Tropical hydro-climatic responses to global warming and solar radiation modification in the Kelantan River Basin, Malaysia

Since the pre-industrial era, human activities have increased greenhouse gas concentrations, which are the primary drivers of global warming (Masson-Delmotte et al. 2021). A direct consequence of global warming is the rise in atmospheric temperatures, leading to enhanced evaporation and an increased capacity of the atmosphere to hold water vapor. The intensification of the hydrological cycle exacerbates heavy precipitation and flooding but may also contribute to prolonged drought periods in terrestrial regions (Masson-Delmotte et al. 2021). In other words, wet areas are expected to become even wetter in the future, while dry regions may become drier (Held & Soden 2006).

With insufficient efforts to reduce CO₂ emissions, solar radiation modification (SRM) has emerged as one of the most discussed strategies for preventing global temperature from rising more than 1.5 °C above the pre-industrial level. SRM aims to rapidly cool the Earth by reducing incoming shortwave radiation to offset the warming effects (Crutzen 2006; Masson-Delmotte et al. 2021). Hydro-climatic SRM studies primarily focus on extreme temperatures, extreme precipitation (Jones et al. 2013), throughflow (Shen et al. 2023), streamflow (Tan et al. 2023b), evaporation (MacMartin et al. 2022), and snowmelt (Ridley & Blockley 2018). Among these hydrological components, studying streamflow provides valuable insights into water availability and related disasters such as floods and droughts. Compared to Representative Concentration Pathways (RCP) 4.5, the GeoMIP G4 scenario results in increased streamflow in western Eurasia and North America, alongside reduced streamflow in their eastern part (Wei et al. 2018). Under the G1 or 4 × CO₂ scenario, Tilmes et al. (2013) reported a global precipitation increase of approximately 6.9%, accompanied by notable zonal and regional differences. For instance, precipitation increased by 10% over Asia and decreased by 7% in the North American summer monsoon. In Africa, the likelihood of dry conditions increased, while the probability of wet conditions declined during the installation and termination stages of Stratospheric aerosol injection (SAI) (Obahoundje et al. 2023).

In recent years, scholars have raised concerns about the validity of simulating SRM scenarios (Visioni et al. 2023). The Geoengineering Model Intercomparison Project (GeoMIP), a sub-project of CMIP6, provides a standardized framework for SRM modeling experiments, enabling more robust comparisons of model responses and uncertainties. In the sixth phase of GeoMIP (GeoMIP6), the G6solar and G6sulfur experiments were proposed to transition radiative forcing from the high-emission scenario (Shared Socioeconomic Pathway 5-8.5; SSP5-8.5) (Meinshausen et al. 2020) to the intermediate-emission scenario (SSP2-4.5) by reducing the solar constant or injecting SO₂ into the stratosphere. Based on the GeoMIP6 guideline, the mean global 2 m air temperature under the G6solar or G6sulfur scenarios must be within 0.2 K of the corresponding decade for each model's SSP2-4.5 simulation from 2021 to 2100 (Kravitz et al. 2021). Visioni et al. (2021) reported that the global mean precipitation response during the 2081–2100 period is projected to vary from −3.79 ± 0.76 % for G6sulfur and −2.07% ± 0.40% for G6solar compared to SSP2-4.5. Similar GeoMIP6 studies have explored regional scales, including China (Wang et al. 2023), East Asia (Liang & Haywood 2023), and Indonesia (Shen et al. 2023). However, there is a notable lack of research on SRM impacts across Southeast Asia, particularly at localized and river basin scales.

The Kelantan River Basin (KRB) in Malaysia experiences frequent floods almost every year during the first phase of the northeast monsoon season, which spans from November to mid-January (Tan et al. 2021). Additionally, the KRB faces dry conditions during the second phase of the northeast monsoon and the first inter-monsoon, typically from February to May. These dry conditions reduce water availability and affect freshwater supply, as the basin lacks reservoirs to store water for the dry seasons. Monthly precipitation and streamflow in the KRB are projected to become more frequent and intense, according to simulations from 36 downscaled climate projections from CORDEX-SEA (Tan et al. 2017a, 2020). Tan et al. (2023b) were the first to investigate the impact of SRM in the KRB using Stratospheric Aerosol Geoengineering Large Ensemble (GLENS) members. They found that SRM could reduce precipitation during flooding months compared to RCP8.5, but it would also decrease precipitation during dry months. However, the GLENS model may not adequately address uncertainties from different climate models. Furthermore, it relies on simplified assumptions about SRM and primarily focuses on temperature control, potentially overlooking other SRM strategies such as G6solar and G6sulfur.

