Hydrological modeling in a Brazilian tropical headwater basin using gauge and satellite rainfall datasets

Abstract

Introduction:

In regions with complex topography and sparse monitoring networks, such as tropical headwater basins, uncertainties in precipitation data can significantly affect hydrological modeling. This study evaluates the Distributed Hydrological Model performance of the National Institute for Space Research using two rainfall inputs. The inputs are rain gauge observations and the MERGE satellite precipitation product, in the Upper Grande River Basin, Brazil.

Methods:

Hydrological simulations were conducted for the period 2000–2017, including warm-up, calibration, and validation phases, and analyzed across ten sub-basins with contrasting drainage areas and slope conditions. Model performance was assessed using multiple statistical metrics and flow-duration curves.

Results and discussion:

Model performance was satisfactory for both rainfall datasets, with average NSE and KGE values ranging from 0.60 to 0.78 and 0.66 to 0.84, respectively. Higher performance was generally observed in larger sub-basins characterized by gentler slopes. While simulations driven by rain gauge data showed superior results during calibration, satellite-based precipitation inputs performed comparably or better during validation, particularly in reproducing low flow conditions. These findings demonstrate the potential of satellite rainfall products to support water resources assessment and decision-making in data-scarce tropical regions, especially in basins of strategic importance for hydropower generation and water supply.

1 Introduction

Hydrological prediction plays a fundamental role in water resources management, particularly in regions affected by increasing climatic variability and growing water demands (Siddique et al., 2023; Souza et al., 2025; Pinto et al., 2026). In this context, hydrological models have become essential tools for understanding, representing, and predicting hydrological processes under different environmental conditions (Alvarenga et al., 2016, 2020; Melo et al., 2022a; Oliveira et al., 2023). Among these models, the Distributed Hydrological Model developed by the National Institute for Space Research (MHD-INPE) is particularly noteworthy (Rodriguez and Tomasella, 2016). Previous studies have successfully applied MHD-INPE to hydrological analyses and simulations (Melo et al., 2022b, 2024), assessments of climate change impacts on hydrological dynamics (Alvarenga et al., 2026), and investigations of hydrological regime alterations associated with land use and land cover changes (Rodriguez and Tomasella, 2016; Pinto et al., 2026). Despite these advances, limited attention has been given to evaluating how different rainfall datasets influence MHD-INPE performance.

Rainfall and drainage network information are key inputs for hydrological models (Devi et al., 2015; Jian et al., 2025). Among these, rainfall represents one of the main sources of uncertainty in hydrological simulations, particularly regarding its spatial variability (Bárdossy et al., 2022; Costa et al., 2024; Alvarenga et al., 2025). In mountainous and poorly monitored regions, conventional rain gauge networks often fail to adequately represent the spatial heterogeneity of precipitation, which can compromise streamflow predictions. In this context, satellite-based rainfall products have emerged as valuable alternatives for complementing or replacing observed data in hydrological applications (Collischonn et al., 2008; Cassalho et al., 2020; Rozante et al., 2010; Nascimento et al., 2022; Romero-Hernández et al., 2024; Mantovani et al., 2025; Huang et al., 2025).

Rainfall estimates derived from satellite imagery, such as those provided by the Tropical Rainfall Measuring Mission (TRMM) and its successor, the Global Precipitation Measurement (GPM) mission, serve as valuable inputs for hydrological models. However, limitations remain in tropical mountainous regions. In areas with complex topography, such as the Mantiqueira Range in the Upper Grande River Basin, satellite-based rainfall products may underestimate localized convective and orographic precipitation. Moreover, these estimates are not direct observations and are subject to uncertainties associated with sampling methods, retrieval algorithms, sensor characteristics, and spatial resolution (Huang et al., 2025). Their performance also depends on the density and spatial distribution of rain gauges used for bias correction. Consequently, in headwater basins with sparse monitoring networks, uncertainties may persist, particularly during extreme rainfall events and short-duration precipitation. Previous studies have reported reduced accuracy of satellite precipitation products in steep terrain due to the difficulty of representing highly heterogeneous rainfall patterns and cloud microphysical processes in mountainous tropical environments (Rozante et al., 2020; Romero-Hernández et al., 2024). These limitations should be considered when interpreting hydrological simulations driven by satellite rainfall data, especially in smaller sub-basins with steeper slopes. Therefore, further investigations of satellite-based rainfall products for hydrological applications remain essential, particularly in regions with limited observational networks (Falck et al., 2015, 2018; Junqueira et al., 2022; Silva et al., 2022).

The Upper Grande River Basin is a strategic region for water resources management in southeastern Brazil, supporting hydropower generation, agriculture, and urban water supply. Due to its complex topography, strong spatial rainfall variability, and sparse monitoring network in headwater areas (0.0013 station per km²), with gaps in the time series, improving precipitation inputs for hydrological modeling remains a major challenge in the basin.

