Model-based analysis of the impact of climate change on hydrology in the Guayas River basin (Ecuador)

Climate change is threatening global food security, which is one of the important challenges in the 21st century (Change 2013). There are substantial impacts of climate change on water resources, hydropower production, food security and human health (Magadza 2000). Particularly, water resources and crop production are affected due to changes in precipitation patterns and increase in temperature, among others (Ashraf Vaghefi et al. 2017). It is expected that there will be a 1°C and over 4°C temperature increase under RCP 2.6 (low emission scenario) and RCP 8.5 (high emission scenario), respectively (Change 2013). Hydrological cycles and crop production are also predicted to be altered due to the projected variations in precipitation patterns and temperature (Paul et al. 2017). Climate change is also expected to severely alter the hydrological regime such as the occurrence and severity of runoff, droughts and floods (IPCC 2007). Moreover, there is a severe and irreversible increase in ecological and socio-economic crises (Smale et al. 2019). Thus, various adaptation strategies have been increasingly developed for water resources and river basins in general.

Several hydrological models were used to estimate the effect of climate change on the water resources in the river basin under different emission scenarios. Several studies were also conducted to evaluate the hydrology of river basins impacted by climate change (Saddique et al. 2019; Paul et al. 2020; Mengistu et al. 2021). Among the hydrological models, the semi-distributed, physical-based model Soil and Water Assessment Tool (SWAT) has been extensively used to simulate the impact of climate change on basin hydrology around the world (Arnold et al. 1998). The hydrological model SWAT was selected in this study since it was used successfully and widely all over the globe (Bhatta et al. 2019; Saddique et al. 2019; Paul et al. 2020; Mengistu et al. 2021). Furthermore, the SWAT has a high computational efficiency to operate at a reasonable time on a large scale, the ability to use available data and predict long-term impacts (Gassman et al. 2007).

Future climate data are mostly derived from general circulation models (GCMs). GCMs are computer-driven models for forecasting the weather, climate understanding and climate change projection (IPCC 2007). These models take various factors into account such as chemical, biological, atmospheric conditions and ocean movement for predicting the future climate. Studies have evaluated the effect of climate change on the water resources in numerous regions using different GCMs' climate projections, and have subsequently predicted the impact of climate change on the hydrological regime using hydrological models (Saddique et al. 2019; Paul et al. 2020; Mengistu et al. 2021). The findings of these studies predicted a rise in temperature for the 21st century. Moreover, Elshamy et al. (2009) analyzed 17 GCMs and reported a decrease in the flow of the Blue Nile River by the end of the 21st century. A similar outcome was generated by Cherie (2013), who predicted a decrease in flow using three other GCMs in the same river. However, other studies using the outputs of GCMs suggest an uncertainty on whether an increase or decrease in river flow and precipitation change will occur (IPCC 2007; Change 2013; Saddique et al. 2019). Mounir et al. (2024) used projections from EURO-CORDEX and predicted an increase in evapotranspiration by 12% of Ouergha catchment, Morocoo. Moreover, Colín-García et al. (2024) projected a decrease in water yield by 61.01% of the Mixteco River basin, Oaxaca, Mexico using the SWAT model. These uncertainties are due to the selection of different GCMs and downscaling techniques, observed data quality and availability for the calibration of the hydrological model (Saddique et al. 2019). Generally, different GCMs and multiple scenarios are recommended rather than using just a single GCM for better foresight into future climate as studies highlight the concern in the assessment of climate change (Bhatta et al. 2019).

Pesticides have been used to destroy, prevent and mitigate pests and increase yield and quality (Delcour et al. 2015). One of the major pollutants associated with agriculture is the presence and accumulation of pesticides in the natural environment. This problem is further exacerbated as it is expected that a higher quantity of pesticides will be used in the future as the world population increases to promote food security (FAO 2009). Particularly in Ecuador, active ingredients an average of 15,630 metric tons annually (2010–2014) were used for agricultural purposes. However, information is scant on the effect of the changes in precipitation and climate change in general on pesticide concentrations in surface waters. Thus, it is worthwhile to investigate how changes in the water balance affect the presence and accumulation of pesticides in the environment, which can be estimated and simulated using a hydrological model coupled with a pesticide transfer model (Mottes et al. 2014).

