Hydrologic modeling in mountainous terrains with monsoonal climate: Do gridded precipitation data improve streamflow simulations?

Abstract

Accurate hydrologic modeling in data-scarce mountainous regions remains challenging due to strong climatic variability, complex topography, and limited ground data. This study evaluates the influence of four precipitation inputs on the performance of the Soil and Water Assessment Tool Plus (SWAT+) model in the monsoon-dominated Humaya River basin, located in the Sierra Madre Occidental, northwestern Mexico. The evaluated inputs include: (1) rain gauge data (Gauge), (2) Daymet gridded data, and two bias-corrected Daymet products using (3) quantile mapping (Qmap) and (4) mean field bias correction (MFBC). Model calibration and validation were conducted independently for each input using monthly streamflow data (1983–2001). Model performance was assessed using NSE, PBias, and complementary diagnostics, including KGE, logNSE, regime-based PBias, and flow-duration curve metrics, enabling a more comprehensive evaluation of both high- and low-flow conditions. An ensemble-based approach was implemented by retaining the 100 best-performing simulations per configuration to account for parameter uncertainty and equifinality, allowing hydrologic fluxes to be analyzed as probability distributions. Spatial analyses, including a quantitative assessment using Moran's I, were conducted to evaluate the spatial representation of hydrologic processes under each forcing. Overall, SWAT+ reproduced basin streamflow dynamics with good to very good performance. The Daymet-MFBC configuration achieved the best overall results (NSE = 0.81; PBias = 6.4%; KGE = 0.85), followed by Daymet and Gauge, whereas Daymet-Qmap showed comparatively lower performance. Daymet and Daymet-MFBC also produced more heterogeneous spatial patterns of hydrologic processes, supported by lower Moran's I values. However, all configurations showed systematic deficiencies in low-flow simulation, as indicated by high PBias in low-flow conditions and negative logNSE values (−0.98 to −0.77), reflecting a known limitation of SWAT+ in representing baseflow dynamics. The ensemble analysis indicates that precipitation forcing influences not only model performance but also constrains parameter uncertainty and the range of hydrologically plausible internal process representations. These findings highlight the value of high-resolution gridded precipitation datasets for improving both the accuracy and spatial realism of hydrologic modeling in mountainous and data-scarce regions.

1 Introduction

Hydrologic modeling is a fundamental tool for understanding and assessing water resources in diverse and dynamic environments (Baran-Gurgul and Rutkowska, 2024; Nath et al., 2024). Accurate simulations of the hydrologic cycle are essential for addressing current challenges such as water scarcity, flood risk, climate change adaptation, and ecosystem sustainability. However, achieving reliable model performance in mountainous regions remains particularly challenging due to the strong spatial variability of climatic and physiographic conditions (Gochis et al., 2006; Mascaro et al., 2014) and the scarcity of ground-based observations (de la Fraga et al., 2024; Khandu et al., 2016; Khatakho et al., 2021; Robles-Morua et al., 2012). These challenges are particularly pronounced in northwestern Mexico, where water scarcity, increasing agricultural demand, dependence on summer monsoon rainfall, and ecosystem fragility [Comisión Nacional del Agua (CONAGUA), 2024a; Loaiza et al., 2021; Nesbitt et al., 2008; Robles-Morua et al., 2015] require robust hydrologic assessments to support adaptive water management.

A key factor in hydrologic modeling is the accurate representation of precipitation, which serves as the primary driver of hydrologic processes (Beck et al., 2017; Gebrechorkos et al., 2024; Tapiador et al., 2012). Traditional rain gauge networks, though widely used, often fail to capture the spatial and temporal variability of rainfall because of their sparse and uneven distribution, especially in mountainous terrain, and also due to the localized nature of precipitation events (Kidd et al., 2012, 2017). To overcome these limitations, high-resolution gridded precipitation products such as the Daily Surface Weather and Climatological Summaries (Daymet; Thornton et al., 2021) have gained popularity for their spatial coverage. Yet, these datasets can exhibit systematic biases and uncertainties (de la Fraga et al., 2024; Muche et al., 2020), raising concerns about their reliability for hydrologic modeling. Consequently, selecting and evaluating precipitation inputs remains a critical step in achieving accurate and physically consistent simulations.

The Soil and Water Assessment Tool Plus (SWAT+) (Bieger et al., 2017) is a widely used semi-distributed model for simulating hydrologic processes in heterogeneous landscapes. Nonetheless, its performance in data-scarce mountainous regions remains underexplored. In particular, the interplay between precipitation input uncertainty, parameter uncertainty, and the representation of dominant hydrologic processes has received limited attention (Sánchez-Gómez et al., 2024; Yimer et al., 2023). Moreover, previous studies have predominantly focused on lowland or temperate environments (e.g., Barresi-Armoa et al., 2023; Tapas et al., 2025), leaving a gap in understanding SWAT+ performance in subhumid mountainous regions characterized by complex topography and intricate precipitation patterns influenced by orographic effects and convective systems, conditions that amplify hydrologic complexity.

Another key research gap concerns the usefulness of spatially distributed precipitation inputs not only for streamflow simulations (e.g., Elsner et al., 2014; Gao et al., 2017; Khatakho et al., 2021; Mei et al., 2022; Ray et al., 2022; Vu et al., 2012; Zhang et al., 2021) but also for the representation of hydrologic processes such as evapotranspiration, surface runoff, lateral flow, and groundwater flow, an aspect that has received limited attention (e.g., Maina et al., 2020; Muche et al., 2020; Shuai et al., 2022). This issue is especially relevant in regions such as the Sierra Madre Occidental in northwestern Mexico, where rain gauge networks are sparse and unevenly distributed across steep terrain, and where precipitation is strongly influenced by convective and orographic processes that vary considerably over short distances. Under these conditions, the performance and reliability of gridded precipitation products remain uncertain. To address this, we evaluate the utility of Daymet V4 (Thornton et al., 2021), the latest version of this dataset, whose application in hydrologic modeling is still emerging (e.g., Ghimire et al., 2025; Shuai et al., 2022).

Our study addresses these research gaps by evaluating the performance of the SWAT+ model in the Humaya River basin, a watershed characterized by steep elevation gradients, fractured igneous geology, and monsoonal precipitation regimes. Specifically, we investigate the impact of four precipitation inputs on model performance: (1) rain gauge data (hereafter Gauge), (2) Daymet gridded data (hereafter Daymet), (3) Daymet bias-corrected via quantile mapping (hereafter Daymet-Qmap), and (4) Daymet bias-corrected using mean field bias correction (hereafter Daymet-MFBC). This work advances our previous evaluation of several gridded precipitation datasets in northwestern Mexico (de la Fraga et al., 2024), which identified Daymet as the most accurate product for daily precipitation estimates. While that work focused on climatic accuracy, the present research provides a comprehensive hydrologic validation by integrating Daymet and its bias-corrected versions into the SWAT+ model. Our approach introduces several novel aspects to the assessment of precipitation inputs in complex terrains: (1) the use of a probabilistic ensemble approach (100 best simulations per forcing) to explicitly quantify parameter uncertainty and equifinality, revealing how precipitation data condition the range of plausible hydrologic mechanisms; (2) a qualitative and quantitative spatial analysis of key processes (precipitation, evapotranspiration, surface runoff, and groundwater flow) to evaluate how input resolution influences the representation of internal basin heterogeneity, a critical but underexplored aspect in mountainous SWAT+ applications; and (3) a direct comparison of two distinct bias-correction methods (quantile mapping and mean field bias correction) for gridded data, assessing their impact not only on streamflow accuracy but also on parametric uncertainty and spatial realism. By moving beyond a point comparison of precipitation to a holistic evaluation of hydrologic performance, internal process representation, and uncertainty, this study addresses key gaps in understanding the value of high-resolution gridded data for watershed modeling in data-scarce, monsoonal mountain regions.

We hypothesize that high-resolution gridded precipitation data can enhance hydrologic simulations relative to gauge-based forcings. To test this hypothesis, we address the following research questions: (i) How well does SWAT+ simulate observed streamflow in a monsoon-influenced mountainous basin with limited ground-based data? (ii) Do gridded precipitation inputs (Daymet, Daymet-Qmap, and Daymet-MFBC) improve streamflow simulation compared to gauge-based forcing? (iii) What are the implications of precipitation input selection on the simulated water balance components? and (iv) What differences arise in the spatial distribution of simulated hydrologic processes under each precipitation input? Through these analyses, we aim to advance the understanding of how spatially distributed precipitation data influence hydrologic model performance and to provide practical guidance for improving model outputs in data-scarce, complex terrains.

2 Materials and methods

2.1 Study area

The Humaya River basin (8,261 km²) is located in northwestern Mexico and spans portions of the states of Chihuahua, Sinaloa, and Durango. It originates in the Sierra Madre Occidental and drains westward into the Culiacán River (Figure 1). Elevations range from 216 to 3,156 m a.s.l., and the watershed is characterized by steep slopes, deep canyons, high plateaus, and highly weathered, fractured and faulted igneous rocks (ignimbrite, rhyolite, andesite, basalt, and granodiorite) [Nesbitt et al., 2008; Servicio Geológico Mexicano (SGM), 1999]. Climatically, the basin area transitions from warm subhumid lowlands (mean annual temperature 22 °C) to semi-cold subhumid highlands (10 °C) [CLICOM, 2015; García, E., and Comisión Nacional para el Conocimiento y Uso de la Biodiversidad (CONABIO) 1998], with an average annual precipitation of 958 mm, primarily concentrated during the June–September monsoon season (50%−70% of total) (de la Fraga et al., 2024; Forzieri et al., 2013; Gebremichael et al., 2007).

Figure 1

The basin outflow supplies the Adolfo López Mateos Dam, an important reservoir with a storage capacity of 4,034.5 hm3 [Comisión Nacional del Agua (CONAGUA), 2024b], which is vital for regional socio-economic activities, including irrigation for 60,000 ha of agricultural land, hydropower generation, and ecosystem services (Loaiza et al., 2021; Moreno-Leon and Ortega-Rubio, 2009; Quevedo-Castro et al., 2018). Downstream, the Culiacán Valley ranks as one of Mexico's most productive agricultural regions [Comisión Nacional del Agua (CONAGUA), 2024a]. However, reduced rainfall and intensive agricultural activities have exacerbated water scarcity in recent years (Loaiza et al., 2021), increasing pressure on local water resources [Comisión Nacional del Agua (CONAGUA), 2024a] and highlighting the need for reliable hydrologic assessments to inform sustainable water resource planning.

2.2 Hydrologic model: SWAT+

SWAT+ is a completely revised version of the original SWAT model (Arnold et al., 1998; Bieger et al., 2017). It is a small-watershed to river-basin scale model used to simulate the quality and quantity of surface and groundwater and to predict the environmental impact of land use, land management practices, and climate change [U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research, 2025a]. Key improvements include a refined input file structure and a more flexible spatial representation of interactions and processes within a watershed (Bieger et al., 2017). Operating on a daily time step, the model integrates data on topography, land use/land cover, soil properties, and climate to compute water fluxes using the following water fluxes using the water balance equation presented in Equation 1 (Neitsch et al., 2011):

where SW_t and SW₀ are the final and initial soil water content (mm), and i denotes the daily time index within the simulation period. P_i is precipitation (mm), Q_{surf, i} is surface runoff (mm), ET_i is evapotranspiration (mm), w_{seep, i} is water percolation through the soil profile (mm), and Q_{gw, i} is return flow (mm).

2.3 Input data and model setup

To ensure full reproducibility of this work, all input datasets, data processing scripts, model setup files, SWAT+ input/output files, and statistical analysis scripts are documented and provided in a public repository (de la Fraga et al., 2026; see Data availability statement). This includes the complete SWAT+ configuration files, forcing datasets, and post-processing routines required to replicate the methodology and the results presented in this study.

2.3.1 Spatial datasets preparation

Three main spatial datasets were prepared as model inputs: topography, land use/land cover, and soils (Table 1).

Topographic information was derived from NASA's Shuttle Radar Topography Mission (SRTM) digital elevation model (DEM) at 30 m resolution (Farr et al., 2007), using the hole-filled seamless SRTM data V4 (Jarvis et al., 2008).

Land use and land cover data were derived from vector information published by Instituto Nacional de Estadística y Geografía (INEGI) (2021) at 1:250,000 scale (Series VII) and reclassified into SWAT+ categories to ensure compatibility with the model. Pine forests accounted for 62% of the basin, oak forests for 14%, and minor proportions included range-brush (9%), deciduous forests (8%), agriculture (6%), pasture (1.5%), and urban areas (0.02%) (Figure 2).

Figure 2

Soil data were obtained from Instituto Nacional de Estadística y Geografía (INEGI) (2024) vector maps at 1:250,000 scale (Series II) and complemented with 17 field soil profiles (Figure 3) from Instituto Nacional de Estadística y Geografía (INEGI) (2013), which included information on depth, texture, sand, silt, and clay content, organic carbon content, electrical conductivity, and other physical and chemical properties. Additional soil parameters required for SWAT+ modeling (hydrologic group, bulk density, available water capacity, saturated hydraulic conductivity, albedo, and erodibility factor) were derived using established tools and equations:

Figure 3

• Soil hydrologic group: Classified according to U.S. Department of Agriculture (USDA)–Natural Resources Conservation Service (NRCS) (2009) criteria.

• Bulk density, available water capacity, and saturated hydraulic conductivity: Calculated based on texture (% sand/clay), organic matter, gravel, and electrical conductivity using the Soil Water Characteristics module of the Soil-Plant-Air-Water (SPAW) software (version 6.02.75), based on Saxton and Rawls (2006).

• Organic matter (OM) content: Estimated from organic carbon (orgC) content using the van Bemmelen conversion factor of 1.724 (Jackson, 1964), as shown in Equation 2:

• Soil albedo: Estimated using Baumer's (1990) equation, as shown in Equation 3:

• USLE soil erodibility (K_USLE): Calculated using (Williams 1995) equation, as shown in Equation 4:

where the factors in Equation 4 are defined according to Equations 5–8:

where m_s, m_silt, and m_c are the percentages of sand, silt, and clay, respectively.