A previous study by Falck et al. (2015) in the Tocantins Basin (a basin of 764,000 km² located in the center-north region of Brazil) applied a two-dimensional stochastic error model to satellite-based precipitation products in streamflow simulations of the MHD-INPE model, demonstrating that the error model improved MHD-INPE simulations. Additionally, streamflow errors decreased with the catchment area for satellite products, revealing the importance of the basin scale and its relief for satellite rainfall estimates.

Besides this study, no other MHD-INPE model application has investigated how different rainfall datasets affect distributed hydrological simulations, especially in headwater sub-basins with contrasting drainage areas and more complex terrain compared to the Tocantins river, such as those located in southeastern Brazil.

We hypothesize that satellite-based rainfall products can provide hydrological simulations comparable to those obtained using rain gauge observations in larger sub-basins, whereas lower performance is expected in smaller and steeper headwater sub-basins due to stronger topographic effects, higher spatial rainfall variability, and sparse monitoring networks. Therefore, this study evaluates streamflow simulations generated by the MHD-INPE model using rain gauge observations and satellite rainfall estimates from the MERGE product in the Upper Grande River Basin. By comparing model performance under different rainfall inputs, sub-basin characteristics, and flow conditions, this study advances the understanding of the applicability of satellite-based rainfall products for distributed hydrological modeling in tropical headwater basins with complex terrain.

2 Material and methods

2.1 Experimental area

The Upper Grande River Basin (UGRB) is part of the Grande River Basin, which drains into the Paraná River Basin, the largest hydrographic region in Brazil for hydroelectric power generation. The three main sub-basins in the UGRB are the Grande River Headwater (GRA), Aiuruoca River (AIU), and Capivari River (CAP); for this study, ten sub-basins were identified in the GRA, AIU, and CAP. The total drainage area of the UGRB is 8,758 km² (Figure 1). The headwater region of the UGRB, located in the south, is characterized by rugged terrain, with elevations ranging from 798 to 2,653 m. The drainage area, average slope, and average annual rainfall of the sub-basins in the UGRB range from 274 to 8,057 km², 13 to 38%, and 1,225.70 to 1,655.08 mm, respectively (Table 1). The UGRB is located in the southern and southeastern regions of the state of Minas Gerais, Brazil, and directly supplies the Camargos and Itutinga hydropower plants, highlighting its strategic importance for hydropower generation and regional water resources management.

Figure 1

According to the Köppen-Geiger climate classification (Beck et al., 2018), the climate in the region near the Mantiqueira Range, in the southern part of the basin, is Cwb, with an average annual rainfall of 2,100 mm and an average temperature of 15°C. In the northern basin, near the Campo das Vertentes region, the climate is classified as Cwa, with an average annual rainfall of 1,500 mm and an average temperature of 18°C. The Cwb climate is characterized by humid, mild summers and dry, cold winters. In contrast, the Cwa climate is distinguished by hotter summers. The rainy season begins in October, with the hydrological year defined from October to September of the following year (Oliveira et al., 2017).

2.2 Distributed Hydrological Model developed by the National Institute for Space Research

The MHD-INPE developed by the National Institute for Space Research (INPE) (Rodriguez and Tomasella, 2016) is a regular grid-cell model that uses hydrological response units in each cell as a function of land use/cover and soil characteristics. Rainfall is interpolated using the inverse square of the distance, considering the altitude in each pixel. In each cell, exchanges of matter and energy between the surface and the atmosphere, as well as storage variations under both saturated and unsaturated conditions are simulated alongside horizontal flows. Horizontal flows are estimated based on the topographic index of the grid cell. Evapotranspiration is estimated using the Penman-Monteith equation while in-cell propagation is performed using the simple linear reservoir method. Flow propagation among cells is based on the connectivity and direction between cells, which is determined by the drainage basin network using the Muskingum-Cunge method. Additional details on the hydrological model can be found in Rodriguez and Tomasella (2016).

2.3 Satellite estimates and observational data of rainfall

The rainfall data obtained from the MERGE satellite product were developed by the Center for Weather Forecast and Climate Studies (CPTEC/INPE) to provide spatially consistent daily rainfall fields by combining satellite-based estimates with ground-based observations (Rozante et al., 2020). This blended approach reduces uncertainties associated with sparse rain gauge networks, particularly in regions with complex terrain or limited observational coverage. MERGE has been widely used in hydrological and climatological studies across Brazil, supporting analyses of extreme rainfall events, water resource management, and regional climate monitoring (Cassalho et al., 2020; Silva et al., 2022). Its application in this study is justified by its proven ability to capture the spatiotemporal variability of rainfall with greater accuracy than products based solely on satellite estimates or rain gauge observations, especially in tropical regions such as South America (Rozante et al., 2020). Moreover, a recent study of Menezes et al. (2026) concluded that MERGE is the most robust alternative for precipitation studies in Brazil when compared to other well-known operational rainfall products.

