Quantified impact of projected climate change on groundwater recharge and river discharge leveraging the use of open access geospatial data

In previous years, water is abundant in the Padsan River Watershed, even in tropical climates. However, with increasing variability due to global warming, the watershed faces disruption of water resources. This requires thorough study, yet with fragmented data management, it becomes challenging. With technological advantages, integrated water resources management (IWRM) becomes possible using open-access data to understand the potential impact of climate variability. The corrected Climate Hazards Group InfraRed Precipitation (CHIRPS) and ten Coupled Model Intercomparison Project Phase 6 (CMIP6) becomes the notable open-access data including request-based institutional data were used as inputs for the Soil and Water Assessment Tools (SWAT) to quantify the impact of climate change on river discharge and groundwater recharge. Results showed that all Shared Socioeconomic Pathways (SSPs) will disrupt river discharge and groundwater recharge, with a prominent increase in river discharge. Furthermore, SSP 585 in the 2090s has a more notable impact on river discharge than others. In contrast, the SSP 126 has a lesser impact but displays higher variability across the rest of the century. This important simulated observation highly supports the Philippine Development Plan (PDP) aims, which is that climate change's impact could disrupt existing infrastructure and recharge conservation while establishing recharge areas to combat water scarcity during the dry period in the watershed. In contrast to the PDP regarding groundwater use, the study also supports the increasing conjunctive use of surface and subsurface resources, given that comprehensive management of the subsurface extraction must be established based on the study’s results.

The Padsan River Watershed (PRW) is one of the key agricultural watersheds in the Philippines, offering essential resources that support crop production and benefit various community sectors. Like other watersheds, PRW serves as a vital water source, enabling settlement and sustaining livelihoods. Water as most crucial natural resources (Stigter et al., 2023) is essential in for sustaining human life (Stigter et al., 2023; Yang et al., 2021) (Anatolia et al., 2021; Kılıç Çetinkaya and Aslandoğan, 2022), supporting socioeconomic development (Alexandratos et al., 2019; Yang et al., 2021), and maintaining healthy ecosystems (Bogardi et al., 2020). For these notions, surface and subsurface freshwater resources are essential for global watershed availability and accessibility, supporting drinking, irrigation, industrial uses, and other terrestrial uses.

However, these essential watershed resources are under threat from the impacts of climate change (Kılıç Çetinkaya and Aslandoğan, 2022), which is expected to alter the Earth’s hydrological cycle significantly (Yang et al., 2021) through changes in precipitation patterns, evapotranspiration rates, and temperature (Kılıç Çetinkaya and Aslandoğan, 2022). Concerns are growing about the potential consequences of climate change (Yang et al., 2021) on the recharge and dynamics of groundwater and surface water systems, as these processes are primary determinants of water availability. These alterations are already observed in PRW. The river discharge volume is projected to increase while the flow variability has enhanced (Tolentino et al., 2016), indicating more extreme hydrologic events. These extreme hydrologic events cause surplus water during the rainy season while deficit during the dry season and increased runoff from increasing rainfall intensity causes water inaccessibility (Alibuyog et al., 2009 as cited by Northwestern University, 2024).

Furthermore, Philippine irrigated agricultural expansions rely heavily on either the construction of surface irrigation systems or the exploitation of groundwater (Inocencio and Barker, 2018). These highlight the conjunctive use of surface and subsurface watersheds in agricultural production. Yet, altered recharge and discharge regimes can lead to the depletion of aquifers, reduced baseflow to rivers and streams, and changes in the frequency and magnitude of extreme hydrological events such as floods and droughts. These impacts can have profound implications for agricultural productivity, infrastructure investment inefficiency, rising water conflicts, economic crises, migration (Kılıç Çetinkaya and Aslandoğan, 2022), and overall resource management (Gleick, 2014; Kundzewicz et al., 2008; Taylor et al., 2013).

Moreover, the PRW agricultural sector has suffered the most in recent years due to extreme climate events. Ilocos Norte, where the PRW provides shelter for half of the province, lost more than 644,000 USD in the agricultural sector due to the excessive drought (Lazaro, 2024). On the other hand, the region experiences six consecutive typhoons from October to mid-November 2024 (World Food Programme Philippines, 2024). One of these typhoons resulted in 283,528 metric tons (MT) of agricultural products being lost in the region, which includes the Province of Ilocos Norte. These notions ranked the nation’s fourth overall global climate risk index from 2000 to 2019 (Eckstein et al., 2021) to first in the 2024 world ranking according to the World Risk Index of 2023 (World Food Programme Philippines, 2024).

Extreme events significantly impact agriculture, while climate-increasing variability exposes water resources’ vulnerability to degradation. Altered rainfall patterns (Rola et al., 2015; Tabios III, 2020; Velasco et al., 2020) could lead to water stress during the dry season, while excessive wet seasons exacerbate water scarcity and flood risks (Alejo et al., 2021a; Tolentino et al., 2016). The sea rise levels (Rola et al., 2015; Tabios III, 2020; Velasco et al., 2020) threaten coastal aquifers with saltwater intrusion, reducing freshwater availability (O’Neill et al., 2016; Tolentino et al., 2016) for the 60% of Filipinos living in coastal areas (D’Agnes et al., 2005). Furthermore, hydrologic studies conducted from various watersheds in the country show that river discharge will be altered, and reservoir management is in a challenging situation, posing a threat to water and food security and potentially increasing casualties during peak seasons, especially widely untested watersheds, e.g., PRW. Addressing these challenges that could impose the same impact on PRW necessitates a comprehensive understanding of the complex interplay between water resource availability and growing vulnerabilities driven by climate variability and extremes (Saedi et al., 2022; Yang et al., 2021). It is crucial for developing effective adaptation strategies and sustainable water management plans beneficial to Philippine watersheds.

However, the formulation of sustainable solutions is constrained by the persistent scarcity of high-quality and reliable data, highlighting the urgent need for improved data collection, management, and analysis. These constraints could underpin comprehensive and effective management strategies (Rola et al., 2015; Valenzuela and Gutierrez, 2019; Velasco et al., 2020). As Tabios III (2020) emphasizes, the lack of sufficient monitoring infrastructure and the prevalence of sparse, discontinuous, and fragmented observatory data (Valenzuela and Gutierrez, 2019; Velasco et al., 2020) impede the development of adequate water resources management frameworks. In this context, leveraging open-access datasets emerges as a vital step towards bridging these gaps and enabling informed decision-making for sustainable watershed management, consistent with the national direction toward strengthened and integrated water resources governance under the (National Water Resources Management Act S. No. 2771, 19th Congress, 2024).

These mentioned challenges apply to the case of the PRW as a mid-prioritized watershed of the nation that is widely untested in water resources. The watershed was heavily affected by recent climate variability and extremes. By quantifying the impacts of future climate change on groundwater recharge and river discharge within the watershed using free-access high-quality data and information, this study aims to inform local and regional water resources management efforts and contribute to the broader understanding of climate change’s influence on the hydrological cycle of the PRW that might be helpful to other nation’s watershed. The study will employ descriptive climate projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6) to achieve this. CMIP6 provides the most up-to-date and comprehensive climate model projections, incorporating the latest scientific understanding of climate processes and their interactions. These projections will be instrumental in assessing future climate scenarios and their potential impacts on hydrological processes in the PRW (Eyring et al., 2016). Recent studies, such as Khoi et al. (2022), highlight the critical importance of integrating climate projections with hydrological models to predict changes in water resources, reinforcing the need for comprehensive climate impact assessments.