In developing countries, water managers plan future strategies based on non-academic expertise because of the limited research capacity in hydrological and climate modeling which lead to insubstantial predictions (Vera et al. 2020). Specifically, the changing climate has an adverse consequence on the climatic conditions in South America, particularly in Ecuador, such as changes in the amount of precipitation and temperature. Nevertheless, studies on the impact of future climate conditions on water resources and the hydrological regime are scarce in these regions. Ecuador's lowland areas' main source of fresh water is the precipitation during the wet season which fulfills water demands for domestic purposes, hydropower generation and agricultural need (Rivadeneira Vera et al. 2020). Thus, there is a need to determine the future climate conditions as well as their effect on water resources for an informed decision regarding water management and to better deal with the potential challenges in the future.

This study aims to assess the impact of projected climate on the hydrology and water quality, i.e. pesticide concentration in the streams of the Guayas River basin, which is one of the major river basins in Ecuador, using the SWAT and future simulations of four GCMs HadGEM2-ES, GDLF-ESM2M, ISPL-CM5A-LR and MIROC5, under RCPs 4.5, 6.0 and 8.5 for the periods of 2021–2030 (near future), 2051–2060 (mid-century) and 2081–2090 (late century). The findings of this study provide insights into potential measures to mitigate the forthcoming impacts of climate change on the water resources and agriculture in the river basin. This research investigates the impact of future climate conditions on water balance, i.e. potential evapotranspiration (PET), surface runoff, groundwater flow and lateral flow in a data-scarce region that is prone to climate change impacts. The Guayas River basin is very important for both local agriculture production, as well as downstream aquaculture and fisheries (De Cock et al. 2022).

The main soil types are loam, clay loam and sandy loam which account for 23, 20 and 15.3% respectively in the basin. Most of the area is covered by agriculture (46%), whereas 29% is forest and 13% is pastures (Frappart et al. 2017). The Guayas River basin provides 70% of crop yield which is the most productive agricultural region of the country (Frappart et al. 2017). The main crops are maize (11.50%), rice (9.34%), cacao (9.19%), sugar cane (3.45%) and banana (3.02%).

The SWAT model calculates water balance daily for evapotranspiration, precipitation, runoff, baseflow and percolation. At the HRU level simulations are performed and summed up in each sub-basin. The surface runoff is estimated from daily precipitation using the soil conservation service curve number (SCS-CN), as applied in this study. Surface runoff is estimated based on soil type, moisture conditions and land use. The SWAT model calculates evaporation from plants and soils separately. The Hargreaves method is used for the calculation of PET. This method is used because of the projected climate data availability.

The model used in this was set up by Cambien et al. (2020). In general, the SWAT model requires a land use/land cover, digital elevation model (DEM) and soil data for the delineation of HRU and sub-basins. The DEM used in this study was gathered from the Shuttle Radar Topographic Mission with a resolution of 30 × 30 m, and the stream network was generated by using that DEM. Based on the generated stream networks, the basin was split into 29 sub-basins with an average area of 1,092 km² and a threshold of 3% was chosen as suggested by Jha et al. (2004). The sub-basins are further divided into HRUs with a threshold of 5% for land use and 20% for slope and soil (Neitsch et al. 2011). The land use (scale 1:100,000) was obtained from the Ministry of Agriculture, Livestock, Aquaculture and Fisheries of Ecuador (MAGAP) and soil data, i.e. soil type and soil texture (scale 1:5,000,000) were obtained from Harmonized World Soil Database (HWSD). The input data used for this model and their description are presented in Table S1, being comprehensively explained in the paper of Cambien et al. (2020). The empirical method, the SCS-CN (Arnold et al. 1998; Neitsch et al. 2011) was used for the estimation of surface runoff. Moreover, in the model, pesticides are transported by dissolving in the water and by adsorbing to the sediment. Therefore, the pesticide simulations required sediment yield and routing (Neitsch et al. 2011).