In addition to MERGE, observed rainfall data were obtained from rain gauge stations available on the HIDROWEB platform of the Brazilian National Water Agency (ANA) (Figure 1B). These two rainfall datasets—ground-based observations and MERGE satellite-derived estimates—constitute the different rainfall inputs evaluated in this research. Therefore, the only varying component in the hydrological model inputs was the rainfall dataset.

2.4 Input data and performance analysis of the MHD-INPE hydrological model

Hydrological simulations using the MHD-INPE require spatial information on topography, soil class, land use and cover, meteorological data, and model parameters as inputs. During the calibration and validation periods, observed streamflow data from each of the ten sub-basins were used to evaluate the performance of the hydrological model. Observed streamflow records were acquired from the Brazilian National Water Agency (Agência Nacional de Águas e Saneamento Básico (ANA), 2024).

Meteorological data for the region were obtained from conventional meteorological stations monitored by the National Institute of Meteorology (Instituto Nacional de Meteorologia (INMET), 2022), as shown in Figure 1. The meteorological inputs included rainfall, mean air temperature, dew point temperature, 10 m wind speed, atmospheric pressure, and incident global radiation.

The Digital Elevation Model (DEM), obtained from SRTM satellite images with a spatial resolution of 30 m, was used to delimit the sub-basins. Soil classes were derived from hydrological environments, i.e., a map that relates the altitude of each point to the altitude of the nearest drainage network, generated by the Height Above the Nearest Drainage model (HAND). The soil classes considered are as follows: the valley class (0 to 15 m)—hydromorphic, which occurs in areas with a predominance of Neosols and silty clay texture; the intermediate class (15 to 80 m), found in areas with a predominance of Latosol and clay texture; and the plateau class (above 80 m)—non-hydromorphic, which occurs in areas with a predominance of Cambisol and sandy clay loam texture (Figure 2A).

Figure 2

Annual land use and cover maps were extracted from the MapBiomas Brasil (2022), with a spatial resolution of approximately 30 m. This project created mosaics with data collected throughout the year, which were used as a source for image classification. The automatic classifier was calibrated with samples obtained from reference maps, previous classifications, or visual analysis of the images (Souza et al., 2020). The main vegetation classes found in the watershed include pasture, agriculture, forest, grassland, and silviculture (Figure 2B).

The fixed soil and vegetation parameters are detailed in Table 2. Soil parameters were defined based on soil characteristics and texture classification (Soil Survey Staff, 1951), along with studies carried out in the region (Junqueira Júnior et al., 2008). The vegetation parameters were defined according to the characteristics of the Atlantic Forest, pasture and the main crops cultivated in the region (Wright et al., 1996; Freitas et al., 2012; Terra et al., 2015). The parameters that change during model calibration include: Thickness of the upper layer (m)-D₁, Thickness of the intermediate layer (m)-D₂, Thickness of the bottom layer (m)-D₃, Saturated soil hydraulic conductivity-fK_ss, Maximum transmissivity of the bottom layer (m² day⁻¹)-T_sub, Decay of transmissivity with the thickness of the saturated zone (m day⁻¹ m⁻¹)-μ, soil anisotropy-α, minimum soil storage capacity (%)-ξ; Routing water storage parameter for surface and subsurface flows (s)-C_sup; and Routing water storage parameter for baseflow (s)-C_sub.

The MHD-INPE model was calibrated using two rainfall datasets and measured streamflow. The warm-up, calibration, and validation periods were from October 2000 to September 2002, October 2002 to September 2014, and October 2014 to September 2017, respectively. The SPEA2 algorithm (Zitzler et al., 2001) was employed for multi-objective parameter optimization considering the Nash-Sutcliffe efficiency coefficient (NSE) and the logarithmic Nash-Sutcliffe efficiency coefficient (LNSE). For each rainfall dataset, hydrographs were analyzed by comparing the simulated and observed streamflow, focusing on their temporal behavior, including peak and minimum streamflow. Streamflow at the 10% (Q10%) and 90% (Q90%) exceedance frequencies, commonly used as indicators of extreme events, were also evaluated in this study.

To assess the goodness of fit between observed and simulated streamflow, five performance coefficients were used: i. the Coefficient of Determination (R²), which evaluates the variance between simulated and observed streamflow; ii. the Nash-Sutcliffe Efficiency coefficient (NSE) and log-Nash-Sutcliffe Efficiency (LNSE), which quantify the residual variance relative to the data variation. These coefficients indicate how closely the simulated data align with a 1:1 line compared to the observed data, with values ranging from –∞ to 1, where 1 represents an optimal fit (Nash and Sutcliffe, 1970); iii. the PBIAS, which quantifies the average tendency of the difference between simulated and observed streamflow; iv. the Kling-Gupta Efficiency (KGE), which decomposes the NSE into three components: correlation coefficient (r), bias (β), and variability (α) (Gupta et al., 2009). Figure 3 summarizes the methodological procedure used to evaluate the influence of different rainfall datasets on hydrological model performance.

