Global warming and associated climate change not only causes disasters such as droughts and floods, but also likely to affect water resources in various ways. Especially in dam reservoirs, climate change can affect water temperature, stratification strength and rainfall patterns, which will affect water quality and aquatic ecosystems by inducing the frequency and intensity of turbid water generation and the transition characteristics of algae species. Numerical modeling techniques are widely used to predict the impact of climate change on water resources, but uncertainty arises at each stage of the modeling process. Uncertainty is cumulative and propagated, and the projected results includes the total uncertainty. Therefore, in addition to assessing the impact of future water resources, it is necessary to quantify the uncertainty of the entire process in order to increase confidence in the prediction.

The purpose of this study was to project the changes in water temperature, stratification structure, and future turbidity flow events in Soyanggang Reservoir located in North Han River of Korea based on the IPCC's climate change scenarios, and to quantify the uncertainty in each step and scenario of the overall modeling processes. For this study, climate data generated by seven GCM models, using two Representative Concentration Pathway (RCP) scenarios were downscaled for Soyanggang Reservoir basin. Daily inflow data were generated using the Soil and Water Assessment Tool. And the long-term water temperature simulations were performed in the reservoir using CE-QUAL-W2, a two-dimensional hydrodynamic and water quality model. The model was evaluated using AME, RMSE, and NSE. The R package UncDecomp (Kim et al., 2019) was used to quantify the uncertainty for the whole process.

In all scenarios, there are differences depending on climate scenario and GCM models, but Soyanggang Reservoir's air temperature, water temperature, and stability of water bodies tended to increase in the future. In particular, the RCP 8.5 scenario showed a greater and faster rise. The epilimnion water temperature was expected to rise by 0.029℃(±0.012)/yr and 0.016℃(±0.009)/yr, respectively in the RCP 4.5 and RCP 8.5 scenarios, which are corresponds to 88.1% and 85.7% of the air temperature rise rate. Meanwhile, the hypolimnion water temperature was expected to rise by 0.016℃(±0.009)/yr and 0.027℃(±0.010)/yr, respectively, which is about 48.6% and 46.3% of the air temperature rise rate.

Some scenarios exceeded the peak discharge concentration and persistent turbidity duration of the largest historical turbidity flow event occurred in 2006. In particular, the event with a peak discharge concentration of 2,739.0 mg/L was projected to occur in the RCP 4.5 and CCSM4 model, which is approximately 10 times that of 2006. In the RCP 8.5 and CCSM4 model, the event with a turbidity duration of 704 days was projected. This is the longest persistent turbidity duration in the predicted period and is about twice the maximum discharge concentration and the turbidity duration experienced in 2006.

Uncertainty estimates for the overall process were found to be 26.4% for the RCP scenario, 59.3% for the GCM model, and 13.0% for the W2 model. The uncertainty in the GCM model was greatest in the overall process. The three models that contributed the most to the uncertainty of the GCM model matched the models representing extreme values in the seasonal water temperature predictions.

This study presents that future climate change may affect the thermal characteristics and extreme turbidity flow events in Soyangang Reservoirs, the most important water resources in Han River. Therefore, further research on the impact of these changes on water quality and aquatic ecosystems and measures to adapt to changes are required.