In order to implement the localized application of ECMWF (European Centre for Medium-Range Weather Forecasting) model well and improve the accuracy of precipitation forecast in Sichuan Province, the systematic deviation characteristics of forecast of precipitation with various magnitudes from ECMWF model were analyzed from July to September during 2020-2021. The result shows that the rain days forecasted by ECMWF model are more than the observations in Sichuan Province from July to September during 2020-2021, especially in Panxi region and western Sichuan Plateau. The heavy rain days forecasted by the model are more than the observations in southwestern Basin and Panxi region, while they are less than the observations in southern Basin. Then, the correction experiment about 24-hour cumulative precipitation forecast was carried out based on quantile mapping method, and it was applied to heavy rainfall forecast. After the correction using quantile mapping method, the TS (Threat Score) of forecast of rainstorm and above is improved by 7%-15%, and the TS of forecast of precipitation with various magnitudes is 2%-4% higher than the multi-model integrated objective forecast products. The POD (Probability of Detection) of forecast of heavy rain, rainstorm and above is improved by 10%-20%. The corrected location of rain belt in particular rainstorm areas is closer to the actual.

In order to deeply understand the water vapor characteristics and sources of persistent rainstorms in the Sichuan Basin and improve the rainstorm forecast capability in this region, the meteorological observation data from 4 955 national and regional automatic meteorological stations in Sichuan Province, the global data assimilation system (GDAS) data, the fifth-generation atmospheric reanalysis (ERA5) from the European Centre for Medium-Range Weather Forecasts (ECMWF) are used to analyze the water vapor transport characteristics of a continuous rainstorm process in August 2020 in the Sichuan Basin by using the Lagrangian method. The results show that the characteristics of water vapor transport at different initial height layers are different before and during the heavy precipitation process. In the middle and high level (from 5 500 to 10 000 m), the low latitude ocean is main source of air mass trajectory before the heavy precipitation occurring, while the southern coast of the Mediterranean is main source of air mass trajectory during the heavy precipitation process and dry and cold air in the middle and high latitude westerlies is brought to the basin. In the middle and lower level (from 1 500 to 5 500 m), during the heavy precipitation process, the source of water vapor trajectory adjusts from the southern coast of the Mediterranean to the low latitude ocean surface and warm and humid air on the low latitude ocean surface is brought to the basin. In the lower level (from ground to 1 500 m), the source of water vapor trajectory first adjusts to the low latitude ocean surface before the heavy precipitation occurring, and warmer and wetter air flow is delivered to the basin compared with the middle and lower level. Quantitative analysis of the water vapor contribution rates of different sources shows that the water vapor from the Bay of Bengal to the Gulf of Thailand is the dominant (66.6%), followed by the Arabian Sea (23.9%), and the South China Sea is the lowest (9.5%).

Based on the daily 2 m maximum and minimum temperature data from 1990 to 2019 in Sichuan Province, the temperature transitional weather processes have been analyzed statistically. Then a correction model of temperature change during transitional processes of temperature has been performed by using of NCEP/NCAR (National Center for Environmental Prediction/National Center for Atmospheric Research) daily reanalysis data and the LightGBM (Light Gradient Boosting Machine) algorithm.The results show that the area with the most temperature transitional processes is the slope transition zone between the plateau and the basin, while the least is in the basin. The number of temperature transitional processes in each region has an obviously seasonal differences with the most in spring and the least in winter, and the temperature transitional processes in spring is significantly more than those in the other three seasons. For the training set from 1990 to 2019，the LightGBM model has good performances with an overall accuracy of 78.64% and a mean absolute error of 1.35 ℃. For the independent testing set in 2020，the LightGBM model has an overall accuracy of 53.60% and a mean absolute error of 2.19 ℃, which are better than those of ECMWF (European Centre for Medium-Range Weather Forecasting), SCMOC and SPCO models.