Table 2 summarizes the soil parameters required by SWAT+. To incorporate these data into the model, we created a usersoil.csv file with the structure and format specified by SWAT+.

The most representative soil types in the basin are Regosols (33%), Luvisols (20%), and Leptosols (18%), with minor proportions of Cambisols (12%), Phaeozems (11%), Umbrisols (6%), and Vertisols (0.4%) (Figure 3).

2.3.2 Climate data configurations

2.3.2.1 Climatological station data

Four climate input configurations were used for the period 1980–2001, consistent with the availability of observed streamflow data. The first configuration relied on gauge-based data. Daily observed precipitation and maximum/minimum temperature were obtained from six climatological stations operated by the Mexican National Meteorological Service [Servicio Meteorológico Nacional (SMN) (2026)]. Following an exploratory analysis, we implemented a quality control protocol based on five main criteria: (i) station location, (ii) period of record and data continuity, (iii) missing data, (iv) detection of anomalous and erroneous data, and (v) time series homogeneity.

Location. We initially examined data from all climatological stations within or near the basin (40 stations in total). Many were no longer operational, some had been active for only a few years, and others contained extensive periods with no data. Based on the subsequent quality control steps, six stations were ultimately selected.

Period of record. Priority was given to maximizing the length of continuous data coverage. The longest continuous period identified spanned from 1969 to 2017. However, because Daymet data are available from 1980 onward and observed streamflow records are continuous only for 1980–2001, the common period 1980–2001 was adopted for all analyses.

Missing data. A threshold of 25% missing data for the entire study period was considered, as recommended by several authors (Campozano et al., 2014; Dong and Peng, 2013; Morales-Martínez et al., 2019; Navarro-Céspedes et al., 2023; Pérez-Morga et al., 2013). Among the six selected stations, only two exhibited 18% and 23% of missing daily data; the remaining stations had substantially lower missing data rates (Supplementary Table S1).

Anomalous and erroneous data. We conducted an additional exploratory analysis to identify anomalous values, obvious data capture errors, duplicate entries, and other inconsistencies. Suspect values were cross-checked with neighboring stations and with documented events (e.g., tropical cyclones, extreme precipitation events). Outliers were rejected only after manual inspection.

Homogenization and missing data estimation. Homogenization and gap-filling were performed using the Climatol package in R (version 3.1.2; Guijarro, 2018). Climatol employs an adaptation of the (Paulhus and Kohler 1952) method, estimating data as a weighted average of the closest normalized series. Normalization is achieved by scaling each series by its respective mean. These estimates are then used to compute anomalies for outlier detection and homogeneity testing, and to fill missing values. The algorithm organizes stations in a tree structure, grouping clusters with high correlations into the same branch (Kessabi et al., 2022). Outlier correction is automated, and predefined criteria monitor the operation's quality by detecting relative in homogeneities using the Standard Normal Homogeneity Test (SNHT; Alexandersson, 1986). Homogeneity was assessed on monthly aggregated data, as daily series exhibit high variability that reduces the sensitivity of inhomogeneity detection. The quality of the homogenization was evaluated using the root mean square error (RMSE) between observed and estimated data.

After applying this quality control protocol, six stations were retained; their spatial distribution is shown in Figure 1; their geographic coordinates, elevation, record length, and percentage of missing data are provided in Supplementary Table S1. The exact cleaned time series used in this study are available in the public repository. Due to the complex terrain and difficult accessibility, these stations are sparse, unevenly distributed, located far from each other, and situated predominantly near the basin perimeter; consequently, their spatial representativeness is limited. These limitations highlight the need for reliable alternative precipitation datasets to improve the understanding of water-related aspects of the whole basin, which is one of the motivations of this study.

2.3.2.2 Daymet precipitation data

The second configuration employed gridded precipitation data from the Daymet V4 dataset (Thornton et al., 2021). Data were downloaded using an automated Python script (Thornton and Devarakonda, 2011) and extracted at 358 grid points spaced at 5 km intervals across the basin (Figure 1). Although Daymet has a native resolution of 1 km, a coarser 5 km resolution was chosen to reduce computational requirements while preserving spatial representation.

2.3.2.3 Bias-corrected Daymet precipitation (Qmap)

The third configuration applied a statistical bias correction to the Daymet precipitation data using the qmap package in R (version 1.0-6; Gudmundsson, 2025). We implemented a non-parametric empirical quantile mapping approach (Gudmundsson et al., 2012). For each Daymet grid cell, the empirical cumulative distribution function (CDF) of the raw Daymet time series was adjusted to match the CDF of the nearest rain gauge station. This was performed using the fitQmapQUANT function, which estimates the empirical CDFs of both observed and modeled time series for regularly spaced quantiles. We set the quantile step (qstep) to 0.001 and the number of bootstrap samples (nboot) to 1 (i.e., the estimation was based on all available data without resampling, see Gudmundsson, 2025 for further details). A wet-day threshold of 0.1 mm was applied to restrict the CDF fitting to the portion of the distributions corresponding to observed wet days (Piani et al., 2010). The quantile mapping was then executed using the doQmapQUANT function with tricubic interpolation (monotonic tricubic spline interpolation via splinefun), applying the transformation previously fitted with fitQmapQUANT. Quantile mapping is a widely applied method (Enayati et al., 2021; Gudmundsson et al., 2012; Gutjahr and Heinemann, 2013; Lafon et al., 2012; Li et al., 2010; Nguyen et al., 2024; Teng et al., 2015) and was selected for its ability to reduce distributional biases; specifically, a non-parametric transformation was implemented due to its superior skill in reducing biases compared to other methods (Gudmundsson et al., 2012). The complete implementation, including input data, parameter settings, and execution script in R, is provided in the repository.

2.3.2.4 Bias-corrected Daymet precipitation (MFBC)

A fourth configuration was implemented using the mean field bias correction (MFBC) method (Yoo et al., 2014). In this case, Daymet precipitation data were adjusted by applying a spatially uniform, multiplicative correction factor on a daily basis. To derive the correction factor for a given day, we considered co-located pairs of Daymet pixels and the six rain gauge stations. Only days with at least two available gauge–Daymet pairs simultaneously recording positive precipitation values were corrected. This procedure maximized the number of corrected days while effectively reducing bias between Daymet and gauge observations; a total of 2,193 days (27% of the 8,036-day simulation period) met this criterion. For each such day, the correction factor was calculated as the slope (β) of a linear regression line forced through the origin, with Daymet precipitation as the independent variable and gauge precipitation as the dependent variable, as expressed in Equation 9: where n is the number of positive gauge–Daymet pairs in a given day, x_i is the Daymet precipitation value, and y_i is the corresponding gauge precipitation value. This single daily factor was then applied uniformly to all Daymet grid cells. The MFBC method was chosen for its computational efficiency and its key property of preserving the original spatiotemporal relative structure (i.e., the spatial patterns and event timing) of the Daymet dataset while correcting its mean field bias. The full implementation of the MFBC method, including the daily pairing of gauge–Daymet data and regression-based factor computation, is available in the repository as a reproducible R script.

2.3.2.5 Temperature and additional meteorological variables

For all configurations, maximum and minimum temperature inputs were taken from the gauge dataset, ensuring uniformity across experiments. Additional meteorological variables required by SWAT+ (solar radiation, relative humidity, and wind speed) were generated using the built-in weather generator in SWAT+ Editor.

2.3.3 Streamflow data

Monthly observed streamflow data used for model calibration and validation were obtained from the Guatenipa II hydrometric station, provided by the Mexico's National Surface Water Database [Comisión Nacional del Agua (CONAGUA), 2016]. This station is located at the basin outlet and provides continuous records for the 1983–2001 period, matching the temporal coverage of the climate datasets. The original dataset consists of daily mean discharge values reported in m³/s. Monthly streamflow values were computed as the arithmetic mean of all daily discharge measurements within each month.

2.3.4 Model setup

The model setup was conducted in QGIS version 3.28.15 using the QSWAT+ interface (version 2.4.7, Dile et al., 2023) and SWAT+ Editor [version 2.3.3, U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research, 2026b]. The DEM was used within QSWAT+ to delineate the basin boundaries, define the stream network, and generate the subbasins (see Dile et al., 2023 for a detailed description of the QSWAT+ interface and step-by-step configuration). We used the burn in existing stream network option to incorporate the INEGI stream network. The threshold sizes for creating channels and streams were the default values initially calculated by the algorithm (18 and 180 km², respectively); the area threshold defines the minimum drainage area required to initiate a channel or a stream in the model.

The basin was discretized into 19 subbasins and 249 landscape units (LSUs), which are finer spatial divisions within a subbasin where the hydrologic response units (HRUs) outputs are aggregated at the LSU level (Bieger et al., 2017). HRUs are lumped land areas within a subbasin defined by a unique combination of land use, soil type, and slope, for simulating hydrologic processes occurring in the landscape (Bieger et al., 2017). To create the HRUs, the DEM, land use, and soil maps were input in raster format. Three slope classes (< 8%, 8%−30%, >30%) (Figure 4) were defined based on Food and Agriculture Organization of the United Nations (FAO) (2003) guidelines and similar studies (Molina-Navarro et al., 2016; Montecelos-Zamora et al., 2018; Sánchez-Gómez et al., 2022, 2024).

Figure 4

Additionally, lookup tables were required to convert the numeric values in the land use and soil maps to SWAT+ land use codes and soil names, respectively. These data were prepared in advance and imported as comma-separated values (.csv) files. Lookup tables for land use must contain the string “landuse” in their names; lookup tables for soils must contain the string “soil,” otherwise they will not appear in the corresponding pull-down menus in the interface. Because our project is not located in the U.S. and we did not use SSURGO or STATSGO databases, a custom soil properties file was required. This traditionally has a name “usersoil,” or one including that name. After deriving the soil parameters described in Section 2.3.1, we created a usersoil.csv file with the structure and format described in the SWAT+ documentation. This file, along with lookup tables for land use and soils, is provided in the repository.

Initially, the model included 5,071 HRUs. To reduce computational demands, the number of HRUs was reduced in each subbasin by eliminating land uses covering less than 10% of the subbasin area, soil types covering less than 10% of the area of the remaining land uses, and slope classes covering less than 10% of the area of each remaining land use and soil type combination. As a final result, a model with 1,636 HRUs was obtained. This simplification enhanced computational efficiency while ensuring that representativeness was maintained without loss of relevant information.

Within SWAT+ Editor, additional adjustments were made, climate data were entered, and the simulation period was defined. For every configuration, the daily precipitation and maximum/minimum temperature records were formatted for integration into SWAT+ following standard protocols (see the Climate section; U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research, 2025b, 2026a). We created individual station files (.txt) for precipitation and temperature, each containing the daily time series. These station files were then listed in the corresponding master files (pcp.txt for precipitation and tmp.txt for temperature) which serve as indices specifying the file names and locations of all available stations. All these files were imported through the weather stations import data function; the editor automatically assigned to each subbasin and HRU the weather station closest to its centroid (U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research, 2025b, 2026a).

Potential evapotranspiration (PET) was estimated using the Penman–Monteith method (Monteith, 1965), surface runoff was simulated with the SCS curve number method [U.S. Department of Agriculture (USDA)–Soil Conservation Service (SCS) (1972)], and channel routing was performed using the Muskingum method (Strupczewski and Kundzewicz, 1980). The simulation period (1980–2001) was divided into three phases: (1) a 3-year warm-up period (1980–1982), (2) a 10-year calibration period (1983–1992), and (3) a 9-year validation period (1993–2001). The warm-up period was not included in performance evaluations. All simulations were executed at a daily time step, with outputs aggregated to monthly intervals for calibration/validation and model evaluation.

2.4 Sensitivity analysis, calibration, and validation

2.4.1 Sensitivity analysis

Sensitivity analysis quantifies the influence of input parameter variations on model outputs, helping to identify the parameters most relevant for calibration (Arnold et al., 2012). This process guides model calibration and validation and informs how to prioritize efforts to reduce uncertainties (Norton, 2015; Pianosi et al., 2016). A global sensitivity analysis was conducted using the Fourier Amplitude Sensitivity Test (FAST) (Cukier et al., 1973) integrated in SWAT+ Toolbox (version 1.0.5; Chawanda, 2021). Twenty parameters commonly calibrated in SWAT studies (Arnold et al., 2012; Mehdi et al., 2018; Meins, 2013; Schürz et al., 2019) were tested within their standard ranges (Supplementary Table S2). For a detailed description of these parameters, see (Neitsch et al. 2011). Based on their sensitivity rankings, documented relevance in previous studies, and importance for basins with similar characteristics, a subset of 12 parameters was selected for calibration.

2.4.2 Calibration and validation

Calibration involves systematically adjusting model parameters to improve agreement between simulated and observed streamflow, while validation assesses the predictive capability of the model by comparing simulated outputs against independent observed data not used during calibration (Arnold et al., 2012).

Automatic calibration was performed on a monthly time step using the Dynamically Dimensioned Search (DDS) algorithm (Tolson and Shoemaker, 2007) within SWAT+ Toolbox. DDS was selected because it has been shown to outperform other stochastic optimization algorithms as model complexity increases (Arsenault et al., 2014). The Nash-Sutcliffe Efficiency (NSE; Nash and Sutcliffe, 1970) was used as the objective function, as it effectively reflects the overall fit of a hydrograph (Servat and Dezetter, 1991) and is adequate for calibrating SWAT models (Molina-Navarro et al., 2017).

Each precipitation configuration (Gauge, Daymet, Daymet-Qmap, Daymet-MFBC) was calibrated independently using identical parameter ranges (Supplementary Table S3) and the same calibration/validation periods. A total of 500 simulations were performed for each configuration, following recommendations for ensuring sufficient simulations (Arnold et al., 2012). This procedure yielded one optimal simulation per configuration, defined as the parameter set producing the lowest discrepancy between simulated and observed streamflow. These optimal parameter sets (Supplementary Table S3) were then used for validation.

In addition, we accounted for parameter uncertainty by retaining the 100 best-performing simulations per configuration. This ensemble-based approach allowed model outcomes to be represented as probability distributions (e.g., medians with interquartile ranges), thereby bracketing equifinality and enabling comparisons of how precipitation inputs constrain or amplify uncertainty in water balance partitioning.

