八角林病虫害遥感识别模型

Remote Sensing Recognition Model of Illicium verum Forest Pests and Diseases

【目的】受病虫害胁迫诱导的微弱光谱信号极易淹没在植被物候引起的光谱变化特征中,探索一种基于植被胁迫光谱弱信号增强的森林病虫害监测方法,为森林病虫害防治与管理提供科学依据。【方法】以八角林为研究对象,选取广西壮族自治区乐业县为试验区,收集2019—2021年试验区Sentinel-2影像数据,首先选取对植被病虫害胁迫响应敏感的红边归一化差异植被指数(NDVI705)、红边位置指数(REPI)、红边叶绿素指数(CI_(red-edge))、植被衰减指数(PSRI)、特征色素简单比值指数(PSSRA)、植被光合有效辐射吸收系数(FAPAR)作为八角林病虫害指数的预选集;其次采用S-G滤波法构建光谱指数时间序列曲线,优选出能够综合表征八角林病虫害胁迫诱导的形态颜色和生理要素变化敏感的PSRI和FAPAR指数;然后应用季节趋势分解法(STL)对FAPAR、PSRI指数进行时间序列分解、季节性分量剥离,综合剥离季节性影响后的FAPAR、PSRI分量构建八角林病虫害敏感指数(IPDI);最后运用随机森林算法建立八角林病虫害胁迫监测模型。【结果】1)与健康植被相比,受病虫害胁迫的八角林FAPAR偏低、PSRI偏高;2)应用STL方法对植被病虫害胁迫响应敏感参数进行时间序列分解,可有效剥离植被物候变化中对植被胁迫弱信息的影响,增强八角林病虫害胁迫响应敏感性;3)基于IPDI的八角林病虫害遥感识别模型精度较高,2019—2021年模型的Kappa系数和总体精度分别为0.81、0.84、0.80和87.59%、88.51%、84.17%,2020年八角林遥感计算的病虫害胁迫受害面积与统计受灾面积的相对误差为2.08%。【结论】基于植被胁迫光谱弱信号增强方法可有效监测八角林病虫害胁迫分布状况,显著提升防治效率,助力八角林的可持续管理与生态保护。

【Objective】Weak spectral signals induced by pest and disease stress are often obscured by spectral variations caused by vegetation phenology.This study investigates a forest pest and disease monitoring method that enhances weak spectral signals related to vegetation stress,aiming to provide a scientific basis for the prevention and management of forest pests and diseases.【Method】Illicium verum was selected as the study species,with Leye County in the Guangxi Zhuang Autonomous Region designated as the experimental area.Sentinel-2 imagery data from 2019 to 2021 were collected for this region.Initially,six vegetation indices sensitive to pest and disease stress responses were selected as preliminary indicators for Illicium verum pest and disease stress:the normalized difference vegetation index(NDVI705),red edge position index(REPI),chlorophyll reflectance rededge index(CIred-edge),plant senescence reflectance index(PSRI),pigment-specific simple ratio chlorophyll index(PSSRA),and fraction of absorbed photosynthetically active radiation(FAPAR).The Savitzky-Golay(S-G)filtering method was then employed to construct time series curves of these spectral indices.PSRI and FAPAR were identified as the most effective indices for comprehensively characterizing morphological color and physiological changes induced by pest and disease stress in Illicium verum.The Seasonal-Trend decomposition using LOESS(STL)method was applied to decompose the time series of FAPAR and PSRI indices,allowing for the isolation of seasonal components.This facilitated the construction of the Illicium verum Pest and Disease Index(IPDI)by integrating the seasonally adjusted FAPAR and PSRI components.Finally,a monitoring model for pest and disease stress in Illicium verum was developed using the Random Forest algorithm.【Result】1)Compared to healthy vegetation,Illicium verum plantations under pest and disease stress exhibited lower FAPAR and higher PSRI values.2)The STL method effectively isolates the influence of phenological changes on parameters sensitive to vegetation stress from pests and diseases,thereby enhancing the sensitivity of Illicium verum to stress detection.3)The remote sensing identification model based on IPDI demonstrated high accuracy,with kappa coefficients and overall accuracies from 2019 to 2021 of 0.81,0.84,and 0.80,and87.59%,88.51%,and 84.17%,respectively.In 2020,the relative error between the remote sensing-calculated damage area of Illicium verum and the statistical disaster area was 2.08%.【Conclusion】The method based on enhancing weak spectral signals from vegetation stress effectively monitors the distribution of pest and disease stress in Illicium verum forests.This approach significantly improves control efficiency and supports the sustainable management and ecological conservation of these forests.