Furthermore, the study uses the Climate Hazards Group InfraRed Precipitation with Station (CHIRPS: Funk et al., 2015) as precipitation input. It was further bias-corrected using Quantile Mapping (Heo et al., 2019) to remove the systematic bias and ensure the monthly precipitation’s variability aligned with the discrete ground-based observations. Other data model inputs were gathered from the institutional website, which provides free-access geospatial data maps dedicated to further scientific studies.

The Soil and Water Assessment Tool (SWAT) model will also be utilized for hydrological modeling. SWAT is a robust, widely used model designed to predict the impact of land management practices on water, sediment, and agricultural chemical yields in large complex watersheds with varying soils, land use, and management conditions over long periods (Arnold et al., 2012). Studies have demonstrated that SWAT can effectively simulate the impacts of climate change on hydrological processes, providing critical insights into future water availability and distribution (Khoi et al., 2022). Hydrological studies using SWAT have been shaping the understanding of water resources in the different watersheds in the country. It shows broad applicability across diverse watersheds (Alejo et al., 2021a; Jamilla et al., 2021; Panondi and Izumi, 2021) for hydrological, climate change (Alejo and Alejandro, 2021; Singson et al., 2023) and land use and land cover (LULC) and urbanization (Boongaling et al., 2018; Jamilla et al., 2021) assessment. It also shows excellent allocations for hydropower potential assessment (Garcia et al., 2016; Tambong et al., 2019) and shaping policy and planning tools (Alejo and Alejandro, 2022; Guiamel and Lee, 2020). Various studies have also been conducted on Integrating CMIP6 climate projections with SWAT modeling, enabling a detailed analysis of how climate change could alter groundwater recharge and river discharge patterns. This comprehensive approach will provide critical insights into the magnitude and timing of these impacts, allowing water managers and policymakers to develop more informed and effective strategies for adapting to climate change and ensuring the long-term sustainability of the region’s water resources.

This study aims to (1) integrate validated CHIRPS and other open-access data into the SWAT model, (2) simulate river discharge and groundwater recharge using the calibrated and validated model, (3) assess the projected impacts of climate change based on CMIP6 scenarios, and (4) recommend strategies for sustainable water resource management and projected climate change mitigation.

Research methods and approach

The study’s conceptual framework (Figure 1) illustrates the integrated research approach for achieving the desired outcomes. The framework provides a detailed systematic approach to quantify the impact of projected climate change on groundwater recharge (GWR) and river discharge (RiverQ), leveraging open-access data. It integrates open-access geospatial data with hydrological modeling to quantify the impact of projected climate change on groundwater recharge and river discharge. The integrated approach is structured into five multi-stage phases. The first phase is the data extraction and preprocessing for hydrological modeling, which includes model data input acquisition for the study area and rainfall bias correction. The second phase is the model simulation, calibration, and validation, which provides for Model Uncertainty Analysis. The third phase is the climate change scenario building, and the fourth is the impact simulation of climate change on river discharge data and groundwater recharge. The last one is visualizing the impact of climate change on selected water resources phenomena. Each phase ensures logical connectivity (Figure 1) while leveraging robust datasets and modeling techniques. Each phase is further discussed in subsequent sections.

Data extraction and preprocessing for hydrological modelling

The study’s first phase focuses on extracting and processing essential data inputs to support accurate hydrological simulations using the Soil and Water Assessment Tool (SWAT). Ensuring data quality and availability is critical for modeling the impacts of climate change (IPCC, 2023) on the Padsan River Watershed (PRW), a vulnerable agricultural watershed in the Philippines (Tolentino et al., 2016). Reliable datasets and robust pre-processing are essential to reduce model uncertainty and improve simulation accuracy (Tabios III, 2020).

Moreover, PRW (Figure 2) in northwestern Luzon covers 1,325 km 2 and extensively supports agriculture and urban development in Ilocos Norte Province (Provincial Government of Ilocos Norte, 2024). It has a climate type 1 based on the Modified Coronas Climate Classifications (Peralta et al., 2020). Climate variability and extreme weather events have increased hydrological risks, including altered river discharge patterns and potential groundwater depletion, threatening water security and agricultural productivity (Singson et al., 2023). Thus, a robust, data-driven integrated approach must be established.

The study leverages open-access geospatial data to overcome data scarcity (Valenzuela and Gutierrez, 2019) in the watershed. The Digital Elevation Model (DEM), soil, Land Use Land Cover (LULC), and meteorological data were extracted during the data extraction, making them essential in hydrological modeling. The high spatial resolution of climate change scenarios was also extracted. The DEM was extracted from the Shuttle Radar Topography Mission (SRTM GL1) with a global resolution of 30 m provided by the OpenTopography portal at https://portal.opentopography.org/raster?opentopoID=OTSRTM.082015.4326.1 (NASA Shuttle Radar Topography Mission, 2013). The LULC and soil data of the National Mapping and Resource Information Authority (NAMRIA) and the Department of Agriculture (DA), respectively, were extracted from the Geoportal Philippines (https://www.geoportal.gov.ph). This is to ensure that recent information from the government was prioritized to ensure that results work alongside the current local classification of LULC and soil data followed by the Philippine government. Furthermore, discrete meteorological data was also extracted from the Philippines Atmospheric, Geophysical and Astronomical Services Administration (PAGASA) at https://www.pagasa.dost.gov.ph/climate/climate-data following the licensing procedure and user agreement. The watershed has only one ground-based station, making the discrete station misinterpret most of the area. The acquired ground-based weather data validated the bias-corrected Satellite Precipitation Data extracted from the Climate Engine web portal (https://www.climateengine.org). Ground-based observation data for RiverQ were acquired from the National Irrigation Administration–Ilocos Norte (NIA-IN) with monthly available data from 2001 to 2021. These ground-based observational data were used to calibrate and validate the hydrological Model. The Satellite Precipitation Estimates (SPEs) used were the Climate Hazards Group InfraRed Precipitations with Station Data (CHIRPS).

The CHIRPS dataset (Funk et al., 2015) is widely used as precipitation inputs due to its high spatial and temporal resolution, but it may contain systematic biases. Bias correction is necessary when satellite precipitation products exhibit consistent distributional offsets relative to ground observations, as such biases can propagate through hydrological simulations and distort projected outputs (Maraun, 2016). To improve data accuracy, this study applies Nonparametric Quantile Mapping (QM), a bias correction method that aligns CHIRPS data with observed rainfall distributions (Heo et al., 2019). Studies show that QM improves hydrological simulations by reducing systematic biases while preserving climate trends (Cannon et al., 2015; Enayati et al., 2021). The QM method uses the empirical cumulative distribution function that Yersaw and Chane (2024) concluded performs well compared to other bias correction methods (BCM). Correcting CHIRPS data enhances its reliability for hydrological modeling, ensuring accurate projections of GWR and RiverQ in the PRW. Corrected CHIRPS augmented the shortage of available rainfall data in the watershed.

The performance of the QM correction method was evaluated using selected statistical performance metrics. These parameters (Table 1) were assessed using four key criteria, including the model’s accuracy, bias and efficiency, data distribution, and uncertainty. The model’s accuracy was tested using the Nash Sutcliff Efficiency (NSE), Coefficient of Determination (R 2), Root Mean Error to Standard Deviation ratio (RSR), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The system bias and efficiency were evaluated using the Percent Bias (PBIAS), Kling-Gupta Efficiency (KGE), and Standard Deviation Ratio (SDR). The data distribution was assessed using the quantiles and standard deviation ratio (SDR), while uncertainty was evaluated using the 95% confidence interval.