2.5 Model performance analysis

Model performance was evaluated through statistical analyses at a monthly scale. Although SWAT+ simulations were run at a daily time step, evaluation was based on monthly aggregations to reduce the influence of observational uncertainties in manually-obtained daily records. Statistical performance was quantified using the NSE and Percent Bias (PBias) (Gupta et al., 1999; Nash and Sutcliffe, 1970) due to their complementary nature.

NSE measures the predictive skill of the model by quantifying how well the simulated values capture the variance of observed streamflow and was calculated according to Equation 10. It ranges from –∞ to 1, where NSE = 1 indicates a perfect fit. Values between 0 and 1 are acceptable, whereas values ≤ 0 indicate that the mean of the observed values is a better predictor than the model (Moriasi et al., 2007; Ritter and Muñoz-Carpena, 2013). PBias quantifies the model tendency to over- or underestimate observed values and was calculated according to Equation 11; a value close to zero indicates minimal bias, while positive (negative) values indicate under- (over-) estimation (Gupta et al., 1999).

where n is the sample size; and Q_{sim, i} are the observed and simulated streamflow at time step i, and is the mean observed streamflow.

Following (Moriasi et al. 2015), results were classified into four skill categories (Table 3) to facilitate objective comparison across configurations. PBias was evaluated at four levels: (1) calibration, validation, and total period; (2) seasonal scale (wet: June–September; transition: October–January; dry: February–May); (3) high flows (observed discharge >75th percentile); and (4) low flows (observed discharge < 25th percentile).

To complement the traditional NSE and PBias metrics and provide a more comprehensive assessment of model performance, particularly regarding low flows and the balance between hydrograph components, we employed additional diagnostics following recommendations in hydrologic modeling (Gupta et al., 2009; Pushpalatha et al., 2012; Yilmaz et al., 2008).

The Kling-Gupta Efficiency (KGE) was used to provide an integrated view of model performance by decomposing the error into three components: correlation, variability, and bias (Gupta et al., 2009). KGE was calculated according to Equation 12:

where r is the Pearson correlation coefficient between simulated and observed streamflow, is the variability ratio (ratio of simulated to observed standard deviations), and is the bias ratio (ratio of simulated to observed means). KGE values range from -∞ to 1, with values closer to 1 indicating better agreement.

The log-transformed Nash-Sutcliffe Efficiency (logNSE) was applied to assess model performance specifically for low-flow conditions (Oudin et al., 2006; Pushpalatha et al., 2012). By applying the NSE formula to log-transformed streamflow values, logNSE gives more weight to errors during low-flow periods, where absolute errors are small but relative errors can be substantial. LogNSE was calculated according to Equation 13:

where ε is a small constant (here taken as 1/100 of the mean observed flows) to avoid issues with zero flows (Pushpalatha et al., 2012; Santos et al., 2018).

Finally, flow-duration curve (FDC) diagnostics (Supplementary Figure S1) were employed to evaluate model performance across different flow regimes (Yilmaz et al., 2008). We computed the percent bias for the 5th and 95th percentiles of the FDC, denoted as Q5 and Q95, respectively. Q5 represents high-flow conditions (exceeded 5% of the time), while Q95 represents low-flow conditions (exceeded 95% of the time). The bias for each flow percentile was calculated using the standard PBias formulation given in Equation 14:

where Q_{obs, p} and Q_{sim, p} are the observed and simulated streamflow values at the p-th percentile of the FDC. These diagnostics provide insights into the model's ability to reproduce peak flows (Q5) and baseflow conditions (Q95), complementing the overall performance metrics. All performance metrics (NSE, PBias, KGE, logNSE, and FDC diagnostics) were computed using reproducible scripts in R, which are provided in the repository along with the processed time series.

2.6 Water balance and spatial analysis

Water balance components were analyzed both deterministically and probabilistically. First, annual average values of water balance components were obtained from the single best-performing simulation for each precipitation configuration. Additionally, a probabilistic analysis was conducted using the 100 best-performing simulations for each configuration. This ensemble-based approach enabled the examination of parameter variability by bracketing the uncertainty introduced by each forcing product. Using the ensemble results, we identified the dominant hydrologic mechanisms (i.e., water balance partitioning) represented in each configuration and evaluated their differences across forcings. In this way, the water balance partitioning served as a means to understand parameter uncertainty that is introduced by the precipitation input. Moreover, this analysis tested the hypothesis that better forcing products reduce parameter equifinality and lead to less variations within the 100 simulations.

Furthermore, the spatial distributions of hydrologic processes from the single best-performing simulation of each configuration were mapped at the LSU level (for groundwater flow, at the aquifer scale) to examine the spatial representation of the outputs with each of the forcings. These maps should not be interpreted as validated representations of basin hydrology, since no independent observations on flow partitioning or distributed fluxes are available. Instead, they illustrate how precipitation inputs condition the spatial variability of model outputs, serving to contextualize relative differences among configurations rather than to identify the “true” dominant processes.

To quantitatively assess the spatial heterogeneity of simulated hydrologic components across the four precipitation forcings, we applied Global Moran's I statistic (Moran, 1950) to the LSU outputs, and for groundwater flow at the aquifer scale. Moran's I measures spatial autocorrelation and was calculated according to Equation 15:

where n is the number of spatial units, x_i and x_j are the values of the hydrologic component at spatial units i and j, is the mean value, and w_ij is a spatial weight matrix defining the neighborhood relationship between spatial unit i and j. We used a queen contiguity-based weights matrix, which considers spatial units as neighbors if they share a boundary or vertex. Moran's I values range from −1 (perfect dispersion) to 1 (perfect clustering), with 0 indicating complete spatial randomness (Cliff and Ord, 1981; Moran, 1950). Statistical significance was evaluated under the null hypothesis of spatial randomness using analytical tests implemented in the spdep package in R (version 1.4-2; Bivand et al., 2026). This analysis was performed separately for each precipitation forcing and each simulated hydrologic component (precipitation, evapotranspiration, surface runoff, and groundwater flow) using the best-performing simulation of each configuration.

3 Results and discussion

3.1 SWAT+ model performance

The SWAT+ model demonstrated strong performance, effectively capturing the temporal variability of observed streamflow despite the challenges posed by data scarcity and the absence of prior hydrologic studies in the basin. Notably, the model accurately captured peak flows, an essential aspect for water resource management, though it consistently underestimated low flows (Figure 5). This limitation is further confirmed by flow-regime analysis and complementary diagnostics (Sections 3.2.1 and 3.2.2), particularly high PBias in low-flow conditions, negative logNSE values, and FDC biases, which indicate persistent deficiencies in low-flow representation. This limitation is likely attributable to the simplified representation of groundwater dynamics in the model (Bailey et al., 2020; Montecelos-Zamora et al., 2018; Sánchez-Gómez et al., 2024; Yimer et al., 2023). The simulation of groundwater processes is recognized as a key limitation of both SWAT and SWAT+ due to its simple parameterization of aquifers (Sánchez-Gómez et al., 2022, 2024; Yimer et al., 2023), constraining its ability to simulate baseflow.

Figure 5

In addition to structural constraints, the calibration process may have further contributed to low-flow underestimation. Calibrating SWAT+ simultaneously for very high and very low flows remains challenging (Pfannerstill et al., 2014; Shrestha et al., 2018). The use of NSE as the objective function prioritizes peak flow accuracy, often at the expense of baseflow representation (Kardhana et al., 2024; Zhang et al., 2015). While alternative objective functions available in SWAT+ Toolbox may better capture low-flow dynamics, NSE was chosen due to its widespread use and relevance to high-flow simulation. Despite this limitation, the overall model performance across the simulation period supports its application in the study region.

3.2 Streamflow simulation performance

Comparisons between simulated and observed monthly streamflows (Figure 5) revealed notable differences in model performance depending on the precipitation dataset used. As summarized in Table 4, the Daymet-MFBC configuration achieved the highest performance over the full simulation period (NSE = 0.81; PBias = 6.4%), indicating both strong predictive skill and low systematic bias. Daymet ranked second, performing slightly better than Gauge. Daymet yielded a NSE of 0.76 and a near-zero PBias (−0.6%), reflecting its ability to reproduce streamflow dynamics with good accuracy while minimizing systematic errors. The improvement of Daymet over Gauge (NSE = 0.73; PBias = −5.8%) highlights the value of gridded precipitation in capturing spatial variability that is not represented by the sparse gauge network.

In contrast, the Daymet-Qmap configuration did not enhance model performance relative to Daymet. Instead, the quantile-mapping correction led to a moderate decline in skill for the full period (NSE = 0.69; PBias = −3.6%) and a marked deterioration during validation (NSE = 0.39; PBias = 14.6%; Table 4). This result is consistent with previous studies indicating that bias correction, while beneficial in some contexts, can also introduce additional uncertainties (Teng et al., 2015; Zhang et al., 2022), which can be attributed to the mismatch between the high spatial resolution of Daymet and the sparse, unevenly distributed network of reference stations used for correction.

Importantly, while the long-term performance differences between Gauge and Daymet were moderate, the added value of Daymet lies in its finer spatial depiction of precipitation, which becomes relevant for analyses involving spatially distributed hydrologic processes and for applications in ungauged basins where ground-based measurements are limited.

These results align with previous studies showing the strong performance of Daymet for hydrologic modeling in diverse settings. For example, (Muche et al. 2020) reported strong SWAT performance with Daymet in a low-relief watershed in Kansas, while (Ray et al. 2022) found Daymet to be among the most reliable datasets for streamflow simulation in north-central Texas. (Shuai et al. 2022) observed excellent results with Daymet in a high-alpine basin in Colorado. Our findings extend these outcomes to a mountainous basin in northwestern Mexico, reinforcing the usefulness of Daymet for hydrologic modeling in complex terrain.

Daymet-MFBC produced the most accurate streamflow simulations, followed by Daymet. These results emphasize the utility of high-resolution gridded precipitation products for hydrologic applications in mountainous regions. Moreover, the strong performance of Daymet-MFBC demonstrates the effectiveness of the mean field bias correction approach in reducing precipitation biases and improving streamflow simulation. While quantile mapping remains a useful bias-correction method, its effectiveness depends strongly on the density and spatial representativeness of the reference gauge network.

The markedly lower performance of the Daymet-Qmap configuration, particularly during the validation period, warrants further discussion. Quantile mapping (Qmap) is highly sensitive to the density and representativeness of the reference gauge network used for correction (Al-Rawas et al., 2025). In our study area, the sparse and unevenly distributed network of six stations likely failed to capture the full spatial variability of precipitation, especially in complex terrain where orographic effects create strong local gradients. Consequently, applying station-derived distributional adjustments to high-resolution grid cells may have introduced localized errors and distorted the spatial precipitation field. In contrast, the mean field bias correction (MFBC) method applies a spatially uniform, daily multiplicative factor. While simpler, this approach preserves the inherent relative spatial structure and variability of the original Daymet product, which our prior evaluation (de la Fraga et al., 2024) found to have a reasonable spatial correlation with observed patterns. In this specific context of a data-scarce mountainous basin, preserving Daymet's spatial patterns while systematically adjusting its magnitude proved more beneficial for streamflow simulation than the more complex, but gauge-network-dependent, distributional mapping of the Qmap method.

3.2.1 Over- and under-prediction patterns

Although total-period PBias values were relatively small (−5.8% to 6.4%), further diagnostic evaluation revealed substantial regime-dependent deviations (Table 5).

At the seasonal scale, all simulations underestimated the streamflow during the dry season (February–May), with PBias ranging from 51.9% to 60%. This indicates that approximately half of the observed dry-season volume was not captured by the model. The mean observed dry-season discharge is approximately 15.3 m³/s; therefore, a 60% PBias indicates that simulated flows are on average about 9 m³/s lower than observed. While the percentage appears large, the magnitude of the flows during this period is small.

During the wet season (June–September), PBias values were comparatively small (−12.3% to 0.4%), suggesting adequate representation of monsoonal runoff dynamics. Given a mean observed wet-season discharge of approximately 86.5 m³/s, a −10% PBias for example implies that simulated discharge exceeds observations by about 8–9 m³/s. Despite the larger absolute magnitudes involved, relative deviations during the monsoon period remain moderate to small, indicating that runoff volumes during the wet season are generally well represented. The Daymet-MFBC configuration showed nearly unbiased performance during this period (0.4%).

Flow-regime analysis further revealed systematic underestimation of low flows (< P25), with PBias between 78.5% and 87.5%. These high percentages reflect the difficulty of the model in reproducing baseflow conditions and are amplified in part by the low magnitude of observed discharge. The average observed discharge in this regime is approximately 4.15 m³/s. For example, in the Daymet-MFBC simulation, mean low-flow discharge is approximately 0.52 m³/s, corresponding to an underestimation of about 3.6 m³/s. Although this translates into a large relative bias (87.5%), the absolute difference is small compared to high-flow conditions. These high percentages are therefore amplified by the low magnitude of observed baseflow discharge.

Conversely, high flows (>P75) were moderately overestimated in most simulations (PBias between −9.1% to −6.5%), except for Daymet-MFBC (−0.8%). Given that the mean observed high-flow discharge is approximately 141.4 m³/s, a −0.8% PBias corresponds to an overestimation of about 1 m³/s. Thus, the relative bias remains very small and indicates a very good simulation of peak-flow volumes.

3.2.2 Complementary performance diagnostics

The inclusion of KGE provides additional insight into the balance between correlation, variability, and bias (Table 6). KGE values ranged from 0.758 (Daymet-Qmap) to 0.851 (Daymet-MFBC), confirming the superior overall performance of the Daymet-MFBC configuration. Correlation coefficients (r = 0.893–0.918) indicate strong temporal agreement across all forcings. The variability ratio (α) exceeded unity in all cases (1.107–1.215), indicating slight overdispersion in simulated variability relative to observations. The bias ratio (β) remained close to unity (0.936–1.058), consistent with previously reported total PBias values.

