Abstract: Based on the correlation analysis between temperature in April and 130 monitoring indexes in Guizhou Province from 1981 to 2019, a fixed monitoring index, optimal monitoring index, and stepwise regression statistical prediction model were established, and the results of the prediction model from 1981 to 2020 were tested. The results show that: in the work of forecasting climate trends, forecasting at anomalous level can improve climate forecasting skills. Among the three types of statistical prediction models, the stepwise regression has the best prediction effect, followed by the best monitoring index, and the fixed monitoring index has the worst effect. Compared with the provincial and national forecast products, the statistical forecast model has shown certain advantages in the return effect of the past 9 years. The regression prediction model established with the 7 optimal indexes has the highest Ps score for the temperature in April 2020, which is 14.1 points higher than that of the provincial forecast Ps score and 66.7 points higher than that of the national guidance forecast.