基于无人机影像和半变异函数的玉米螟空间分布预报方法

Prediction method for spatial distribution of corn borer based on unmanned aerial vehicle images and semivariance function

空间变异拟合是实现小区域亚洲玉米螟变量防治和药剂减施的工作基础。该文以建立玉米螟空间分布最优半变异拟合函数和预测模型为研究目的,选择半径为2 km的圆形区域作为核心研究区,开展无人机高清影像获取、受虫害株率人工调查、越冬虫源基数调查、村庄网格划分等数据获取与处理,研究玉米螟空间扩散影响因子及空间变异特征。研究表明,受虫害株率空间变异显著,服从对数正态分布;经地统计学交叉验证,0阶指数模型为拟合受虫害株率空间分布的最优半变异函数模型;越冬玉米螟虫源数量与受虫害株率相关系数为0.61。因此,通过受虫害株率人工调查,基于指数模型利用克里格插值方法,可以生成受虫害株率空间分布图;由于影响因素众多,仅依据越冬虫源数量尚难以准确预测玉米螟发生严重程度。该文所探索的技术路线可以为玉米螟小区域变量防治提供实现途径,据此可以达到药剂节约和环境保护的目的。

Serious pest of Asiatic corn borer in China would lead to yield reduction by more than 30%. Precise prediction and variable rate prevention of corn borer are required for cite-specific management nowadays, since too much chemical usage lowers the quality of agricultural product as well as deteriorates the pollution of agricultural environment. The farmers producing silage corn are in face of serious problem of sustainable development. Semivariance function for spatial distribution prediction is the basis of cite-specific management for corn borer. The objective of the research was to find the optimum semivariance function and prediction model for Asiatic corn borer in Shenyang. A circular area around Dayanghe Village with a radius of 2 km was selected as core experimental area. Five other villages were located around Dayanghe Village. Silage corn was planted by China Huishan Dairy Holding Company Limited in the fields among the 6 villages. Grain corn was planted in some small fields around the houses by farmers. PHANTOM 4, unmanned aerial vehicle （UAV） with the camera of FOV 9420 mm, was used to collect the high resolution image. The ground resolution was 0.04 m. The villages, corn straws, and corn fields were interpreted through the UAV images. Both percentage of pest damaged plants （91 sampling points） and percentage of overwintering worms were investigated manually in 2016. And the villages in the research area were divided into grids. The life cycles of Asiatic corn borer, the impact factors of spread, and the spatial variability of percentage of pest damaged plants were studied. Quantity of worm source and spread distance were the key impact factors of the spread behavior of corn borer. Only the larvae and adults have the spread capability, and the adults of corn borer could spread for a large area. The results showed that the spatial variability of pest damaged plants was significant, whose standard deviation was 12.03% and coefficient of variation was 47.38%. It proved that variable rate fertilization for corn borer cite-specific management was necessary. The distribution of percentage of pest damaged plants fitted log-normal distribution, whose K-SP was 0.084. Furthermore, zero-order exponential model was the optimum semivariance function for the fitting of pest damaged plants distribution by cross-validation. Correlation coefficient between overwintering insects and pest damaged plants was 0.61. Therefore, the spatial distribution maps of Asiatic corn borer could be produced through investigating the percentage of pest damaged plants and Kriging interpolation with zero-order exponential model. Only worm sources could not precisely predict the distribution and influence degree of corn borer because of numerous complex impact factors. But the investigation of overwinter worm sources was meaningful for occurrence degree for the current year. Compared with traditional spraying in the research area, the dosage of variable rate spraying reduced by 28.1%, which supposed the spraying dosage of level II was 70% of that of level III. Therefore, the reduction of spraying dosage was obvious, and the variable rate spraying was valuable and should be encouraged. In conclusion, the technical route of this paper by using UAV and ground investigation, can be the implementation approach for cite-specific management of Asiatic corn borer, and can reduce pesticide waste and environmental pollution.