基于MODIS数据地表温度反演劈窗算法的比较研究

选取QIN和SOB两种代表性劈窗算法对辽宁地区地表温度进行反演，并分析二者的精度和误差分布。结果表明：QIN和SOB算法反演的地表温度（TS）与地面气象台站准同步观测的气温和地温的线性拟合显著，SOB算法线性拟合更好；从误差分布直方图上看，两种算法的反演结果与地温更接近，SOB算法与同步气温和地温在±2 ℃之间的误差比例略高于QIN算法；在野外开展与卫星遥感空间尺度一致的地表温度观测试验，QIN和SOB算法与实测值的平均绝对误差均为1.5 ℃；与NASA官网发布的地表温度产品对比发现，QIN和SOB算法的平均绝对误差分别为1.75 ℃、1.70 ℃；因此QIN、SOB算法在辽宁地区均适用，SOB算法误差更小。

Comparative study on land surface temperature retrieval using split-window methods based on MODIS data

Using QIN and SOB representative algorithms for retrieving land surface temperature (LST) based on the MODIS in Liaoning province, both precision and error were analyzed. The results show that LST retrieved by QIN and SOB algorithms has a good linear fitting with the observed air temperature and surface temperature, especially for a SOB algorithm. According to the error histograms, retrieved LST by two methods is close to the observed surface temperature. Errors of the air and surface temperature between ±2 ℃ calculated by two methods are compared, and error ratio of the SOB algorithm is slightly higher than that of the QIN algorithm. Field experiment, being a same resolution with remote sensing data, suggests that mean absolute errors between two the retrieved and observed temperature both are 1.5 ℃. Compared with the LST from NASA website, mean absolute error of the QIN and SOB algorithms are 1.75 ℃ and 1.70 ℃, respectively. Thus, the two algorithms both are suitable in Liaoning province, and the SOB algorithm has less error.