以青海省作为研究区,利用MODIS每日地表反射率产品MOD09GA和逐日雪被合成产品MOD10A1,通过调整NDSI阈值,合成积雪分类图像,根据气象台站实测雪深数据,评价积雪分类精度,探索研究了适合该地区的NDSI阈值.研究结果表明,1)NSIDC发布的全球MODIS积雪产品MOD10A1在青海高原的积雪分类精度较低,在晴空下雪深大于3 cm的积雪分类精度为86.01％.2)研究区适合的NDSI阈值为0.37.在晴空下雪深大于3cm时,合成雪被图像的积雪分类精度可达90.37％,总精度99.51％,多测误差0.22％,漏测误差9.63％.3)同MODIS逐日雪被产品MOD10A1进行雪深分段精度比较,发现整体上自定义雪被图像的积雪分类精度较高,合成图像更符合青海高原积雪空间分布的真实情况.%Using the MODIS (moderate resolution imaging spectroradiometer) surface reflectance product of MOD09GA, the daily composite product of MOD10A1, and snow depth data observed in climate stations, the snow classification accuracy under different NDSI (normalized difference snow index) thresholds were studied. The snow classification accuracy of these snow cover maps were evaluated to find a credible NDSI threshold in Qinghai Province. 1) With a clear sky, the snow classification accuracy of MOD10A1 was 86. 01% when snow depth was more than 3 cm. 2) In the study areas, the credible NDSI threshold value was 0. 37. When snow depth was more than 3 cm, a snow classification accuracy of user-defined images under clear skies was 90. 37%, with an overall accuracy of 99. 51%, commission error of 0. 22%, and omission error of 9. 63%. 3) Compared with the MOD10A1 product, the user-defined snow cover images had higher overall accuracy and snow classification accuracy for different snow depths. Thus, the new NDSI threshold was more suitable for this study area in snow cover mapping.
展开▼
