric station “El Saltito” has records of the average daily discharge for this area from 1956 to 2019 (

Table 1. Rain gauge stations located within the El Saltito sub-basin.

Key	Name	State	Latitude	Longitude	Height (msnm)	Topography
10002	Canatlán/SMN	Durango	24.55	104.74	1960	Plain
10016	Chinacates	Durango	25.01	105.21	2050	Mountain
10022	El Pino	Durango	24.62	104.87	2100	Mountain
10024	El Saltito	Durango	24.03	104.35	1847	Plain
10027	Francisco I. Madero	Durango	24.4	104.32	1960	Plain
10030	Guadalupe Victoria	Durango	24.45	104.12	2000	Plain
10051	Otinapa	Durango	24.05	105.01	2400	Mountain
10066	San José de Acevedo	Durango	23.81	104.27	2100	Mountain
10076	Santiago Bayacora	Durango	23.9	104.6	2150	Mountain
10083	Tejamén	Durango	24.81	105.13	1930	Plain
10090	Canatlán	Durango	24.52	104.78	2000	Plain
10092	Durango	Durango	24.02	104.67	1900	Plain
10103	Santa Barbara	Durango	23.82	104.93	2260	Mountain
10110	Hacienda La Pila	Durango	24.12	104.29	1890	Plain
10137	Guatimape	Durango	24.81	104.92	1974	Plain
10006	Cendradillas	Durango	26.28	106.01	2270	Mountain

) [40,41].

2.2. Data

2.2.1. Satellite Precipitation Products

We utilized the fusion data of the CHIRPS, CMORPH, and TRMM satellite precipitation products to analyze the suitability of using these platforms’ SPPs in monitoring extreme precipitation events, precipitation–runoff relationships, and seasonal/year-to-year variability. The selection of TRMM (instead of IMERG) was based on its greater historical availability (1997–2015) and its well-established validation in climate studies, particularly in tropical regions [42]. The SPPs in 16 sites were contrasted point to point with data from rainfall gauge stations and with a daily temporal resolution for four years (2015–2019). In hydrological and climatic studies, the robust validation of satellite data (such as CMORPH, TRMM, and CHIRPS) requires extended periods to capture the natural climate variability, systematic errors, and extreme events. The literature suggests that at least 5 years of data are needed, according to the practical examples in [43,44,45,46,47]. We obtained an estimation of the daily precipitation values for CMORPH (global, ~8 km, daily), TRMM (tropics, ~25 km, daily), and CHHIRPS in sixteen on-site rainfall stations. For CMORPH and TRMM, we used the GEONETCast-Toolbox (Food and Agriculture Organization [FAO], 2020), and, for CHIRPS, we used the get command to connect and download them from the CHIRPS website [19]. The original format for these SPP layers was HDF4/NetCDF. We converted them from the original format, NetCDF grids, to GeoTIFF using the Argis function. Finally, we extracted the spatial coordinates of the sixteen rainfall site stations and the GeoTIFF precipitation raster using the ArcGIS tool.

Table 2. General parameters for SPPs used in this study.

Name	Spatial Coverage	Reference	Time Span	Website
CHIRPS	50° S−50° N	[19]	1981–present	https://data.chc.ucsb.edu/products/CHIRPS-2.0/
CMORPH	60° S−60° N	[14]	1998–present	Index of /precip/global_CMORPH
TRMM	35° S−35° N	[51]	1997–present	https://gpm.nasa.gov/missions/trmm/mission-end

provides a brief description of the three SPPs. CHIRPS emerged from a collaborative effort involving scientists at the USGS Earth Resources Observation and Science (EROS) Center [48,49]. It is a critical tool in providing accurate, up-to-date, and extensive datasets that aid in early warning systems, as well as evaluating trends and tracking drought conditions across seasons. The platform delivers global precipitation information at a high resolution of 0.05°, with data available at daily, monthly, and yearly intervals [17,50]. CHIRPS datasets span from 1981 to the present. Daily CHIRPS data with a spatial resolution of 0.25° were utilized for this research to evaluate its suitability for hydrological-related analyses.