Figure 3

3 Results

The most sensitive parameters of the MHD-INPE model are shown in Table 3A and Table 3B. These final values are related to the surface, groundwater, and soil phases of the hydrological cycle and were subsequently used for model calibration. The largest amplitudes were observed for the following parameters: saturated soil hydraulic conductivity (fKss), decay of transmissivity with the thickness of the saturated zone (μ), soil anisotropy (α), and the routing water storage parameter for baseflow (Csub) (Table 3A and Table 3B). The parameters that exhibited the greatest variation among the ten sub-basins are those associated with soil water infiltration and baseflow, indicating their significant influence on runoff generation.

The observed and simulated hydrographs, obtained from rain gauge and MERGE rainfall data, for each sub-basin are shown in Figure 4 for the calibration and validation periods and for the period from October 2013 to September 2015, in the Supplementary Material (Supplementary Figure 1). In each sub-basin, there was good agreement between observed and simulated streamflow. Although some discrepancies were noted in peak streamflow values, it was observed that the daily streamflow simulations, based on satellite rainfall estimates, were comparable to those obtained from rain gauges. In general, streamflow data simulated by MHD-INPE were closer to observed values during dry months. This is particularly relevant for analyzing water availability in regions focused on hydroelectric power generation, as simulations that accurately capture streamflow during dry periods are crucial.

Figure 4

Table 4A and Table 4B summarize that, overall, there is a good agreement between observed and simulated streamflow in the ten sub-basins. Positive/negative PBIAS values indicate overestimate/underestimate of the streamflow. On average, the results overestimate the streamflow, except in the daily validation with the MERGE dataset input. The model's performance across all sub-basins is satisfactory during calibration and validation periods, with NSE, LNSE and KGE exceeding 0.60, except in the AIU1, AIU2, AIU3 and AIU4 sub-basins, where we found the lowest values for these coefficients. These sub-basins are located in headwater regions with a low density of observed rainfall data. More details about the sub-basin AIU2 are available in the Supplementary Material (Supplementary Figure 2).

The influence of orographic effects on rainfall should be highlighted as an additional difficulty in rainfall-runoff modeling in Brazilian headwater basins. In agreement with Falck et al. (2018), another potential source of uncertainty is the extrapolation of peak flows from the rating curves, particularly in headwater basins, where measuring streamflow during severe floods is logistically challenging. These aspects explain the lower performance of the hydrological model for AIU1, AIU2, AIU3 and AIU4 sub-basins (Table 4A and Table 3B). Therefore, the model's performance was better in the larger sub-basins than small sub-basins (Table 1), also due to the fact that smaller areas are more affected by topographic complexity, which increases the spatial-temporal variability of rainfall.

The authors Falck et al. (2018) and Melo et al. (2022b) obtained NSE and LNSE values ranging from 0.249 to 0.892 and 0.68 to 0.88, respectively, in calibration and validation periods of the MHD-INPE. According to Moriasi et al. (2015), NSE and LNSE above 0.5 are considered satisfactory for hydrological modeling using continuous simulation. Thus, the calibration and validation model performance in this study were satisfactory, considering the two rainfall input datasets.

Limitations were observed in the hydrological modeling of the smaller, more rugged sub-basins, as also noted by Falck et al. (2018) and Cassalho et al. (2020), when using rainfall data obtained from satellite products. On average, simulated streamflow (rain gauge) showed better results than the simulated streamflow (MERGE) in the calibration, however, the opposite occurs in the validation. These results indicate that rainfall obtained from satellite products is a good alternative for hydrological modeling, especially in regions with scarce rainfall datasets. Furthermore, satellite rainfall estimates can be a practical tool for identifying damaged rain gauges at a basin scale. Considering the difficulty in obtaining rainfall inputs necessary for distributed hydrological models, the results obtained in this study are of great importance for future MHD-INPE users, as well as for studies in regions with complex topography in southeastern Brazil, especially in the headwater sub-basins.

The daily streamflow variability is depicted in the flow-duration curves (Figure 5). Compared to observed streamflow, both simulated curves exhibit a flatter slope, indicating the presence of significant natural streamflow regulation. The main differences between observed and simulated flows occur at the extremes (Q10% and Q90%). Overall, the results indicate better performance in the simulation of high flows (Q10%) than low flows (Q90%).