The corrected CHIRPS dataset has a slight reduction in accuracy. The NSE values slightly reduced from 0.87 to 0.85, while R 2 decreased from 0.87 to 0.86. Nevertheless, the corrected CHIRPS maintained its strong predictive performance. The RMSE has also indicated a slight increase from 4.20 mm while the MAE decreases by 2.88 mm. The slight rise in RMSE demonstrates that the correction has a slight impact on the overall error magnitude of the corrected compared to the uncorrected. At the same time, the increase in MAE produces a slight reduction in the average error magnitude. The RSR increased marginally from 0.36 to 0.38 (dimensionless), reflecting a slight reduction in relative model accuracy, while the MAE decreased by 2.88 mm (from 55.27 to 52.39 mm), indicating a modest improvement in average absolute error.

On the other hand, the BIAS and system efficiency can significantly improve by the QM correction. PBIAS significantly improves to nearly zero, while the MBE of approximately zero demonstrated the significant reduction of the average difference between CHIRPS and observed precipitation from a negative bias, which indicates considerable underestimation of rainfall by the uncorrected CHIRPS. Furthermore, the overall evaluation parameter KGE has a more significant increase from 0.86 to 0.93, indicating that when bias, correlation, and variability are comprehensively evaluated at once, the correction makes a significant impact, making bias and variability dominate the slight decrease of accuracy parameters.

The data distribution metrics also indicate an improved dataset for the CHIRPS. The SDR of corrected CHIRPS turns 1 from 0.879, suggesting that it efficiently reflects the variability of the ground-based rainfall. These justify that the variability and bias dominated the overall enhancement of the dataset over the slight change in accuracy. Furthermore, the 25 th, 50 th, and 75 th quantile significantly decreases from the uncorrected CHIRPS. These significant value reductions represented the monthly rainfall variability in the watershed, which had two pronounced seasons: wet and dry, where both give frequent extremes, vast rainfall during the rainy season, and scarcity of rainfall during the dry season.

The results provide an overall enhancement of the corrected CHIRPS data. The reduction of uncertainty, aligning the variability of the corrected CHIRPS to the observed rainfall, and the reduction of systematic bias of the CHIRPS justify the suitability of the corrected CHIRPS as rainfall input for the Soil and Water Assessment Tool (SWAT). After the correction, all essential model inputs were processed into the desired format required by the SWAT model.

Model simulation, uncertainties analysis, calibration and validation

The SWAT model has become a valuable tool in assessing the impacts of climate change on hydrological processes, providing flexibility to simulate complex watershed dynamics under changing climatic conditions. Widely used in climate change studies, SWAT integrates global and regional projections to quantify future water availability, including river discharge and groundwater recharge. For instance, SWAT was applied by the Mekong River Commission (MRC) to evaluate water resources management strategies in the Lower Mekong River Basin (LMRB) under different climate scenarios (MRC, 2015). Its long-term simulation capabilities make it suitable for assessing altered precipitation drivers of water resource dynamics in agricultural watersheds like PRW. SWAT facilitates integrated surface and subsurface water simulations (Arnold et al., 2012), providing a rapid and robust foundation for quantifying RiverQ and GWR for conjunctive use management. Its extensive user community has developed various coupled systems, enabling diverse hydrological applications. Given the study’s focus on utilizing open-access data, only the SWAT model was used, excluding more complex coupled models such as SWAT-MODFLOW. This approach aims to deliver reliable insights while emphasizing the need for greater transparency to achieve comprehensive water resource management outcomes. The developed model will be a foundation for broader applications in future studies, including integrating coupled systems.

The ArcSWAT 2012.10_8.25 software, compatible with ArcMap 10.8.2, was used to simulate the SWAT model for the PRW over 20 years (2000–2020) using pre-processed spatial and climate data. SWAT-CUP with the SUFI-2 algorithm (Abbaspour et al., 2017; Khalid et al., 2016) was employed for uncertainty analysis, calibration, and validation, noted for its efficiency in handling large-scale models (Pandey et al., 2021). 16 sensitive SWAT parameters (Anaba et al., 2017) were considered during the uncertainty analysis (Abbaspour et al., 2017). The model was calibrated for river discharge from 2003 to 2014 and validated from 2015 to 2021 using the most sensitive parameters fitted values, ensuring reliability through uncertainty analysis and performance evaluation. Model performance was evaluated using statistical parameters, including NSE, PBIAS, R 2, and RMSE, with sensitivity indicated by t-stat and p-values.

In the uncertainty analysis, out of 16 identified sensitive SWAT parameters, four were revealed to be highly sensitive (Table 2). The baseflow factor ratio, SCS Runoff Curve Number for moisture condition II, Effective hydraulic conductivity, and moisture bulk density remain highly sensitive (p < 0.00), affecting the comparative analysis between simulated and observed RiverQ. Singson et al. (2023) also identified these sensitive parameters as sensitive parameters and related studies. Their sensitivity is rooted directly in their direct influence on critical hydrological processes (Araza et al., 2021), such as surface runoff, infiltration, baseflow, and soil-water interactions, essential for accurately simulating RiverQ and GWR in PRW.

Afterward, the fitted values (Table 2) of the identified sensitive parameters yielded satisfactory calibration and validation results (Table 3). During the calibration period (2003–2014), the model achieved an NSE of 0.57, indicating that the model captured 57% of the variability in river discharge compared to observed values. An R 2 of 0.66 shows a moderate correlation between simulated and observed discharge, confirming the model’s ability to replicate trends in streamflow. The PBIAS of 2.80% indicates a minimal overestimation, which is acceptable within hydrological modeling standards (Moriasi et al., 2015). In the validation period (2015–2021), the model performance remained satisfactory with an NSE of 0.54, an R 2 of 0.71, and a PBIAS of 14.0%, although there was a slight decrease in NSE and an increase in PBIAS.

The slight decline in NSE and rise in PBIAS can be attributed to extreme discharge misestimations commonly observed in SWAT simulation studies (Singson et al., 2023). NSE reductions indicate that the model’s ability to capture variability in extreme events has limitations, particularly during high-flow periods. PBIAS increases suggest that the model tends to overestimate streamflow, particularly in extreme months, which is consistent with findings from Du et al. (2024) and Alejo et al. (2021b), who emphasized the need for local validation and multi-dataset comparisons to improve data accuracy in hydrological models. These results underscore the importance of interpreting the model’s outputs based on trends rather than absolute values to ensure reliable insights, particularly when assessing extreme hydrological events in data-scarce regions. With the increasing impact of climate variability, the Padsan River Watershed (PRW) has experienced significant changes in extreme events, necessitating further validation and refinement of hydrological models to account for these variabilities. More discussions on the works of Jimenez et al. (2025).

