新疆沙漠地区地表宽波段比辐射率遥感估算

地表比辐射率是估算地表温度以及地表长波辐射的一个重要参数,为了解决遥感影像反演地表比辐射率在裸地的精度不足问题,本文在新疆沙漠地区利用2类数据：① 傅里叶变换热红外光谱仪（FTIR）数据,在2013年、2014年秋天沿2条穿越塔克拉玛干沙漠的沙漠公路测量得到25个点的地表比辐射率数据;② 与FTIR数据同时期的MODIS温度/比辐射率数据MOD11A1、MOD11B1和反射率数据MOD09GA以及反照率数据MCD3A3,利用这2类数据为数据源,估算新疆沙漠地表比辐射率。首先,重新估算了基于MODIS宽波段比辐射率（BroadBand Emissivity, BBE）方程的系数和基于GLASS（Global L And Surface Satellite）BBE方程的系数,由此获得了GLASS BBE和MODIS BBE的修正方程。其次,将修正前后的GLASS BBE与FTIR和MODIS BBE作对比,发现其精度显著提高：① 与FTIR数据对比,修正前后的GLASS BBE方程的决定系数R²值从0.42增加到0.95,均方根误差（RMSE）和偏差（Bias）分别减少了1和3个数量级;② 与MODIS BBE方程数据对比,修正前后的GLASS BBE的R²值从0.69增加到0.91,RMSE和Bias分别减少了1和2个数量级。因此,修正后的基于GLASS和MODIS的BBE方程,极大地提高了遥感影像对裸地尤其是沙漠地区地表比辐射率的反演精度。使用修正后的GLASS BBE方程反演出新疆3个沙漠地区的BBE分布特征。结果表明,塔克拉玛干沙漠由于土地类型较为单一,其BBE值主要为0.88~0.92,而古尔班通古特沙漠以及库姆塔格沙漠受到地形、植被等的影响,BBE值稍微偏高,分别为0.89~0.95和0.89~0.94,沙漠周边稀疏植被区及其边缘地区的值范围为0.95~1.00。本文基于GLASS和MODIS的适用于新疆沙漠的BBE方程,为陆面过程的研究与模拟提供了支持。

Remote Sensing Estimation of Surface Broadband Emissivity over the Deserts in Xinjiang

Surface Broadband Emissivity (BBE) is a key variable for estimating surface longwave net radiation, which is a component of the surface radiation budget and an important parameter in climate, weather, and hydrological models. A constant land surface longwave emissivity, or simple parameterization, has been adopted by many land-surface models because of the lack of reliable observations. Moreover, of all the various Earth surface types, bare soil has the largest variation in BBE. Thus, accurate estimation of land surface emissivity for bare soil is important for retrieval of surface temperature and calculation of longwave surface energy budget. In order to retrieval accurate emissivity from remote sensing in the bare-soil area, two types of data were obtained indeserts of Xinjiang: (1) Land surface emissivity at 25 sites along two highways across the Taklimakan Desert. The spectral of broadband emissivity were measured in the fall of 2013 and 2014 by portable FTIR ( Fourier Transform thermal InfraRed spectroscopy), (2) MODIS (Moderate Resolution Imaging Spectroradiometer) temperature and emissivity data ( MOD11A1 and MOD11B1 ), reflectance data ( MOD09GA ), and albedo data (MCD43A3) of the same period.The two types of data were combined to estimate the surface emissivity of the Xinjiang deserts. Firstly, we re-estimated the coefficients of the MODIS BBE equation and the GLASS (Global Land Surface Satellite) BBE equation. The MODIS and GLASS BBE equations were both optimized with the new coefficients. Secondly, we compared with the optimized GLASS BBE equation with the FTIR and MODIS BBE equations. By comparison, the accuracy of optimized GLASS BBE equation was significantly improved, which was proved by: (1) According to the error analysis against FTIR data, the value of R² (coefficient of determination) increased from 0.42 to 0.95, the RMSE ( Root Mean Square Error ) and the Bias reduced by 1 and 3 orders of magnitude, respectively; (2) Compared to MODIS BBE data, the value of R² increased from 0.69 to 0.91, the RMSE and Bias reduced by 1 and 2 orders of magnitude, respectively. In our study, the BBE in Xinjiang desertswasfinally calculated using the optimized GLASS BBE equation. Our results show that the BBE in Taklimakan Desert ranged from 0.88 to 0.91, which was due to the single type of terrain, soil, and particularly aridity.While the Gurbantunggut Desert and the Kumtag Desert were more affected by topography and vegetation, their BBE values (0.89~0.95 and 0.89~0.94, respectively) were slightly higher than that of the Taklimakan Desert. The sparse vegetated area around the deserts and the edge area had the highest BBE(0.95~1.00).The BBE equationsdeveloped for Xinjiang desertsbased on GLASS and MODIS provides useful reference forfuture land-surface process models.