The Tropical Rainfall Measuring Mission (TRMM) tropical satellite precipitation program covers the global range of 38° N~38° S [51,52]. The TRMM satellite was launched into space in 1997 in a joint mission between the National Aeronautics and Space Administration (NASA) and Japan Aerospace and Exploration Agency (JAXA) [53,54]. The TRMM satellite was the first satellite with the specific objective of monitoring tropical precipitation [55,56]. The TRMM satellite’s orbit is a polar orbit with an inclination of 35° and an altitude of 403 km (as of 2001), with a period of 92.5 min (about 16 times per day). This orbit provides detailed spatial resolutions to capture the variations in the diurnal cycle of tropical rainfall [55,57].

Onboard the TRMM satellite, different sensors for different spectral bands exist. The sensors onboard the satellite are the TRMM Microwave Imager (TMI), Precipitation Radar (PR), Visible and Infrared Radiometer System (VIRS), and TMPA [58]. All of them are converted to the TRMM Science Data and Information System (TSDI) precipitation estimation with the GPROF version of the algorithm [59,60]. In the case of rainfall AMSU estimation, the estimate is converted by the National Environmental Satellite Data and Information Service (NESDIS) with the algorithm proposed in [61].

The most recent CMORPH V1.0 release includes three distinct versions: CMORPH RAW, CMORPH CRT, and CMORPH BLD [62,63]. CMORPH RAW is a satellite-derived precipitation dataset that integrates passive microwave (PMW) estimates from various low-orbiting satellites with infrared (IR) data collected from multiple geostationary satellites [49,64]. The CMORPH CRT version is developed by calibrating CMORPH RAW against the CPC unified daily gauge analysis over land and the Global Precipitation Climatology Project (GPCP) pentad data over oceans, utilizing a probability density function technique for bias correction [50,64]. The CMORPH BLD product is further refined by incorporating optimal interpolation (OI) methods, combining additional gauge analyses to enhance the accuracy [51,65].

2.2.2. Rain and Hydrometric Gauge Data

From 2014 to 2019, daily rainfall data were collected at 16 weather stations (Table 1) and the average daily flow was obtained by the hydrometric station “El Saltito”, located in the outlet for the sub-basin El Saltito (saS). Climate data were provided by the National Water Commission (CONAGUA). The digital elevation model [66] maps were provided by the National Institute of Statistics and Geography (INEGI; https://www.inegi.org.mx/temas/edafologia/). The comparison between on-site measurements and satellite precipitation data was performed point by point, processed with the help of the free-to-use software Java (JDK 17 LTS; http://jdk.java.net/), RStudio (2023.12.0; https://github.com/rstudio/rstudio), and Python (3.10.12; https://www.python.org/).

2.3. Methods

2.3.1. Extreme Rainfall and Runoff Indices

The Expert Team on Climate Change Detection and Indices (ETCCDI), established by the World Meteorological Organization, has created a series of metrics to track extreme climate events, particularly emphasizing temperature and precipitation extremes. These metrics are essential in evaluating climate extremes [52,67]. The indices related to extreme precipitation are calculated using daily precipitation data and are designed to be simple in concept, broadly applicable, and statistically reliable. They demonstrate minimal noise and high significance, facilitating a thorough examination of the intensity, frequency, and duration of extreme temperature and precipitation events. These indices provide a robust framework for the study and interpretation of the patterns of extreme climate events [68,69].

These indicators are commonly employed to evaluate precipitation products’ (SPPs) ability to represent extreme precipitation events accurately. This study focuses specifically on intense precipitation due to its direct association with flooding, which significantly impacts water resource management, natural ecosystems, and human livelihoods [53,55]. Additionally, extreme runoff events were analyzed. This research utilized widely recognized extreme runoff indices, as referenced in prior studies. Researchers globally have shown significant interest not only in extreme meteorological events but also in extreme runoff phenomena. These include high-runoff events linked to flood risks and low-runoff events associated with drought conditions [70,71].

To study extreme runoff characteristics, methods such as the annual maximum sampling, peak-over-threshold sampling, and quantile sampling are typically employed. High-runoff events are often described using attributes like the peak flow, flood volume, and duration, while low-runoff events are characterized by the flow frequency [72,73]. The choice of sampling methods and attribute indicators depends on the specific research context. In this study, the maximum runoff over 1, 3, 5, and 7 days was selected to analyze extreme runoff patterns in the watershed [74,75].

Extreme runoff events include high runoff, which can cause flooding, and low runoff, which may lead to drought conditions. This research employed the maximum annual flow rates over 1, 3, and 5 days, referred to as Qx1d, Qx3d, and Qx5d, respectively (

Table 3. Extreme precipitation and flow rate definitions.