Statistical criteria were used for the evaluation of the goodness of fit between measured and simulated values: PBIAS and NSE. PBIAS the fitting value is 0, and low extent values of PBIAS indicate accurate model simulation (Equation (2)). A positive value shows the model is underestimating while a negative value shows the model is overestimating (Moriasi et al. 2007). The NSE (Nash & Sutcliffe 1970) is the ratio between residual variance and measured data variance (Equation (3)). However, it shows how a good plot fits the 1:1 line between the observed and simulated data. The value of NSE ranges from minus infinity to 1.0, with a value of 1 indicating a perfect fit. Model performance were classified as very good for PBIAS < ±5 and NSE > 0.80, good ±5 < PBIAS < ±10 and 0.70 < NSE < .80, Satisfactory ±10 < PBIAS < ±15 and 0.50 < NSE <0.70 and not satisfactory PBIAS > ±15 and NSE < 0.50 (Moriasi et al. 2007).

Four GCMs were used for the climate projections that participated in the World Climate Research Programme Coupled Model Intercomparison Project phase 5 (CMIP5). The Climatic data were downloaded from four GCMs of the Inter-Sectoral Impact Model Intercomparison Project phase 2b (ISIMIP2b; Table 1) (Frieler et al. 2017). Three representative concentration pathways (RCPs): RCP 4.5, RCP 6.0 and RCP 8.5 were used for the historical and future periods. Based on various radiative forcing (2.6–8.5 Watt/m²) RCP is defined, in which RCPs 4.5 and 6.0 are intermediate scenarios with a radiative forcing of 4.5 and 6.0 Watt/m² and a potential increase in temperature of 2.4 and 2.8°C by 2100, respectively. On the other hand, RCP 8.5 has the highest radiative forcing, i.e. 8.5 Watt/m² and the potential increase in temperature of 4.3°C by 2100 (Change 2013). The future climatic data were available at a grid resolution of 0.5° × 0.5° and at a daily timestep. The climatic data were reformatted from NetCDF into text files as formatted in the SWAT. The four GCMs which were used in the climate change scenarios include HadGEM2-ES, GDLF-ESM2M, IPSL-CM5A-LR and MIROC5 (Table 1) with RCP 4.5, RCP 6.0 and RCP 8.5, as these GCMs are extensively used in climate change studies around the globe (Saddique et al. 2019; Lotfirad et al. 2021; Adib et al. 2023). The data obtained from the GCMs require bias correction because when it is projected it does not follow the trend with the observed data. The climate model data for the hydrologic modeling (CMhyd) tool (Rathjens et al. 2016) was used for bias correction of the daily maximum temperature, minimum temperature and precipitation, where the delta change correction method was selected. In which the observed data (1990–2000) and historical climatic data of GCMs (1986–2005) were used as input data for the bias correction of the projected data from the GCMs at each station.

The highly sensitive parameters were determined by sensitivity analysis. The result showed four groundwater parameters, i.e. (ALPHA_BF), (GW_DELAY), (GWQMN) and (GW_REVAP), influencing runoff, i.e. curve number (CN2), streamflow routing (CH_K2) and one parameter influencing actual evapotranspiration (AWC). In this study, calibration (1993–2000) and validation (2001–2009) of the model were done manually using the monthly streamflow data of five gauging stations in the basin. This manual calibration allows a better understanding of the effects and dynamics of the parameters which simulate the streamflow. Table 2 illustrates the efficiencies (calibration and validation) for the monthly streamflow The model performance for the gauging stations ‘Daule en La Capilla’ and ‘Quevedo en Quevedo’ during the calibration period was assessed as very good. The model performance for the calibration period, gauging station ‘Baba dam’ and ‘Vinces en Vinces’ were classified as satisfactory, whereas for ‘Zapotal en Lechugal’ it was unsatisfactory. On the other hand, the validation (2001–2009) for the gauging stations ‘Vinces en Vinces’ and ‘Zapotal en Lechugal’ was assessed as very good, for ‘Daule en La Capilla’ as good and for ‘Baba dam’ and ‘ Quevedo en Quevedo’ as unsatisfactory. During the calibration period, an El Nińo event occurred, which resulted in a notable variance in average streamflow for the period of calibration and validation. Figure S1 illustrates the measured and simulated monthly average streamflow for the ‘Daule en La Capilla’ and ‘Zapotal en Lechugal’ for the calibration and validation period. The graphical comparison (Figure S1) indicates that simulated streamflow follows the measured trends quite well for the ‘Daule en La Capilla’ and ‘Zapotal en Lechugal’. The model performance for monthly streamflow was assessed using the historical climatic data of the GCMs and results are described in the supplementary part.

