用MODIS遥感数据监测东亚飞蝗灾害——以河北省南大港为例

Use of MODIS data to detect the Oriental migratory locust plague: A case study in Nandagang, Hebei Province

东亚飞蝗自 2 0世纪 80年代以来在我国再度猖獗危害。本文选择国家一类蝗区河北省南大港农场水库为实验区 ,以2 0 0 2年东亚飞蝗 (夏蝗 )大发生为背景 ,测定了不同受害程度芦苇的光谱曲线 ,并用MODIS遥感数据分析蝗灾危害范围和程度。结果表明 ,健康芦苇 (未受害 )呈现出典型植被光谱的“峰和谷”特征 ;而受害芦苇 ,其红光区处反射率增加 ,且受害程度越重 ,增加愈显著 ,近红外区的反射率则显著降低 ,且受害程度越重 ,反射率愈低。比较多时相归一化植被指数(NDVI)受灾前后的变化 ,NDVI值增加表明未受蝗虫危害 ,而其下降则表明是受灾区域 ,结合地面数据 ,找出了不同危害程度的NDVI临界值 ;其次根据像元累计法 ,确定了不同受灾程度的面积。结果显示 ,NDVI小于 0 2 15 6和 0 2 389分别为严重受灾区和中等受灾区 ,其发生面积依次为 16 8 74hm2 和 337 4 8hm2 ,判对率分别为 72 97%和 6 8 35 %。MODIS遥感监测蝗灾发生面积 (中等受灾及严重受灾 )占实际发生面积的 82 76 %。研究结果为实时、快速、大面积监测蝗虫种群动态奠定了基础 ,并为合理。

Outbreaks of the Oriental migratory locust, Locusta migratoria manilensis (Meyen) have once again become a serious problem in the last two decades of the 20th century in China. One of the coastal breeding areas of the locust, Nandagang reservoir of Hebei Province was chosen as study area in this paper. Against the background of the locust outbreak (summer locust) of this area in 2002, spectral reflectance curves of reeds for different damaged degrees (heavy, moderate or no damage) were measured and determined. The results showed that healthy reeds manifested the “peak and valley” configuration as that of other healthy green vegetation. But for damaged reeds the red reflectance increased, and the heavier the damage, the higher the red reflectance values, whereas the reflectance in the near infrared bands decreased dramatically with the damage escalating. Additionally, the locust damaged areas and degrees were determined using MODIS data. Comparisons of NDVI changes between pre and post damage showed that the increase for the same site in NDVI relative to the pre damage situation corresponded to the undamaged areas whereas the decrease in NDVI corresponded to the damaged areas. Coupled with the ground truth data, critical values of NDVI for heavy and moderate damage were determined as 0 2156 and 0 2389 respectively. Namely, areas with NDVI less than 0 2156 were identified as with heavy damage; while areas with NDVI less than 0 2389 as with moderate damage. By calculating the numbers of pixels, areas discriminated from MODIS images as with heavy and moderate damage were 168 74 hm 2 and 337 48 hm 2 respectively, with 72 97% and 68 35% accuracy respectively evaluated by the ground truth data. The total damaged areas identified from MODIS data amounted to 82 67% of actual occurrence areas. These findings contributed to the scientific foundation for quickly monitoring the dynamics of locust populations in real time and at large scale and efficiently controlling locust plague.