Category	ID	Definition	Unit
Extreme precipitation indices	RR95p	95th percentile of daily precipitation on wet days (days with daily precipitation ≥1 mm)	mm/day
	R25mm	Annual count of days when daily precipitation is ≥25 mm	days
	Rx1d	Maximum annual precipitation of 1 day	mm
	Rx2d	Annual maximum of 5 days of consecutive precipitation	mm
Extreme flow rates	Qx1d	Maximum annual flow rate of 1 day	m3/S
	Qx2d	Maximum annual flow rate of 2 days	m3/S
	Qx3d	Maximum annual flow rate of 3 days	m3/S

[76,77]).

2.3.2. Accuracy Assessment Methods

To assess the effectiveness of SPPs in modeling extreme rainfall and runoff, Table 3 [78,79] outlines these indices in detail. Statistical measures such as Pearson’s correlation coefficient (R) and the relative bias (BIAS), mean error (ME), and root mean square error (RMSE) were utilized to compare the SPPs with observations from gauges. R evaluates the linear relationship’s strength, BIAS identifies systematic deviations, and the RMSE and ME measure the average magnitude of errors [80,81]. Perfect agreement is indicated by an R value of 1, while BIAS, RMSE, and ME values of 0 signify optimal performance; see

Table 4. Understanding the performance statistics.

Statistical Index	Description
Bias volume (%)	$V_{bias}={\displaystyle \frac{\left|Q_{vo}-Q_{vs}\right|}{Q_{vo}}}\times 100$	(1)
Mean square error	$RMSE=\sqrt{{\displaystyle \frac{{\sum }_{i=1}^n{\left(Q_{O\left(i\right)}-Q_{s\left(i\right)}\right)}^2}{n}}}$	(2)
Correlation	$R^2={\displaystyle \frac{\left({\sum }_{i=1}^n\left(Q_{o\left(i\right)}-{\underset{\_}{Q}}_S\right)\right)2}{{\sum }_{i=1}^n\left(Q_{o\left(i\right)}-{\underset{\_}{Q}}_0\right)2·{\sum }_{i=1}^n\left(Q_{s\left(i\right)}-{\underset{\_}{Q}}_s\right)2}}$	(3)
Medium bias	$ME={\displaystyle \frac{{\sum }_{i=1}^n\left(Q_{O\left(i\right)}-Q_{s\left(i\right)}\right)}{n}}$	(4)
	where Qv is the volume of observation, Qvs is the volume of simulation, Qpo is the observation peak, Qps is the simulation peak, Qo is the data of observation, Qs is the simulation data, and n is the total sample number.

. These metrics were computed using the programming tool RStudio to ensure data consistency and accuracy.

2.3.3. Hydrologic Model Calibration

The Storm Water Management Model (SWMM) is a real-time computational tool designed to simulate and analyze surface hydrology and stormwater systems. It functions as a dynamic model that captures the entire rainfall–runoff process, including the response of stormwater pipe networks [82,83]. Developed in the United States during the 1970s, the SWMM provides detailed outputs for surface and pipe network conditions. For surface hydrology, it generates data on rainfall patterns, infiltration losses, evaporation rates, and the final storage of surface water [84,85]. Stormwater networks offer insights into lateral inflows, the hydraulic head, the water depth, and other critical parameters related to pipes, nodes, and discharge points [41,86].

A notable feature of the SWMM is its capability to function as a distributed, physically based hydrological model at the basin scale [87,88]. The system segments the study area into smaller sections called hydrological response units (HRUs), which are determined by factors like land use, vegetation cover, the soil characteristics, and the natural drainage network [66,89]. These HRUs enable the simulation of water balance components at the sub-basin level [90,91]. The runoff generated from these units is then combined into stream networks and directed toward the watershed outlet [82,92].

A notable extension of the SWMM is the PCSWMM, which integrates advanced functionalities for the simulation of flood-prone areas, water depths, and flood risks during extreme rainfall events; see

Table 5. Input parameters for PCSWMM model calibration.