Beyond calibration performance metrics, the uncertainty structure established through SUFI-2 carries direct implications for the projected hydrological changes that is reported in the following sections. SUFI-2 quantifies parameter uncertainty through the 95% prediction uncertainty band (95PPU), assessed via the p-factor (fraction of observations bracketed by the 95PPU, target > 0.70) and r-factor (normalized band width, target < 1.5; Abbaspour et al., 2017). The feasible parameter space defined for the four most sensitive parameters (V_ALPHA_BNK, R_CN2, R_SOL_K, R_SOL_BD) during calibration is preserved when future climate forcing is applied, meaning residual parameter uncertainty propagates forward into projected outputs. The narrow and consistent standard deviations across all SSPs (±3.88% to ±4.91%) suggest that parameter uncertainty contributes a relatively stable background variability to projected discharge rather than a scenario-dependent amplification, consistent with evidence that in well-constrained SWAT applications, climate forcing uncertainty dominates over parameter uncertainty for long-term mean discharge projections (Abbaspour et al., 2017; Pandey et al., 2021). However, the validation PBIAS of 14.0% indicates a systematic tendency to overestimate during high-flow periods, which may be amplified under projected precipitation intensification, warranting conservative interpretation of the upper discharge bounds in Table 4. All projections in this study are therefore best interpreted as directional trend estimates rather than precise absolute values. Future work should explicitly propagate the SUFI-2 parameter ensemble through the climate change simulation to formally quantify the 95PPU of projected discharge and separate parameter uncertainty from GCM structural uncertainty, the dominant source of total projection spread in CMIP6-driven hydrological assessments (Wu et al., 2024).

Climate change model Selection and scenario building

Climate change scenarios are essential for understanding future climatic conditions and their impacts on hydrological processes. These scenarios project changes in temperature and precipitation under various greenhouse gas emissions and socioeconomic pathways, providing projections for assessing potential impacts on water resources and ecosystems. The study utilized the Couple Model Intercomparison Project Phase 6 (CMIP6), which offers advanced climate simulations with improved model resolution and representation of physical processes compared to the previous version (CMIP5), offering reliable projections for hydrological studies (Eyring et al., 2016; O’Neill et al., 2016).

To ensure consistency and robustness, the study adopts the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP) framework, specifically the ISIMIP3b version, which integrates CMIP6 projections to downscaled and bias-corrected climate data (Lange and Büchner, 2022). The ISIMIP framework ensures that the climate projections are standardized and comparable across sectors, providing a consistent basis for multi-sectoral climate impact assessments (Frieler et al., 2017) including water resources.

The study extracted provincial-level downscaled monthly relative change data for precipitation and temperature from the ClimoCast platform (https://a-plat.nies.go.jp/ap-plat/cmip6/global.html) to represent the mean monthly changes in the PRW. The ClimoCast platform serves as the data provider for the ISIMIP3b (Lange and Büchner, 2022) downscaled climate data from ISIMIP3b ensures high-resolution projections with bias correction, improving the accuracy of hydrological modeling for RiverQ and GWR in the PRW.

This study considers climate projections from 10 models (Table 5) based on the latest CMIP6 data. These model are internationally recognized for their robust simulation of climate processes and have been validated through historical performance assessment. These models composed of 5 primary model that includes GFDL-ESM4, IPSL-CM6A-LR, MPI-ESM1-2HR, MRI-ESM2-0 and UKESM1-0-LL while the secondary model are CanESM5, CNRM-CM6-1, CNRM-ESM2-1, EC-Earth, and MIROC6 (Lange, 2021). These four climate scenarios (SSP126, SSP245, SSP370, SSP585) (O’Neill et al., 2016) represent different future greenhouse gas (GHG) emissions pathways for the hydrologic impact simulations, providing the derive insight for broader applications and efficient information.

By using state-of-the-art climate from CMIP6 and ISIMIP3b, the study ensures a comprehensive and consistent impact assessment, providing reliable insights for policy and decision-making regarding climate resilience in conjunctive water resources management.

Climate change impact simulation and output visualization

The calibrated and validated model simulated the watershed hydrology from 2000 to 2021, including a three-year warming period (2000 to 2002) to establish a baseline for the simulation. This baseline supports the evaluation of climate change scenarios using the Shared Socioeconomic Pathways (SSPs) framework available from the ClimoCast data provider. SSP1 envisions a sustainable future with a shift toward inclusive development and environmental awareness, prioritizing education, health, and clean technology investments to reduce inequalities and enhance climate resilience. SSP2, labeled “Middle of the Road,” describes a world where current trends continue, leading to moderate progress in sustainability with uneven regional development and persistent social inequalities. SSP3, known as “Regional Rivalry (A Rocky Road),” portrays a fragmented world focusing on regional interests, resulting in strong national identities, limited international cooperation, and weak environmental initiatives. SSP4, termed “Inequality (A Road Divided),” depicts a scenario of pronounced inequalities, where a technological elite thrives while the majority remains marginalized, leading to low social cohesion and significant environmental degradation. Finally, SSP5, “Fossil-Fueled Development (Taking the Highway),” illustrates a world characterized by rapid economic and technological growth driven by fossil fuels, resulting in high emissions but improved adaptive capacity through economic gains. These scenarios provide a comprehensive framework for quantifying the impacts of climate change on river discharge and groundwater recharge, as detailed by O’Neill et al. (2017). This study aimed to utilize these diverse pathways to understand the potential impact of future environmental changes and inform effective policy strategies for managing water resources under varying socioeconomic conditions.

Integration of validated CHIRPS and other open-access model data inputs

The fragmented management of data in the watershed presents challenges in delivering rapid and robust strategies for preserving water resources in the context of climate change and depleting resources. However, the increasing availability of open-access satellite data (such as CHIRPS) and high-resolution, bias-corrected CMIP6 climate projections offer opportunities to address those gaps. Similar to several studies highlighting the works of Alejo et al. (2021a), the use of CHIRP as rainfall input for SWAT watershed modeling in various climate types in the Philippines (Alejo and Alejandro, 2021), achieving satisfactory calibration and validation results (Table 3). On the other hand, CHIRPS performance varies across regions and requires local validation and bias correction to improve its reliability for hydrological modeling (Du et al., 2024). By integrating bias-corrected CHIRPS data and other open-access datasets into the SWAT model, this study achieved satisfactory performance in simulating river discharge and assumed to satisfactorily simulate the groundwater recharge, demonstrating the model’s potential to provide scientific insights for conjunctive surface and subsurface water management, particularly in data-scarce regions, like PRW.

Moreover, the calibrated and validated SWAT model supports the study’s claims that it can accurately simulate river discharge (RiverQ) and groundwater recharge (GWR) in the Padsan River Watershed (PRW). The model effectively captures RiverQ variability, which aligns with the precipitation patterns observed in the CHIRPS dataset (Figure 3). However, extreme hydrological events were overestimated, consistent with the findings of Du et al. (2024), who highlighted variability in CHIRPS performance across regions and its limitations in capturing extreme precipitation events while mentioning its strength in representing extremely low conditions. Similarly, Alejo et al. (2021b) noted that CHIRPS may misrepresent localized rainfall patterns, particularly during high-intensity rainfall events, which can lead to biases in hydrological simulations. These discrepancies suggest that applying the study’s findings to other contexts requires additional validation and analysis, particularly for extreme hydrological conditions, to ensure accurate and reliable applications.

Climate change projections in the PRW

The monthly relative change of precipitation (Figure 4) was extracted from the ISIMIP3b framework through the ClimoCast platform. Time series divided into four future timeframes. All scenarios indicate fluctuating rainfall patterns, with the 2030s showing minor variability while variability increases by the 2090s, which makes the projected impact of greenhouse gas emissions distinct.

Furthermore, SSP126 shows moderate fluctuations in a low-emission scenario, indicating stable precipitation patterns. In contrast, SSP245 and SSP370 display intermediate to increasing variability along with higher timeframe. At the same time, SSP585, the high-emission scenario, exhibits extreme rainfall variations, with a pronounced peak by the 2090s, highlighting the potential for severe hydrological impacts.

These findings correspond to the highlights provided by Tebaldi et al. (2021), which could lead to a heightened risk of extreme rainfall events that could lead to more frequent and severe flooding due to higher river flow while significantly altering groundwater recharge. The anticipated escalation in rainfall variability underscores the necessity for adaptive water resources management strategies that could mitigate the adverse effects on groundwater recharge cycles and river discharge patterns, enhancing the climate resilience of the watershed. This growing climate variability over time enables Philippine policies, e.g., the Integrated Water Resources Management Plan (IWRMP), to recommend actions with strict timelines for sustainable water use (Department of Environment and Natural Resources (DENR, 2024).

Inter-model variability in projected monthly precipitation anomaly

Figure 5, which plots the ten-member CMIP6 ensemble (Figure 5), reveals substantial inter-model spread in projected mean monthly precipitation anomaly (mm month -1) over the PRW across all four SSP scenarios and future periods, underscoring that characterize the full range of plausible hydrological futures must be considered. Under SSP1-2.6, the Current Future (CF, 2021–2040) ensemble yields a median anomaly of +8.86 mm month -1 relative to the 1981–2000 baseline (IQR: +4.06 to +16.23 mm month -1; total range: 30.38 mm month -1), indicating a general wetting tendency but with meaningful model disagreement even in the near-term.

The spread widens through the Near Future (NF, 2041–2060; IQR-width: 14.13 mm month -1) and peaks in the Mid Future (MF, 2061–2080; median: +15.77 mm month -1; IQR-width: 20.78 mm month -1), before partially converging in the Far Future (FF, 2081–2100; median: +12.03 mm month -1; IQR-width: 15.02 mm month -1), a pattern consistent with the stabilizing radiative forcing trajectory of the low-emission pathway. Under SSP2-4.5, inter-model spread amplifies progressively: IQR-width grows from 8.47 mm month -1 in the CF to 21.90 mm month -1 in the MF, before settling at 18.42 mm month -1 in the FF, where the ensemble mean reaches +13.83 mm month -1, suggesting a moderate but increasingly uncertain wetting signal over the latter half of the century.

Inter-model spread of projected mean monthly precipitation anomaly (mm month -1)

The most pronounced departure from ensemble consensus occurs under SSP3-7.0 in the CF, where the ensemble median is negative (−2.60 mm month -1) and the IQR spans the widest range of any CF period across all scenarios (26.73 mm month -1), with six of ten models projecting drying (MRI-ESM2-0: −18.32 mm month -1; CNRM-CM6-1: −17.70 mm month -1; MPI-ESM1-2-HR: −14.04 mm month -1; CNRM-ESM2-1: −13.62 mm month -1) against four projecting wetting of up to +23.63 mm month -1 (CanESM5). This near-term bimodal response under an intermediate forcing pathway reflects the dominance of natural internal variability over the forced signal when cumulative radiative forcing has not yet imposed a clear precipitation trend, a behaviour documented across tropical monsoon regions by Tebaldi et al. (2021). By the FF under SSP3-7.0, the total model range expands dramatically to 73.40 mm month -1 — the largest across all scenario-period combinations in this study, driven by the sharply divergent trajectories of EC-Earth3 (+49.03 mm month -1) and MRI-ESM2-0 (−24.37 mm month -1), two models with contrasting representations of convective parameterization and western Pacific sea surface temperature feedbacks (Wu et al., 2024). Under SSP5-8.5, the ensemble mean increases monotonically from +3.48 mm month -1 in the CF to +22.48 mm month -1 in the FF, with the IQR widening to 26.45 mm month -1 and the total FF range reaching 56.24 mm month -1, anchored at the upper bound by EC-Earth3 (+58.47 mm month -1) and MPI-ESM1-2-HR (+43.98 mm month -1). Notably, the CF under SSP5-8.5 exhibits the tightest IQR of any scenario-period combination (2.16 mm month -1), reflecting a transient phase of model agreement before forced signals amplify divergence in later periods. Across all scenarios, MRI-ESM2-0 is the only model projecting negative FF anomalies under SSP2-4.5 (−3.96 mm month -1) and SSP3-7.0 (−24.37 mm month -1), confirming that a drying outcome, while not the ensemble consensus, remains within the physically plausible range and must not be excluded from water resources planning. This pervasive inter-model divergence is consistent with Wu et al. (2024), who demonstrated that model structural uncertainty accounts for approximately 73% of total CMIP6 projection spread for precipitation globally, a proportion that may be further elevated in monsoon-influenced watersheds such as the PRW where local orographic forcing and land-sea interactions amplify GCM sensitivity differences. Collectively, these results indicate that groundwater recharge management and irrigation planning in the PRW should be designed against a broad range of precipitation futures rather than anchored to a single ensemble trajectory.

Projected impact of climate change on river discharge

The calibrated and validated SWAT model simulated future river discharge in the Padsan River Watershed under various climate change scenarios, comparing these projections with historical discharge data from 2000 to 2020. Figure 6 (Top Line Graph) shows that the historical river discharge displays seasonal fluctuations driven by annual precipitation cycles. From 2020 onward, all SSP scenarios predict an increase in river discharge, with higher-emission pathways such as SSP585 and SSP245 showing more substantial increases. The peaks in discharge are more pronounced under SSP585, reflecting more frequent and intense precipitation events, whereas SSP126 shows a moderate increase, closer to historical values, indicating that lower emissions result in less drastic changes.

The relative changes in river discharge compared to the historical baseline (Table 4; Figure 6) show a consistent upward trend across all SSP scenarios. SSP585 exhibits the steepest increase, indicating more extreme shifts in water flow over time. Even under low-emission pathways like SSP126, the gradual rise in discharge highlights that future river flow will exceed historical norms, with relative changes ranging from 116% to 119%.

Furthermore, Table 4 provides the average net river discharge values (net and absolute) and each scenario’s standard deviation for four future decades (2020 to 2100). High river discharge averaged around 723.77 m 3 /s, while the projected future discharge substantially increased across all SSP scenarios. The average discharge ranges from 844.75 m 3 /s under SSP370 to 866.83 m 3 /s under SSP126, reflecting the overall increase in water flow at the watershed outlet regardless of the emission pathways. Notably, the standard deviation remains relatively stable across all scenarios, indicating that while the absolute discharge values increase, the discharge variability (or fluctuation) is consistent with historical patterns. This suggests that while more water will flow through the watershed on average, the seasonal fluctuations and interannual variability will remain within a similar range, though at higher discharge levels. The relatively stable standard deviation values also imply that the watershed will continue to experience predictable seasonal cycles. Still, these will occur at a higher baseline discharge, particularly in higher-emission scenarios. The average relative changes in Table 4 support these claims. The average relative change remains consistent across all scenarios, with values ranging between 116% (SSP370) and 119% (SSP126 and SSP585). These relative changes indicate that river discharge will be approximately one-fourth higher than the historical baseline on average across the projection period. While the increase is relatively consistent, the standard deviation values remain predictable. SSP126 and SSP585 exhibit slightly higher averages than SSP245 and SSP370, suggesting that the lowest and highest emission scenarios will lead to more significant relative discharge increases than the medium-emission scenarios. The overall predictability of relative changes increases in river discharge over time, driven by climate change.