Figure 5

For the rainfall input based on rain gauge observations, the simulated streamflow overestimation/underestimation (%) at the 10% exceedance probability (Q10%) was −2.7, + 0.9, + 1.5, + 1.9, + 0.8, + 2.2, + 6.4, −6.1, + 0.9, and + 3.0% for the GRA1, GRA2, AIU1, AIU2, AIU3, AIU4, AIU5, CAP1, CAP2, and CAM sub-basins, respectively (Table 5). At the 90% exceedance probability (Q90%), the corresponding values were −7.5%, + 6.1%, −16.1%, −1.6%, + 19.9%, + 8.2%, 0.0%, + 20.1%, + 15.2%, and + 5.9%. For the MERGE rainfall input, the simulated streamflow overestimation/underestimation (%) at Q10% was −9.9%, −3.8%, −5.7%, −10.1%, + 0.9%, + 0.3%, −1.8%, + 1.4%, + 1.6%, and + 0.2% across the same sub-basins. At Q90%, the respective values were + 8.0%, −4.6%, −18.4%, −12.6%, + 6.5%, + 28.8%, −16.6%, −6.4%, −6.5%, and + 10.8% (Table 5).

The spatial variability in low-flow bias among the sub-basins (underestimate/overestimate) may be associated with differences in basin size, topography, soil properties, groundwater contribution, and land use patterns (Alvarenga et al., 2016). Additionally, the performance of the MERGE product may vary among sub-basins depending on how well the satellite component captures the spatial and temporal variability of precipitation. In some basins, satellite estimates may better represent rainfall distribution patterns, improving low-flow simulations, whereas in other basins the precipitation variability may not be adequately captured, contributing to underestimation or overestimation of Q90% flows.

4 Discussion

This study demonstrates that the MHD-INPE model is capable of satisfactorily representing streamflow dynamics in a Brazilian tropical headwater basin when driven by both rain gauge and satellite rainfall datasets. The results reinforce the robustness of the MHD-INPE for simulating hydrological processes in headwater basins (Falck et al., 2018; Melo et al., 2022a). Although rain gauge data generally provided better calibration performance, satellite rainfall inputs produced comparable or even superior results during the validation period, particularly under low-flow conditions. This behavior suggests that satellite-based precipitation products may better capture the spatial distribution of rainfall over large areas and extended periods, improving the model's capacity for temporal generalization during validation.

Similar findings have been reported in other studies conducted in tropical regions, where satellite precipitation products contributed to satisfactory hydrological simulations despite limitations associated with complex terrain and heterogeneous rainfall regimes (Cassalho et al., 2020; Rozante et al., 2020; Silva et al., 2022). Furthermore, the sensitivity of MHD-INPE to rainfall inputs observed in this study is consistent with the behavior reported for other hydrological models commonly applied in Brazil, such as SWAT, DHSVM, and VIC (Alvarenga et al., 2016, 2020; Oliveira et al., 2023).

The results further demonstrate that the choice of rainfall dataset plays a significant role in hydrological model performance, particularly in smaller headwater sub-basins with complex relief. In these regions, orographic effects are responsible for higher spatial rainfall variability, increasing uncertainties in precipitation estimates, while rapid hydrological responses create additional challenges to hydrological models due to the non-linear effects of the rainfall runoff transformation (Melo et al., 2022b). The spatial resolution of the MERGE estimates may partially explain the lower performance observed in smaller headwater catchments, where localized rainfall variability and orographic effects are more difficult to represent. In addition, gridded precipitation products from rain gauges may smooth intense rainfall events and consequently affect the hydrological responses. These factors tend to reduce model performance, especially in small and heterogeneous catchments, as discussed by Bárdossy et al. (2022) and Alvarenga et al. (2016). In contrast, larger sub-basins exhibited improved model performance, likely due to the attenuation of the effect of localized rainfall variability and the greater linearity of hydrological processes at broader spatial scales, as also noted by Oliveira et al. (2023). Additionally, the satisfactory representation of low-flow conditions using satellite rainfall inputs is particularly relevant for water resources management, since these periods directly affect hydropower generation and water supply planning corroborating previous findings reported by Falck et al. (2018), Rozante et al. (2020), Huang et al. (2025), and Jian et al. (2025).

Despite the satisfactory overall performance, some limitations remain evident. Differences in MHD-INPE performance among the analyzed sub-basins indicate that model sensitivity remains strongly influenced by local catchment characteristics and input data uncertainties. These findings highlight the need for further investigations on parameter regionalization, uncertainty propagation, and the transferability of calibrated parameter sets among neighboring catchments to improve model robustness and support broader hydrological applications.

5 Conclusion

The results demonstrate that the MHD-INPE model adequately represents the hydrological characteristics of the study region in terms of streamflow behavior. Overall, the model successfully reproduced the magnitude and variability of observed streamflows when forced with the two rainfall input datasets. Qualitative assessments also indicated that the simulated streamflows, derived from both observed and satellite rainfall, captured the temporal variability observed across the sub-basins.

