Development of teleconnection-based seasonal rainfall forecasting models for reservoir watersheds in Taiwan

In order to support decision making for preparing drought-resistance actions in advance during the dry period, the study developed the teleconnection-based seasonal rainfall forecasting models for the 18 reservoir watersheds and 3 dam watersheds in Taiwan. The monthly climatic teleconnection indices, including circulation, sea surface temperature, wind field, Niño3, Southern Oscillation Index (SOI), and Pacific Decadal Oscillation (PDO), and antecedent monthly rainfall of watershed were used as the candidates of model input variables and the seasonal rainfall as the output variable. Multiple regression analysis, support vector machine, and random forests were adopted for model construction and different combinations of the input variables were used for finding the optimal combination for each watershed and each forecast-beginning month (form October to March). Comparison results show that the RF-based forecasting models perform the best. Further, for application convenience, the regional optimal teleconnection-based seasonal rainfall forecasting models were developed for northern, central, and southern Taiwan. It reveals that the proposed models perform better than the models which use (1) the antecedent monthly rainfall as input variable and (2) circulation, sea surface temperature, wind field, and Niño3 as input variables. © 2021, Taiwan Agricultural Engineers Society. All rights reserved.