In recent years, the extreme social and economic damages caused by flooding and drought have increased due to extreme weather. Extreme drought can cause great economic damage to agricultural productivity due to the reduction of the agricultural water supply, as well as an increase in water consumption by crops. Agricultural facilities for water supply include reservoirs, pumping stations, weirs, and infiltration galleries. According to a survey in 2011, there are 17,505 reservoirs in Korea, accounting for 25% of the total agricultural water facilities. However, 772,108 ha are irrigated by reservoirs, which accounts for 58% of the entire irrigated area. Therefore, as a major supplier of agricultural water in Korea, it is important to look into the behavior of reservoirs in relation to climate change. This requires estimation of the changes in inflow amounts from upstream areas, water consumption by crops in irrigated areas, and water storage level of reservoirs. Finally, agricultural drought vulnerability should be analyzed by considering water demand and available water supply via agricultural facilities for each crop growth stage, taking into account temporal and spatial changes. With this information, an adaptation plan for agricultural reservoir drought can be provided. Therefore, the objective of this study is to evaluate the impacts of climate change on agricultural water resources in Korea by considering the agricultural water supply through reservoirs, water demand from irrigated crop land, and changes in storage level and related agricultural drought according to the characteristics of the reservoirs.

104 reservoirs were selected based on the maximum available data. The selected reservoirs were classified into five clusters based on storage ratio (lowest storage / max. storage) and time of minimum storage level. Among the 104 study reservoirs, 5 reservoirs which have stream flow gauge stations downstream of the reservoir, were selected to evaluate the applicability of SWAT on ungauged watersheds.

This study consisted of 3 sub‐components: 1) generating climate change scenario data through bias correction and statistical downscaling of multiple Global Climate Models (GCMs), 2) reservoir modeling using SWAT and HOMWRS, and 3) estimating the changes in watershed characteristics, including cluster analysis of study reservoirs and spatial analysis of the reservoir‐irrigated paddy area. The reservoir modeling included 4 sub‐items, including 1) evaluating the impacts of upstream inflow on available water supply, 2) evaluating the impact of ET changes on water demand within an irrigation district, 3) estimating the daily storage level of reservoirs based on the estimated demand and supply, and 4) providing an agricultural drought index based on the estimated storage level. This reservoir modeling approach was applied within the selected 5 representative reservoirs and then extended to all the remaining 99 study reservoirs.

Among 34 available GCMs, the KMA 12.5km resolution RCM and eight and ten GCMs for the RCP4.5 and 8.5 scenarios, respectively, were selected because the selected GCMs provided the 6 weather variables required for SWAT application. The selected GCMs were bias corrected and downscaled for historical (1976‐2005), future 2020s (1911‐2040), 2050s (2041‐2070), and 2080s (2071‐2100) periods using the non‐parametric quantile mapping method based on observed data from 76 Korea Meteorological Administration (KMA) stations. Compared to the observed data, the bias corrected data appropriately reflected the temporaltrends of the selected weather variables, precipitation characteristic index, and the three reservoir modeling outputs, namely inflow, water demand, and storage level. It was decided that the quantile mapping method is suitable for agricultural reservoir analysis in which reproducing temporal trends is important for management purposes.

The bias‐corrected climate data for future periods showed the highest uncertainties in the precipitation variable according to the selection of GCMs by showing different (increasing or decreasing) trends compared to the historical period. However, both the minimum and maximum temperature variables increased in all GCMs, regardless of RCP scenarios. The future scenario data of other weather variables showed a tendency to converge closely to the past observations.

In the case of the RCP8.5 scenario, most of the eleven GCM data, including the KMA 12.5km RCM data, showed a tendency to increase during most months. As a result, the 30‐year monthly mean of the multi‐model ensemble (MME) showed a 5.6% increase in total precipitation.

When the MME was used, inflow to reservoirs in the future period (2011~2040) increased by 7.8% and 9.3% for the RCP4.5 and 8.5 scenarios, respectively, mainly due to the increase in precipitation. Similarly, irrigation water demands in 2020s increased 0.7% and 0.5% for RCP4.5 and 8.5, respectively, due to the increase in temperature. As a result, the water storage level increased by 2.3% and 1.6%, respectively, for RCP4.5 and 8.5 due to the combined influence of the increase in inflow and in water demand.

Clustering reservoirs based on characteristics such as storage capacity and ratio of watershed area to benefitted area cannot explain the responses of reservoirs within each cluster which show a wide range of variations in storage levels. When storage rate and time with minimum storage level were used for reservoir clustering, variations in each cluster decreased and it has been considered that clusters with the lowest storage rate can be most vulnerable to severe drought under climate change. However, inflow and water demand showed similar temporal patterns among clusters by showing an increasing trend, regardless of clustering methods. As far as the evaluation of the applicability of SWAT for integrated watershed management by including agricultural reservoirs in the watershed modeling frame, SWAT showed limitations in representing the processes of ponded paddy fields and linking required water demands in benefitted areas to reservoir storage. However, SWAT showed reasonable performance for the ungauged watershed conditions and the difference in inflow amounts between SWAT and HOMWRS showed a range from‐11.8% and 15.5% uncertainty envelope.

As an adaptation plan, structural, non‐structural (management‐based), and institutional measures are available. However, it was suggested that a no‐regret or low‐regret approach, which is based on non‐structural and institutional measures will be appropriate, considering the high uncertainty in climate change impact analysis results. Non‐structural measures may include efficient water management to minimize the loss of agricultural water supplies by adjusting the amount and time of irrigation and increasing the use of effective rainfall. This may require the development of a drought forecasting system, decision support system for effective irrigation, and an automatic water management system and monitoring system. Institutional solutions may include the inclusion of a governing body within the integrated watershed management concept for maximizing the use of surplus water and return‐flow by linking water resources‐related facilities. Also, the quality of agricultural water, environmental water, and the ecosystem should be considered in the concept of integrated water resources management.

1. 서론

2. 연구 자료 및 방법

2.1 연구 재료

2.2 연구 방법

3. 연구결과

3.1 기후변화 시나리오 분석 결과

3.2 SWAT의 적용성 평가 및 유출 모델링에 따른 불확실성 평가

3.3 저수지 가뭄지표 분석

4. 결론 및 토론