Parameter or Attribute	Description
Area	Watershed area (hectares or acres)
Width	Characteristic width of the flow path due to superficial runoff
%Slope	Average slope of the basin in %
%Imprev	Percentage of watershed whose soil is impermeable
N-Inprev	Manning’s n coefficient for surface flow over the waterproof area of the basin
N-Prev	Manning’s N coefficient for surface flow over the permeable area of the basin
Dstore-Imperv	Storage height in depression above the impermeable area of the basin
Dstore-Prev	Storage height in depression above the permeable area of the basin
%Zero-Imprev	Percentage of impermeable soil that does not have depression storage
Percent Routed	Percentage of runoff between different areas
Infiltration	The process of precipitation that penetrates the soil surface in the unsaturated soil zone of permeable catchment sub-areas. The SWMM offers five models as an option in infiltration modeling: Horton’s classical method; Horton’s modified method; Green–Ampt method; modified Green–Ampt method; curve number method.

[93,94]. The PCSWMM also incorporates ArcGIS capabilities, providing a user-friendly interface and enhanced operational efficiency [95,96]. Additionally, it supports 2D flood analysis, offering a more comprehensive assessment of flood dynamics [97,98]. In a recent application, the PCSWMM was utilized with data from 2015 to 2019, split into a calibration phase (2015–2017) and a validation phase (2018–2019). During calibration, the necessary model parameters were fine-tuned to ensure accurate simulations. The model was deliberately calibrated with fixed parameters across all precipitation sources to enable direct an inter-PPS comparison, following established protocols for input-focused hydrological evaluations [99,100]. While this approach omits a comprehensive parameter uncertainty analysis, it ensures that any performance differences reflect precipitation product characteristics, rather than parameter tuning variations.

3. Results and Discussion

3.1. Satellite Product Comparison

For a large sub-basin such as the present study area, the systematization of the spatial–temporal distribution of precipitation is of importance, due to the spatial variability of precipitation. The Saltito sub-basin has high mountainous zones to the northwest and the southwest and valleys to rest of the area, which have different precipitation ranges [101,102]. If the other variables, such as the soil moisture and temperature, are similar within these different zones, the spatial difference in precipitation seems to be the main factor that explains the spatial variability of surface runoff [103,104]. It is possible to see correspondence among the high precipitation in the high mountainous zone and lower precipitation in the valley zones; this correspondence is observed with the SPPs and the observations of gauge rainfall stations; see Figure 2. For the 31 surface meteorological observation stations, calculations are performed based on surface observation data and their corresponding grid fusion data.

According to Figure 3, the satellite products overestimate precipitation in flat areas and underestimate it in mountainous regions. We confirm this tendency with a statistical comparison, which is included in

Table 6. Statistical comparison of TRMM, CHIRPS, and CMORPH rainfall estimates across mountain and plain stations.

Metric	Mountain Stations	Plain Stations
Correlation (Rs)
CHIRPS	0.63 **	0.75 **
CMORPH	0.58 *	0.71 *
TRMM	0.51 **	0.59 **
RMSE (mm)
CHIRPS	28	22
CMORPH	30	19
TRMM	35	33
Relative Bias (%)
CHIRPS	−29	40
CMORPH	−22	30
TRMM	−27	35
Mean Ratio
CHIRPS	0.78	1.30
CMORPH	0.68	1.40
TRMM	0.80	1.20

All correlation values represent Spearman’s rank coefficients (rₛ). Statistical significance (p-value) (* for p < 0.05, ** for p < 0.01).

. The satellite rainfall products (SSP) exhibit systematic underestimation in mountainous stations and overestimation in plains, as evidenced by the mean ratios (MR < 1 and MR > 1, respectively), and the relative biases are average negative values for mountain stations and positive for plain stations. In mountainous regions, this underestimation likely stems from the topographic obstruction of satellite sensors, leading to missed orographic precipitation [105,106], as well as cold cloud biases, where infrared-based algorithms fail to detect shallow warm rain processes that are common in highlands [107], and gauge underrepresentation in complex terrain, reducing the calibration accuracy [108,109]. Conversely, overestimation in plains may arise from anvil contamination in convective systems, where satellites misclassify non-precipitating ice clouds as rain [110,111]; surface emissivity errors over flat, homogeneous landscapes [112]; and temporal sampling gaps, as infrequent satellite passes amplify errors during short-duration storms [111,113]. These findings align with previous evaluations of SPPs in topographically complex regions. For instance, studies in the Andean Altiplano [35] and Himalayan basins (Prakash et al., 2018) [114] similarly reported underestimations of orographic precipitation by TRMM (30–40% bias) and CMORPH (20–30% bias), attributed to cold cloud algorithm limitations. However, our results contrast with those of evaluations in the Ethiopian Highlands [115], where CHIRPS showed minimal bias due to its gauge integration. This discrepancy suggests that SPPs’ performance in mountains may depend on the local calibration density and the dominant precipitation mechanism. These findings highlight the need for region-specific bias correction and hybrid gauge–satellite approaches to mitigate spatial dependencies in the error structure.