On average, model performance was considered satisfactory for the analyzed period and rainfall inputs. However, some limitations were identified in the headwater sub-basins, where the largest uncertainties occurred in the simulation of extreme flows (Q10% and Q90%). The lowest model performance occurred mainly in the AIU1, AIU2, AIU3, and AIU4 headwater sub-basins, where NSE values ranged from 0.31 to 0.55, LNSE values ranged from 0.34 to 0.62, and KGE values ranged from 0.35 to 0.7 during validation analyses with rain gauge datasets. In contrast, larger sub-basins with gentler slopes showed improved performance in the calibration and validation, suggesting that the model is more effective in areas with less complex topography and more stable hydrological responses.

Rainfall data derived from satellite-based products satisfactorily represented the seasonal variability of streamflow. MERGE-driven simulations showed performance comparable to, and in some cases better than, those based on rain gauge observations during validation. These findings indicate that, in regions where rain gauge networks are sparse or unevenly distributed, MERGE rainfall estimates can serve as a valuable alternative data source for distributed hydrological modeling. However, in smaller headwater basins characterized by complex terrain and strong orographic effects, ground-based rainfall observations remain essential to improve the representation of localized rainfall patterns and hydrological extremes.

The integration of satellite rainfall estimates into hydrological modeling has the potential to support continuous monitoring and near real-time hydrological applications, particularly in data-scarce regions where rapid rainfall information is needed. In addition, despite the higher reliability of rain gauge data, a significant number of stations in Brazil are still operated manually (conventional stations), precluding their use in early warning systems since their information is not available in real-time. However, issues related to data latency, availability, and operational real-time performance were not specifically assessed in this study. For the Camargos and Itutinga hydropower plants, the results suggest that satellite rainfall products (MERGE) can support operational water resources management, inflow monitoring and seasonal water availability assessment in areas with limited rain gauge coverage. However, in smaller headwater sub-basins characterized by complex terrain and strong rainfall spatial variability, the integration of satellite estimates with ground-based rainfall observations needs improvements in terms of their spatial resolution and in the representation of rainfall extremes.

References

• 1

  Agência Nacional de Águas e Saneamento Básico (ANA), (2024). Portal HidroWeb. Available online at: https://www.snirh.gov.br/hidroweb/serieshistoricas (Accessed April 15, 2024).

  • Google Scholar
  • View reference in article

• 2

  AlvarengaC. M. D.AlvarengaL. A.MeloP. A.TomasellaJ.PintoP. R. F.MelloC. R. (2025). Meteorological Droughts in the Paraopeba river Basin: current scenarios and future projections. Land14:2093. doi: 10.3390/land14102093

  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  AlvarengaC. M. D.AlvarengaL. A.MeloP. A.TomasellaJ.PintoP. R. F.MelloC. R.et al. (2026). Impacts of climate change on the hydrology of a highly disturbed tropical river basin. Earth7:52. doi: 10.3390/earth7020052

  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  AlvarengaL. A.CarvalhoV. S. O.OliveiraV. A.MelloC. R.ColomboA.TomasellaJ. (2020). Hydrological simulation with SWAT and VIC models in the verde river watershed, minas gerais. Ambient. Água15:e2492. doi: 10.4136/ambi-agua.2492

  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  AlvarengaL. A.MelloC. R.ColomboA.CuartasL. A.BowlingL. C. (2016). Assessment of land cover change on the hydrology of a Brazilian headwater watershed using the distributed hydrology-soil-vegetation model. Catena143, 7–17. doi: 10.1016/j.catena.2016.04.001

  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  BárdossyA.KilsbyC.BirkinshawS.WangN.AnwarF. (2022). Is precipitation responsible for the most hydrological model uncertainty?Front. Water4:836554. doi: 10.3389/frwa.2022.836554

  • CrossRef
  • Google Scholar
  • View reference in article

• 7

  BeckH. E.ZimmermannN. E.McVicarT. R.VergopolanN.BergA.WoodE. F. (2018). Present and future Köppen-Geiger climate classification maps at 1-km resolution. Sci. Data5:180214. doi: 10.1038/sdata.2018.214

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  CassalhoF.Daleles RennóC.Bosco Coura dos ReisJ.Cláudio da SilvaB. (2020). Hydrologic validation of MERGE precipitation products over anthropogenic watersheds. Water12, 1268. doi: 10.3390/w12051268

  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  CollischonnB.CollischonnW.TucciC. E. M. (2008). Daily hydrological modeling in the Amazon basin using TRMM rainfall estimates. J. Hydrol.360, 207–216. doi: 10.1016/j.jhydrol.2008.07.032

  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  CostaA. R.deS.AlvarengaL. A.ThebaldiM. S.LimaL. A.ColomboA.et al. (2024). Characterization of artificial rainfall produced by a portable rainfall simulator using a rotating dynamic rainfall gauge system. Hydrol. Sci. J.69, 1218–1227. doi: 10.1080/02626667.2024.2368266

  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  DeviG. K.GanasriB. P.DwarakishG. S. (2015). A review on hydrological models. Aquat. Procedia4, 1001–1007. doi: 10.1016/j.aqpro.2015.02.126