In contrast, logNSE values were consistently negative (−0.978 to −0.770), reflecting poor performance under logarithmic transformation and confirming substantial deficiencies in low-flow simulation. FDC diagnostics further support this interpretation. High-flow signature bias (Q5) ranged from −17.9% to −10.6%, indicating moderate overestimation of peak-flow magnitudes. In contrast, low-flow signature bias (Q95) approached 100% across all simulations, indicating that simulated 5th-percentile discharge values are near zero relative to observations. This confirms that baseflow recession and groundwater contributions are insufficiently represented.

Together, these complementary diagnostics provide a more comprehensive interpretation of model performance than NSE and PBias alone. While NSE, PBias, and KGE indicate good overall agreement, logNSE, regime-based PBias, and FDC metrics reveal systematic deficiencies in low-flow representation. This divergence demonstrates that satisfactory performance in traditional metrics does not necessarily imply an adequate representation of underlying hydrologic processes, particularly groundwater contributions. Accordingly, internal process interpretations are framed in terms of hydrologic plausibility rather than validation.

3.3 Water balance analysis

The water balance analysis offered additional insights into the hydrologic functioning of the Humaya River basin across precipitation inputs. Table 7 presents annual average values of water balance components for the best-performing simulations over the full period (1983–2001), while results for calibration and validation periods are presented in Supplementary Table S4. In addition, distribution-based results from the 100 best simulations in the calibration period (1983–1992) are shown in Figure 6, which provide a probabilistic perspective of water balance partitioning.

Figure 6

Evapotranspiration (ET) was the dominant flux, accounting for 49%−55% of annual precipitation across all configurations. This outcome is consistent with the extensive forest cover (~84%) and warm monsoonal climate in the basin, both of which promote high evaporative demand. Furthermore, in the North American Monsoon region, ET represents a major source of atmospheric moisture that contributes to convective precipitation events (Hu and Dominguez, 2015). The simulated mean potential evapotranspiration (PET) of 1,668 mm/yr closely agrees with estimates reported by (Monterroso-Rivas and Gómez-Díaz 2021) and (Shamir et al. 2024) for the study region, indicating that the model adequately captures regional dynamics. The contrast between PET (1,668 mm/year) and actual ET (477–524 mm/year) reflects a large moisture deficit, suggesting that, despite considerable precipitation (947–1,031 mm/year), the atmospheric demand remains unmet due to soil moisture constraints (Kurc and Small, 2004; Vivoni et al., 2008, 2010), characteristic of subhumid mountains with shallow, low-retention soils.

Differences in annual precipitation between Gauge and Daymet (958 and 1,031 mm) are consistent with (de la Fraga et al. 2024), who identified a systematic tendency of Daymet to overestimate precipitation in northwestern Mexico. Bias correction reduced these discrepancies, bringing corrected Daymet values closer to gauge observations (Daymet-Qmap = 947 mm and Daymet-MFBC = 1,000 mm).

Total streamflow represented 22%−28% of annual precipitation, though streamflow partitioning differed markedly across configurations. Surface runoff contributed between 38%−62%, groundwater between 31%−56%, and lateral flow remained consistently minor (6%−7%). These differences reflect the complex hydrologic setting, influenced by steep slopes, deep canyons, and the predominance of shallow, low-infiltration soils (e.g., Leptosols, Regosols), which favor rapid drainage and surface flow over deep infiltration (Descroix et al., 2007). These characteristics, combined with moderate to high precipitation, support the simulated partitioning patterns (Custodio and Llamas, 1976; Sánchez-Gómez et al., 2022).

Additionally, the basin geological setting, dominated by highly weathered, fractured, and faulted igneous rocks [Servicio Geológico Mexicano (SGM), 1999], enhances secondary permeability, thereby influencing groundwater flow dynamics. In this context, the simulated groundwater contributions in the Gauge model (56%) and the values in the Daymet and Daymet-corrected models (31%−42%) are hydrologically plausible, aligning with expected values for mountainous terrains with intermediate permeability substrates, which are neither fully impervious (as would be the case in compact shales and schists) nor highly permeable (as in carbonate-dominated formations) (Carroll et al., 2018, 2024; Custodio and Llamas, 1976; Döll and Fiedler, 2008; Miller et al., 2016; Muñoz-Villers and McDonnell, 2012; Rumsey et al., 2015; Winter, 2007).

Although the hydrologic model incorporated spatial datasets of topography, soils, and land use/cover, and key soil and aquifer parameters were carefully calibrated to reflect the hydrogeological characteristics of the basin, it is important to interpret the partitioning of internal components with caution. In the absence of field validation (e.g., groundwater levels, soil moisture), these results represent process-based approximations rather than verified flux estimates.

The model-derived flux ratios can be interpreted in the context of previous hydrologic studies conducted in Mexico. Reported ET/P ratios for a neighboring basin located south of the study area and for a humid forested mountainous basin in central Veracruz range between 48%−77% (e.g., Aragón-Hernández et al., 2024; López-Ramírez et al., 2021). In contrast, in drier and warmer regions to the north, such as the state of Sonora, reported ET/P values reach about 80% (e.g., Robles-Morua et al., 2012) and can be even higher in arid zones like the Chihuahuan Desert (e.g., Mielnick et al., 2005). At the national scale, the mean ET/P ratio for Mexico has been estimated to be approximately 72% (Arreguín et al., 2011). Reported Q/P ratios for the study region and other montane forest basins with similar hydroclimatic characteristics typically range between 10%−30% [e.g., Comisión Nacional del Agua (CONAGUA), 2009; Descroix et al., 2007; Instituto Nacional de Estadística y Geografía (INEGI), 1995; Muñoz-Villers and McDonnell, 2012]. Specifically, for the Humaya River basin, the Q/P ratios range from 15%−45% between July and October (Gochis et al., 2006). The values obtained in this study (ET/P = 49%−55%; Q/P = 22%−28%) fall within these ranges, supporting the hydrologic plausibility of the simulated flux partitioning.

While these results provide insights into the basin hydrologic functioning, they also reflect the influence of model calibration and parameterization on water balance partitioning (Huang and Liang, 2006). Different precipitation inputs may guide the calibration process toward distinct parameter combinations that reproduce streamflow with comparable accuracy but diverge in terms of underlying processes. To better capture this dimension of uncertainty, we further analyzed ensembles of the best-performing parameter sets for each forcing, thereby framing model outcomes as probability distributions rather than single deterministic values. This approach allows a more robust assessment of parameter uncertainty and equifinality (Beven, 2006; Ficklin and Barnhart, 2014; Huang and Liang, 2006), and sheds light on how precipitation inputs constrain (or expand) the range of plausible hydrologic mechanisms.

3.4 Parameter uncertainty and equifinality

The ensemble analysis based on the 100 best-performing simulations for each precipitation input revealed clear differences in water balance partitioning. Boxplots of ET, surface runoff (SURQ), lateral flow (LATQ), and groundwater flow (GWQ) (Figure 6) showed that, while median values were broadly consistent with the deterministic results reported for the calibration period in the Supplementary Table S4, the interquartile ranges highlighted the extent of parameter-driven uncertainty, particularly under Daymet forcing. These results highlight the classic problem of equifinality in hydrologic modeling, where multiple parameter sets yield similar performance at the outlet but imply distinct internal hydrologic processes (Beven, 2006; Beven and Freer, 2001).

Groundwater contributions in the Gauge-based ensemble were consistently higher, whereas Daymet and bias-corrected ensembles produced lower values, suggesting that Gauge-forced runs tend to produce parameter sets that favor baseflow generation. This outcome is linked to the behavior of parameters such as flo_min (threshold depth from surface to water table for groundwater flow to occur), revap_co (re-evaporation coefficient), and alpha (baseflow alpha factor) (Figure 7). Increasing flo_min directly favors the contribution of groundwater flow to the stream. Similarly, a lower revap_co decreases the upward flux from shallow aquifers to the unsaturated zone, which indirectly sustains groundwater availability and, consequently, enhances subsurface contributions to streamflow. In addition, higher values of alpha accelerate aquifer response and increase the rate at which groundwater contributes to streamflow (Arnold et al., 2012; Guse et al., 2014; Neitsch et al., 2011).

Figure 7

A key question is whether the observed differences in parameters (e.g., flo_min) are the cause of the distinct water balance partitioning, or rather a consequence of the different precipitation forcings. Our ensemble-based approach allows us to address this. Although parameters such as flo_min exhibit noticeable differences among precipitation configurations (Figure 7), these variations should be interpreted in the context of the calibration process rather than as independent drivers of the simulated water balance. Because each precipitation configuration was calibrated independently using identical parameter ranges and objective functions, the precipitation forcing constrains the region of the parameter space that yields acceptable model performance. Consequently, parameter values adjust to compensate for differences in precipitation magnitude, timing, and spatial distribution. The resulting parameter distributions therefore represent the model's response to the forcing characteristics. This relationship explains why differences in water balance partitioning (Table 7 and Supplementary Table S4, and Figure 6) are primarily associated with the precipitation inputs, while parameter values act as the mechanisms through which the model adjusts internal fluxes in order to reproduce the observed streamflow dynamics.

Surface runoff displayed the opposite tendency. Specifically, Daymet and its bias-corrected variants consistently yielded higher contributions, whereas Gauge simulations favored lower values. Consequently, this suggests that Daymet-derived simulations produce parameter sets that favor rapid surface responses. This tendency is linked to awc (available soil water capacity), bd (bulk density), chk (channel bottom conductivity), cn2 (curve number), and surlag (surface runoff lag coefficient). Higher values of bd and lower values of awc and chk limit infiltration capacity, promoting greater overland flow. Higher values of cn2 directly increase runoff potential by reducing infiltration, while lower values of surlag shorten the delay between rainfall and surface runoff peaks, further enhancing direct flow responses (Arnold et al., 2012; Neitsch et al., 2011; Van Liew et al., 2007). This outcome is consistent with the geomorphological setting of the Humaya River basin, where steep slopes and shallow soils promote fast drainage and surface flow.

ET remained comparatively stable across ensembles, although systematic differences emerged among forcings. Gauge-forced runs produced higher ET (~570 mm) with narrower uncertainty ranges, whereas Daymet ensemble yielded lower values (~550 mm) and greater variability. The Daymet-Qmap ensemble resulted in the lowest ET (~525 mm), consistent with its reduced precipitation input, while Daymet-MFBC produced slightly higher ET (~562 mm). These outcomes reflect the combined influence of precipitation inputs and soil–vegetation parameters such as esco (soil evaporation compensation factor) and epco (plant uptake compensation factor). Lower esco values increase the contribution of deeper soil layers to evaporation, while epco values approaching 1 enable greater root water uptake from deeper horizons (Neitsch et al., 2011). The Gauge and bias-corrected Daymet ensembles tended toward lower esco and near-unit epco values, which favored higher ET rates. Importantly, the trade-off between ET and surface runoff suggests that ensembles with higher surface runoff tended to allocate less water to ET. Conversely, configurations with greater groundwater contributions often sustained higher ET rates. This pattern aligns with (Robles-Morua et al. 2012), who reported that infiltration-excess runoff promotes limited infiltration and rapid drying of shallow soils, thereby constraining ET. In contrast, sustained ET under drier surface conditions can occur when transpiration is supported by deep root zones that remain moist due to recharge. Such behavior highlights the role of water availability in maintaining ET during dry periods.

The application of bias correction reduced the variability evident in Daymet ensembles, both in the distribution of parameter values and in water balance partitioning. This indicates that bias correction in precipitation inputs reduces parameter uncertainty and equifinality by constraining the range of plausible parameter values, revealing which parameters truly influence the hydrologic response rather than merely compensating for input errors.

Overall, the ensemble-based framework provides a probabilistic context for evaluating water balance partitioning. By presenting outcomes as distributions rather than single deterministic simulations, the analysis captures both forcing and parameter uncertainty, enabling a more cautious and transparent interpretation of internal fluxes (Beven, 2006; Huang and Liang, 2006). Furthermore, the results support the hypothesis that more consistent forcings (e.g., Daymet-MFBC) reduce the range of plausible hydrologic mechanisms, whereas less consistent inputs (e.g., Daymet) expand the parameter space and yield more divergent outcomes.

3.5 Spatial distribution of hydrologic processes

Most SWAT model assessments have traditionally relied on statistical comparisons between simulated and observed streamflow at the watershed outlet (Arnold et al., 2015). However, evaluating model performance solely based on watershed outlet streamflow may mask the spatial heterogeneity of hydrologic processes occurring within the basin. To address this limitation, spatial maps of key variables (Figures 8–10, Supplementary Figures S2, S3) were generated to examine the spatial representation of the outputs with each of the forcings.

Figure 8

Figure 10

Model outputs driven by Daymet and Daymet-MFBC demonstrated a superior capability to capture the spatial variability of the basin hydrology, offering a more detailed representation of its heterogeneity. Precipitation fields from these datasets more clearly delineated high- and low-rainfall zones, whereas simulations based on Gauge and Daymet-Qmap inputs yielded more spatially homogeneous patterns, constrained by the sparse and uneven distribution of meteorological stations (Figure 8).

Spatial patterns of ET were better represented in the Daymet-based configurations (Figures 9B–D), with higher ET values in the forested, humid uplands of the Sierra and lower values in agricultural and rangeland zones and valleys. This spatial coherence indicates that SWAT+ effectively captured the main land cover and climate controls on ET. In contrast, the Gauge-driven model produced a more uniform ET distribution (Figure 9A), reflecting the spatial limitations of the input data.

Figure 9

The finer spatial resolution of Daymet facilitated a more detailed representation of the spatial variability of runoff across the basin. This can be seen in the model outputs related to Daymet and Daymet-MFBC (Figures 10B, D). Conversely, the Gauge and Daymet-Qmap configurations displayed smoother and less differentiated runoff fields (Figures 10A, C), again constrained by the limited number and distribution of meteorological stations. These results are consistent with recent studies (e.g., Chen et al., 2024; Nguyen et al., 2026; Sang et al., 2022), which emphasize the importance of high-resolution precipitation data for realistic runoff simulation.

