Abstract: Based on the long-time series of wind observation data and blueberry growth period data of 9 national meteorological stations in Liaoning province during 1991−2020, the optimal machine learning model was selected to extend the maximum wind speed time series, using the maximum wind speed and disaster information data of the corresponding stations, considering the frequency and duration comprehensively, the meteorological index thresholds of disaster risk caused by wind of different grades during the ripening period of blueberries were determined, and the risk levels of mild, moderate and severe wind disaster were formulated, and the time-space distribution characteristics of blueberry gale-induced disaster risk were analyzed by the frequency of disasters and the ratio of stations. Based on the information diffusion theory, the risk probability of gale disaster is evaluated. The results show that the random forest model had higher simulation accuracy and could better extend the time series of extreme wind speed. The risk threshold of gale disaster in the mature period of blueberry with different maturity was maximum wind speed ≥13.9m·s−1, and different grades of gale disaster were further divided, and the verification results were more in line with the actual situation. There were differences in the trend of the climatic tendency rate of the average maximum wind speed in the study area. The early and middle maturity periods showed an upward trend, while the late maturity period showed a downward trend, and the risk of gale disaster also decreases. The change trend of the number of gale days in the mature period of blueberry with different maturity was not completely consistent. Among them, Kuandian and Benxi station kept the increasing trend unchanged, and Xiuyan, Qingyuan and Xifeng stations kept the decreasing trend unchanged. The high wind disaster risk index experienced by different maturity blueberry maturity was not much different, ranging from 14 to 17 days, but the high wind disaster risk index between stations was quite different. Among them, Xifeng station had the most days in each maturity period, ranging from 22 to 31 days, while Kuandian station had less days in each maturity period, ranging from 4 to 6 days. The exceeding probability of gale disaster risk in different blueberry varieties at maturity stage in the study area was 42.6%, 40.4% and 32.3% from high to low, respectively. The greater the probability, the higher the risk of gale disaster. In general, the northwest region of Liaodong Green Economic Zone was a high-risk area for gale-induced disasters, with a wide range of disasters, high frequency and heavy degree. The return period of gale-induced disaster risk was more than 3 years. Among them, Xifeng and Qingyuan stations had the highest probability of gale risk in each mature period, followed by Fushun station and Benxi station. The whole study area is more suitable for planting late-maturing varieties of blueberry, planting early-maturing varieties and medium-maturing varieties of blueberry area, need to do a good job of gale prevention measures.

Key words: Blueberry, Maturity, Wind disaster, Machine learning, Risk probability