Furthermore, the consistent and substantial increase in projected river discharge across SSPs reflects the compound action of three interconnected hydrological processes within the SWAT water balance framework. First, intensified precipitation amplifies surface runoff through the nonlinear SCS Curve Number mechanism, the second most sensitive calibrated parameter (R_CN2; fitted value: +0.08) (Table 2). In PRW’s Climate Type I regime, the extended wet season rapidly saturates agricultural soils, reducing infiltration capacity and routing a disproportionately large fraction of rainfall to overland flow (Arnold et al., 2012; Neitsch et al., 2011; Singson et al., 2023). Second, while rising temperatures increase potential evapotranspiration (PET), this does not offset the precipitation-driven runoff gains. In tropical monsoon systems, enhanced convective rainfall during peak wet-season months consistently outpaces marginal PET increases, sustaining a net water surplus that feeds both surface runoff and subsurface percolation (Tebaldi et al., 2021; Khoi et al., 2022). Third, the calibrated baseflow alpha factor (V_ALPHA_BNK; fitted value: 0.03) reflects a slow groundwater recession characteristic of PRW’s alluvial and volcanic geology. Greater precipitation increases aquifer percolation, sustaining baseflow between storm events and explaining why discharge variability remains relatively stable (Table 4) even as mean discharge rises. Collectively, CN-mediated runoff amplification, precipitation–ET imbalance during the monsoon peak, and sustained slow-release baseflow constitute the dominant drivers of the projected discharge increases, with their combined effect expected to intensify under higher emission pathways as both the wetting signal and thermal forcing strengthen toward the far future (Tebaldi et al., 2021; Wu et al., 2024).

These findings that capture the impact of CMIP6 on river discharge of PRW are comparable with the findings by Tolentino et al. (2016), which generally concluded a general increase of water flow for most Philippine watersheds. Similar to the findings of Jimenez et al. (2022), the impact of climate change disrupts the seasonal pattern of discharge, affecting water availability. This comparison potentially aggravated current drought (Alonzo et al., 2023) and flooding (Graciela et al., 2018) events in the watershed and the province. On the other hand, the findings are comparable to the improvement of previous studies, which used the previous version of climate change. CMIP6 projects a broader range of future climate outcomes with higher radiative forcing scenarios and increased climate sensitivities leading to greater projections (Tebaldi et al., 2021). This boosts its significant application in the watershed, which is highly vulnerable to flow variability due to its geographical location and challenging adaptation measures (Ignacio-Reardon and Luo, 2023).

Seasonal variability of RiverQ across SSPs and futures

In addition to the annual long-term impact of climate change, the seasonal response (Figure 7) carries distinct implications for water resources management. PRW operates under a pronounced monsoonal regime, with a wet season from May to October and a dry season from November to April, a distinction that determines irrigation viability, flood exposure, and aquifer replenishment windows. Disaggregating projected discharge by season reveals that climate change will not affect both periods equally (Figure 7).

Wet-season discharge increases across all scenarios, all timeframes. Under SSP585 it reaches 30.26% above baseline in the far future, the highest seasonal signal in the dataset. SSP126 and SSP245 both stay above 24%, peaking around 28% in the long future. SSP370 is the outlier — 20.20%–24.19%, still a meaningful departure from baseline but the most conservative of the four. These amplified wet-season flows are not unexpected; intensified monsoonal precipitation under greenhouse forcing has been well-documented for the region (Tebaldi et al., 2021). What they mean operationally is that PRW’s river system, already peaking at 762.84 m 3 /s in September under baseline conditions, will see higher and faster peak flows. The River Irrigation System infrastructure was not designed around those numbers.

Dry-season discharge is a different problem. The percentage increases are smaller — 12.97%–17.79% across scenarios—but the direction is consistent and the implications are more immediately actionable. SSP585 shows a steady climb from 12.97% in the near future to 17.79% by the end of the century. SSP126 and SSP245 hold between 15% and 17%, peaking in the long future. This moderate increase in dry-season flow is the clearest opportunity in the dataset: if captured through retention infrastructure or updated NIA operating rules, it reduces groundwater extraction pressure during the months when irrigation demand is highest. That said, higher dry-season flows without deliberate storage simply pass through as baseflow—the opportunity does not capture itself.

One structural point worth flagging for planning: SSP126 and SSP245 peak in the long future (2061–2080), while SSP585 continues intensifying through the far future (2081–2100). Moderate-emission adaptation strategies have a mid-century target window. High-emission scenarios require infrastructure designed for conditions that do not materialize until late century.

Projected impact of climate change on groundwater recharge

Considering the increasing importance of subsurface resources in the face of climate change, the fragmented institutional management (Valenzuela and Gutierrez, 2019) leads to scarce frameworks that challenge adaptation measures (Ignacio-Reardon and Luo, 2023) for the watershed. With this data scarcity in the watershed management, the reliance on the process-based model to provide the foundation in establishing governance of the subsurface resources recharge in the face of climate variability is essential, highly recommending the significant contribution of consolidated data management for institutions.

While the SWAT model was calibrated and validated against observed river discharge data from the NIA-IN, the simulated GWR estimates remain physically grounded through the model’s process-based water balance framework. Within SWAT, GWR is computed as the net percolation flux through the soil profile into the shallow aquifer (Arnold et al., 2012; Neitsch et al., 2011), governed by parameters that are simultaneously constrained during discharge calibration. Notably, the baseflow alpha factor (V_ALPHA_BNK, ranked most sensitive; fitted value: 0.03), the SCS Curve Number (R_CN2), and effective hydraulic conductivity (R_SOL_K) directly regulate the partitioning of precipitation between surface runoff, baseflow, and subsurface recharge. By achieving satisfactory calibration of river discharge—which integrates both surface and subsurface contributions—the model’s internal water balance, including recharge, is implicitly constrained (Arnold et al., 2012; Abbaspour et al., 2017). This approach is consistent with comparable tropical SWAT applications where direct groundwater observation data are unavailable; Khoi et al. (2022), working in a similarly monsoon-influenced tropical watershed in Vietnam, applied an analogous discharge-calibrated SWAT framework to derive CMIP6-driven GWR projections without independent recharge validation, and Singson et al. (2023) adopted the same approach within Philippine watershed contexts. Nevertheless, the absence of direct GWR validation introduces uncertainty inherent to parameter equifinality—where multiple parameter sets may yield similar discharge performance while producing divergent subsurface flux partitioning (Abbaspour et al., 2017). This uncertainty is further compounded by SWAT’s simplified lumped representation of the shallow aquifer, which may not fully capture the spatial heterogeneity of subsurface dynamics in the PRW (Taylor et al., 2013; Stigter et al., 2023). Given these constraints, the GWR projections in this study are best interpreted as directional trend estimates within the reported range of variability across SSPs and future periods, rather than precise absolute values. The directional consistency between the projected GWR increases (12.21%–64.34% across scenarios) and the wetting tendency identified in the precipitation anomaly analysis further supports the physical plausibility of these estimates. Future studies should prioritize installing groundwater monitoring infrastructure in the PRW and applying coupled SWAT-MODFLOW frameworks to enable direct subsurface calibration and reduce structural model uncertainty (Valenzuela and Gutierrez, 2019; Tabios III, 2020).

The calibrated SWAT model simulated groundwater recharge from 2020 to 2100 under four SSP scenarios. The general trend (Figure 8) shows periodic fluctuations in recharge driven by seasonal variability and a long-term increase across all scenarios. Higher emission scenarios (SSP585 and SSP245) exhibit steeper trends, indicating more significant increases in recharge rates, while SSP126 and SSP370 show moderate increases. This suggests that climate change, particularly under high-emission pathways, will lead to increased groundwater recharge, primarily due to intense precipitation events that may pose flooding risks rather than contributing to sustainable groundwater replenishment.

