山地冰川表面温度反演算法对比——以祁连山七一冰川为例

Comparison of Algorithms for Retrieving Mountain Glacier Surface Temperature from Remote Sensing Data:A Case Study on the Qiyi Glacier in the Qilian Mountains,China

冰川表面温度决定冰川表面热力状况分布和消融状态,其对建立冰川能量—物质平衡模型和研究气候变化背景下的冰川响应具有重要意义。Landsat系列卫星提供了海量免费高空间分辨率遥感数据,广泛应用于表面温度的时空变化研究。本文使用2013年5月至2014年9月的10景Landsat 8 TIRS热红外数据,对比三种单通道温度反演算法(覃志豪单窗算法、Juan C.Jiménez-Munoz普适性单通道算法和Jordi Cristóbal普适性单通道改进算法)和两种劈窗温度反演算法(Juan C.Jiménez-Munoz劈窗算法和Offer Rozenstein劈窗算法)在祁连山七一冰川表面的准确度和适用性,结果表明:(1)劈窗算法反演精度高于单通道算法,其中Offer Rozenstein劈窗算法误差最小,均方根误差为1.75 K、平均绝对误差为1.49 K,而Jordi Cristóbal普适性单通道改进算法误差较大,均方根误差和平均绝对误差分别为3.35 K和2.72 K;(2)夏季消融期各算法均有较大偏差,冬季是各算法反演误差最小的季节;(3)覃志豪单窗算法对水汽含量敏感性最低,Offer Rozenstein劈窗算法对水汽含量和发射率的敏感性较高;Jordi Cristóbal普适性单通道改进算法在低大气含水量的高海拔冰川地区存在一定局限性。研究结果为不同表面温度反演算法在高海拔山地冰川区域的对比和适用性研究提供了科学依据。

Changes in glacier surface temperature can affect the thermal and the surface ablation statuses of a glacier.Landsat series satellites provide a huge and high spatial resolution remote sensing data,which has been widely used for the study on the temporal and spatial changes of the Earth surface temperature,including the glacier surface temperature.Some algorithms for retrieving the land surface temperature from the remote sensing data have been developed,such as single-channel algorithms(for examples,Qin Zhihao's algorithm,Juan C.Jiménez-Munoz's algorithm and Jordi Cristóbal's algorithm)and split window algorithms(for examples,Juan C.Jiménez-Munoz's algorithm and Offer Rozenstein algorithm).In this paper,we evaluated the accuracies and the adaptabilities of these algorithms for estimations of the glacier surface temperature based on the Landsat8 Thermal Infrared Sensor data and the observed glacier surface temperature data by the autonomous weather stations on the Qiyi Glacier in the Qilian Mountains in China over the period of May 2013 through September 2014.The results indicated that:(1)The accuracy of the split window algorithm was higher than that of the single channel algorithm,among which,the error of Offer Rozenstein SW_R algorithm was the smallest,with a root-mean-square error(RMSE)of 1.75 K and mean absolute error(MAE)of 1.49 K,while that of Jordi Cristóbal SC_T algorithm was higher,with RMSE of 3.35 K and MAE of 2.72 K.(2)All the retrieval algorithms had larger errors in summer melting season but smaller in winter.(3)The Qin Zhihao MW algorithm had the lowest sensitivity to the variations in water vapor,while the Offer Rozenstein SW_R algorithm had a higher sensitivity to water vapor and emissivity,and there were some limitations of Jordi Cristóbal SC_T algorithm for high altitude glacier areas.This comparative analysis provides a basis for selecting a suitable algorithm for retrieving the glacier surface temperature by using the remote sensing data.