| Popis: |
This study proposes a deep neural network (DNN) as a downscaling framework to compare original variables and nonlinear data features extracted by kernel principal component analysis (KPCA). It uses them as learning data for DNN downscaling models to assess future regional rainfall trends and uncertainties in islands with complex terrain. This study takes Taichung and Hualien in Taiwan as examples. It collects data in different emission scenarios (RCP 4.5, RCP 8.5) simulated by two Global Climate Models: ACCESS and CSMK3, in the Fifth Assessment Report (AR5) of the Intergovernmental Panel on Climate Change (IPCC), and monthly rainfall data of case regions from January 1950 to December 2005 in the Central Weather Bureau in Taiwan. DNN model parameters are optimized based on historical scenarios to estimate the trends and uncertainties of future monthly rainfall in the case regions. A multivariate linear regression is used as a baseline model to compare their effectiveness. The simulated results show that by both ACCESS and CSMK3, the dimensionless root mean squared error (RMSE) of KPCA was higher than that of the original variables in Taichung and Hualien. According to the analysis of three-class classification (according to the arrangement in descending power of historical rainfall, the predicted rainfall is divided into three ranges, low, normal, and high, marked by 30% and 70% of monthly rainfall), the wet season rainfall at the two stations is concentrated in the normal range. The probability of rainfall increase will improve in the dry season and will reduce in the wet season in the mid-term to long-term. The future wet season rainfall in Hualien has the highest variability. It ranges from 201 mm to 300 mm, with representative concentration pathways RCP 4.5 much higher than RCP 8.5. The median percentage increase and decrease in RCP 8.5 are higher than in RCP 4.5. This indicates that RCP 8.5 has a greater impact on future monthly rainfall. |