Previous research has demonstrated that SRM could effectively lower mean temperature; however, its impact on tropical precipitation and streamflow, which exhibit greater variability than other variables, remains unclear. Therefore, this study aims to develop a comprehensive and up-to-date hydro-climatic assessment framework for evaluating the impact of SRM on tropical hydro-climatic changes. Three specific objectives of this study are (1) to apply bias correctly to GeoMIP6 models, including CNRM-ESM2-1, IPSL-CM6A-LR, MPI-ESM1-2-LR, and UKESM1-0-LL; (2) to calibrate and validate the Soil and Water Assessment Tool plus (SWAT+) model for simulating streamflow in the KRB; and (3) to assess potential hydro-climatic changes in the KRB for the near-term (2025–2044), mid-term (2045–2064), and end-term (2065–2084) of the 21st century compared to the baseline period (1985–2004) under the SSP2-4.5, SSP5-8.5, G6solar, and G6sulfur scenarios. This study is among the first to apply SRM scenarios from GeoMIP6 at the river basin scale in Southeast Asia, providing new insights into the potential impacts of SRM on local climate dynamics and the hydrological cycle. Understanding these localized effects is crucial for developing countries that are particularly vulnerable to climate change, as it allows them to assess how SRM might affect their specific water resource systems, thereby supporting decision-making and policy development. Additionally, the findings can serve as a reference for the GeoMIP community in formulating more realistic SRM strategies for the upcoming GeoMIP7 scenarios.

The SWAT model, widely used in Southeast Asia (Tan et al. 2019), requires long-term, continuous daily data for precipitation, maximum temperature, and minimum temperature. Among the GeoMIP6 models, only CESM2-WACCM, IPSL-CM6A-LR, UKESM1-0-LL, MPI-ESM1.2-LR, and CNRM-ESM2-1 simulate these daily meteorological variables. However, the CESM2-WACCM model does not provide temperature data for historical periods (Visioni et al. 2023; Tan et al. 2024), preventing the use of observed data for bias correction. Therefore, this study utilizes the other four GeoMIP models listed in Table 1. The general findings are reported based on the ensemble mean data from these four models.

Kravitz et al. (2015) proposed G6solar and G6sulfur as part of the standard experiments in GeoMIP6. These experiments aim to reduce global mean temperatures from the high-emission SSP5-8.5 scenario to the moderate-emission SSP2-4.5 scenario. IPSL-CM6A-LR and UKESM1-0-LL are adjusted every 10 years, based on changes in the solar constant or specified increases in stratospheric SO₂ or aerosol, to achieve a surface temperature reduction aligning with SSP2-4.5. In contrast, MPI-ESM1.2-LR and CNRM-ESM2-1 implemented these adjustments annually (Visioni et al. 2021, 2023).

Similarly, and represent the corrected GeoMIP6 daily temperature data from 1985 to 2004 and 2021 to 2100, respectively, whereas and represent the raw GeoMIP6 and observed daily temperature data from 1985 to 2004, respectively. Besides that, the climatological characteristics of the original and bias corrected data from the four GeoMIP6 datasets are evaluated against observations.

The SWAT model is one of the most widely used hydrologic models in basins worldwide for the past decades (Arnold et al. 1998). In this study. the SWAT+ model, an upgraded version of the SWAT model, was used. The SWAT is based on the hydrological response unit (HRU), which consists of areas with similar soil, slope, and land use features. In the SWAT, HRUs are defined within a sub-basin, while in SWAT + , HRUs are specified within a smaller unit called the landscape unit, which is a sub-unit that forms the sub-basin. The SWAT+ model retains the same fundamental methods for calculating physical processes but introduces significant changes in the structure and organization of the input code and files (Andaryani et al. 2021; Rathjens et al. 2022).

The soil type and land use maps were used to generate the HRUs. The digital elevation model (DEM) was used to generate river networks and the slope of the basin. Observed daily precipitation, maximum temperature, and minimum temperature data collected from 1985 to 2004 were used to run the model. Monthly streamflow data obtained from the Department of Irrigation and Drainage Malaysia were used to calibrate and validate the SWAT+ model for the same period, from January 1985 to December 2004. The SWAT+ toolbox was used for sensitivity analysis, model calibration, and verification. Based on the parameters employed in the calibration of the SWAT+ model by Tan et al. (2023b) and Yang et al. (2022), a total of 14 parameters were identified, as shown in Table 2. Among them, curve number condition II (CN2) and lateral flow coefficient (LATQ_CO) were identified as the most sensitive parameters for the KRB, which is consistent with the findings reported by Tan et al. (2023b). The fitted values indicate the optimal parameter values for accurately simulating streamflow in the KRB. This study utilized two popular statistical methods, Nash–Sutcliffe Efficiency (NSE) and coefficient of determination (R²), to assess the performance of the SWAT in simulating streamflow. The performance of the SWAT can be categorized as ‘very good’ if NSE > 0.8 and R² > 0.85, ‘good’ if NSE > 0.7 and R² > 0.75, ‘satisfactory’ if NSE > 0.5 and R² > 0.6 (Tan et al. 2021). Similarly, the SWAT+ model was evaluated using these standards.

