气温与陆地表面温度和光谱植被指关系的研究

Pre-study on Reverse Air Temperature from Remote Sensing——Relationship Between Vegetation Index,Land Surface Temperature and Air Temperature

Prihodko和Goward(1997)在假设气温与浓密植被冠层温度近似的基础上,利用空间分辨率为1 km NOAA影像13×13像素窗口的植被指数和LST的线性外推求得NDVI=0.86时LST,作为中间像素气温的估计值,这种方法假设了两个前提,即13×13像元窗口的植被指数和LST呈负线性相关;高植被覆盖条件下的LST与气温相等。根据Parton和Logan(1981)提出的气温时间尺度转换模型将297个气象观测站获得的最高和最低气温资料计算MODIS Terra卫星过境时刻的气温,利用MODIS每天陆地表面温度(LST)产品、16 d合成植被指数产品,探讨气温、植被指数和LST之间的关系,对Prihodko&Goward法的两个前提进行调研。结果表明:1在晚上,LST与植被指数之间相关性很小;方差分析的结果表明晚上LST与晚上气温差异不显著,因此晚上的气温基本可以由LST代替;2在白天,在地形平坦的平原地区,植被覆盖度范围较大的情况下,LST与植被指数呈负相关关系,但是在地形复杂的青藏高原地区和植被覆盖度范围较小(如在沙漠地区)的情况下,植被指数与LST的关系很不明确;3在白天,LST与气温的关系随着植被生长状况差异而不同,在稀疏植被覆盖条件下,LST大于气温;当植被指数>0.7时,获得的LST与气温差异不显著,这与前人研究成果一致。根据结果2和3,我们认为Prihodok&Goward模型应用于区域尺度上计算白天气温存在一定局限性,特别是应用于我国地形复杂的青藏高原地区和植被稀少的西北荒漠地区。

On the assumption that air temperature approximately equal to canopy temperature of dense vegetation and the negative linear relation between NDVI and LST in a 13 ×13 window, Prihodko and Goward （1997） estimate LST when NDVI=0. 86 as Ta. In this paper, we first estimated the air temperature at the time of Terra satellite passing through from the maximum and minimum Ta measured from meteorological station with Parton ＆ Logan model and took it as ＂true air temperature＂. The relation between NDVI, LST and ＂true air temperature＂ were analyzed to verify the two assumptions that Prihodko and Goward used. The following conclusion were reached.①In night, the relationship between LST and NDVI cannot be identified but the air temperature can be represented by LST obtained from satellite approved by variance analysis; ②In day, the negative linear correlation between NDVI and LST can be observed in a plain when NDVI has a wide range. But the relation between NDVI and LST can＇t be defined in the Qingzang plateau which has a complex topographty and in desert which has a very narrow NDVI range; ③In day, ＂true air temperature＂ are smaller than LST in the pixels with low NDVI because of the higher soil background temperature. But when NDVI〉0.7, there was no difference between ＂true air temperature＂ and LST statistically. Based on results ② and ③, the method proposed by Prihodok Goward for air temperature reversion from remote sensing has some limits when used in regional scale especially in desert and mountainous region.