Abstract: Tea gardens are predominantly located in low mountain and hilly areas with complex terrain and variable weather conditions, making the prediction and early warning of frost damage quite difficult. To enhance the accuracy of frost damage prediction for tea trees under such complex terrain conditions, we utilized the Google Earth Engine (GEE) cloud platform and multi-temporal Sentinel-1, Sentinel-2 optical imagery. An accurate extraction of tea planting information was performed using a Support Vector Machine (SVM) classifier resulting in a spatial distribution map of tea plantation with a 10 m resolution for Xuancheng City, Anhui Province, China, in 2023. Then, a precise prediction method for tea frost damage was developed by integrating numerical weather prediction products, a Digital Elevation Model (DEM) of Xuancheng City and meteorological prediction grade indicators for tea frost damage. Using Geographic Information System (GIS) technology and ANUSPLIN, an analysis of a frost damage event that occurred in Xuancheng City on March 13, 2023 was conducted. The results showed that the tea plantation area in 2023 was 28012 hm² in Xuancheng City, with 81% distributed at altitudes between 100 and 500 m. Compared with the data from the Anhui Provincial Bureau of Statistics, the relative error was 2.23%. The effectiveness of refined forecast for frost damage was 10 days and the spatial resolution was 100 m. Based on the current planting zone, the precise frost damage forecast revealed that 76.2% of tea gardens in Xuancheng experienced frost damage on March 13, 2023. Among these, the areas affected by mild, moderate, and severe frost damage were 17564, 3586, and 196 hm², accounting for 62.7%, 12.8%, and 0.7% of the total tea garden area, respectively. The prediction accuracy for tea frost damage (with a grade difference ≤1) reached 83.4%, 83.3%, 87%, and 90% for forecasts made 7, 5, 3, and 1 days in advance, respectively. This study realizes refined and quantitative prediction of tea frost damage at the municipal level, demonstrating high value for practical applications.

Key words: