Assessing the impact of climate change on surface runoff of the Ubolratana Reservoir, Thailand

Assessing surface runoff in river basins is paramount for effectively managing water resources. Climate change significantly impacts the availability of water within these basins. The Ubolratana reservoir is vital for sustainable water supply, aquaculture, agricultural practices, and domestic needs. This research aims to evaluate the runoff availability of the reservoir by using the Hydrologic Engineering Center’s Hydrologic Modeling System (HEC-HMS) under distinct climate change scenarios. The HEC-HMS model was calibrated based on observed rainfall and runoff data. To project future runoff dynamics, calibrated model parameters were coupled with the bias-corrected rainfall data of Can-ESM5, CESM2, and GFDL-ESM4 models for two Shared Socioeconomic Pathways (SSPs) climate scenarios (SSP 245 and SSP 585). The results show upward trends in both projected rainfall and runoff within the drainage area of the Ubolratana reservoir between 2022 and 2061. In summary, the findings of this research may be useful for regional water resources management and strategic planning endeavors.


Introduction
Water resources are essential to human survival, social development, and a sustainable economy.In many parts of the world, climate change affects the water cycle.These conditions increase challenges to water resources management.Responding to the hydrological processes under climate change is one of the critical aspects of hydrological and water resources research [1].Assessing the impact of climate change on the surface runoff of a reservoir is to recognize how the changes in rainfall may affect the water availability and quality of the reservoir.This information can help a step plan effectively reduce the potential risks of flooding, sedimentation, water scarcity, erosion, and pollution.
The anticipation and evaluation of evolution of hydrological processes' evolution under future climate change have become increasingly substantial.Typically, researchers employ forthcoming climate models and scenarios, coupled with mathematical model computations [2] and downscaling techniques [3], to process and analyse the forthcoming climate's evolution.However, the most challenge in the research lies in the uncertain elements surrounding future climate scenarios and hydrological models.The World Climate Research Program (WCRP) established the International Coupled Model Comparison Program (CMIP), which has been extensively adopted for climate change IOP Publishing doi:10.1088/1755-1315/1311/1/012045 2 modelling and predictive data analysis to address this issue.The CMIP6 pilot program was introduced in 2019, and the outcomes of its analysis are to underpin forthcoming climate assessments [4].In a study by Ding, et.al.[5], twelve CMIP6 models were selected under SSP 245 and SSP 585 climate scenarios to predict alterations in rainfall patterns in the Yellow River Basin of China from 2015 to 2100.The findings revealed that a majority of CMIP6 models exhibit a more increasing trend in precipitation under the SSP 585 scenario compared to the SSP 245 scenario.Tian,et.al. [6] used a bias-corrected method, specifically the Equidistant Cumulative Distribution Functions, to simulate precipitation fluctuations across eight CMIP6 models encompassing five SSP scenarios for 2015 and 2099 over China.In addition, Try, et.al.[7] found that the performance of CMIP6 in precipitation and discharge simulations is significantly increased compared to CMIP5 for future projections of flooding in the Mekong River Basin using SSP 245 and SSP 585.
Hydrological models stand as effective tools for the comprehensive management of water resources.They can be used to model the intricate water circulation processes within a river basin, considering the effects of climate change and human activities.One prominent hydrological model, the Hydrological Engineering Center's Hydrologic Modeling System (HEC-HMS), was developed by the U.S. Army Corps of Engineers.This model employs a physical conceptualization and utilizes distribution data to facilitate basin-scale hydrological simulations.Recent research attempts frequently involve a combination of methods within the HEC-HMS model for simulation purposes.Shakarneh,et.al. [8] used HEC-HMS to simulate the rainfall runoff events in the Daraja and Al Gahr catchments of Palestine.Their study showcased the robust performance of the model in both the calibration and validation phases.In a different context, Fanta and Sime [9] conducted hydrological modelling for runoff simulation model in the Toba catchment of Ethiopia.Notably, the HEC-HMS model outperformed the SWAT model, yielding superior results.Similarly, Lohpaisankrit, et.al.[10] undertook a study involving the HEC-HMS model, adopting a combination of methods such as the Soil Conservation Service-Curve Number (SCS-CN) loss method, SCS unit hydrograph, and Muskingum method.This comprehensive approach was employed to simulate floods in the Huai Sangka watershed, a small agricultural subbasin within the Chi River basin of Thailand.The application of these combined techniques demonstrated the model's efficiency in flood simulation.
Ubolratana reservoir is the largest reservoir system in Thailand's northeastern region.It serves as a vital water source for Khon Kaen Province and the broader northeastern area.This reservoir fulfills a multitude of roles, functioning as a basis for electricity generation, water supply, irrigation, ecological preservation, transportation, agriculture, aquaculture, and domestic consumption.The tropical climate in the basin area produces approximately four months of pronounced seasonal variation, with rainfall exhibiting a notably uneven distribution throughout the year.Mainly, rainfall concentrates within the wet season spanning from May to October.Currently, limited research exists regarding the development of local water resources.This study seeks to address this research gap by inquiring about the dynamics of rainfall-runoff interactions within the Ubolratana reservoir.Additionally, it aims to evaluate the correlation between regional climate patterns and surface runoff behaviour, employing various climate models and scenarios to forecast potential future outcomes.This analysis holds the potential to provide substantial insights into supporting water resources management strategies, thereby contributing to climate change mitigation efforts and the preservation of ecological stability within the regional area.