Figure 3 shows that the original GeoMIP6 models underestimated the mean maximum monthly temperature and overestimated the mean minimum monthly temperature compared to the observed data. The temperature differences between the original GeoMIP6 models and observations range from 1 to 4 °C. The peak mean monthly maximum temperature in April, consistent with the observed data, is captured by most of the original models, except for CNRM-ESM2-1, which records it in June. Regarding the mean minimum monthly temperature, the original UKESM1-0-LL failed to capture the peak in April, whereas MPI-ESM1-2-LR recorded the peak 1 month later. After bias correction, most of the models successfully captured the climatological temperature characteristics over the basin. However, UKESM1-0-LL differed from the other three models in terms of the mean minimum monthly temperature, with the highest corrected values in October, while the other models showed the highest values in April. In contrast, the corrected CNRM-ESM2-1 simulated the lowest mean monthly minimum temperature in January, while the other models and observations indicated this occurrence in February. Similar to monthly precipitation, the bias-corrected mean monthly maximum and minimum temperatures now exhibit a closer match with the observed data.

The ensemble mean projected an increasing trend in annual precipitation in the middle of the 21st century, followed by a decreasing trend toward the end of the 21st century under the evaluated emission scenarios, except for the G6sulfur scenario, which shows a continuous decrease in precipitation from 2025 to 2084 (Figure 4 (a)–(d)). In particular, the reduction in precipitation is most pronounced under the two G6 scenarios, with the G6sulfur scenario showing the greatest decrease (slope = −8.32). The precipitation projections from the four models are consistent with previous studies, indicating a reduction in mean precipitation when compared to the SSP5-8.5 scenario (Henry & Mendonca 2020). The drier conditions projected under the G6sulfur scenario are particularly evident in the 2065–2084 period, as indicated by the IPSL-CM6A-LR and UKESM1-0-LL models, with a decrease of up to 8.66% as detailed in Table 3. The mean annual precipitation is projected to increase by 7.55–14.27% under the SSP5-8.5 scenario. These increases could be reduced by 1.9 to 12.72%, −2.8 to 5.09%, and 5.21 to 9.42% under the G6solar, G6sulfur, and SSP2-4.5 scenarios, respectively. CNRM-ESM2-1 and IPSL-CM6A-LR projected decreases of annual precipitation in the 2045–2064 period under the SSP2-4.5 and G6sulfur scenarios, while UKESM1-0-LL projected decreases across all four emission scenarios in the 2065–2084 scenario. Overall, based on the ensemble mean, the total precipitation in most scenarios remains higher than the historical levels (1985–2004), except for the G6sulfur scenario in 2065–2084, which shows a slight decrease compared to the historical period.

Simultaneously, the multi-model ensemble mean shows a substantial increasing trend in annual maximum and minimum temperatures over the KRB from 2025 to 2084, particularly under the SSP5-8.5 scenario (Figure 4). Under the SSP5-8.5 scenario, there is a notable rise in both maximum and minimum temperatures, ranging from 1.04 to 3.52 and 0.75 to 3.09 °C, respectively (Table 3). The G6solar and G6sulfur scenarios limit the dramatic increases in maximum temperature projected under SSP5-8.5, ranging from 1.01 to 2.07 and 1.31 to 2.33 °C, respectively. For minimum temperature, these scenarios also limit the increases compared to SSP5-8.5, ranging from 0.76 to 1.83 °C (G6solar) and 0.82 to 1.72 °C (G6sulfur). These changes in maximum and minimum temperatures under SRM correspond closely with those under SSP2-4.5, ranging from 1.25 to 2.22 and 0.81 to 1.96 °C, respectively. In the SSP5-8.5 scenario, the temperature rise is driven by a significant increase in greenhouse gas concentrations in the atmosphere. In contrast, the SRM scenario mitigates this temperature increase by reducing climate forcing through human intervention. The primary cause of the cooling effect from SRM is the reflection of shortwave radiation back to space by the aerosols that are injected into the stratosphere. While this process helps to reduce the intense solar radiation reaching the equatorial region and consequently lowers temperatures, the increases remain below those observed in the SSP5-8.5 scenario.

