Daily temperature prediction with CMIP6 in the historical period of Hong Kong

Abstract

An ensemble forecast of historical climate simulations consisting of CMIP6 models was used to project years during World War II for which temperature data were missing for Hong Kong. In multi-model ensemble prediction, since the deep learning in this paper considers that GCM has dependencies between continuous values, the overall effect of prophecy may be better than machine learning. The daily temperature recorded by the Hong Kong Observatory and the extreme temperature index are used to evaluate the daily temperature forecast effect of CMIP6 in the historical period. In the 139-year record span of the Hong Kong Observatory, comparing the forecast results with and without CMIP6, the results show that except for the TN10p and TX10p cold indicators, all other indicators are on the rise, indicating that the temperature is rising.