Research area
The Ubolratana reservoir is one of Thailand's largest reservoirs, located in Khon Kaen Province.This reservoir is divided into upstream and downstream areas.The upstream of Ubolratana reservoir consists of five sub-basins of Chi Basin, including Nam Phuai, Lam Phaniang, Nam Choen, Nam Phrom, and Upper Phong [11].The reservoir covers a geographical area of approximately 12,024 km 2 , with 101⁰24' -102⁰41' east longitude and 17⁰36'-16⁰5' north latitude., as shown in Figure 1.Due to the IOP Publishing doi:10.1088/1755-1315/1311/1/0120453 agriculture-deforestation land use and insufficient water storage problems often found in this area, the Ubolratana reservoir was selected as the study location.
It has a mid-temperate monsoon climate, with an annual average temperature of 27.6ºC and annual average precipitation of 1,102 mm.Rainfall is most frequently concerted in the wet season, May to October.However, the climate in Thailand and the selected study area is changeable.Furthermore, river floods in the catchment are primarily caused by torrential rains and tropical storms.The occurrence of heavy rains, tropical storms, and floods mostly occur from August to September.

B. Meteorological and Hydrological Data
The hydrometeorological information for establishing the HEC-HMS model comes from the Water Resources Engineering Program, Department of Civil Engineering, Khon Kaen University, Thailand.Runoff data is provided by the Royal Irrigation Department, while rainfall data from the Thailand Meteorological Department.Actual measurement daily precipitation data is selected from 2010 to 2019 from 8 stations: Phu Kradueng, Si Bun Rueang, Chum Phae, Phu Wiang, Ubolratana, Phu Kiao, Kaset Sombun, and Ban Thaen.Intending to forecast the projected runoff trend in the research area, CMIP6 experimental data with applied three models, namely Can-ESM5, CESM2, and GFDL-ESM4 was selected.As one of the organizations participating in the CMIP6, the Hydro Informatics Institute, Thailand, provides the models to utilize in this project.This paper applied to the Can-ESM5, CESM2, and GFDL-ESM4 models.The Canadian Earth System Model version 5 (Can-ESM5) is a universal model established to model former climate change and variability, construct climate-scale projections, and generate initialized seasonal and decadal forecasting [12].The Community Earth System Model version 2 (CESM2) is the current generation of the integrated climate models generated by the National Center for Atmospheric Research (NCAR) [13].In contrast, the Geophysical Fluid Dynamics Laboratory Earth System Model version 4 (GFDL-ESM4) is part of the fourth generation coupled chemistrycarbon-climate model development, which contributed publicly to the CMIP6 [14].
The daily rainfall from 2022 to 2061 is a sequence of two climate scenarios, SSP 245 and SSP 585.The SSP 245 scenario aims to achieve an effective radiative forcing value of 4.5 W/m 2 by the end of the 21st century while maintaining a medium range of greenhouse gas emissions which is aligned with economic and social growth [15].The scenario with the uppermost concentration of greenhouse gas, SSP5-8.5, with the effective radiative forcing value, is projected to stabilize at 8.5 W/m 2 by the end of the 21st century [4].

