基于遥感温度植被干旱指数的小蠹虫害预警

针对小蠹虫对森林的危害隐蔽强,症状滞后性明显,在其早期发生时进行遥感识别非常困难。该文基于干旱和虫害存在一定的时滞相关性的假设,提出基于温度植被干旱指数(temperature vegetation dryness index,TVDI)预报小蠹危害的方法。以遭受大面积连续干旱和小蠹危害的云南省中部的石林县为案例区,利用Landsat数据,建立归一化植被指数(normalized difference vegetation index,NDVI)-地表温度(land surface temperature,Ts)特征空间,估算逐像元TVDI。基于地面小班调查的虫害等级数据(健康、轻度、中度和重度4个等级),比较不同虫害等级斑块TVDI差异。同时,以持续干旱2011年轻度受害区为例,结合受害前后云南松林NDVI差值(difference of NDVI before and after the forest attacked by bark beetles,d NDVI)表征实际受害程度的方法,建立TVDI与d NDVI的关系,对2012年进行预测。结果表明,2010-2015年,受害区整体呈下降趋势,TVDI由西向东逐渐变大。健康云南松林TVDI显著高于虫害云南松林(P0.05),且虫害越严重,TVDI越小;2011年,TVDI与d NDVI呈显著负相关(P0.05),可以用线性模型进行拟合,拟合决定系数R2为0.322。采用模型对2012年实际发生情况进行预测,得到预测与实测d NDVI均方根误差RMSE为0.237。在整体干旱的环境下,相对湿润的地方小蠹虫害更严重。因此,可以根据TVDI空间分布特征,找出TVDI相对较小的区域,作为虫害可能发生的重点关注区域,该研究对及时发布虫情监测信息有建设性的意义。

Prediction of bark beetles pests based on temperature vegetation dryness index

Abstract: Bark beetles (Scolytidae) are widely distributed in the world. In Yunnan Province, the southwest of China, Yunnan pine (Pinus yunnanensis) has been seriously damaged by bark beetles. Because Yunnan pine is one of the most important afforestation tree species in Yunnan Province, the government has implemented a series of measures to protect Yunnan pine from being damaged by bark beetles. However, it is very difficult to identify the bark beetle pest in the early stage due to the hysteresis symptoms. Based on the hypothesis of delay effect between drought and insect pests, one of the remote sensing drought indices TVDI (temperature vegetation dryness index) was proposed to predict the damage of bark beetles. TVDI was estimated by NDVI (normalized difference vegetation index) and brightness temperature in Shilin County where large area was damaged by drought and bark beetles. NDVI and brightness temperature were derived from long-term Landsat images from 2009 to 2014. The damage rating of Yunnan pine forests attacked by bark beetles was divided into 4 categories, including the healthy forests, lightly damaged forests, moderately damaged forests and severely damaged forests. TVDI was related to different damage ratings, and the relationship between TVDI and the difference of NDVI (dNDVI) before and after the Yunnan pine forest attacked by bark beetles was analyzed to effectively predict the possible occurrence of pests in the future. The dNDVI was used to stand for the real damage degree of Yunnan pine forests. To evaluate the relationship, lightly damaged forest polygons were selected in this study. However, only the pixels with dNDVI greater than 0 were extracted to eliminate those areas without being affected by insects in each polygon, and the larger polygon was split into many small polygons. In order to eliminate the difference of dNDVI range between different years, the maximum and minimum normalization method was used to normalize dNDVI. Results showed that the area and the number of attacked forest patches by beetles were declining from 2010 to 2015. TVDI in the healthy forest patches (0.657±0.114) was higher than that in the pest infected forest patches. Whereas, the value of TVDI (0.530±0.112) of the lightly damaged Yunnan pine forest patches was not significantly higher than that of the moderately damaged Yunnan pine forest patches (0.498±0.097) (P>0.05), but it was significantly higher than that of the severely damaged Yunnan pine forest patches (0.449±0.113) (P<0.05). And no significant difference was found between moderately and severely damaged Yunnan pine forests (P>0.05). Smaller TVDI generally corresponded to more serious degree of pest damages. The time series of TVDI spatial distribution also showed that TVDI was gradually increasing in recent years, while the pest infected area and the mean damage degree declined. Furthermore, TVDI was negatively correlated with dNDVI during the period of 2011, and a linear regression model could successfully express the relationship between them with the R2 of 0.322. The linear model could predict the dNDVI in 2012 with the root mean square error (RMSE) of 0.237. However, the regression line deviated from 1:1 line, and other factors should be considered in the future, including site conditions, stand structure, and forest growth. Overall, the relatively humid areas were more likely to have bark beetle pests during the drought periods. According to the spatial distribution of TVDI, the area with lower TVDI would be prone to be infected by insect pests. However, the prediction of the spatial distribution of bark beetles pest still needs further study. Other than TVDI, the model should involve other factors. This study proposes constructive suggestion on the timely releasing pest monitoring information.