To further evaluate whether model differences were amplified in steeper terrain, surface runoff was analyzed across three slope classes (< 8%, 8%−30%, and >30%) using the mean annual values (1983–2001) at the LSU scale (Supplementary Table S5). Across all precipitation forcings, surface runoff increased slightly with slope steepness; however, the magnitude of this increase was modest. The ratio between the steepest (>30%) and the flattest (< 8%) classes was consistently ~1.05 for all configurations, indicating only a 5 % increase in runoff between low- and high-slope zones. Although Daymet-based configurations produced higher absolute runoff magnitudes across all slope classes, the relative slope-driven gradient remained nearly identical among forcings. This suggests that differences in runoff partitioning are primarily associated with the precipitation input itself rather than an enhanced sensitivity to steep terrain.

Lateral flow (Supplementary Figure S2) exhibited similar general patterns among forcings, with greater values along canyon and valley zones, although Daymet and Daymet-MFBC runs offered finer spatial contrast. Groundwater flow displayed marked differences across datasets (Supplementary Figure S3). The Gauge model yielded the highest overall groundwater flow rates, while the Daymet and Daymet-MFBC simulations produced more localized recharge zones, suggesting a closer correspondence between precipitation heterogeneity and percolation patterns. The Daymet-Qmap configuration resulted in intermediate magnitudes and spatial structures between the Gauge and Daymet models.

To quantify the spatial patterns observed in Figures 8–10 and Supplementary Figures S2, S3, we computed Global Moran's I statistic (Moran, 1950) for each hydrologic component under the four precipitation forcings (Supplementary Table S6). Moran's I measures spatial autocorrelation, with values near 1 indicating strong spatial clustering (smooth, homogeneous patterns), values near 0 indicating spatial randomness, and negative values indicating dispersion (Cliff and Ord, 1981; Moran, 1950).

For precipitation, evapotranspiration, and surface runoff, all configurations exhibited positive and highly significant Moran's I values (p < 0.001), confirming that the simulated hydrologic fields are spatially structured rather than randomly distributed. However, clear differences emerged among forcings. Gauge and Daymet-Qmap consistently produced higher Moran's I values (e.g., precipitation: 0.875 and 0.880, respectively) compared to Daymet and Daymet-MFBC (0.792 and 0.789). Lower Moran's I values do not imply poorer performance; rather, they indicate greater spatial heterogeneity and increased variability between neighboring LSUs. These results statistically confirm our visual interpretation of Figures 8–10, where Gauge and Daymet-Qmap generate more spatially smoothed patterns, whereas Daymet and Daymet-MFBC yield greater heterogeneity, i.e., more spatially differentiated hydrologic fields.

Groundwater flow exhibited moderate spatial autocorrelation. The Gauge configuration showed the highest spatial organization (I = 0.41, p < 0.001), followed by Daymet (I = 0.32, p = 0.006) and Daymet-MFBC (I = 0.29, p = 0.010). This suggests that Daymet and Daymet-MFBC have slightly lower spatial organization, i.e., greater spatial heterogeneity or variability. In contrast, Daymet-Qmap produced weak and statistically non-significant spatial autocorrelation (I = 0.13, p = 0.091), suggesting a more spatially diffuse groundwater representation. The reduction in Moran's I at the aquifer scale compared to LSUs reflects a scale effect, as spatial aggregation inherently attenuates autocorrelation strength.

Overall, these results demonstrate that while all forcings generate spatially structured hydrologic fields, Daymet and Daymet-MFBC enhance internal heterogeneity, whereas Gauge and Daymet-Qmap promote stronger global clustering. The spatial resolution and structure of precipitation inputs therefore influence not only magnitude but also the spatial organization of hydrologic processes.

Our findings also align with (Maina et al. 2020) and (Shuai et al. 2022), who highlight that the spatial resolution of meteorological inputs is essential to preserve heterogeneity and improve the representation of distributed hydrologic processes. The improved spatial depiction of precipitation enhances the model ability to capture in greater detail the spatial distribution of key processes, including ET, surface runoff, lateral flow, and groundwater flow, an important advancement over models based on data from a few meteorological stations that tend to homogenize spatial patterns and, consequently, overlook localized hydrologic phenomena.

It is important to note that the spatial maps presented here cannot be directly validated due to data scarcity. However, the simulated patterns (e.g., higher ET in forested uplands, greater runoff in steep deforested slopes and higher precipitation zones) are consistent with known hydroclimatic controls reported for other mountainous basins in Mexico (e.g., Descroix et al., 2007; Muñoz-Villers and McDonnell, 2012; Robles-Morua et al., 2012; Vivoni et al., 2010), providing indirect support for the realism of the outputs.

These spatial insights are highly relevant for water resource management and land-use planning, as they allow the identification of hydrologically sensitive zones, recharge areas, and regions vulnerable to runoff; thus, supporting the implementation of targeted, site-specific interventions. For example, in high-runoff zones, such as steep, deforested slopes, reforestation, or soil conservation efforts can help mitigate erosion and reduce flood risks. In groundwater-dependent valleys, managed aquifer recharge initiatives may be prioritized to counteract over-extraction. In precipitation hotspots, optimizing reservoir storage can buffer dry-season shortages. In summary, incorporating high-resolution gridded precipitation products such as Daymet not only improves streamflow simulation but also enhances the spatial fidelity of hydrologic process representation, providing a stronger basis for site-specific management and adaptive water resource planning, particularly valuable in mountainous and data-scarce environments facing growing climatic and anthropogenic pressures.

3.6 Limitations and applicability

While this study demonstrates the utility of high-resolution gridded precipitation data for hydrologic modeling in a data-scarce mountainous basin, several limitations must be acknowledged to contextualize the findings and their broader applicability.

3.6.1 Model and methodological limitations

The SWAT+ model employs simplified representations of key subsurface processes, particularly the parameterization of groundwater flow and baseflow generation, which can affect the simulation of low flows and the partitioning of streamflow components (Bailey et al., 2020; Sánchez-Gómez et al., 2024). Furthermore, our calibration relied exclusively on streamflow data at the basin outlet. This discharge-only calibration, while common, constrains our ability to uniquely identify internal process representations, as evidenced by the equifinality captured in our ensemble analysis. The resulting water balance partitioning should therefore be interpreted as a set of hydrologically plausible scenarios consistent with the outlet hydrograph, rather than a definitively validated account of internal fluxes. The absence of spatially distributed validation data, such as soil moisture, groundwater levels, or evapotranspiration fluxes, limits conclusive verification of the simulated spatial patterns.

3.6.2 Input data limitations

The performance of any gridded precipitation product is inherently tied to its underlying algorithms and the density of ground observations used for its creation or correction. Our findings on the effectiveness of Daymet and the mixed results from bias correction are specific to the climatic and topographic context of the Sierra Madre Occidental in northwestern Mexico. The quantile mapping approach (Daymet-Qmap), for instance, was constrained by the sparse and uneven distribution of the reference gauge network, which may not adequately represent precipitation variability at the scale of the Daymet grid.

3.6.3 Universal applicability

The conclusions of this study are most directly relevant to tropical-subtropical, monsoon-influenced mountainous watersheds with limited gauge networks, similar to our study area. The superior performance of the mean field bias correction (MFBC) over quantile mapping (Qmap) in this setting highlights that the optimal method for adjusting gridded precipitation may depend heavily on the density and representativeness of the available reference data. In regions with different dominant precipitation mechanisms or topography, the relative performance of Daymet and the efficacy of different bias-correction techniques could differ. Therefore, for applications in new regions we recommend to conduct a preliminary evaluation of available precipitation products, similar to (de la Fraga et al. 2024), followed by a focused hydrologic validation to identify the most reliable forcing dataset for local modeling objectives.

Despite these limitations and context-specific considerations, the consistency of results across multiple independent diagnostics reinforces the robustness of the main conclusions. Furthermore, our core methodological approach, using ensemble simulations to evaluate how precipitation inputs affect both integrated streamflow performance and the uncertainty of internal process representation, provides a transferable framework for improving hydrologic modeling in other data-scarce, complex terrains.

4 Conclusions

This study evaluated the influence of four precipitation input configurations: Gauge, Daymet, Daymet-Qmap, and Daymet-MFBC, on the performance of the SWAT+ model in the mountainous, data-scarce, monsoon-influenced Humaya River basin in northwestern Mexico. The SWAT+ model demonstrated strong performance and applicability in this complex and topographically diverse setting. Among the forcings, Gauge, Daymet, and Daymet-MFBC produced good to very good monthly streamflow simulations, with Daymet-MFBC achieving the best overall performance (NSE = 0.81, PBias = 6.4%, KGE = 0.85) and the most consistent spatial representation of hydrologic processes. In contrast, Daymet-Qmap showed comparatively lower performance, suggesting that the effectiveness of this bias correction method depends on the availability and representativeness of ground observations, and that methods preserving spatial structure (e.g., MFBC) may be more suitable in data-scarce contexts.

Despite good overall performance according to NSE, PBias, and KGE, all configurations exhibited systematic deficiencies in low-flow simulation, as indicated by high PBias in low-flow conditions and negative logNSE values (−0.98 to −0.77). These results point to a structural limitation of SWAT+ in representing baseflow processes under the studied conditions, rather than differences attributable solely to precipitation forcing.

Water balance analyses identified evapotranspiration as the dominant flux (49%−55% of annual precipitation). Streamflow represented 22%−28% of precipitation, with clear differences in the partitioning of surface and subsurface flows among forcings, i.e., the Daymet derived models produced higher surface runoff and lower groundwater contributions relative to the Gauge-based model. These variations illustrate the sensitivity of the calibration of internal hydrologic mechanisms to the choice of precipitation inputs. These differences are interpreted in terms of hydrologic plausibility rather than direct validation of internal processes. Ensemble-based analyses further showed that precipitation forcing influences not only model performance but also constrains the range of plausible parameter combinations and associated process representations, directly addressing equifinality. This ensemble framework provides a transparent approach for disentangling the effects of forcing uncertainty and parameter uncertainty in hydrologic modeling.

Spatial analyses, including Moran's I, indicated that Daymet and Daymet-MFBC produce more heterogeneous and spatially differentiated representations of basin hydrology, whereas Gauge and Daymet-Qmap tend to yield smoother spatial patterns, likely reflecting the limitations imposed by sparse station coverage.

Overall, the results highlight the importance of high-resolution gridded precipitation datasets for hydrologic modeling in mountainous, data-scarce environments. In such regions, uncorrected or minimally corrected products that preserve spatial structure, such as Daymet or Daymet-MFBC, may offer the best compromise between spatial realism and predictive accuracy. The findings contribute to a more comprehensive understanding of how precipitation input selection influences not only streamflow simulation, but also the internal dynamics, parametric uncertainty, and spatial representation of hydrologic processes. The insights gained here are not only applicable to northwestern Mexico but also to other regions characterized by complex terrain, data scarcity, and strong precipitation gradients. Improving the accuracy and spatial realism of hydrologic models in such settings is essential for advancing water resource assessment, enhancing climate resilience, and supporting sustainable watershed management.

References

• 1

  AlexanderssonH. (1986). A homogeneity test applied to precipitation data. J. Climatol.6, 661–675. doi: 10.1002/joc.3370060607

  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  Al-RawasG.NikooM. R.SadraN.MousaviF. (2025). Integrated quantile mapping and spatial clustering for robust bias correction of satellite precipitation in data-sparse regions. Sustainability17:8321. doi: 10.3390/su17188321

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  Aragón-HernándezJ. L.Hernández-GonzálezL. A.de LeónD. A. S.González-VillarrealF. J. (2024). Water availability: spatially distributed analysis and evaluation of input variables. H2Open J.7, 333–350. doi: 10.2166/h2oj.2024.016

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  ArnoldJ. G.MoriasiD. N.GassmanP. W.AbbaspourK. C.WhiteM. J.SrinivasanR.et al. (2012). SWAT: model use, calibration, and validation. Trans. ASABE55, 1491–1508. doi: 10.13031/2013.42256

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  ArnoldJ. G.SrinivasanR.MuttiahR. S.WilliamsJ. R. (1998). Large area hydrologic modeling and assessment part I: model development. J. Am. Water Resour. Assoc.34, 73–89. doi: 10.1111/j.1752-1688.1998.tb05961.x

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  ArnoldJ. G.YoussefM. A.YenH.WhiteM. J.SheshukovA. Y.SadeghiA. M.et al. (2015). Hydrological processes and model representation: impact of soft data on calibration. Trans. ASABE58, 1637–1660. doi: 10.13031/trans.58.10726

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  ArreguínF. I.LópezM.MarengoH. (2011). “Mexico's water challenges for the 21st century,” in Water Resources in Mexico. Scarcity, Degradation, Stress, Conflicts, Management, and Policy, ed. Ú. O. Spring (Berlin; Heidelberg: Springer-Verlag), 21–38. doi: 10.1007/978-3-642-05432-7

  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  ArsenaultR.PoulinA.CôtéP.BrissetteF. (2014). Comparison of stochastic optimization algorithms in hydrological model calibration. J. Hydrol. Eng.19, 1374–1384. doi: 10.1061/(ASCE)HE.1943-5584.0000938

  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  BaileyR. T.BiegerK.ArnoldJ. G.BoschD. D. (2020). A new physically-based spatially-distributed groundwater flow module for SWAT+. Hydrology7, 1–23. doi: 10.3390/hydrology7040075

  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  Baran-GurgulK.RutkowskaA. (2024). Water resource management: hydrological modelling, hydrological cycles, and hydrological prediction. Water16:3689. doi: 10.3390/w16243689

  • CrossRef
  • Google Scholar

• 11

  Barresi-ArmoaO. L.SauvageS.HouskaT.BiegerK.SchürzC.Sánchez-PérezJ. M. (2023). Representation of hydrological components under a changing climate—a case study of the Uruguay River basin using the new version of the Soil and Water Assessment Tool Model (SWAT+). Water15:2604. doi: 10.3390/w15142604

  • CrossRef
  • Google Scholar

• 12

  BaumerO. W. (1990). Predictions of Soil Hydraulic Parameters. Nebraska: WEPP Data Files for Indiana SCS National Soil Survey Lab., Lincoln.