Hydrological Modelling Procedure
This study employs the HEC-HMS model to simulate runoff on open-channel routing by analysing weather and climate data.The HEC-HMS distributed rainfall-runoff model with the physical concept, which includes four components: watershed, control specification, weather, and time-series management.To generate the hydrological processes of the reservoir, the operation of HEC-HMS can be used for different calculation schemes.The DEM (Digital Elevation Model) data were evaluated by HEC-Geo HMS, and water systems features and topographic indicators of the catchment are acquired as well as categorized into 8 sub-basins.The HEC-HMS project information is prepared using the Thiessen polygon technique to determine the weight of the rainfall station in every sub-basin.The constructed HEC-HMS project data is on the premise, as seen in Figure 4.The runoff producing element uses the initial and constant method, the transform section employs the Snyder Unit Hydrograph method, the baseflow section employs the recession method, and the river routing section employs the Muskingum method.According to these methods, the HEC-HMS in the Ubolratana reservoir was created.Daily runoff data were selected from 2010 to 2014 as calibration model variables, and daily runoff data between 2015 and 2019 validated the simulation significance of the model.

Evaluation Metrics
The Root Mean Square Error (RMSE) and Nash-Sutcliffe efficiency (NSE) have been widely used in evaluating the efficiency of the baseline model between simulated and observed flow, given by: where  means the simulated data values,  is the actual data scores, and n is the amount of data.A low RMSE signified that the model has a small error, while a high RMSE value indicates that the model has a large error.
where  means the simulated data score,  is the actual data scores and  is the average of actual data score for the whole-time duration.The higher NSE scores equate to better model performance.

Results and Discussion
The HEC-HMS is a hydrological model to analyse a reservoir system and predict in order to forecast the hydrological response of the system to a user specified rainfall event.The HEC-HMS model needs to be calibrated and validated to acquire the optimal parameter values.Results of the below graphs (Figure 5 and Figure 6) are obtained from the daily observed runoff at the Ubolratana gauge location.Measured daily runoff values acquired from the Ubolratana gauge point are linked with runoff result values extracted from the HEC-HMS model after calibration to further evaluate the model during the validation part.From the comparison of both graphs, it can be noticed that the model simulated rainfall-runoff process from 2010 to 2019 is essentially in line with the measured process trend.
Figure 5 shows the time series hydrograph of the runoff variance between the predicted and observed flow throughout the calibration period from 2010 to 2014.The line chart illustrates that the maximum flow occurred in the monsoon period (May to October), moderate flow occurred in the postmonsoon season, and less runoff within the pre-monsoon season based on the rainfall.During the tropical region monsoon time period, the discharge fluctuates drastically, varying from 200 to 2000 m 3 /s due to the high rainfall event.The calibration hydrograph depicts the annual steep peak for all the years, which in 2011 has the highest peak.The RMSE and NSE values throughout the calibration process are 0.6 m 3 /s and 0.64, respectively.Figure 6 presents the time trends hydrograph of the runoff difference among the predicted and actual flow throughout the verification period from 2015 to 2019.The chart figures a significant correlation among the predicted and actual flow at the Ubolratana outlet.Furthermore, the RMSE and NSE scores within the validation period have been raised after the modelling calibration period.The discharge variance measured within the monsoon season ranges from 350 to 2000 m 3 /s, which has been decreased than the calibration period.It also might be because of the lower rainfall during the verification period.The model's efficiency is good with the RMSE and NSE values of 0.6 m 3 /s and 0.65, respectively.

