Climate (Apr 2020)
Prediction of Autumn Precipitation over the Middle and Lower Reaches of the Yangtze River Basin Based on Climate Indices
Abstract
Autumn precipitation (AP) has important impacts on agricultural production, water conservation, and water transportation in the middle and lower reaches of the Yangtze River Basin (MLYRB; 25°–35° N and 105°–122° E). We obtain the main empirical orthogonal function (EOF) modes of the interannual variation in AP based on daily precipitation data from 97 stations throughout the MLYRB during 1980–2015. The results show that the first leading EOF mode accounts for 30.83% of the total variation. The spatial pattern shows uniform change over the whole region. The variance contribution of the second mode is 16.13%, and its spatial distribution function shows a north-south phase inversion. Based on previous research and the physical considerations discussed herein, we include 13 climate indices to reveal the major predictors. To obtain an acceptable prediction performance, we comprehensively rank the climate indices, which are sorted according to the values of the new standardized algorithm of information flow (NIF, a causality-based approach) and correlation coefficient (a traditional climate diagnostic tool). Finally, Tropical Indian Ocean Dipole (TIOD), Arctic Oscillation (AO), and other four indicators are chosen as the final predictors affecting the first mode of AP over the MLYRB; NINO3.4 SSTA (NINO3.4), Atlantic-European Circulation E Pattern (AECE), and other four indicators are the major predictors for the second mode. In the final prediction experiment, considering the time series prediction of principal components (PCs) to be a small-sample problem, the Bayesian linear regression (BLR) model is used for the prediction. The experimental results reveal that the BLR model can effectively capture the time series trends of the first two modes (the correlation coefficients are greater than 0.5), and the overall performance is significantly better than that of the multiple linear regression (MLR) model. The prediction factors and precipitation prediction results identified in this study can be referenced to rapidly obtain climatological information for AP over the MLYRB and improve the regional prediction of AP elsewhere, which will also help policymakers prepare appropriate adaptation and mitigation measures for future climate change.
Keywords
