植被调节下的干旱区不透水面覆盖率遥感估算方法

基于DMSP/OLS夜间灯光数据的居住区指数模型（HSI）广泛应用于区域尺度城市不透水面扩张监测。但是，在干旱区由于受到裸岩、沙漠、戈壁等低植被覆盖区干扰，HSI算法的精度和适应性受到了一定的影响。为解决这一问题，本文利用植被覆盖度作为调节系数，对灯光数据与植被指数进行动态调整，构建了适用于干旱区的城市植被调节不透水指数（VAISI）；然后采用SVR模型，通过机器学习的方法构建了城市不透率参考数据与VAISI之间的非线性关系模型，实现对干旱区区域尺度不透水面覆盖率估算；最后，对模型估算结果进行了精度验证和比较分析。试验结果表明：在干旱区，VAISI解决了由于灯光溢出问题及城市周边裸土等低植被覆盖等因素导致的城市周边裸土像元不透率估算过高问题，一定程度上提高了城市内部不透水面空间分布信息的表达能力，有效克服了非灯光区估算结果高于背景值的现象。平均相关系数R由0.69提升到0.79，RMSE由0.17降至0.14。

Method for Arid Land Impervious Surface Percentage Estimation by Vegetation Index Adjustment

DMSP/OLS nighttime light data has been widely used in reginal impervious surface percentage estimation and monitoring. But, in the arid area, because of bare soil, desert and other the low rate of vegetation covered area effect, the accuracy and robustness of the existed algorithm was decreased seriously. In order to solve this issue, this paper used the vegetation coverage as the adjustment coefficient to adjust the light data and NDVI dynamically and build vegetation adjusted impervious surface index ( VAISI) . Impervious surface reference data were extracted from Landsat image, which was used as the sample data and validation data of model. And then, the non-linear relationship was built between impervious surface reference data and VAISI by SVR model to estimate the impervious surface percentage. The result indicated that the VAISI model solved the problem of higher estimated in arid land because of the bare soil around the city and the low rate of vegetation covered area, improved the impervious surface's space distribution information inside city, and overcame the obstacle of that the non-light area is higher than background values. The average correlation coefficient between the estimated result by the VAISI and the reference data were increased from 0.68 to 0.79 and RMSE was decreased from 0.17 to 0.13.