The result showed a rise in the streamflow during most months for all the GCMS and RCPs (Figure S2). The variability in streamflow was higher between February to June compared to the baseline. Peak streamflow was observed in April of the 2080s for all the GCMs and RCPs. The highest flow increase was during the low-flow months (June, July and August). The change in streamflow concerning the baseline is noticeable in the 2080s for June-August while in the 2020 and 2050s, it was observed that there is no relevant difference except for MIROC5 and probably by 2050s.

Surface runoff simulations show the increasing trend for all the GCMs expected for the ISPL-CM5A-LR. In the 2020s, the changes range from −18% for ISPL-CM5A-LR projection under RCP 8.5 to 23% for MIROC5 projection under RCP 8.5 (Table 3). The changes vary from −6% for GDLF-ESM2M-ES projection under RCP 6.0 to 39% for MIROC5 projection under RCP 4.5 in the 2050s. In the 2080s, surface runoff changed from 9% ISPL-CM5A-LR projection under RCP 4.5 to 57% for MIROC5 projection under RCP 8.5.

The maximum monthly average surface runoff increased by 44% (July), 24% (February) and 29% (August) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2020s. Moreover, the maximum monthly average surface runoff is expected to increase by 74% (August), 58% (July) and 93% (August) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2050s. Moreover, the maximum monthly average surface runoff will increase by 104% (August), 142% (July) and 208% (August) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2080s.

There is a greater change in surface runoff caused by the increase in precipitation (Figures 5, 6 and S3). Future precipitation projections are observed to increase, which consequently increases the surface runoff. Those later variations are also related to temperature change.

The water yield simulations for all the GCMs and RCPs showed an increasing trend except for the ISPL-CM5A-LR and GDLF-ESM2M-ES 2050s projection under RCP 6.0 and HadGEM2-ES 2050s projection under RCP 6.0. The increase in precipitation results in the increase of surface runoff, lateral flow and groundwater flow for most of the GCMs simulation. The total water yield of the basin is projected to increase with the rise of all these water balance components.

The changes in water yield of the Guayas River basin in the 2020s vary from −14% for ISPL-CMSA-LR projection under RCP 8.5 to 19% for MIROC5 projection under RCP 8.5. On the other, the changes vary from −5% for HadGEM2-ES projection under RCP 6.0 to 34% for MIROC5 projection under RCP 4.5. The changes in the 2080s vary from −5% for ISPL-CMSA-LR projection under RCP 4.5 to 49% for MIROC5 projection under RCP 8.5.

The maximum monthly average water yield increased by 9% (December), 21% (February) and 19% (December) for the 2020s under the RCPs 4.5, 6.0 and 8.5, respectively in 2020s. Moreover, the maximum monthly average water yield increases by 26% (December), 20% (December) and 33% (December) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2050s. Moreover, the maximum monthly average water yield increases 31% (December), 49% (May) and 64% (April) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2080s. With the increase in the precipitation, surface runoff, base flow and lateral flow, the water yield increased.

General observation on pesticide simulation is related to simulated runoff and pesticide output of a certain sub-basin. Figures 7, 8 and S4 show the simulated pesticide concentration in the ‘Vinces en Vinces’ station, which is situated downstream of a corn cultivation area where the pesticide is applied. The pesticide (pendimethalin) outputs are linked to the values of simulated runoff (Figures 7(b), 7(c), 8(b) and 8(c)). In the Vinces en Vinces station, pendimethalin was simulated showing a maximum concentration. As the distance from the pesticide source increases, the relationship between pesticide load and surface runoff becomes less clear. It is observed that pesticide concentration shows a downward trend from January until June and then a slight increment in July; then followed a downward trend up to November; and finally again increasing in December. The maximum pesticide output was observed during the wet season. This trend was followed by all the GCMs and RCPs of the future projection.

