Development of Drought Response Actions Based on Water Shortage Prediction and SDF Curves

Abstract

This study aims to quantify the properties of drought events and provide some suggestions for drought adaptions based on a drought prediction model Firstly the operation of Tsengwen-Wushantou Water Resources System was simulated to find potential water shortage events Secondly the model “water shortage prediction of reservoir” with the inputs of “reservoir storage” or “standardized precipitation index (SPI)” was developed by applying the linear regression backpropagation neural network (BPN) and support vector machine (SVM) Three performance indices (e g normalized root mean square error) were applied and revealed that BPN and SVM provide better prediction results than linear regression model Thirdly the “drought severity-duration-frequency (SDF) curves” were constructed by applying the statistical frequency analysis methods The “annual maximum water shortage for public use” was defined as drought severity Then Kolmogorov-Smirnov Test was used to identify the best-fit probability distribution (i e GEV) from several candidate probability distributions The SDF curves can quantify the severity duration and frequency of water shortage events By mapping historical drought evens on SDF curves the information about drought frequency can be derived Finally this study provided four-level drought response actions based on SDF curves Given a predicted water shortage the predicted drought event can be mapped onto the SDF curves so that the properties of potential drought events can be identified and the suggestions for drought response actions can also be found

Date of Award	2019
Original language	English
Supervisor	Pao-Shan Yu (Supervisor)