利用NDVI估算云覆盖地区的植被表面温度研究

Study on Estimation of Vegetation Surface Temperature in Cloudy Region by NDVI

干旱监测等实际应用都需要全面掌握地表温度(LST)的空间分布,而云覆盖是这种应用的重要阻碍。试图根据地表温度变化与地表植被之间的相互关系,研究遥感影像中云覆盖区域植被表面温度的估算方法。由于植被的蒸腾作用,植被茂密程度对其表面温度的空间分布有较大影响。这种影响不仅在晴朗无云区域存在,同样适用于云覆盖区域。因此,首先分析云覆盖区域周边无云植被像元的LST与植被指数NDVI之间的关系,建立方程式,然后再利用NDVI在短时间内相对稳定的特点用另一幅图像来获取云覆盖区域的NDVI值,最后根据NDVI与LST之间的关系估计云覆盖植被像元的表面温度。将这一方法应用到山东省聊城市的Landsat ETM+图像,结果表明:当云覆盖范围≤2 000个像元(约1.72km2)时,通过NDVI来估计云覆盖区域植被表面温度的平均绝对误差<0.7℃,均方根误差<1.2℃。为了验证其实用性,又将该方法应用于安徽省蚌埠地区的TM图像,云覆盖范围在300个像元以下时,平均绝对误差小于0.1℃。因此,可以认为,当云覆盖范围不是很大时,利用NDVI估算云覆盖地区的植被表面温度,具有一定的可行性。

Drought monitoring and other practical applications all need to obtain comprehensive spatial distribution of LST,but cloud cover is a major barrier of this process.We attempt to study the method of estimating vegetation surface temperature of cloudy areas in the remote sensing images by the relationship between changes of LST and ground vegetation.Because of the vegetation transpiration,the density of vegetation has a great effect on changes of spatial distribution of LST.This effect not only exists in cloudless areas,but also can be applied to the cloudy areas.Therefore,we first analyzed the relationship between LST and NDVI in cloudless areas which were nearby cloudy areas and established the equation;Then we used the feature that NDVI was stable within a short time to acquire NDVI value of cloudy areas.At last,we estimated the LST of cloudy areas according to the relationship between the NDVI and LST.We applied this method to Landsat ETM＋ images of Liaocheng city in Shandong province.The results show that,when the cloudy area is within or equal to 2 000 pixels（about 1.72 km2）,the mean absolute error（MAE） of LST in cloudy area which estimated through NDVI is little than 0.7 ℃,the RMS is1.2 ℃.In order to verify its practicality,we also applied the method to TM images of Bengbu in Anhui province,when cloudy area is within 300 pixels,the MAE is less than 0.1 ℃.Hence,it can be argued that when the range of cloudy area is not very large,using NDVI to estimate the LST of cloudy area has certain feasibility.