关键词: 稻纵卷叶螟, 气象条件, 田间种群发生量, 机器学习, 余庆县

Abstract:

Taking Yuqing county of Guizhou province as the study area, data from the rice leaffolder (RLF) field surveys and meteorological stations during the main local RLF occurrence period (June-August) from 2011 to 2020 and methods such as correlation analysis, mathematical statistics, and machine learning were used to analyze the effects of meteorological conditions on the population abundance of larvae and adults of RLF in Guizhou province, screen for key meteorological influence factors, and explored the predictive effects of different modeling methods. The results showed that: (1) meteorological factors that favored the increase in RLF larvae were mean minimum temperature, precipitation, mean relative humidity, and minimum relative humidity. Mean temperature, minimum temperature, mean precipitation, mean relative humidity, mean minimum relative humidity, sunshine hours, and mean 0cm ground temperature were found to favor an increase in RLF adults. (2) The lagged effect of meteorological conditions on RLF’s field populations in Guizhou province was greater. The RLF field population abundance was mainly influenced by mean temperature, mean minimum temperature, precipitation, minimum relative humidity, mean wind speed, and mean 0cm ground temperature, which were per-five-day daily averaged between June to August each year. The longest lag effect could be up to about 30d, with significant effect period of 3, 4, and 5 pentads in advance. (3) The results of the model simulation varied considerably between the different modeling schemes. Non-linear models (R²=1.00, MAE=2.62 individual, RMSE=3.89 individual) were more effective than linear models (R²=0.46, MAE=164.98 individual, RMSE=240.66 individual), and the simulation effect for the adult population (R²=0.68, MAE=81.29 individual, RMSE=117.98 individual) was better than that for the larva population (R²=0.67, MAE=118.78 individual, RMSE=173.92 individual).

Key words: Rice leaffolder（RLF）, Meteorological condition, Field population abundance, Machine learning, Yuqing county