混合像元对遥感干旱指数监测能力的影响

Effect of Mixed Pixel on Monitoring Ability of Remote Sensing Drought Index

混合像元是卫星遥感中常见的现象,也是影响卫星遥感地表参数精度的重要因子。为了解不同植被盖度的混合像元对遥感干旱指数监测能力的影响,利用一次陇东黄土高原卫星—地面准同步的观测数据,以常用的垂直干旱指数(Perpendicular Drought Index,PDI)为例,研究不同植被盖度下混合像元对PDI监测表层含水量能力的影响,提出了一个考虑混合像元的表层含水量(SMsur),并在此基础上建立表层含水量的遥感监测模型。结果表明:(1)在不考虑混合像元的情况下,PDI与裸土和植被区表层土壤含水量的相关系数(R2)最高分别为0.5、0.05;(2)考虑混合像元后,PDI与SMsur相关性较好,相关系数(R2)最高达0.6,说明在遥感指数监测能力评估中考虑混合像元问题,可以更加准确地衡量遥感干旱指数在研究区的适用性及监测效果;(3)基于PDI建立的模型反演研究区表层含水量与考虑混合像元的SMsur接近,可以反映不同植被盖度下地表的水分状况,相关系数(R2)达到0.85,相对偏差低于8.0%,均方根误差最小达到2.17%。模型估算的研究区SMsur区域分布特征与实地考察结果基本一致,植被覆盖浓密的农田湿度最大,SMsur最高可达到30%以上。

Mixed pixel is not only a common phenomenon in remote sensing but also an important factor influencing accuracy of surface parameters retrieval. To understand the influence of mixed pixel on monitoring ability of remote sensing drought index,using the Perpendicular Drought Index（ PDI） as an example,the ground-satellite experiment data were used to verify the applicability of PDI on surface moisture monitoring under different coverage conditions in Longdong loess plateau,and a new surface water content（ named SMsur） considering mixed pixel was proposed,and on this basis,the remote sensing monitoring model for surface water content was established. The results are as follows：（ 1） Without considering mixed pixel,the maximum correlation coefficients between PDI and soil water content were 0. 5 and 0. 05 for bare soil and vegetation conditions,respectively.（ 2） Considering mixed pixel,the maximum correlation coefficient between PDI and SMsurwas 0. 6,which indicated that it could describe the applicability and monitoring effect more accurately about remote sensing indexes considering mixed pixel in monitoring ability evaluation in study area.（ 3） The surface water content estimated by PDI model considering mixed pixel could reflect the status of surface moisture under different vegetation coverage underlying surface,and their values were close to the observed data,the correlation coefficient of them was 0. 85,the mean absolute percent error was less than 8. 0%,and the minimum root mean square error was 2. 17%. The regional distribution of the estimated SMsurwas consistent with the investigation result in the study area,and the moisture of the farmland was largest with SMsurmore than 30%.