  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  FalckA. S.MaggioniV.TomasellaJ.DinizF. L. R.MeiY.BenetiC. A.et al. (2018). Improving the use of ground-based radar rainfall data for monitoring and predicting floods in the Iguaçu river basin. J. Hydrol.567, 626–636. doi: 10.1016/j.jhydrol.2018.10.046

  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  FalckA. S.MaggioniV.TomasellaJ.VilaD. A.DinizF. L. R. (2015). Propagation of satellite precipitation uncertainties through a distributed hydrologic model: a case study in the tocantins-araguaia basin in Brazil. J. Hydrol. 527, 943–957. doi: 10.1016/j.jhydrol.2015.05.042

  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  FreitasA. V. L.KaminskiL. A.MilkeO. H. H.BarbosaE. P.Silva-BrandãoK. L. (2012). A new species of Yphthimoides (lepidoptera: nymphalidae: satyrinae) from the southern atlantic forest region. Zootaxa3526, 31–44. doi: 10.11646/zootaxa.3526.1.3

  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  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
  • View reference in article

• 16

  HuangY.ChenX.WangL.WuH.GuG.LiC.et al. (2025). Hydrological applicability and uncertainty assessment of satellite precipitation products: a global review. J. Hydrol. 651:133214. doi: 10.1016/j.jhydrol.2024.131310

  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  Instituto Nacional de Meteorologia (INMET), (2022). Banco de Dados Meteorológicos para Ensino e Pesquisa (BDMEP). Available online at: https://bdmep.inmet.gov.br/ (Accessed April 4, 2022).

  • Google Scholar
  • View reference in article

• 18

  JianJ.HeS.LiuW.LiuS.GuoL. (2025). A refined method for the simulation of catchment rainfall–runoff based on satellite–precipitation downscaling. J. Hydrol. 653:132795. doi: 10.1016/j.jhydrol.2025.132795

  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  Junqueira JúniorJ. A.SilvaA. M.da MelloC. R.de PintoD. B. F. (2008). Spatial continuity of soil physical-hydric attributes in a headwater sub-basin. Ciênc. Agrotec. 32, 914–922. doi: 10.1590/S1413-70542008000300032

  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  JunqueiraR.ViolaM. R.AmorimJ.daS.CamargosC.MelloC. R. (2022). Hydrological modeling using remote sensing precipitation data in a Brazilian savanna basin. J. South Am. Earth Sci.115:103773. doi: 10.1016/j.jsames.2022.103773

  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  MantovaniJ. R.AlcântaraE.BaiãoC. F.PampuchL.CurtarelliM. P.Mariano RibeiroJ. V.et al. (2025). Unprecedented flooding in porto alegre metropolitan region (southern brazil) in may 2024: causes, risks, and impacts. J. South Am. Earth Sci.160:105533. doi: 10.1016/j.jsames.2025.105533

  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  MapBiomas Brasil, (2022). Coleções MapBiomas. Available online at: https://brasil.mapbiomas.org/colecoes-mapbiomas/ (Accessed April 4, 2022).

  • Google Scholar
  • View reference in article

• 23

  MeloP. A.AlvarengaL. A.TomasellaJ.MelloC. R.MartinsM. A.CoelhoG. (2022a). Analysis of hydrological impacts caused by climatic and anthropogenic changes in upper grande river basin, Brazil. Environ. Earth Sci.81:504. doi: 10.1007/s12665-022-10631-2

  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  MeloP. A.AlvarengaL. A.TomasellaJ.SantosA. C. N. (2024). Uncertainty analysis on long-term runoff projection from the Budyko framework and a conceptual hydrological model. J. Water Clim. Change15, 3850–3866. doi: 10.2166/wcc.2024.127

  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  MeloP. A.AlvarengaL. A.TomasellaJ.SantosA. C. N.MelloC. R.ColomboA. (2022b). On the performance of conceptual and physically based modelling approach to simulate a headwater catchment in Brazil. J. South Am. Earth Sci.114:103683. doi: 10.1016/j.jsames.2021.103683

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  MenezesP. C. M.de SouzaD. C.TavaresM. G.MarquesR. A. G. (2026). Comparative analysis of the accuracy of temperature and precipitation data in Brazil. Meteorology5:3. doi: 10.3390/meteorology5010003

  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  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
  • View reference in article

• 28

  NascimentoC.DemattêJ. A.MelloF. A.RosasJ. T.TayebiM.BellinasoH.et al. (2022). Soil degradation detected by temporal satellite image in São Paulo state, Brazil. J. South Am. Earth Sci.120:104036. doi: 10.1016/j.jsames.2022.104036

  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  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
  • View reference in article

• 30

  OliveiraC.deM. M.deL. A.BeskowS.da CunhaZ. A.VargasM. M.MeloP. A.et al. (2023). Hydrological model performance in the verde river basin, minas Gerais, Brazil. Resources12:87. doi: 10.3390/resources12080087