CMORPH presents the best performance at the daily scale, when there is little or close to zero precipitation, while the performance of all three is similar and better when providing precipitation estimations on a monthly basis [116,117]. It can be observed that the days without precipitation are overestimated by the satellite products [118,119]. This is also true for the 0–1 and 1–10 ranges. For the ranges of 10–25 to 25–50 and over 50, the satellite values are more in line with the ground station values, and, in the ranges of 25–50 and over 50, the satellite estimates start to underestimate with reference to the reference values; see Figure 4. The daily-scale superiority of CMORPH mirrors findings in the Mekong Basin [120], where its morphing technique reduced false alarms for light rain (<1 mm/day). However, our observed overestimation of dry days conflicts with the findings of studies in the Amazon [121], where CMORPH underestimated drizzle events. This divergence may reflect regional differences in cloud microphysics or the threshold definitions of ‘no precipitation’ (e.g., <0.1 mm vs. <0.5 mm). Notably, the improved monthly performance of all SPPs echoes global assessments [122], where temporal aggregation to 30-day scales reduced the RMSE by 50–60% across the products.

3.2. The Efficiency of SPPs in Monitoring Extreme Precipitation Events

The evaluation of the data information corresponding to 16 stations is used to assess the efficiency of SPPs in monitoring extreme precipitation events through extreme rainfall indices. The statistical measures of the extreme precipitation indices for the three SPPs, spanning 16 points, are provided in

Table 7. Error evaluation statistics of extreme climatic indices (RR95p, R25mm, Rx1d, and Rx5d), derived from satellite precipitation products in the study area.

Extreme Precipitation Index	Statistical Results	TRMM	CHIRPS	CMORPH
RR95p	Rs	0.55 *	0.67 **	0.51 *
	RMSE (mm)	150.00	120.00	160.00
	ME (mm)	12.00	3.00	13.00
	BIAS (%)	−8.00	−0.50	−9.00
R25mm	Rs	0.46 *	0.22 *	0.51 **
	RMSE (mm)	3.50	3.20	3.40
	ME (mm)	2.20	2.00	2.20
	BIAS (%)	30.00	10.00	32.00
Rx1d	Rs	0.42 *	0.25 *	0.31 *
	RMSE (mm)	18.00	16.50	18.50
	ME (mm)	12.00	11.50	12.50
	BIAS (%)	−10.00	−18.00	−0.50
Rx3d	Rs	0.39 *	0.52 *	0.54 **
	RMSE (mm)	30.00	28.00	31.00
	ME (mm)	25.00	24.00	25.50
	BIAS (%)	−30.00	−22.00	−35.00
Rx5d	Rs	0.46 *	0.53 **	0.58 *
	RMSE (mm)	29.50	27.50	30.00
	ME (mm)	25.00	24.00	25.50
	BIAS (%)	−30.00	−22.00	−35.00

All correlation values represent Spearman’s rank coefficients (Rs). Statistical significance (p-value) (* for p < 0.05, ** for p < 0.01).

. The analysis reveals that CHIRPS and CMORPH demonstrate superior performance in detecting extreme precipitation events, exhibiting lower RMSE, ME, and BIAS values, along with higher R values, for most extreme precipitation indices when compared to TRMM [123]. CHIRPS performs the best overall, with the lowest RMSE, ME, and bias for most indices (RR95p, Rx1d, Rx3d, and Rx5d). It also has competitive correlation coefficients, particularly for RR95p and Rx3d/Rx5d. CMORPH shows the strongest correlation for RR95p and Rx1d but has higher RMSE, ME, and bias values compared to CHIRPS, making it less reliable overall. TRMM performs poorly for most indices, with negative correlations for Rx1d, Rx3d, and Rx5d and higher RMSE and bias values compared to CHIRPS. The difference in performance between these precipitation products (SPPs) can be attributed to the varying data sources and calibration algorithms used in their development [19,47]. For CMORPH, the probability distribution function (PDF) matching the unified daily calibration analysis of the CPC MORPHing technique was applied to correct biases [97,98]. This correction enhances CMORPH’s performance for precipitation probability estimates, although it may still exhibit limitations in capturing extreme precipitation events, as evidenced by the higher RMSE and bias values compared to CHIRPS. On the other hand, CHIRPS benefits significantly from its integration of IR-based precipitation estimates and the combination of NASA and NOAA satellite-based precipitation data. This multi-source approach, supported by robust calibration and validation techniques [124,125], contributes to its superior performance in estimating precipitation, particularly for extreme indices [126,127].