When comparing the four models, UKESM1-0-LL stands out from the other models in most of the evaluated periods and scenarios. For example, in 2025–2044, except for UKESM1-0-LL, the other three models exhibit minor fluctuations compared to the baseline period. From December to April, there is an increase in monthly mean precipitation in UKESM1-0-LL, followed by a decrease from June to October. The most significant increase in precipitation during this period exceeds 100% of the baseline period, particularly in February. From June to October, there is a reduction in monthly mean precipitation compared to the baseline period, with the most significant reduction in July across all scenarios, ranging from 48.72 to 58.31%. MPI-ESM1-2-LR shows more precipitation over the spring season under certain scenarios, such as G6solar during 2045–2064, where the monthly mean precipitation in February is 241.26% greater than the baseline period. In the period of 2065–2084, UKESM1-0-LL projected an increase in monthly mean precipitation from February to August, while the other three models show a decrease in precipitation. UKESM1-0-LL shows a decrease in monthly mean precipitation from September to January, while the other three models show an increasing trend. In the MPI-ESM1-2-LR model, monthly mean precipitation under SSP5-8.5 shows an increase from January to April, with the most substantial increase in February, which is 163.37% greater than the baseline period. In summary, the precipitation simulations for the KRB from the UKESM1-0-LL and MPI-ESM1-2-LR models exhibit significant inconsistencies, potentially exacerbating extreme precipitation events.

Regarding model uncertainty, the increases and decreases in monthly mean maximum temperature for UKESM1-0-LL remain the most significant across the models. In the period of 2025–2044, it shows a notably larger temperature rise compared to the other three models from August to January and also the lowest value in April and May, as shown in Figure 6. In 2045–2064, the temperature variation of UKESM1-0-LL is consistently higher than that of the other three models from December to May. During the 2065–2084 period, it shows a larger temperature rise from November to March. MPI-ESM1-2-LR demonstrates the smallest increases in monthly mean maximum temperature across all scenarios, occasionally even displaying negative changes.

The multi-model ensemble mean projected the greatest increases in monthly mean minimum temperature during the 2065–2084 period across all four emission scenarios, as shown in Figure 7, similar to the trends in the monthly mean maximum temperature. In the 2025–2044 period, the monthly mean minimum temperature shows a significant rise at the start and end of the year. The smallest increase is observed in July under G6sulfur and in June for the other three scenarios, with increases ranging from 0.79 to 0.87 °C. The monthly mean minimum temperature is projected to increase moderately during the 2045–2064 period, ranging from 1.61 to 2.32 °C. In contrast, the 2065–2084 period records the highest rise in monthly mean minimum temperature in March, ranging from 3.09 to 4.57 °C. Meanwhile, the smallest temperature increases occur in September, ranging from 1.73 to 3.04 °C. The temperature variations observed during the 2025–2044 and 2065–2084 periods can largely be attributed to the influence of UKESM1-0-LL, particularly its decline seen in June during the 2025–2044 period. Meanwhile, the monthly mean minimum temperature had a substantial rise during the 2065–2084 period. A comparison among models reveals notable regional variations in simulation outcomes when considering climate change within the context of geoengineering scenarios (Cao et al. 2017).

This study employs four GeoMIP6 models and SWAT+ to evaluate the impact of SRM on hydro-climatic changes in the KRB, Malaysia, using a baseline period from 1985 to 2004 and three future periods of 2025–2044, 2045–2064, and 2065–2084, Four distinct scenarios are considered, including two SRM scenarios (G6solar and G6sulfur) and two Shared Socioeconomic Pathways (SSP2-4.5 and SSP5-8.5). The main findings of this study are as follows:

• The original MPI-ESM1-2-LR and UKESM1-0-LL models were unable to accurately capture the peak monthly precipitation in the KRB. Additionally, all four GeoMIP6 models showed an underestimation of maximum temperature and an overestimation of minimum temperature. The LSC approach is effective in reducing biases in precipitation and temperature within the original GeoMIP6 models, bringing them into closer alignment with the observations.

• G6solar and G6sulfur demonstrate effective cooling effects on both maximum and minimum temperatures, thereby limiting the temperature rise projected under SSP5-8.5, which could reach up to 3.52 °C by the end-21st century. This aligns with the temperature projections of SSP2-4.5. Monthly maximum and minimum temperatures are projected to increase in almost all the evaluated periods, except in April and May in the 2025–2044 period. This situation is attributed to the decrease in monthly temperature as projected by UKESM1-0-LL. Additionally, MPI-ESM1-2-LR demonstrates the most modest temperature increases compared to the other three models.