  • Google Scholar

• 13

  BeckH. E.VergopolanN.PanM.LevizzaniV.van DijkA. I. J. M.WeedonG. P.et al. (2017). Global-scale evaluation of 22 precipitation datasets using gauge observations and hydrological modeling. Hydrol. Earth Syst. Sci.21, 6201–6217. doi: 10.5194/hess-21-6201-2017

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  BevenK. (2006). A manifesto for the equifinality thesis. J. Hydrol.320, 18–36. doi: 10.1016/j.jhydrol.2005.07.007

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  BevenK.FreerJ. (2001). Equifinality, data assimilation, and uncertainty estimation in mechanistic modelling of complex environmental systems using the GLUE methodology. J. Hydrol.249, 11–29. doi: 10.1016/S0022-1694(01)00421-8

  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  BiegerK.ArnoldJ. G.RathjensH.WhiteM. J.BoschD. D.AllenP. M.et al. (2017). Introduction to SWAT+, a completely restructured version of the soil and water assessment tool. J. Am. Water Resour. Assoc.53, 115–130. doi: 10.1111/1752-1688.12482

  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  BivandR.AltmanM.AnselinL.AssunçãoR.BeraA.BerkeO.et al. (2026). spdep: Spatial Dependence: Weighting Schemes, Statistics (R Package Version 1.4-2). doi: 10.32614/CRAN.package.spdep

  • CrossRef
  • Google Scholar

• 18

  CampozanoL.SánchezE.AvilésÁ.SamaniegoE. (2014). Evaluation of infilling methods for time series of daily precipitation and temperature: the case of the Ecuadorian Andes. Maskana5, 99–115. doi: 10.18537/mskn.05.01.07

  • CrossRef
  • Google Scholar

• 19

  CarrollR. W. H.BearupL. A.BrownW.DongW.BillM.WilllamsK. H. (2018). Factors controlling seasonal groundwater and solute flux from snow-dominated basins. Hydrol. Process.32, 2187–2202. doi: 10.1002/hyp.13151

  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  CarrollR. W. H.NiswongerR. G.UlrichC.VaradharajanC.Siirila-WoodburnE. R.WilliamsK. H. (2024). Declining groundwater storage expected to amplify mountain streamflow reductions in a warmer world. Nat. Water2, 419–433. doi: 10.1038/s44221-024-00239-0

  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  ChawandaC. J. (2021). SWAT+ Toolbox user manual. Available online at: https://www.openwater.network/assets/downloads/SWATplusToolboxUserMannual.pdf (Accessed February 15, 2026).

  • Google Scholar
  • View reference in article

• 22

  ChenJ. L.LuoW. B.WangW. J.YuL.LiY. L.ZouZ. K.et al. (2024). Influence of rainfall spatial variability on runoff simulation in hilly irrigation areas. Appl. Ecol. Environ. Res.22, 3347–3359. doi: 10.15666/aeer/2204_33473359

  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  CLICOM (2015). Climate computing program system. Available online at: http://clicom-mex.cicese.mx/ (Accessed February 15, 2026).

  • Google Scholar
  • View reference in article

• 24

  CliffA. D.OrdJ. K. (1981). Spatial Processes: Models and Applications. London: Pion.

  • Google Scholar
  • View reference in article

• 25

  Comisión Nacional del Agua (CONAGUA) (2009). Programa Hídrico Visión 2030 del Estado de Durango. Available online at: https://www.cmic.org.mx/comisiones/sectoriales/infraestructurahidraulica/estadisticas/estadisticas%202008/PHV2030EDURANGO2009%20feb%2009.pdf (Accessed March 30, 2026).

  • Google Scholar
  • View reference in article

• 26

  Comisión Nacional del Agua (CONAGUA) (2016). Banco Nacional de Datos de Aguas Superficiales (BANDAS). Available online at: https://app.conagua.gob.mx/bandas/ (Accessed February 15, 2026).

  • Google Scholar
  • View reference in article

• 27

  Comisión Nacional del Agua (CONAGUA) (2024a). Estadísticas del Agua en México 2023. Available online at: https://sinav30.conagua.gob.mx:8080/port_publicaciones.html (Accessed March 30, 2026).

  • Google Scholar

• 28

  Comisión Nacional del Agua (CONAGUA) (2024b). Sistema Nacional de Información del Agua (SINA 3.0). Available online at: https://sinav30.conagua.gob.mx:8080/SINA/?opcion=monitoreo (Accessed February 15, 2026).

  • Google Scholar

• 29

  CukierR. I.FortuinC. M.ShulerK. E.PetschekA. G.SchaiblyJ. H. (1973). Study of the sensitivity of coupled reaction systems to uncertainties in rate coefficients. I theory. J. Chem. Phys.59, 3873–3878. doi: 10.21236/AD0762420

  • CrossRef
  • Google Scholar

• 30

  CustodioE.LlamasM. R. (1976). Hidrología Subterránea: Vol. Tomo 1 (Second Edition). Barcelona: Omega.

  • Google Scholar

• 31

  de la FragaP.Del-Toro-GuerreroF. J.VivoniE. R.CavazosT.KretzschmarT. (2024). Evaluation of gridded precipitation datasets in mountainous terrains of Northwestern Mexico. J. Hydrol. Reg. Stud.56:102019. doi: 10.1016/j.ejrh.2024.102019

  • CrossRef
  • Google Scholar

• 32

  de la FragaP.VivoniE. R.Montecelos-ZamoraY.Del-Toro-GuerreroF. J.CavazosT.KretzschmarT. (2026). Data and code for: Hydrologic modeling in mountainous terrains with monsoonal climate: Do gridded precipitation data improve streamflow simulations? Zenodo. doi: 10.5281/zenodo.18820320

  • CrossRef
  • Google Scholar

• 33

  DescroixL.ViramontesD.EstradaJ.Gonzalez BarriosJ. L.AsselineJ. (2007). Investigating the spatial and temporal boundaries of Hortonian and Hewlettian runoff in Northern Mexico. J. Hydrol.346, 144–158. doi: 10.1016/j.jhydrol.2007.09.009

  • CrossRef
  • Google Scholar

• 34

  DileY.SrinivasanR.GeorgeC. (2023). QGIS Interface for SWAT+: QSWAT+. College Station and Temple, TX: Texas A&M University.

  • Google Scholar

• 35

  DöllP.FiedlerK. (2008). Global-scale modeling of groundwater recharge. Hydrol. Earth Syst. Sci.12, 863–885. doi: 10.5194/hess-12-863-2008

  • CrossRef
  • Google Scholar

• 36

  DongY.PengC.-Y. J. (2013). Principled missing data methods for researchers. SpringerPlus2:222. doi: 10.1186/2193-1801-2-222

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  ElsnerM. M.GangopadhyayS.PruittT.BrekkeL. D.MizukamiN.ClarkM. P. (2014). How does the choice of distributed meteorological data affect hydrologic model calibration and streamflow simulations?J. Hydrometeorol.15, 1384–1403. doi: 10.1175/JHM-D-13-083.1

  • CrossRef
  • Google Scholar

• 38

  EnayatiM.Bozorg-HaddadO.BazrafshanJ.HejabiS.ChuX. (2021). Bias correction capabilities of quantile mapping methods for rainfall and temperature variables. J. Water Clim. Change12, 401–419. doi: 10.2166/wcc.2020.261

  • CrossRef
  • Google Scholar

• 39

  FarrT. G.RosenP. A.CaroE.CrippenR.DurenR.HensleyS.et al. (2007). The shuttle radar topography mission. Rev. Geophys. 45. doi: 10.1029/2005RG000183

  • CrossRef
  • Google Scholar

• 40

  FicklinD. L.BarnhartB. L. (2014). SWAT hydrologic model parameter uncertainty and its implications for hydroclimatic projections in snowmelt-dependent watersheds. J. Hydrol.519, 2081–2090. doi: 10.1016/j.jhydrol.2014.09.082

  • CrossRef
  • Google Scholar

• 41

  Food and Agriculture Organization of the United Nations (FAO) (2003). The Digital Soil Map of the World, Version 3.6.Rome: Food and Agriculture Organization of the United Nations.

  • Google Scholar

• 42

  ForzieriG.VivoniE. R.FeyenL. (2013). Ecosystem biophysical memory in the southwestern North America climate system. Environ. Res. Lett.8:44016. doi: 10.1088/1748-9326/8/4/044016

  • CrossRef
  • Google Scholar

• 43

  GaoJ.SheshukovA. Y.YenH.WhiteM. J. (2017). Impacts of alternative climate information on hydrologic processes with SWAT: a comparison of NCDC, PRISM and NEXRAD datasets. CATENA156, 353–364. doi: 10.1016/j.catena.2017.04.010

  • CrossRef
  • Google Scholar

• 44

  GarcíaE. Comisión Nacional para el Conocimiento y Uso de la Biodiversidad (CONABIO) (1998). “Climas” (clasificación de Koppen, modificado por García). Escala 1:1000000. Mexico. Available online at: http://geoportal.conabio.gob.mx/metadatos/doc/html/clima1mgw.html (Accessed March 30, 2026).

  • Google Scholar

• 45

  GebrechorkosS. H.LeylandJ.DadsonS. J.CohenS.SlaterL.WortmannM.et al. (2024). Global-scale evaluation of precipitation datasets for hydrological modelling. Hydrol. Earth Syst. Sci.28, 3099–3118. doi: 10.5194/hess-28-3099-2024

  • CrossRef
  • Google Scholar

• 46

  GebremichaelM.VivoniE. R.WattsC. J.RodríguezJ. C. (2007). Submesoscale spatiotemporal variability of North American monsoon rainfall over complex terrain. J. Clim.20, 1751–1773. doi: 10.1175/JCLI4093.1

  • CrossRef
  • Google Scholar

• 47

  GhimireG. R.KaoS.-C.GangradeS. (2025). Enhancing streamflow reanalysis across the conterminous US leveraging multiple gridded precipitation data sets. Water Resour. Res.61:e2024WR038256. doi: 10.1029/2024WR038256

  • CrossRef
  • Google Scholar

• 48

  GochisD. J.Brito-CastilloL.ShuttleworthW. J. (2006). Hydroclimatology of the North American Monsoon region in northwest Mexico. J. Hydrol.316, 53–70. doi: 10.1016/j.jhydrol.2005.04.021

  • CrossRef
  • Google Scholar

• 49

  GudmundssonL.BremnesJ. B.HaugenJ. E.Engen-SkaugenT. (2012). Technical note: downscaling RCM precipitation to the station scale using statistical transformations - a comparison of methods. Hydrol. Earth Syst. Sci.16, 3383–3390. doi: 10.5194/hess-16-3383-2012

  • CrossRef
  • Google Scholar

• 50

  GudmundssonL. (2025). qmap: Statistical Transformations for Post-Processing Climate Model Output (R Package Version 1.0-6).Vienna: CRAN (Comprehensive R Archive Network) / The R Foundation. doi: 10.32614/CRAN.package.qmap

  • CrossRef
  • Google Scholar

• 51

  GuijarroJ. A. (2018). Homogenization of Climatic Series with Climatol Version 3.1.1. State Meteorological Agency (AEMET), Balearic Islands Office, Spain. doi: 10.13140/RG.2.2.27020.41604

  • CrossRef
  • Google Scholar

• 52

  GuptaH. V.KlingH.YilmazK. K.MartinezG. F. (2009). Decomposition of the mean squared error and NSE performance criteria: implications for improving hydrological modelling. J. Hydrol.377, 80–91. doi: 10.1016/j.jhydrol.2009.08.003

  • CrossRef
  • Google Scholar

• 53

  GuptaH. V.SorooshianS.YapoP. O. (1999). Status of automatic calibration for hydrologic models: comparison with multilevel expert calibration. J. Hydrol. Eng.4, 135–143. doi: 10.1061/(ASCE)1084-0699(1999)4:2(135)

  • CrossRef
  • Google Scholar

• 54

  GuseB.ReusserD. E.FohrerN. (2014). How to improve the representation of hydrological processes in SWAT for a lowland catchment - temporal analysis of parameter sensitivity and model performance. Hydrol. Process.28, 2651–2670. doi: 10.1002/hyp.9777

  • CrossRef
  • Google Scholar

• 55

  GutjahrO.HeinemannG. (2013). Comparing precipitation bias correction methods for high-resolution regional climate simulations using COSMO-CLM. Theor. Appl. Climatol.114, 511–529. doi: 10.1007/s00704-013-0834-z

  • CrossRef
  • Google Scholar

• 56

  HuH.DominguezF. (2015). Evaluation of oceanic and terrestrial sources of moisture for the North American monsoon using numerical models and precipitation stable isotopes. J. Hydrometeorol.16, 19–35. doi: 10.1175/JHM-D-14-0073.1

  • CrossRef
  • Google Scholar

• 57

  HuangM.LiangX. (2006). On the assessment of the impact of reducing parameters and identification of parameter uncertainties for a hydrologic model with applications to ungauged basins. J. Hydrol.320, 37–61. doi: 10.1016/j.jhydrol.2005.07.010

  • CrossRef
  • Google Scholar

• 58

  Instituto Nacional de Estadística y Geografía (INEGI) (1995). Estudio hidrológico del estado de Sinaloa.Aguascalientes: Instituto Nacional de Estadística, Geografía e Informática.

  • Google Scholar

• 59

  Instituto Nacional de Estadística y Geografía (INEGI) (2013). Conjunto de datos de perfiles de suelos. Escala 1:250000 Serie II (Continuo Nacional). Available online at: https://www.inegi.org.mx/app/biblioteca/ficha.html?upc=702825266707 (Accessed February 15, 2026).

  • Google Scholar

• 60

  Instituto Nacional de Estadística y Geografía (INEGI) (2021). Uso del suelo y vegetación. Escala 1:250000 Serie VII (Continuo Nacional). Available online at: http://geoportal.conabio.gob.mx/metadatos/doc/html/usv250s7gw.html (Accessed February 15, 2026).

  • Google Scholar

• 61

  Instituto Nacional de Estadística y Geografía (INEGI) (2024). Conjunto de datos vectorial edafológico. Escala 1:250000 Serie II (Continuo Nacional). Available online at: https://www.inegi.org.mx/app/biblioteca/ficha.html?upc=794551131916 (Accessed February 15, 2026).