Figure 6. Simulated runoff during the validation process
To acquire the historical rainfall simulation value from 1981 to 2022 in the Ubolratana reservoir under three models (Can-ESM5, CESM2, and GFDL-ESM4) and two climate scenarios (SSP 245 and SSP 585), the downloaded CMIP6 future climate scenario data are provided by Hydro Informatics Institute (HII) Thailand.The range projected rainfall is from 2021 to 2061, which is projected into each rainfall gauge in the basin using spatial interpolation and input into the formerly developed HEC-HMS model, and the calibrated parameters are applied to simulate and forecast future in SSP 245 and SSP 585 climate scenarios.These flow conditions are at the outlet of Ubolratana reservoir, which makes it possible to estimate the transformation of flow in the following 40 years to efficiently prevent flood catastrophe and completely grasp the future basin-runoff.

Conclusions
The hydrological modelling of the Ubolratana reservoir's drainage area, facilitated by the HEC-HMS model, has yielded significant insight into surface runoff dynamics.This hydrological model was suitable for simulating surface runoff within the study area over the period from 1981 to 2022 with good performance.The calibration process produced RMSE and NSE values of 0.6 m 3 /s and 0.64, respectively, while the validation process yielded corresponding values of 0.6 m 3 /s and 0.65.Under the three models (Can-ESM5, CESM2, and GFDL-ESM4), combined with the SSP 245 and SSP 585 climate scenarios, a discernible trend of upward rainfall emerges across the basin.This suggests a more equitable distribution of rainfall over the forthcoming four decades.It is revealed that the Can-ESM5 model depicts a 5.22% increase under SSP 245 and a 5.50% decrease under SSP 585 scenarios.Similarly, the CESM2 model shows a decline of 5.53% and 5.27% for SSP 245 and SSP 585, respectively.On a parallel note, the GFDL-ESM4 model projects growth of 5.13% and 5.27% for SSP 245 and SSP 585, respectively.This might correlate to the projected decline in rainfall as contrasted with the historical period and will affect the future climate of water scarcity.Analyzing the simulated runoff dynamics under the purview of two climate scenarios, SSP 245 and SSP 585, distinct peak future runoff periods emerge.Specifically, the Can-ESM5 model foresees these peaks occurring in 2044 and 2033, the CESM2 model anticipates them in 2025 and 2054, while the GFDL-ESM4 model projects events in 2056 and 2050, respectively.These projected variations may be attributed to multifaceted factors, including shifts in climate patterns and potential influence on rainfall.

Figure 1 .
Figure 1.Research area of Ubolratana Reservoir 2.2.Data Collection A. Geospatial Data Geospatial information includes information on land use, soil distribution, and digital elevation for watersheds.The digital elevation model 30 m resolution is provided from https://www.usgs.gov/.The land use and soil distribution data come from the Land Development Department of Thailand, which is extracted by the basin vector boundary as displayed in Figure 2 and Figure 3, respectively.

Figure 2 .Figure 3 .
Figure 2. Land use map of the Ubolratana reservoir

Figure 5 .
Figure 5. Simulated runoff during the calibration process

Figure 7 8 Figure 7 .
Figure 7. Average annual runoff based on the Can-ESM5 model

Figure 8 .
Figure 8.Average annual runoff based on the CESM2 model

Figure 10 .10
Figure 10.Average annual runoff under the SSP 245 climate scenario

Figure 11 .
Figure 11.Average annual runoff under SSP 585 climate scenario