• The multi-model ensemble mean projected an increasing trend in annual precipitation from the early to mid-21st century, followed by a decline toward the end of the 21st century under SSP2-4.5, SSP5-8.5 and G6solar, while G6sulfur shows an overall decreasing trend over the same period. In general, annual precipitation in the KRB is projected to increase in the future compared to the baseline period, except for G6sulfur, which showed a decline in the 2065–2084 periods. The decreases in monthly precipitation under G6sulfur, particularly during the dry months, suggest that this strategy may exacerbate the water shortage issue in the KRB.

• The highest increase in mean annual streamflow is projected by SSP5-8.5, followed by SSP2-4.5, G6solar, and G6sulfur. In the 2025–2044 period, the increases in monthly streamflow are generally higher during the dry months from February to May compared to the other two periods. In contrast, the rise in monthly streamflow is higher during the 2045–2064 period, showing the potential for serious flooding in the mid-21st century in the KRB. Decreases in monthly streamflow were projected in the 2045–2064 period under G6sulfur, particularly in the middle and upper parts of the basin.

However, some limitations need to be addressed in future studies. MPI-ESM1-2-LR and UKESM1-0-LL show greater uncertainties compared to the other two models in the KRB. The model uncertainties may arise from factors such as model structure, regional topography, and atmospheric circulation (Ongoma et al. 2018; Berhanu et al. 2023; Xiao et al. 2023), where differences in initial and boundary conditions, along with the poor representation of convective schemes and model parameterizations, hinder models from capturing precipitation spatial patterns accurately. Climate models that are unable to accurately represent the climate system in the study area may not contribute to improvements in the ensemble mean (Berhanu et al. 2023). While the ensemble mean from a certain number of climate models is commonly accepted as a widely used approach to minimize the model uncertainty, it may not significantly enhance projection capabilities. The upcoming GeoMIP7 should include more climate models and climate variables, such as precipitation and maximum and minimum temperatures, that are suitable for running hydro-climatic modeling at the basin scale. This would allow the selection and ensemble of more appropriate climate models, thereby increasing the robustness of the findings.

Both G6solar and G6sulfur involve the artificial injection of aerosols into the stratosphere or the reduction of the solar constant in the model. Furthermore, there are uncertainties in stratospheric dynamics, such as radiative interactions and stratospheric chemistry interactions, which influence local radiation transmission, leading to varied local heating of the atmosphere and consequent impacts on surface climate and the stratosphere (Niemeier & Schmidt 2017; Kleinschmitt et al. 2018). Consequently, model simulations may not be as useful in reducing some uncertainties. New experiments should be designed to enhance the accuracy of modeling interactions in the atmosphere or improve model parameters for regional-scale simulations.

The SWAT+ model tends to overestimate low flows in the KRB, possibly due to the plant growth and land use models in the SWAT being developed for temperate regions, which differ from the characteristics of tropical regions like Malaysia (Strauch & Volk 2013; Tan et al. 2017a). In this study, hydrological data were solely sourced from a single site, lacking validation at multiple sites. Further investigation is required to determine if integrating groundwater and soil moisture data may enhance the reliability of SWAT+. CMIP6 models are primarily available at global and continental scales (Tan et al. 2020). Before applying them to local-scale assessments, downscaling to finer scales can increase the reliability of future climate assessments (Tan et al. 2020). Improvements in model resolution may also aid in more accurately simulating precipitation (Berhanu et al. 2023). Thus, it would be great to consider employing more robust downscaling techniques (Duan et al. 2017) or possibly include some GeoMIP models in the CORDEX-SEA downscaling initiatives for future SRM hydro-climatic studies.

This research was funded by the Ministry of Higher Education Malaysia through the Long-term Research Grant Scheme Project 2 (Grant No. LRGS/1/2020/UKM-USM/01/6/2), as part of the program of LRGS/1/2020/UKM/01/6. Additional support for this work was provided by the Degrees Initiative (Grant No. RGA-DMF23MYS). L.X. is supported by the US National Science Foundation (Grant No. AGS-2017113). The authors extend their gratitude to the editors and reviewers for their valuable and constructive comments, which have greatly improved this manuscript.

H.D. carried out methodology, software, validation, formal analysis and wrote the original draft preparation. M.L.T. performed conceptualization, supervision, data curation, visualization, project adminstration, funding acquisition, and wrote the reviewed and edited the manuscript. L.X. wrote the reviewed and edited manuscript. Y.L.T. performed data curation. Z.M.Y. wrote the reviewed and edited manuscript.

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

The authors declare there is no conflict.