In addition, the relative changes in mm/mm are also plotted in the projected futures. The relative changes highlight that SSP126 has the highest variability, with peaks reaching nearly 0.6, especially towards the latter half of the century. The climate low emission projection has the highest mean relative change (27.68%) over the rest of the 21 st century (Table 6), which can fluctuate more compared with other climate change scenarios, as evidenced by the standard deviation to mean (±11.42). It can increase from 12.21% to 64.34% from the current annual recharge, resulting in an estimated average annual groundwater change of 116.01 ± 218.79 mm SSP126. Compared to SSP126, the other three climate change scenarios exhibit more stable relative changes, with SSP245 and SSP370 staying consistently lower. SSP245, the moderate emission, posts a relative change to the groundwater recharge of 27.61% ± 3.98%, which can increase the average annual groundwater recharge by 20.86%–39.13%. This resulted in a 70.95 mm to 153.32 increase in the watershed groundwater recharge. SSP370, considered a high emission, has the lowest fluctuation of changes. The average relative change is 24.32% ± 3.98%, which resulted in an increase of 57.49–134.20 (mean of 103.28 mm ± 134.20) from the current annual groundwater recharge. While the high emission had the lowest average relative change, SSP585 was labeled as the extreme (or very high) emission, followed closely by SSP 245, having a relative change of 27.26% ± 4.44% with lesser fluctuations ranging from a minimum of 18%–40.13%. In response, the projected increase in the current annual groundwater recharge was 77.59 mm–157.23 mm.

The relative changes reflect significant deviations from the baseline recharge, driven by precipitation variability under climate forcing (Garcia-Menendez et al., 2017). The response of groundwater recharge rate on lower emissions exhibits extreme sensitivity to rainfall shifts compared to higher emissions with stable recharge rates within a narrower range. These findings highlight the differential sensitivity of groundwater recharge to precipitation variability under low-emission pathways, where natural internal variability dominates the recharge signal, compared to higher-emission scenarios where human-induced forcing progressively narrows the range of plausible recharge outcomes (Reinecke et al., 2021; Taylor et al., 2013) compared to higher emissions with enhanced human-induced forcing. On the other hand, all SSPs indicate an increasing groundwater recharge rate, indicating the potential increase of the groundwater supply that can serve as a buffer against the increasing variability of climate, considering the sustainable extraction by incorporating the variability across time and space in the PRW, which could be implemented with thorough governing policies.

Furthermore, these local findings are contextualized within the growing body of CMIP-SWAT studies in tropical and monsoon-dominated watersheds across the region. Khoi et al. (2022) have a methodological proximate analogue where they applied a discharge-calibrated SWAT framework driven by a seven-model CMIP6 ensemble across 4 SSPs in the tropical monsoon watershed in Ho Chi Minh City of Vietnam. They projected an increase in both GWR and riverQ that are consistent with the local PRW results. In a similar Philippine context, Singson et al. (2023) demonstrated comparable discharge sensitivity to CMIP6 climate forcings in the Magat Reservoir watershed, while Panondi and Izumi (2021) reported a maximum mean annual precipitation increase of 39.10 under CMIP5 scenarios in the Pulang River Basin, reinforcing that upward discharge trajectories are a recurring hydrological signal across Philippine watersheds exposed to intensifying monsoon precipitation.

In the broader Southeast Asia context, Khadka et al. (2023) applied SWAT with an eight-model CMIP6 ensemble under SSP5-8.5 in the Mun River Basin of Northeast Thailand, a tropical agricultural watershed with a hydrometeorological regime broadly similar to the PRW. They projected an increased annual water yield and runoff, highlighting flood risk that also characterizes the PRW’s projected discharge dynamics under higher-emission pathways. In contrast, Kumar et al. (2022), applying SWAT with a three-model CMIP6 ensemble across various timeframe in the semi-arid, monsoon-driven upper Betwa River Basin of central India, projected mean monsoon streamflow decreases of 39%–47% during 2023–2060 and 31%–47% during 2061–2100 relative to the 1982–2020 baseline. Singh et al. (2025), a divergent outcome that underscores the region-specificity of CMIP6-SWAT projections and confirms that the PRW’s consistent discharge increase reflects local orographic and hydrometeorological conditions rather than a universal tropical response. This regional contrast is consistent with Wu et al. (2024), who attributed approximately 76% (runoff) of the total CMIP6 projection spread to model structural differences, a source of uncertainty that operates alongside basin-scale physiographic controls in shaping divergent hydrological outcomes across tropical watersheds globally.

Seasonal variability of groundwater recharge across SSPs and futures

The annual GWR projections in Table 6 are useful for headline planning and supplementing a seasonal dynamic (Figure 9) that determines whether the projected recharge increases are actually harvestable. Wet-season recharge increases consistently across all scenarios. SSP126 ranges from approximately 29% in the near future to 31.15% at peak; SSP245 follows the same arc, topping at 30.73% in the long future. SSP585 starts modestly but reaches 32.00% in the far future—the highest wet-season value in the dataset. SSP370 stays between 25% and 28%, consistent with its conservative hydrological response across other variables. These gains are driven by amplified monsoon precipitation. But the water balance tells the rest of the story: wet-season outflows also increase substantially, from approximately 1,318 mm at baseline to over 1,410 mm under SSP585 in the far future. A portion of the additional precipitation runs off rather than infiltrates, especially under high-intensity events that exceed the soil’s infiltration capacity. The gross recharge increase is real; the net gain available to the aquifer depends on how much is intercepted before it leaves the watershed.

The dry-season projections carry the more important finding. PRW’s dry season has historically received around 30 mm of rainfall in agricultural areas, and aquifer replenishment during these months runs almost entirely on delayed percolation from wet-season infiltration. That pattern is projected to change. Dry-season recharge increases by 29%–31% under SSP126, and comparably under SSP245 and SSP370. Under SSP585, it reaches 32.16% in the far future—marginally exceeding the wet-season value of 32.00% for the same period. A dry season with higher recharge than the wet season has no precedent in the baseline record. The most plausible explanation is episodic extreme precipitation outside the conventional monsoon window, consistent with tropical rainfall seasonality changes documented under high-forcing scenarios. These off-season pulses add recharge volume, but they arrive in short bursts rather than sustained infiltration events. Without pre-positioned MAR infrastructure—percolation ponds, check dams, recharge wells in the high-potential zones this study identified—the additional recharge potential dissipates through lateral drainage before the aquifer can capture it.

Two practical implications follow. First, wet-season MAR infrastructure should be sized against projected intensified peaks, not historical baselines—a recalibration of the design standard, not just an incremental adjustment. Second, dry-season water budgeting for NIA irrigation planning should account for the projected recharge increase but cannot treat it as reliable. Interannual variability under SSP126 (±11.42% SD at the annual scale) translates to high uncertainty in any given dry season. Planning to the median without accounting for that spread systematically underestimates drought-year deficits.

Strengthening national policies with data-driven hydrological changes

The Philippines has long struggled with fragmented water governance (Tabios III, 2020), a structural problem that the proposed Department of Water Resources (National Water Resources Management Act S. No. 2771, 19th Congress, 2024) is designed to address. The PDP 2023–2028 (NEDA, 2023) responds to the same problem by calling for science-based, integrated water planning that combines hydrological modeling with satellite-derived data. The projections from this study provide exactly that evidence base for PRW, but the implications are more operationally specific than general policy alignment suggests.

