In this study, A fully distributed hydrological model based on the coupling of Soil and Water Assessment Tool (SWAT) and modular groundwater flow model (MODFLOW) is set up to simulate the long-term river flow and groundwater head of the Dee River catchment in the UK, using over the period of 1970-2003. Surface water abstractions represented by maximum amount and purpose of use are analysed to estimate daily water withdraws in addition to the daily release from reservoirs. A quantile regression analysis is then conducted on the model simulated river flow and groundwater head over the same period to investigate the trend of both high flow (represented by 95% quantile) and low flow situation (5% quantile) designated for flooding and droughts. The two scenarios of highly managed watershed and natural flow condition are constructed to study how the human activities (e.g. reservoirs) influence on the trend of high and low flow conditions.
It has been shown that the quantile regression is capable of revealing the patterns of high and low flows without having to suffer from non-constant variance. The analysis of 95 % quantile reveals an increase trend for both highly managed and natural flow scenario that is significant trend in part of the river basin. Additionally, the result reveals that management practices trend to reduce the likelihood of flooding in the catchment. For 5 % quantile, increases trend has been observed under both the highly managed and the natural flow scenarios. The management practice for low flow condition reduces the probable drought in the downstream part of Dee River basin and managed to provide water uses for industrial and public water supply and interestingly, this is due to the reduction of the water resources in the upstream part as revealed by the analysis.