  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  OliveiraV. A.de MelloC. R.ViolaM. R.SrinivasanR. (2017). Assessment of climate change impacts on streamflow and hydropower potential in the headwater region of the Grande river basin, Southeastern Brazil. Int. J. Climatol.37, 5005–5023. doi: 10.1002/joc.5138

  • CrossRef
  • Google Scholar
  • View reference in article

• 32

  PintoP. R. F.AlvarengaL. A.TomasellaJ.MeloP. A.MelloC. R.SantosA. C. N. (2026). Effects of land use and land cover changes and water uses on water security in an anthropized basin. Front. Water8:1806322. doi: 10.3389/frwa.2026.1806322

  • CrossRef
  • Google Scholar
  • View reference in article

• 33

  RodriguezD. A.TomasellaJ. (2016). On the ability of large-scale hydrological models to simulate land use and land cover change impacts in Amazonian basins. Hydrol. Sci. J. 61, 1831–1846. doi: 10.1080/02626667.2015.1051979

  • CrossRef
  • Google Scholar
  • View reference in article

• 34

  Romero-HernándezC. M.Avila-DiazA.QuesadaB.MedeirosF.CerónW. L.Guzman-EscalanteJ.et al. (2024). Bias-corrected high-resolution precipitation datasets assessment over a tropical mountainous region in Colombia: a case of study in upper cauca river basin. J. South Am. Earth Sci.140:104898. doi: 10.1016/j.jsames.2024.104898

  • CrossRef
  • Google Scholar
  • View reference in article

• 35

  RozanteJ. R.GutierrezE. R.FernandesA.deA.VilaD. A. (2020). Performance of precipitation products obtained from combinations of satellite and surface observations. Int. J. Remote Sens.41, 7585–7604. doi: 10.1080/01431161.2020.1763504

  • CrossRef
  • Google Scholar
  • View reference in article

• 36

  RozanteJ. R.MoreiraD. S.de GoncalvesL. G. G.VilaD. A. (2010). Combining TRMM and surface observations of precipitation: technique and validation over South America. Wea. Forecasting25, 885–894. doi: 10.1175/2010WAF2222325.1

  • CrossRef
  • Google Scholar
  • View reference in article

• 37

  SiddiqueR.SharmaS.McCreightJ. (2023). Editorial: hydrological modeling, analyses, and predictions: opportunities and challenges. Front. Water4:1121534. doi: 10.3389/frwa.2022.1121534

  • CrossRef
  • Google Scholar
  • View reference in article

• 38

  SilvaE. H.deL.SilvaF. D.dosS.JuniorR. S.daS.PintoD. D. C.CostaR. L.et al. (2022). Performance assessment of different precipitation databases (gridded analyses and reanalyses) for the new Brazilian agricultural frontier: SEALBA. Water14:1473. doi: 10.3390/w14091473

  • CrossRef
  • Google Scholar
  • View reference in article

• 39

  Soil Survey Staff, (1951). Soil Survey Manual. Washington, DC: U.S. Department of Agriculture Handbook.

  • Google Scholar
  • View reference in article

• 40

  SouzaC. M.Z. ShimboJ.RosaM. R.ParenteL. L.A. AlencarA.RudorffB. F. T.et al. (2020). Reconstructing three decades of land use and land cover changes in brazilian biomes with landsat archive and earth engine. Remote Sens.12:2735. doi: 10.3390/rs12172735

  • CrossRef
  • Google Scholar
  • View reference in article

• 41

  SouzaÍ. A.SilvaV.deA. (2025). Flood prediction in Brazil: a review of trends, gaps and methodological perspectives. J. South Am. Earth Sci.164:105637. doi: 10.1016/j.jsames.2025.105637

  • CrossRef
  • Google Scholar
  • View reference in article

• 42

  TerraM. D. C. N. S.MelloJ. M.de MelloC. R.de SantosR. M.dos NunesA. C. R.RaimundoM. R. (2015). Topo-edapho-climatic influence on the vegetation of an atlantic forest fragment in the mantiqueira mountains, MG. ambient. e agua—an interdiscip. J. Appl. Sci.10:1705. doi: 10.4136/ambi-agua.1705

  • CrossRef
  • Google Scholar

• 43

  WrightI. R.GashJ. H. C.RochaH. R.da RobertsJ. M. (1996). “Modelling surface conductance for Amazonian pasture and forest,” in Amazonian Deforestation and Climate, eds. J. H. C. Gash, C. A. Nobre, J. M. Roberts, and R. L. Victoria (New York: John Wiley and Sons), 437–458.

  • Google Scholar

• 44

  ZitzlerE.LaumannsM.ThieleL. (2001). SPEA2: Improving the strength pareto evolutionary algorithm. TIK Rep.103, 1–21. doi: 10.3929/ethz-a-004284029

  • CrossRef
  • Google Scholar
  • View reference in article