  • Google Scholar

• 62

  JacksonM. L. (1964). “Determinaciones de Materia Orgánica en los Suelos” in Análisis Químico de Suelos. Ediciones Omega, Barcelona, Spain (Barcelona: Ediciones Omega), 282–310.

  • Google Scholar

• 63

  JarvisA.ReuterH. I.NelsonA.GuevaraE. (2008). Hole-filled seamless SRTM data V4. International Centre for Tropical Agriculture (CIAT). Available online at: https://srtm.csi.cgiar.org (Accessed February 15, 2026).

  • Google Scholar

• 64

  KardhanaH.LihawaA. W. I.RohmatF. I. W.WulandariS.HarjupaW.AdiprawitaW.et al. (2024). Water balance analysis in the Majalaya watershed: two-step calibration and application of the SWAT+ model for low-flow conditions. Water16:3498. doi: 10.3390/w16233498

  • CrossRef
  • Google Scholar

• 65

  KessabiR.HanchaneM.GuijarroJ. A.KrakauerN. Y.AddouR.SadikiA.et al. (2022). Homogenization and trends analysis of monthly precipitation series in the Fez-Meknes Region, Morocco. Climate10:64. doi: 10.3390/cli10050064

  • CrossRef
  • Google Scholar

• 66

  KhanduK.AwangeJ.ForootanE. (2016). An evaluation of high-resolution gridded precipitation products over Bhutan (1998–2012). Int. J. Climatol.36, 1067–1087. doi: 10.1002/joc.4402

  • CrossRef
  • Google Scholar

• 67

  KhatakhoR.TalchabhadelR.ThapaB. R. (2021). Evaluation of different precipitation inputs on streamflow simulation in Himalayan River basin. J. Hydrol.599:126390. doi: 10.1016/j.jhydrol.2021.126390

  • CrossRef
  • Google Scholar

• 68

  KiddC.BauerP.TurkJ.HuffmanG. J.JoyceR.HsuK. L.et al. (2012). Intercomparison of high-resolution precipitation products over Northwest Europe. J. Hydrometeorol.13, 67–83. doi: 10.1175/JHM-D-11-042.1

  • CrossRef
  • Google Scholar

• 69

  KiddC.BeckerA.HuffmanG. J.MullerC. L.JoeP.Skofronick-JacksonG.et al. (2017). So, how much of the Earth's surface is covered by rain gauges?Bull. Am. Meteorol. Soc.98, 69–78. doi: 10.1175/BAMS-D-14-00283.1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 70

  KurcS. A.SmallE. E. (2004). Dynamics of evapotranspiration in semiarid grassland and shrubland ecosystems during the summer monsoon season, central New Mexico. Water Resour. Res.40:W09305. doi: 10.1029/2004WR003068

  • CrossRef
  • Google Scholar

• 71

  LafonT.DadsonS.BuysG.PrudhommeC. (2012). Bias correction of daily precipitation simulated by a regional climate model: a comparison of methods. Int. J. Climatol.33, 1367–1381. doi: 10.1002/joc.3518

  • CrossRef
  • Google Scholar

• 72

  LiH.SheffieldJ.WoodE. F. (2010). Bias correction of monthly precipitation and temperature fields from Intergovernmental Panel on Climate Change AR4 models using equidistant quantile matching. J. Geophys. Res. Atmos.115:D10101. doi: 10.1029/2009JD012882

  • CrossRef
  • Google Scholar

• 73

  LoaizaJ. G.Rangel-PerazaJ. G.Sanhouse-GarcíaA. J.Monjardín-ArmentaS. A.Mora-FélixZ. D.Bustos-TerronesY. A. (2021). Assessment of water quality in a tropical reservoir in Mexico: seasonal, spatial and multivariable analysis. Int. J. Environ. Res. Public Health 18. doi: 10.3390/ijerph18147456

  • CrossRef
  • Google Scholar

• 74

  López-RamírezS. M.MayerA.SáenzL.Muñoz-VillersL. E.HolwerdaF.LookerN.et al. (2021). A comprehensive calibration and validation of SWAT-T using local datasets, evapotranspiration and streamflow in a tropical montane cloud forest area with permeable substrate in central Veracruz, Mexico. J. Hydrol.603:126781. doi: 10.1016/j.jhydrol.2021.126781

  • CrossRef
  • Google Scholar

• 75

  MainaF. Z.Siirila-WoodburnE. R.VahmaniP. (2020). Sensitivity of meteorological-forcing resolution on hydrologic variables. Hydrol. Earth Syst. Sci.24, 3451–3474. doi: 10.5194/hess-24-3451-2020

  • CrossRef
  • Google Scholar

• 76

  MascaroG.VivoniE. R.GochisD. J.WattsC. J.RodriguezJ. C. (2014). Temporal downscaling and statistical analysis of rainfall across a topographic transect in northwest Mexico. J. Appl. Meteorol. Climatol.53, 910–927. doi: 10.1175/JAMC-D-13-0330.1

  • CrossRef
  • Google Scholar

• 77

  MehdiB.SchulzK.LudwigR.FerberF.LehnerB. (2018). Evaluating the importance of non-unique behavioural parameter sets on surface water quality variables under climate change conditions in a mesoscale agricultural watershed. Water Resourc. Manag.32, 619–639. doi: 10.1007/s11269-017-1830-3

  • CrossRef
  • Google Scholar

• 78

  MeiX.SmithP. K.LiJ.LiB. (2022). Hydrological evaluation of gridded climate datasets in a Texas urban watershed using soil and water assessment tool and artificial neural network. Front. Environ. Sci.10:905774. doi: 10.3389/fenvs.2022.905774

  • CrossRef
  • Google Scholar

• 79

  MeinsF. M. (2013). Evaluation of spatial scale alternatives for hydrological modelling of the Lake Naivasha basin, Kenya. [Master's thesis]. University of Twente. Available online at: https://filetransfer.itc.nl/pub/naivasha/Meins2013.pdf (Accessed February 15, 2026).

  • Google Scholar

• 80

  MielnickP.DugasW. A.MitchellK.HavstadK. (2005). Long-term measurements of CO2 flux and evapotranspiration in a Chihuahuan desert grassland. J. Arid Environ.60, 423–436. doi: 10.1016/j.jaridenv.2004.06.001

  • CrossRef
  • Google Scholar

• 81

  MillerM. P.ButoS. G.SusongD. D.RumseyC. A. (2016). The importance of base flow in sustaining surface water flow in the Upper Colorado River Basin. Water Resour. Res.52, 3547–3562. doi: 10.1002/2015WR017963

  • CrossRef
  • Google Scholar

• 82

  Molina-NavarroE.AndersenH. E.NielsenA.ThodsenH.TrolleD. (2017). The impact of the objective function in multi-site and multi-variable calibration of the SWAT model. Environ. Model. Softw.93, 255–267. doi: 10.1016/j.envsoft.2017.03.018

  • CrossRef
  • Google Scholar

• 83

  Molina-NavarroE.Hallack-AlegríaM.Martínez-PérezS.Ramírez-HernándezJ.Mungaray-MoctezumaA.Sastre-MerlínA. (2016). Hydrological modeling and climate change impacts in an agricultural semiarid region. Case study: Guadalupe River basin, Mexico. Agric. Water Manag.175, 29–42. doi: 10.1016/j.agwat.2015.10.029

  • CrossRef
  • Google Scholar

• 84

  Montecelos-ZamoraY.CavazosT.KretzschmarT.VivoniE. R.CorzoG.Molina-NavarroE. (2018). Hydrological modeling of climate change impacts in a tropical river basin: a case study of the Cauto River, Cuba. Water10:1135. doi: 10.3390/w10091135

  • CrossRef
  • Google Scholar

• 85

  MonteithJ. L. (1965). Evaporation and environment. Symp. Soc. Exp. Biol., 19, 205–234

  • Pubmed Abstract
  • Google Scholar

• 86

  Monterroso-RivasA. I.Gómez-DíazJ. D. (2021). Impact of climate change on potential evapotranspiration and growing season in Mexico. Terra Latinoam. 39. doi: 10.28940/terra.v39i0.774

  • CrossRef
  • Google Scholar

• 87

  Morales-MartínezJ. L.Horta-RangelF. A.Segovia-DomínguezI.Robles-MoruaA.HernándezJ. H. (2019). Analysis of a new spatial interpolation weighting method to estimate missing data applied to rainfall records. Atmosfera32, 237–259. doi: 10.20937/ATM.2019.32.03.06

  • CrossRef
  • Google Scholar

• 88

  MoranP. A. P. (1950). Notes on continuous stochastic phenomena. Biometrika37, 17–23. doi: 10.1093/biomet/37.1-2.17

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 89

  Moreno-LeonM. A.Ortega-RubioA. (2009). First record of Craspedacusta sowerbyi Lankester, 1880 (Cnidaria: Limnomedusae: Olindiidae) in Mexico (Adolfo Lopez Mateos reservoir), with notes on their feeding habits and limnological dates. Biol. Invasions11, 1827–1834. doi: 10.1007/s10530-008-9361-8

  • CrossRef
  • Google Scholar

• 90

  MoriasiD. N.ArnoldJ. G.Van LiewM. W.BingnerR. L.HarmelR. D.VeithT. L. (2007). Model evaluation guidelines for systematic quantification of accuracy in watershed simulations. Trans. ASABE50, 885–900. doi: 10.13031/2013.23153

  • CrossRef
  • Google Scholar

• 91

  MoriasiD. N.GitauM. W.PaiN.DaggupatiP. (2015). Hydrologic and water quality models: performance measures and evaluation criteria. Trans. ASABE58, 1763–1785. doi: 10.13031/trans.58.10715

  • CrossRef
  • Google Scholar

• 92

  MucheM. E.SinnathambyS.ParmarR.KnightesC. D.JohnstonJ. M.WolfeK.et al. (2020). Comparison and evaluation of gridded precipitation datasets in a Kansas agricultural watershed using SWAT. J. Am. Water Resour. Assoc.56, 486–506. doi: 10.1111/1752-1688.12819

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 93

  Muñoz-VillersL. E.McDonnellJ. J. (2012). Runoff generation in a steep, tropical montane cloud forest catchment on permeable volcanic substrate. Water Resour. Res.48:W09528. doi: 10.1029/2011WR011316

  • CrossRef
  • Google Scholar

• 94

  NashJ. E.SutcliffeJ. V. (1970). River flow forecasting through conceptual models part I — A discussion of principles. J. Hydrol.10, 282–290. doi: 10.1016/0022-1694(70)90255-6

  • CrossRef
  • Google Scholar

• 95

  NathN. K.DasP.MishraL. R.KumarA.SuryawanshiS. L.GautamV. K. (2024). “Hydrological modeling and simulation for water resource assessment,” in Integrated Mmanagement of Water Resources in India: A Computational Approach. Water Science and Technology Library, Vol. 129, eds. A. K. Yadav, K. Yadav, and V. P. Singh (Cham: Springer Nature Switzerland), 43–58. doi: 10.1007/978-3-031-62079-9_3

  • CrossRef
  • Google Scholar

• 96

  Navarro-CéspedesJ. M.HernándezJ. H.Alcántara-ConcepciónP. C.Morales-MartínezJ. L.Carreño-AguileraG.PadillaF. (2023). A comparison of missing value imputation methods applied to daily precipitation in a semi-arid and a humid region of Mexico. Atmósfera37, 33–52. doi: 10.20937/ATM.53095

  • CrossRef
  • Google Scholar

• 97

  NeitschS. L.ArnoldJ. G.KiniryJ. R.WilliamsJ. R. (2011). Soil and Water Assessment Tool Theoretical Documentation Version 2009. Texas Water Resources Institute Technical Report No. 406. Available online at: https://swat.tamu.edu/media/99192/swat2009-theory.pdf (Accessed February 15, 2026).

  • Google Scholar

• 98

  NesbittS. W.GochisD. J.LangT. J. (2008). The diurnal cycle of clouds and precipitation along the Sierra Madre Occidental observed during NAME-2004: implications for warm season precipitation estimation in complex terrain. J. Hydrometeorol.9, 728–743. doi: 10.1175/2008JHM939.1

  • CrossRef
  • Google Scholar

• 99

  NguyenC.VazeJ.MateoC. M. R.HutchinsonM. F.TengJ. (2026). Elevation dependent spatial interpolation of hourly rainfall for accurate flood inundation modelling. Hydrol. Earth Syst. Sci.30, 45–66. doi: 10.5194/hess-30-45-2026

  • CrossRef
  • Google Scholar

• 100

  NguyenN. Y.AnhT. N.NguyenH. D.DangD. K. (2024). Quantile mapping technique for enhancing satellite-derived precipitation data in hydrological modelling: a case study of the Lam River Basin, Vietnam. J. Hydroinform.26, 2026–2044. doi: 10.2166/hydro.2024.225

  • CrossRef
  • Google Scholar

• 101

  NortonJ. (2015). An introduction to sensitivity assessment of simulation models. Environ. Model. Softw.69, 166–174. doi: 10.1016/j.envsoft.2015.03.020

  • CrossRef
  • Google Scholar

• 102

  OudinL.AndréassianV.MathevetT.PerrinC.MichelC. (2006). Dynamic averaging of rainfall-runoff model simulations from complementary model parameterizations. Water Resour. Res.42:W07410. doi: 10.1029/2005WR004636

  • CrossRef
  • Google Scholar

• 103

  PaulhusJ. L. H.KohlerM. A. (1952). Interpolation of missing precipitation records. Monthly Weather Rev.80, 129–133. doi: 10.1175/1520-0493(1952)080<0129:IOMPR>2.0.CO;2

  • CrossRef
  • Google Scholar

• 104

  Pérez-MorgaN.KretzschmarT.CavazosT.SmithS. V.Munoz-ArriolaF. (2013). Variability of extreme precipitation in coastal river basins of the southern Mexican Pacific region. Geofís. Int.52, 277–291. Available online at: http://www.scielo.org.mx/scielo.php?script=sci_arttext&pid=S0016-71692013000300006&lng=es&tlng=en (Accessed February 15, 2026).