Distribution patterns in extreme rainfall analysis play an important role. To this end, we conduct an in-depth study of the extreme events in the Saltito River basin based on fused data. From the spatial distributions for the Rx1d, Rx2d, Rx25mm, and RR95p indices for the years 2015–2019, calculated from satellite observations (TRMM, CHIRPS, CMORPH) and ground-based observations (CNA), several conclusions can be drawn with respect to the topography and time series (Figure 5).

The topography (flat vs. mountainous area) has a significant impact on the accuracy of satellite precipitation estimates compared to ground-based observations (CNA). In flat areas, we observe overestimation for the satellite products (TRMM, CHIRPS, CMORPH), which tend to overestimate precipitation compared to ground-based observations (CNA). This is evident at stations such as Tejamen and Hacienda la Pila, where the values of Rx1d, Rx2d, and RR95p are consistently higher in the satellites than in CNA. This could be attributed to the fact that, in flat areas, satellites may have difficulties in distinguishing between the precipitation signal and surface reflectivity, leading to overestimation. For mountainous areas, there is underestimation and more variability than in flat areas. In some stations, such as Santa Barbara and Santiago Bayacora, the satellites underestimate precipitation compared to CNA.

At others, such as El Pino, the values are closer to the ground observations. This may be attributed to the topographic complexity in mountainous areas, which can distort satellite measurements, especially due to effects such as mountain shadowing or snow reflectivity.

The satellite products (TRMM, CHIRPS, CMORPH) show differences in their performance. TRMM tends to overestimate precipitation at most stations, especially in flat areas, and presents the worst performance in the area for the monitoring of extreme precipitation events. CHIRPS show mixed behavior, with overestimation in some stations and underestimation in others. In general, it performs better in mountainous areas than TRMM. CMORPH tends to underestimate precipitation at some stations, especially in mountainous areas. In flat areas, its performance is closer to the ground observations, although still with some overestimation, and CMORPH presents the best performance in general.

3.3. The Performance of SPPs in Extreme Flow Capture

3.3.1. Validation of the PCSWMM Model

With the observations measured by the runoff stations as input, we obtain a hydrological simulation for the calibration (2015–2018) and validation (2018–2019) periods. Figure 5 shows the statistical metrics provided by hydrological model; they indicate good performance in describing runoff (streamflow) across all rainfall data sources (gauge precipitation, TRMM, CHIRPS, and CMORPH). Gauge precipitation performs better than the satellite products; this is expected, as ground-based data are typically more accurate than satellite-based data. All rainfall data sources achieve Nash–Sutcliffe efficiency (NSE) [7] values above 0.6, with those of gauge precipitation and CMORPH exceeding 0.7, indicating that the model acceptably simulates runoff [128,129]. All rainfall data sources achieve R² values above 0.6, with those of gauge precipitation and CMORPH exceeding 0.7, indicating that the model explains a significant portion of the variability in the observed streamflow. All rainfall data sources have BIAS values below 16%, indicating that the model does not significantly overestimate or underestimate streamflow. The slight overestimation (positive BIAS) is consistent across all datasets. The CMORPH, CHIRPS, TRMM, and gauge values show good correspondence with the runoff values in terms of trends and variability [130,131]. This suggests that the models adequately capture the runoff dynamics over time (Figure 6). However, there are times when the modeled values differ significantly from the reference values, which could indicate errors in the models or specific events that are not being captured correctly. The data show clear annual seasonality; it could be noted that the area is characterized by strong mid- to late-summer precipitation maxima in June, July, August, and September, with considerably less precipitation during the rest of the year, which is typical in time series related to hydrology in such arid zones [132,133]. Runoff tends to be higher in certain months of the year (e.g., during the rainy season) and lower in others (during the dry season). This seasonality can be observed in the runoff values, which show recurring peaks and valleys throughout the years.

