气候变化背景下的多尺度松材线虫病害监测与预测

The frequency and intensity of the disturbance events in the forest ecosystem are obviously increasing under the strong influence of climate change and human activity. One disturbance event is caused by pine wood nematode, one of the main exotic pests in China. And the nematode is spreading from south to north in recent years which has caused serious disturbance to the forest ecosystem. So it is of great signification to study the monitoring and prediction of pine wood nematode disease in the context of climate change. This study aims at the key technologies of multi-scale monitoring and predicting the nematode disaster from bottom to top, with the complementarity of the data synergism which contains satellite, airborne and ground. To achieve the goal, this study is based on the theory of insect ecology and geo-statistics, and supported by remote sensing and machine learning technology. Two prediction models and an early identification method for the nematode would be studied by combining the biological and ecological characteristics of pine wood nematode with the relationship between climate change, human factors and the spatial pattern of host plants. One model is disaster risk prediction model in regional scale. The other is disaster degree monitoring and prediction model in high risk area. The models and method would achieve risk prediction under climate change, high efficiency, low cost monitoring and prediction in fragile zone, and early detection of the nematode disaster. As a result, this study would improve the accuracy and efficiency of the nematode monitoring and forecasting, would provide data support for the interpretation of the nematode disease propagation mechanism. Furthermore, this study would provide the basis for the prevention and control of the nematode disease, would lay the foundation for the rational control of ecological safety and the health of the ecosystem.

气候变化和人类活动强烈影响下，森林生态系统扰动事件发生的频度和强度明显呈增加之势。松材线虫是我国主要的外来有害生物之一，近年来出现由南向北蔓延的趋势，对我国森林生态系统造成严重的扰动，开展气候变化背景下的松材线虫病害监测预测研究， 具有重要的意义。本研究以昆虫生态学、地统计学为理论基础，遥感和机器学习为支撑技术，利用星机地协同观测数据特征的互补性，研究自下而上的多尺度灾害监测预测关键技术。将松材线虫的生物和生态学特征与气候变化、人为因素及寄主植物空间格局的关系结合，研建区域尺度灾害发生风险预测模型、高风险区灾害程度监测预测模型以及林分尺度受害木早期识别方法，实现气候变化条件下的风险预测，脆弱区灾害程度高效率低成本监测预测，以及灾害早期发现。提高监测预测精度和效率，以期为松材线虫病害发生蔓延机制的解释提供数据支撑，为松材线虫病害防控提供决策依据，为生态安全和生态系统健康的合理调控奠定基础。

松材线虫病作为一种极具破坏性的病害，对中国松林带来了巨大的经济损失和生态破坏。随着全球气候变暖，该病害出现由南向北蔓延的趋势，因此，开展气候变化背景下的松材线虫病监测预测研究具有重要意义。本研究选择山东省、安徽省、辽宁省、长江经济带为主要实验区，叶片高光谱、无人机高光谱、Landsat等遥感数据为主要数据源，辅以外业调查数据、各地发生防治数据、气象数据及其他基础数据，将松材线虫病的生物生态学特征同遥感技术结合，在小尺度上， 明确了松材线虫病不同感病阶段诊断光谱波段、特征，发展了基于深度学习的受害木识别模型，构建了松材线虫病监测模型4个，其中：基于逐步判别分析的感病阶段识别模型精度达93.33%，基于3D-CNN模型的早期识别精度达96.62%，基于Faster R-CNN模型的监测精度为74.63%，基于Mask R-CNN模型的监测精度为84.06%，在中尺度上，明确了县域尺度环境因素对松材线虫病害发生的影响规律，提出了一种新的松材线虫病害监测指数，构建了中尺度松林枯死率监测模型，精度为75.96％。在大尺度上，明确了长江经济带松材线虫病时空扩散影响因素，提出了一种新的用以提取松材线虫寄主植物范围的空间分布的植被指数，构建了松材线虫病监测预测模型3个，其中：松材线虫病疫区监测模型精度为81.67%，松材线虫病害GAM预测模型能够解释预测松材线虫病受害率的80%的变化，CA-Markov松材线虫病灾害预测模型精度为93.19%。本研究实现了高精度的松材线虫病早期识别和感病阶段监测，松材线虫在中国的适生区分析，以及气候变化背景下的病害预测。为病害防控和森林科学经营管理提供快速、高效、准确的技术支撑。