  • Google Scholar

• 105

  PfannerstillM.GuseB.FohrerN. (2014). Smart low flow signature metrics for an improved overall performance evaluation of hydrological models. J. Hydrol.510, 447–458. doi: 10.1016/j.jhydrol.2013.12.044

  • CrossRef
  • Google Scholar

• 106

  PianiC.WeedonG.BestM.GomesS.ViterboP.HagemannS.et al. (2010). Statistical bias correction of global simulated daily precipitation and temperature for the application of hydrological models. J. Hydrol.395, 199–215. doi: 10.1016/j.jhydrol.2010.10.024

  • CrossRef
  • Google Scholar

• 107

  PianosiF.BevenK.FreerJ.HallJ. W.RougierJ.StephensonD. B.et al. (2016). Sensitivity analysis of environmental models: a systematic review with practical workflow. Environ. Model. Softw.79, 214–232. doi: 10.1016/j.envsoft.2016.02.008

  • CrossRef
  • Google Scholar

• 108

  PushpalathaR.PerrinC.Le MoineN.AndréassianV. (2012). A review of efficiency criteria suitable for evaluating low-flow simulations. J. Hydrol. 420–421, 171–182. doi: 10.1016/j.jhydrol.2011.11.055

  • CrossRef
  • Google Scholar

• 109

  Quevedo-CastroA.Rangel-PerazaJ. G.BandalaE.Amabilis-SosaL.Rodríguez-MataA.Bustos-TerronesY. (2018). Developing a water quality index in a tropical reservoir using a measure of multiparameters. J. Water Sanit. Hyg. Dev.8, 752–766. doi: 10.2166/washdev.2018.049

  • CrossRef
  • Google Scholar

• 110

  RayR. L.SishodiaR. P.TeferaG. W. (2022). Evaluation of gridded precipitation data for hydrologic modeling in north-central Texas. Remote Sens.14:3860. doi: 10.3390/rs14163860

  • CrossRef
  • Google Scholar

• 111

  RitterA.Muñoz-CarpenaR. (2013). Performance evaluation of hydrological models: statistical significance for reducing subjectivity in goodness-of-fit assessments. J. Hydrol.480, 33–45. doi: 10.1016/j.jhydrol.2012.12.004

  • CrossRef
  • Google Scholar

• 112

  Robles-MoruaA.CheD.MayerA. S.VivoniE. R. (2015). Hydrological assessment of proposed reservoirs in the Sonora River Basin, Mexico, under historical and future climate scenarios. Hydrol. Sci. J.60, 50–66. doi: 10.1080/02626667.2013.878462

  • CrossRef
  • Google Scholar

• 113

  Robles-MoruaA.VivoniE. R.MayerA. S. (2012). Distributed hydrologic modeling in northwest Mexico reveals the links between runoff mechanisms and evapotranspiration. J. Hydrometeorol.13, 785–807. doi: 10.1175/JHM-D-11-0112.1

  • CrossRef
  • Google Scholar

• 114

  RumseyC. A.MillerM. P.SusongD. D.TillmanF. D.AnningD. W. (2015). Regional scale estimates of baseflow and factors influencing baseflow in the Upper Colorado River Basin. J. Hydrol. Reg. Stud.4, 91–107. doi: 10.1016/j.ejrh.2015.04.008

  • CrossRef
  • Google Scholar

• 115

  Sánchez-GómezA.SchürzC.Molina-NavarroE.BiegerK. (2024). Groundwater modelling in SWAT+: considerations for a realistic baseflow simulation. Groundwater Sustain. Dev.26:101275. doi: 10.1016/j.gsd.2024.101275

  • CrossRef
  • Google Scholar

• 116

  Sánchez-GómezA.Martínez-PérezS.Pérez-ChaveroF. M.Molina-NavarroE. (2022). Optimization of a SWAT model by incorporating geological information through calibration strategies. Optim. Eng.23, 2203–2233. doi: 10.1007/s11081-022-09744-1

  • CrossRef
  • Google Scholar

• 117

  SangL.ZhuG.QiuD.ZhangZ.LiuY.ZhaoK.et al. (2022). Spatial variability of runoff recharge sources and influence mechanisms in an arid mountain flow-producing zone. Hydrol. Process.36:e14642. doi: 10.1002/hyp.14642

  • CrossRef
  • Google Scholar

• 118

  SantosL.ThirelG.PerrinC. (2018). Technical note: Pitfalls in using log-transformed flows within the KGE criterion. Hydrol. Earth Syst. Sci.22, 4583–4591. doi: 10.5194/hess-22-4583-2018

  • CrossRef
  • Google Scholar

• 119

  SaxtonK. E.RawlsW. J. (2006). Soil water characteristic estimates by texture and organic matter for hydrologic solutions. Soil Sci. Soc. Am. J.70, 1569–1578. doi: 10.2136/sssaj2005.0117

  • CrossRef
  • Google Scholar

• 120

  SchürzC.HollosiB.MatullaC.PresslA.ErtlT.SchulzK.et al. (2019). A comprehensive sensitivity and uncertainty analysis for discharge and nitrate-nitrogen loads involving multiple discrete model inputs under future changing conditions. Hydrol. Earth Syst. Sci.23, 1211–1244. doi: 10.5194/hess-23-1211-2019

  • CrossRef
  • Google Scholar

• 121

  ServatE.DezetterA. (1991). Selection of calibration objective fonctions in the context of rainfall-runoff modelling in a Sudanese savannah area. Hydrol. Sci. J.36, 307–330. doi: 10.1080/02626669109492517

  • CrossRef
  • Google Scholar

• 122

  Servicio Geológico Mexicano (SGM) (1999). Carta Geológico-Minera Pericos, G13-7 Sinaloa, Durango y Chihuahua, Esc. 1:250,000. 1st Edition, Servicio Geológico Mexicano, Pachuca, Hidalgo. Mexico. Available online at: https://mapserver.sgm.gob.mx/Cartas_Online/metadatos_geol/39_GL_pericos.html (Accessed February 15, 2026).

  • Google Scholar

• 123

  Servicio Meteorológico Nacional (SMN) (2026). Información estadística climatológica. Available online at: https://smn.conagua.gob.mx/es/climatologia/informacion-climatologica/informacion-estadistica-climatologica (Accessed February 15, 2026).

  • Google Scholar

• 124

  ShamirE.Mendoza-FierroL.Mohsenzadeh-KarimiS.PelakN.TarouillyE.ChangH. I.et al. (2024). Climate change projections of potential evapotranspiration for the North American Monsoon region. Hydrology11:83. doi: 10.3390/hydrology11060083

  • CrossRef
  • Google Scholar

• 125

  ShresthaS.ShresthaM.ShresthaP. K. (2018). Evaluation of the SWAT model performance for simulating river discharge in the Himalayan and tropical basins of Asia. Hydrol. Res.49, 846–860. doi: 10.2166/nh.2017.189

  • CrossRef
  • Google Scholar

• 126

  ShuaiP.ChenX.MitalU.CoonE. T.DwivediD. (2022). The effects of spatial and temporal resolution of gridded meteorological forcing on watershed hydrological responses. Hydrol. Earth Syst. Sci.26, 2245–2276. doi: 10.5194/hess-26-2245-2022

  • CrossRef
  • Google Scholar

• 127

  StrupczewskiW.KundzewiczZ. (1980). Muskingum method revisited. J. Hydrol.48, 327–342. doi: 10.1016/0022-1694(80)90124-9

  • CrossRef
  • Google Scholar

• 128

  TapasM. R.EtheridgeR.TranT. N. D.LeM. H.HinckleyB.NguyenV. T.et al. (2025). Evaluating combinations of rainfall datasets and optimization techniques for improved hydrological predictions using the SWAT+ model. J. Hydrol. Reg. Stud.57:102134. doi: 10.1016/j.ejrh.2024.102134

  • CrossRef
  • Google Scholar

• 129

  TapiadorF. J.TurkF. J.PetersenW.HouA. Y.García-OrtegaE.MachadoL. A. T.et al. (2012). Global precipitation measurement: methods, datasets and applications. Atmos. Res. 104–105, 70–97. doi: 10.1016/j.atmosres.2011.10.021

  • CrossRef
  • Google Scholar

• 130

  TengJ.PotterN. J.ChiewF. H. S.ZhangL.WangB.VazeJ.et al. (2015). How does bias correction of regional climate model precipitation affect modelled runoff?Hydrol. Earth Syst. Sci.19, 711–728. doi: 10.5194/hess-19-711-2015

  • CrossRef
  • Google Scholar

• 131

  ThorntonP. E.ShresthaR.ThorntonM.KaoS.-C.WeiY.WilsonB. E. (2021). Gridded daily weather data for North America with comprehensive uncertainty quantification. Sci. Data8:190. doi: 10.1038/s41597-021-00973-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 132

  ThorntonM. M.DevarakondaR. (2011). Daymet Single Pixel Extraction Tool (Version 1). Oak Ridge: ORNL Distributed Active Archive Center. doi: 10.3334/ORNLDAAC/2361

  • CrossRef
  • Google Scholar

• 133

  TolsonB. A.ShoemakerC. A. (2007). Dynamically dimensioned search algorithm for computationally efficient watershed model calibration. Water Resour. Res.43:W01413. doi: 10.1029/2005WR004723

  • CrossRef
  • Google Scholar

• 134

  U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research. (2025a). About SWAT+. SWAT+ Installation and Help. Available online at: https://swatplus.gitbook.io/docs (Accessed March 20, 2026).

  • Google Scholar

• 135

  U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research. (2025b). Climate. SWAT+ Documentation. Available online at: https://swatplus.gitbook.io/io-docs/introduction-1/climate (Accessed March 20, 2026).

  • Google Scholar

• 136

  U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research. (2026a). Climate. SWAT+ Installation and Help. Available online at: https://swatplus.gitbook.io/docs/user/editor/inputs/climate (Accessed March 20, 2026).

  • Google Scholar

• 137

  U.S. Department of Agriculture (USDA)–Agricultural Research Service (ARS) and Texas A&M AgriLife Research. (2026b). SWAT+ Editor Documentation. SWAT+ Installation and Help. Available online at: https://swatplus.gitbook.io/docs/user/editor (Accessed March 23, 2026).

  • Google Scholar

• 138

  U.S. Department of Agriculture (USDA)–Natural Resources Conservation Service (NRCS). (2009). National Engineering Handbook, Part 630 Hydrology, Chapter 7, Hydrologic Soil Groups.Washington, DC: USDA.

  • Google Scholar

• 139

  U.S. Department of Agriculture (USDA)–Soil Conservation Service (SCS). (1972). National Engineering Handbook, Section 4, Hydrology.Washington, DC: USDA. p. 762.

  • Google Scholar

• 140

  Van LiewM. W.VeithT. L.BoschD. D.ArnoldJ. G. (2007). Suitability of SWAT for the conservation effects assessment project: comparison on USDA agricultural research service watersheds. J. Hydrol. Eng.12, 173–189. doi: 10.1061/(ASCE)1084-0699(2007)12:2(173)

  • CrossRef
  • Google Scholar

• 141

  VivoniE. R.MorenoH. A.MascaroG.RodriguezJ. C.WattsC. J.Garatuza-PayanJ.et al. (2008). Observed relation between evapotranspiration and soil moisture in the North American monsoon region. Geophys. Res. Lett.35:L22403. doi: 10.1029/2008GL036001

  • CrossRef
  • Google Scholar

• 142

  VivoniE. R.RodríguezJ. C.WattsC. J. (2010). On the spatiotemporal variability of soil moisture and evapotranspiration in a mountainous basin within the North American monsoon region. Water Resour. Res.46:W02509. doi: 10.1029/2009WR008240

  • CrossRef
  • Google Scholar

• 143

  VuM. T.RaghavanS. V.LiongS. Y. (2012). SWAT use of gridded observations for simulating runoff - A Vietnam river basin study. Hydrol. Earth Syst. Sci.16, 2801–2811. doi: 10.5194/hess-16-2801-2012

  • CrossRef
  • Google Scholar

• 144

  WilliamsJ. R. (1995). “Chapter: 25. The EPIC model” in Computer Models of Watershed Hydrology, ed. V. P. Singh (Highlands Ranch, CO: Water Resources Publications), 909–1000.

  • Google Scholar

• 145

  WinterT. C. (2007). The role of ground water in generating streamflow in headwater areas and in maintaining base flow. J. Am. Water Resour. Assoc.43, 15–25. doi: 10.1111/j.1752-1688.2007.00003.x

  • CrossRef
  • Google Scholar

• 146

  YilmazK. K.GuptaH. V.WagenerT. (2008). A process-based diagnostic approach to model evaluation: application to the NWS distributed hydrologic model. Water Resour. Res. 44:W09417. doi: 10.1029/2007WR006716

  • CrossRef
  • Google Scholar

• 147

  YimerE. A.BaileyR. T.PiepersL. L.NossentJ.Van GriensvenA. (2023). Improved representation of groundwater–surface water interactions using SWAT+gwflow and modifications to the gwflow module. Water15:3249. doi: 10.3390/w15183249

  • CrossRef
  • Google Scholar

• 148

  YooC.ParkC.YoonJ.KimJ. (2014). Interpretation of mean-field bias correction of radar rain rate using the concept of linear regression. Hydrol. Process.28, 5081–5092. doi: 10.1002/hyp.9972

  • CrossRef
  • Google Scholar

• 149

  ZhangD.ChenX.YaoH.LinB. (2015). Improved calibration scheme of SWAT by separating wet and dry seasons. Ecol. Model.301, 54–61. doi: 10.1016/j.ecolmodel.2015.01.018

  • CrossRef
  • Google Scholar

• 150

  ZhangL.XinZ.ZhangC.SongC.ZhouH. (2022). Exploring the potential of satellite precipitation after bias correction in streamflow simulation in a semi-arid watershed in northeastern China. J. Hydrol. Reg. Stud.43:101192. doi: 10.1016/j.ejrh.2022.101192

  • CrossRef
  • Google Scholar

• 151

  ZhangY.YeA.NguyenP.AnaluiB.SorooshianS.HsuK. (2021). Error characteristics and scale dependence of current satellite precipitation estimates products in hydrological modeling. Remote Sens.13:3061. doi: 10.3390/rs13163061

  • CrossRef
  • Google Scholar