Climate change is among the main challenges for water resources management on planet Earth. Studies show that climate change causing changes in precipitation patterns could affect the worldwide hydrological regime (Held & Soden 2006). It is projected to increase the severity and repetition of extreme hydrological events, i.e. droughts (Puri et al. 2011) and floods (Mosquera-Machado & Ahmad 2007). Concerning water resources, there would be a significant effect on the accessibility of water for agriculture, domestic purposes and hydropower (IPCC 2007). The main source of freshwater in Ecuador is mainly precipitation which fulfills domestic purposes and agriculture requirements (Rivadeneira Vera et al. 2020). In the wet season (December–April), more than 80% of precipitation occurs (Mera et al. 2018). Thus, it is significant to know the future climatic conditions in the basin to facilitate future basin management.

The change in projected climate, i.e. precipitation and temperature have an adverse consequence on the climatic condition for the Guayas River basin. It was noticeable that there is a high variation amongst all scenarios of the average monthly temperature and precipitation compared to the baseline. In that context, the 2080s showed the highest variation for precipitation and the temperature shows a more evident increase in the 2080s as compared to the 2020s and 2050s. The variability in streamflow was higher between February and June compared to the baseline. Peak streamflow is observed in April of the 2080s for all the GCMs and RCPs. The results of this study indicate that with the increase in the precipitation, there will be an increase in the hydrological regime (precipitation, PET, surface runoff, groundwater flow, lateral flow and water yield) and vice versa. The findings of this study are generally in line with the results of other studies in the same and neighboring basins. Specifically, a rising trend in precipitation (7%) is also observed in the study of Ilbay-Yupa et al. (2021a) for the Guayas River basin, which could result in an increase of river streamflow by 69%, although another method was used, i.e. lumped hydrological modeling (GR2M water balance model) and only the changes in precipitation and stream flow were investigated. In another study on the neighboring basins (Chone and Portoviejo River basins), Vera et al. (2020) used the Lumped Témez hydrological model and predicted an increase (6%) of runoff by the end of this century.

PET tends to rise (following temperature rising) for all the GCMs under all RCPs. A similar increasing trend was found by Vera et al. (2020) for PET (12%) in the Chone and Portoviejo River basins. Conway (1996) estimated that with the increase in temperature of 1°C, there will be a 4% increase in the PET and Elshamy et al. (2009) also estimated throughout the wet season there will be 3.75% annually and 5.60%. The rise in water yield of the basin is caused by the rise in precipitation, surface runoff groundwater flow and lateral flow (Table 3). Based on the results of pesticide simulation, the soluble pesticide is dependent on surface runoff; moreover, the surface runoff is dependent on the precipitation. From this, we can conclude that with an increase in precipitation, there will be an increment on the amount of soluble pesticides due to the increase in surface runoff at a certain station. Moreover, there will be an increase in the surface runoff due to the increase in precipitation, leading to low pesticide concentration due to dilution. Conversely, as precipitation decreases, the amount of soluble pesticides will also reduce, which is the same pattern as surface runoff. Consequently, pesticide transport to rivers will decrease in response to less precipitation, while the associated concentrations in rivers will depend on the dilution and breakdown processes as well, what shows the need and added value of models to investigate this systematically. Pesticide outputs are maximum during the months of December to April (wet season) and minimum during the months of May to November (dry season).

Our findings suggest that there will be variations in the hydrological cycle of the basin. An increase in precipitation during the wet season would cause high flows along the streams resulting in large rates of soil erosion in the upper and middle part of the basin; whereas in the lower part, it will cause sedimentation problems (Ilbay-Yupa et al. 2019). Specifically, the construction of effective drainage systems is crucial for managing excess water, which is particularly important for crops such as sugarcane and banana. Proper drainage helps prevent waterlogging, which can otherwise harm these crops. As a result, investing in drainage infrastructure could enhance agricultural productivity and reduce potential losses. A rise in precipitation decreases maize yield (Li et al. 2019). Therefore, it would have a significant impact on agricultural production due to climate change which is already under stress through issues linked to water resources and population growth. The outcomes of this study may contribute to the assessment of water resources and pesticide fate due to climate change, and can be helpful for the planning of the optimal adaptation measures in the Guayas River basin as well as similar systems worldwide.

The SWAT model was used in this study to simulate the water balance components and pesticide trends in the Guayas River basin while using four GCMs projections and RCPs 4.5, 6.0 and 8.5 to analyze the impact of climate change. In this study, despite the complexity and size of the watershed with limited available data, the SWAT model can sufficiently represent the hydrological regime.

