Challenge of rainfall network design considering spatial versus spatiotemporal variations

Abstract

Listen

Translate article

Precipitation data plays an important role in investigation of water-related fields of research such as water resources management, hydraulic structure design and groundwater quantity/quality parameters due to its high variability in space and time. To evaluate and investigate precipitation pattern, a set of well-designed rain gauge stations can substantially reduce the cost and increase the estimation accuracy. In the present research, a new methodology of spatiotemporal optimization is developed for a rain gauge monitoring network and the results are compared to the optimized network based on spatial variations of precipitation. The optimization process consists of two main steps of application of multi attribute decision making and heuristic approaches. The entropy is chosen as the multi attribute decision making approach to determine optimum number of stations. Then, the optimization process is implemented via coupling of Genetic Algorithm (GA) and geostatistical methods to identify the best rain gauge network configuration. To determine the spatiotemporal structure of precipitation, spatiotemporal variography and geostatistical methods known as Ordinary Kriging (OK) and Bayesian Maximum Entropy (BME) have been undertaken. Thirty years of annual precipitation data from 105 rain gauge stations within and near Namak Lake watershed in the central part of Iran are utilized in this research to optimize rain gauge stations spatially and temporally. Results showed that spatiotemporal network design considerably differs from spatial optimal rain gauge stations. Optimal locations of rain gauge stations resulted from spatiotemporal approach are almost two times better than spatial configuration to reduce rain gauge costs (installation, operation, maintenance, etc.) by avoiding data redundancy more efficiently. In addition, BME-based network design method outperformed OK-based network.