A learning-based approach to regression analysis for climate data-A case of Northeast China

Global climate change is an important issue that all of humanity needs to address together. Precipitation is an important climatic feature for agricultural development and food security, and the study of precipitation and its associated climatic factors is important for the analysis of global change. As an important part of China's food production, Northeast China has a temperate monsoon climate with simultaneous rain and heat, which is favorable for crop growth. In this paper, a scientific workflow for climate data analysis with a learning-based method is designed. Using climate data from typical models in CMIP6, a machine learning-based approach is used to establish regression relationships between precipitation and climate variables such as temperature, humidity and wind speed in Northeast China, which is validated through a time series approach. We design a weight-based model ensemble method and a learning-based bias correction method, so that the ensemble model can achieve better performance. We also analyze the precipitation trends in Northeast China under the three Shared Socio-economic Pathways (SSPs). This will help researchers to analyze the long-term evolution and factors of climate. Based on the 22 models in CMIP6, this paper conducts regression analysis on the climate data of Northeast China, explores the relationship between precipitation and other variables, and proposes a weight-based model ensemble and learning-based bias correction.image