Based on the SWAT model's streamflow calibration and validation, on the monthly basis, the model performs well. The availability of spatial data was limited, and the presence of complex stream networks and the absence of outlet stations make calibration challenging (Cambien et al. 2020). Nevertheless, a hydrological calibration was still possible. The watershed is divided into 29 sub-basins and it is not possible to calibrate separately. Therefore, a good validation strategy is considered to be a multi-site approach rather than a split sample in time methodology (Refsgaard 1997). Concerning enhancing the calibration, the spatial variability data should be refined by increasing the streamflow stations by collecting more data and using smaller threshold drainage areas in the delineated network. Furthermore, to enhance the performance of the model, smaller time steps such as daily should be performed in the calibration period.

The SWAT model application in the Guayas River basin shows case-specific strengths and weaknesses. The first strength of the SWAT model is to be semi-distributed. The balance between data requirements, spatial variations, model complexity and computational efficiency was experienced as very convenient for this research. Moreover, it allows the categorization of various land use and slopes that are essential for the implementation of field practices. The second strength is the accessibility of extensive documentation, user support and databases (Bannwarth et al. 2014). The ArcGIS interface tool was used for this study, however, many new tools are being developed such as using an open-source interface tool QGIS (Bieger et al. 2017). Moreover, the SWAT model has a broad scope of potential applications, but the limitation is constrained by a case study. The main weakness of the SWAT model is that it required various and numerous data for the simulation; however, in developing countries it is challenging to gather high quality and quantity data (Gassman et al. 2007). The second limitation is the time needed for the model development and calibration and they are also challenging (Ghafoor et al. 2022). High technical expertise is required to run and calibrate the SWAT model as it required high input variables. Regardless of the consequence of these weaknesses, particularly the foremost, the SWAT allows obtaining insights into the system's functioning. The recommendations for the improvement of the SWAT model developed for the Guayas River basin are the following: climatic input data quality must be enhanced; and up-to-date high spatial resolution soil map with should be used; and pesticide input data quality should also be improved.

Simulations of this study imply that the hydrological regime of the Guayas River basin would significantly change with the projected/future climate, and mainly this change is caused by the increase in precipitation. The study area (Guayas River basin) is one of the significant watersheds of the western-coastal region in Ecuador because of its high agricultural production and economic contributions (Cambien et al. 2020). An increase in the hydrological regime (precipitation, PET, surface runoff, groundwater flow and water yield) may affect the management and development of the basin. Stakeholders and decision-makers should focus on the effect of climate change in the basin. With the increase in the precipitation, the streamflow would increase in the basin which results in soil erosion in the upstream and middle parts of the basin and sedimentation issues downstream of the basin. Thus, mitigation measures such as riparian buffer strips should be further investigated to mitigate erosion risks during heavy precipitation (Forio et al. 2022; Witing et al. 2022). An increase in precipitation during the wet season also potentially affects the agricultural food production cost due to the establishment of drainage networks as well as a potential additional need for pesticides to control fungi, for instance. Furthermore, it is important to identify the high-flood risk areas so necessary disaster mitigation measures will be considered as illustrated in the study by Glas et al. (2020). Also, special attention should be given to areas that are moderate to high risk of flooding in managing their hydrological regimes. In the river basin, there will be a change in the flow regimes of the rivers/streams due to the impact of climate change. Therefore, there is a need for constructing hydraulic structures within the basin to minimize the effect of flooding and to supply water in the dry periods for agricultural and domestic purposes. The river basin had already been affected by the floods in 1965, 1972–1973, 1982–1983, 1987, 1992 and 1997–1998 which caused losses to humans, infrastructure and the economy (Demoraes & d'Ercole 2001). The environment of Ecuador should be considered as a special case because it had extreme climate variability, as influenced by El Niño, the position of the intertropical convergence zone, trade wind dynamics, Humboldt current and Andes mountain range presences (Ilbay-Yupa et al. 2021b).

