利用MODIS数据反演地表温度的研究

Retrieving Land Surface Temperature from MODIS Data

地表温度(LST)是气象、水文、生态等研究中一个重要的参数,目前国内的研究大多使用NOAA/AVHRR数据来获取地表温度,应用MODIS数据获取LST基本上还是空白。MODISLST反演算法精度较高但是计算复杂,在很大程度上限制了其应用。采用简单的统计方法和神经网络方法,得出了内蒙古东北地区的LST计算公式。该公式计算简单而且精度很高,完全能够满足一般的研究需要。

Land surface temperature (LST) is an important factor in meteorology, hydrology and zoology etc, most researches in these areas in China use LST data calculated from NOAA/AVHRR and MODIS data is seldom used. MODIS LST algorithm of NASA is precise and its error is less than 1K while it's complex and it's not easy to use. In this paper two simple methods, statistical method and artificial neuronal network method, according to split window theory, are used to retrieve LST. The result shows that the latter is more precise than the former and the error of 98.8% LST data is within〔-0.5 0.5〕K. While the precision of neuronal network can't be promoted any more because of the complexity of land surface type, meteorological conditions and so on. A LST formula T=24.393*b31-19.831*b32-00014*θ+245.145 which can be applied to Northeast area of Inner Mongolia is provided. A coarse validation shows that the result of this formula is reliable in common use.