In addition to seasonality, it is possible that there are long-term trends or interannual patterns that could be related to climatic phenomena such as El Niño or La Niña. It would be interesting to perform a more detailed analysis to identify these trends. There are also some outlier peaks in the data, which could be related to extreme events such as floods or droughts.

3.3.2. Extreme Flow Assessment of the Three SPPs

To assess the ability of the four precipitation products (SPPs) to capture extreme flow events, three extreme flow indices—Qx1d, Qx3d, and Qx5d—were chosen for evaluation (see

Table 8. Error evaluation statistics of extreme runoff indices (Qx1d, Qx3d, Qx5d), using various rainfall inputs (CONAGUA, TRMM, CHIRPS, and CMORPH), generated by the PCSWMM software.

Extreme Runoff Index	Statistical Results	TRMM	CHIRPS	CMORPH	Precipitation Gauge (CONAGUA)
Qx1d	R	0.40 **	0.45 **	0.50 **	0.95 **
	RMSE (mm)	174.80	153.74	185.78
	ME (mm)	17.25	5.38	17.35
	BIAS (%)	−11.40	−1.54	−12.40	3.0
Qx3d	R	0.35 *	0.14 *	0.35 **	0.94 **
	RMSE (mm)	4.10	4.088	4.089
	ME (mm)	2.83	2.85	2.83
	BIAS (%)	35.86	14.36	36.16	3.5
Qx5d	R	−0.183 *	0.006 *	0.123 *	0.93 *
	RMSE (mm)	21.15	20.81	21.17
	ME (mm)	15.34	15.14	15.42
	BIAS (%)	−15.48	−23.88	−2.50	3.2

All correlation values represent Pearson’s rank coefficients (R). Statistical significance (p-value) (* for p < 0.05, ** for p < 0.01).

). The bias volume (%) is the lowest (+3.0% to +3.5%), indicating that CONAGUA’s estimates are closest to the observed values. CHIRPS performs better than CMORPH and TRMM but not as well as CONAGUA. The bias volume (%) ranges from +7.5% to +8.5%, indicating consistent but acceptable overestimation. TRMM’s performance is very close to that of CHIRPS, with a slightly higher bias volume (%) and MSE. CMORPH has the worst performance, with the highest bias volume (%), MSE, and medium bias among all sources. It was mentioned earlier that there is generally little or no precipitation in the area, and the SPPs tend to overestimate these measurements, similarly to the performance of SPPs in modeling streamflow, as noted above. These results are only observed for arid zones; a tendency to underestimate the observed averages has been reported in other studies performed in arid zones [134,135].

4. Conclusions

The raw inter-comparison of satellite products (SPPs) allows us to identify intrinsic biases. Satellite products (SPPs) tend to overestimate precipitation in flat areas, while, in mountainous, areas their performance is more variable, with cases of underestimation. The topographic complexity in mountainous areas affects the accuracy of satellite estimates. The CMORPH and CHIRPS sensors generally perform better than TRMM for the study area (semiarid zone).

Although these sensors tend to overestimate precipitation levels, given their spatial and temporal resolutions, they represent a good alternative for the analysis of spatial–temporal patterns of precipitation in this zone, with moderate correlation agreement. However, for the analysis of climate change precipitation indices to evaluate the intensity of climate change in the region, their performance changes to weak agreement. This performance can be improved by creating a correction algorithm to increase their correspondence, which can be considered in a future study.

The precipitation index values show interannual variability but no clear trend of an increase or decrease during the period of 2015–2019. Extreme precipitation events (Rx1d, Rx2d) and days with intense precipitation (Rx25mm) are relatively rare and do not follow a defined pattern.

We found a discrepancy between the good performance of raw SPP rainfall estimations and in modeling runoff and its lower accuracy in estimating extreme precipitation indices. This can be attributed to several factors related to the characteristics of the satellite data, the nature of extreme events, and the ways in which hydrological models process information. The SPPs performed well in describing runoff (streamflow). Without precipitation and hydrometric information, which is usually very scarce in most semiarid areas, they can be used as a source of analysis for these hydrological processes.