This study investigated the climate change impacts on the hydrological regime of the Guayas River basin with 12 climate projections combining 4 GCMs (i.e. GDLF-ESM2M-ES, HadGEM2-ES, ISPL-CM5A-LR and MIROC5) and three emission scenarios (RCP 4.5, RCP 6.0 and RCP 8.50). The model was simulated for 8 years (1993–2000) as calibration and 9 years (2001–2009) as validation with observed streamflow data from five measuring stations, which were located in the basin. The SWAT model was simulated for a period of 10 years as baseline (1991–2000) and three projected periods (2021–2030, 2051–2060 and 2081–2090) climate scenarios using the projections of four GCMs.

The results showed that the hydrological regime of the Guayas River basin is susceptible to climate change. The main findings of this research are as follows:

• In the 2080s climate scenario, average monthly precipitation is predicted to rise in the basin under all the RCPs, as compared to the baseline period. The expected change in the precipitation is expected to heavily alter the water balance of the basin.

• In the 2080s, the monthly average PET is projected to rise to 15.1% for June relative to the baseline largely because of an increase in temperature.

• The maximum monthly average surface runoff increases 104.4% (August), 142.1% (July) and 207.7% (August) under the RCPs 4.5, 6.0 and 8.5 respectively in the 2080s.

• With the increase in the monthly average precipitation and surface runoff, there will be an expected increase in the monthly average water yield of the basin. The maximum monthly average water yield increases 31.4% (December), 49.3% (May) and 64.1% (April) under the RCPs 4.5, 6.0 and 8.5, respectively in the 2080s.

• At the Vinces en Vinces station, with the increase in precipitation in the basin, there will be an increase in the quantity of soluble pesticides due to the increase in surface runoff.

Land use in this study was assumed to be uniform for all the simulations. However, it is expected that land use will change in the future, the hydrological cycle of the basin may probably be affected even more, combined with the rise in quantity and intensity of precipitation because of climate change. We believe that the gain in surface runoff is due to an increase in precipitation. Furthermore, there may be erosion risks at the upstream and sediment issues downstream of the basin because of an increment in average monthly streamflow. Therefore, to tackle the negative impacts of climate change there is an essential need for developing a water management framework incorporating the probable impacts of future climate conditions on water resources.

We suggest future studies address the limitation of the research approach which we adopted. The input data which are used to simulate the model plays an important role. The projected precipitation and temperature would play significant roles in the assessment of the hydrological regime. Thus, obtaining reliable projected climatic data is essential for predictions. In this research, the future climate projections of four GCMs from CMIP5 were adopted to investigate the climate change impacts on the hydrological regime in the Guayas River basin. Accurate projections of climate remain a problem for most GCMs. Meanwhile, the CMIP has progressively entered CMIP6 and it is expected to remedy this problem. The SWAT model application requires more observed data which are a limitation in this study due to data scarcity. The Guayas River basin has limited observed data and pesticide application data. The better calibration of the SWAT model requires more streamflow station data. We recommend considering SWAT + , the latest version of the SWAT model. The SWAT+ offers enhanced modeling capabilities, including more detailed hydrological simulations and dynamic watershed configurations. It also supports a broader range of advanced management practices, allowing for more accurate and flexible modeling of land use and conservation strategies.

Finally, we recommend future studies address the limitations of the research approach which we adopted, such as the bias correction method, GCMs and assumption of constant land use. Also, it will be thought-provoking to further evaluate climate change scenarios of hydrological components (precipitation, surface runoff, PET, etc.) for sub-basins and HRUs to get an understanding of spatial variation. Adaptation policies and management should be established to help mitigate the impact of climate change in the Guayas River basin and safeguard both the housings and food production in the basin and downstream estuary (De Cock et al. 2022). Integrated environmental models can be useful tools for this in the context of sustainable development (Forio & Goethals 2020).

Jawad Ghafoor is supported by Higher Education Commission (HEC) Pakistan. Marie Anne Eurie Forio is supported by the H2020 Projects OPTAIN and MERLIN and the Horizon Europe project OneAquaHealth.

J.G., M.A.E.F. and P.L.M.G. contributed to conceptualization and methodology; J.G., I.N. and M.A. participated in data curation; J.G. wrote the original draft; M.A.E.F., I.N, M.A., P.L.M.G. wrote the original draft and reviewed and edited the manuscript; M.A.E.F. and P.L.M.G. supervised this work.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

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

The authors declare there is no conflict.