On the discharge side, the projections indicate that wet-season river discharge in PRW will intensify substantially across all SSPs, with annual discharge increasing by 108.76%–134.25% relative to the baseline period. This is not a uniform increase that existing infrastructure can absorb incrementally. NIA diversion structures in PRW were designed against historical discharge norms; the projected intensification, especially under SSP5-8.5, where forcing continues through the Far Future (2081–2100), implies that design floods used for infrastructure sizing are systematically underestimated. The PDP (NEDA, 2023) calls for reassessing existing water infrastructure under climate variability, and this study’s discharge projections provide a quantified basis for that reassessment, namely,: flood retention capacity, spillway design standards, and irrigation diversion thresholds at PRW should be recalibrated against SSP5-8.5 upper-bound projections rather than historical extremes. Under lower-forcing pathways (SSP1-2.6 and SSP2-4.5), where discharge peaks in the Middle Future before stabilizing, a phased infrastructure investment schedule, with a mid-century trigger, is financially and technically defensible. Under SSP3-7.0 and SSP5-8.5, intensification continues beyond mid-century, and infrastructure designed only for Near or Middle Future conditions will be undersized by the time the Far Future arrives.

On the recharge side, the findings introduce a more nuanced management problem. GWR increases across all scenarios—from 12.21% under SSP1-2.6 in the Near Future to 40.13% under SSP5-8.5 in the Far Future—and the dry-season recharge increase of 29%–32% across SSPs represents a departure from the baseline pattern, where aquifer replenishment during the dry season depended almost entirely on delayed percolation from the preceding wet season. While the PDP (NEDA, 2023) cautions against groundwater overexploitation, the projections here suggest a more differentiated picture: recharge is increasing, but it arrives in episodic pulses rather than sustained infiltration events, particularly under high-forcing scenarios. Without MAR infrastructure pre-positioned at watershed scale—percolation ponds and check dams in recharge-favorable zones—these pulses dissipate through lateral drainage before the aquifer captures them. Protecting and demarcating recharge zones in PRW is therefore not only consistent with PDP goals for aquifer protection; it is the prerequisite for translating projected recharge increases into actual groundwater storage gains. The MRI-ESM2-0 drying signal under SSP2-4.5 and SSP3-7.0, while not the ensemble consensus, is physically plausible and provides a further argument for locking in recharge zone protections before conditions shift: once upland cover loss or land conversion removes the infiltration capacity, no amount of favorable rainfall translates into groundwater storage.

The high interannual variability projected under SSP1-2.6 (±11.42% SD at annual scale) raises a separate operational concern for NIA irrigation planning in PRW. Current irrigation release schedules, calibrated to historical baseflow, do not account for the expanded range of likely annual recharge. Planning to the projected median without accounting for the spread of outcomes systematically underestimates drought-year deficits; in years at the lower end of the SSP1-2.6 recharge distribution, irrigation water availability could fall well below what median projections suggest. The proposed IWRM framework under PDP (NEDA, 2023) and the mandate of the proposed DWRMA provide the institutional basis for a rule-based adjustment: NIA operating schedules for PRW should incorporate SSP-conditioned recharge projections as an adaptive input, with recalibration at each transition between projection periods (CF → NF → MF → FF) as updated CMIP7 or successor ensembles become available. This is not an unprecedented approach; Alejo and Alejandro (2022) demonstrated that irrigation planning parameters in Philippine watersheds shift materially under projected climate, requiring systematic updates rather than single-point design assumptions.

Taken together, the projections from this study translate into three concrete watershed management recommendations for PRW: (1) recalibrate the design standards for NIA diversion infrastructure against SSP5-8.5 Far Future discharge bounds, not historical baselines; (2) delineate and legally protect MAR-suitable zones across PRW as a precondition for capturing projected recharge gains, with MAR structures sized to projected wet-season peak intensification; and (3) integrate SSP-conditioned hydrological projections into NIA irrigation operating rules as a dynamic, periodically updated input rather than a fixed planning parameter. These recommendations are grounded directly in the quantified outputs of the SWAT-CMIP6 framework applied here and operationalize the science-based planning mandate of both the PDP (NEDA, 2023) and the proposed DWRMA (2024). Implementing them within PRW would also generate a replicable model for other Climate Type I watersheds in Northern Luzon facing the same combination of intensifying wet-season discharge, episodic dry-season recharge pulses, and high interannual variability under low-emission stabilization pathways.

This study integrated validated open-access geospatial data into a calibrated SWAT framework for the PRW. The SWAT model performed satisfactorily during calibration. The decline in NSE and the increase in PBIAS during validation reflect the model’s known limitation in representing extreme discharge events in a single-station, data-scarce watershed, a constraint consistent with findings from comparable Philippine SWAT applications, underscoring the need for future studies in modeling extreme events. Despite these limitations, the model’s satisfactory performance provided a reliable basis for simulating river discharge and groundwater recharge under current and projected climate conditions.

Driven by ten CMIP6 models across four SSP scenarios in four futures, the projections indicate that all emission pathways will intensify hydrological variability in PRW through the 21st century. Annual river discharge is projected to increase by up to 1.34 times across SSPs and future periods, with wet-season intensification posing the dominant threat to existing irrigation infrastructure and flood-control structures. Groundwater recharge is projected to increase by 12.21%–64.34% across the same range, with dry-season recharge rising by around 30 percent under all SSPs. These hydrological shifts are neither uniform nor monotonic across scenarios: the low-emission pathway produces high interannual variability, while the high-emission pathway projects continued intensification through the Far Future, compounding infrastructure and water security risks beyond mid-century.

Yet, these results carry inherent limitations that must be considered and carried to future studies. Three categories of uncertainty are acknowledged. First, regarding climate model assumptions: CMIP6 projections via ISIMIP3B bias correction assume stationarity in the bias correction function across the projection period. Inter-model structure differences in cloud microphysics, ocean-atmosphere coupling, and land-surface parameterization produce the inter-model spread documented across SSPs and periods. This spread is a feature of the ensemble, not a deficiency. This means that results must be interpreted in probabilistic terms rather than as deterministic trajectories.

Second, it is in the input datasets. CHIRPS exhibits known limitations in capturing localized high-intensity precipitation events, and the PRW’s single ground-based meteorological station constraints spatial validation of the bias-corrected product. Third, there is the structural uncertainty of the SWAT model. The model’s simplified linear reservoir representation of the shallow aquifer does not resolve some of the advanced aquifer model’s features, which introduce structural uncertainty into the GWR projections. Land use and land cover are held constant over the 2020–2100 simulation horizon, a simplification that may progressively underestimate or overestimate hydrological responses as agricultural, urban, or reforestation dynamics evolve. Future research should address these constraints through SWAT-MODFLOW coupling, multi-station precipitation networks, and dynamic LULC scenarios.

Despite these constraints, the study’s projections enable comprehensive water resources management that considers the impact of future climate variability. The discharge intensification establishes a quantified basis for recalibrating NIA infrastructure design standards. The consistent GWR increase justifies pre-positioning MAR infrastructure at the watershed scale. The high variability under low-emission pathways argues for SSP-conditioned adaptive operating rules, the science-based mandate of the PDP Plan, and the proposed NWRM Act, offering a replicable open-access framework for data-scarce Climate Type I watersheds across the country.