Ensemble Boosting and Bagging Based Machine Learning Models for Groundwater Potential Prediction

Water Resources Management - Due to the rapidly increasing demand for groundwater, as one of the principal freshwater resources, there is an urge to advance novel prediction systems to more...     

Development of a New Integrated Framework for Improved Rainfall-Runoff Modeling under Climate Variability and Human Activities

The relationship between rainfall and runoff is a complex phenomenon and understanding the physical processes, hydrological components and their impacts on response of watershed to precipitation is one of the challenging issues in watershed hydrology and planning. There is still a need to improve conceptual hydrological models in water scarce regions, such as Iran mainly because in many cases there is not enough data to fully describe this phenomenon. In this research, we aimed to present an improved and parsimonious framework that increases the performance of a conceptual model in water balance and discharge modeling for Delichay watershed located in Hablehroud basin, Iran as one of the main source of water supply for downstream fertile agricultural areas that produce a considerable amount of cereals and play a major role for food and water security of the region. In areas where data for water cycle components are not available or limited, it is recommended to use parsimonious approach in order to have an acceptable level of understanding of the system with minimum possible predictor variables. The Salas model used in current research to model water balance over the period 1983–2012 and evaluation of the results indicated an unsatisfactory performance when the entire period was modeled altogether (NSE?=?0.35, d?=?0.70, R²?=?0.63, RSR?=?0.80, PBIAS?=?4.96 and RMSE?=?41.87). A key reason is that this watershed is intensively impacted by human activities and homogeneity analysis confirmed a sudden shift in runoff data during 1998–1999. Such a sudden shift reveals the role of human activities impacts on the watershed with a total reduction of 58 mm of runoff per year while the climate variability has not occurred in the region. Thus, the entire period (i.e. 1983–2012) was divided into two homogenous sub-periods of before and after the change point (i.e., pre-change and post-change periods). The results indicated that modeling performance in the sub-periods improved (e.g. the NSE was 0.77 and 0.66 for pre-change and post-change, respectively, vs. 0.35 for entire period). Meanwhile, it is revealed that water balance affected by human activities over the time and application of historical data for water balance modeling cannot be reliable without considering the homogeneous data. Since, many watersheds in the world have been affected by human activities or climate variability, it is recommended to consider the homogeneity of observed data before any application.