基于MODIS和AMSR-E资料的青海省旬合成雪被图像精度评价

Accuracy Analysis of Snow Cover Based on MODIS and AMSR-E in Qinghai Province

利用2002年10月1日至2008年3月31日青海省MODIS/Terra-Aqua每日雪被产品(MOD10A1、MYD10A1)和AMSR-E/Aqua每日雪水当量产品,研究了MODIS和AMSR-E旬数据的融合算法,合成了AMSR-E旬积雪分类图像(AE_10D)、MODIS旬积雪分类图像(MOYD_10D)及二者再次合成的旬积雪分类图像MDAE_10D;结合气象台站的雪情数据,对比分析这3种图像的分类精度.结果表明:合成的AE_10D,因受空间分辨率等因素影响,当积雪深度在4cm以上时,积雪分类精度(Sa)仅为48.44%,总精度(Oa)为55.43%;积雪深度对MOYD_10D图像的分类精度有较大影响.MOYD_10D图像对浅层的积雪识别率不高,当雪深在1～3cm时,Sa仅为20.63%;当积雪深度在4cm以上时,Sa为53.56%,Oa为93.54%;当雪深≥11cm时,Sa为88.57%.旬合成图像MDAE_10D在雪深为1～3cm时Sa为20.86%,当雪深在4cm以上时,Sa随雪深的增加而最大,Sa达到59.28%,Oa为93.66%;当雪深≥11cm时,Sa达到最高91.43%.由于MDAE_10D图像结合了AE_10D和MOYD_10D的特点,Sa和Oa均有提高.因此,该合成图像在牧区雪灾监测及评价等方面具有重要的作用和意义.

By use of MODIS/Terra-Aqua daily snow cover products（MOD10A1,MYD10A1） and AMSR/Aqua daily snow water equivalent from October 1,2002 to March 31,2008 in Qinghai province,the ten-days composite algorithms of MODIS and AMSR-E data were studied,and the snow classification images of MOYD_10D,AE_10D and MDAE_10D were composed during the study period.The classification accuracies of the three kinds of images were analyzed using climate station data.The results showed that： 1） when the snow depth is more than 4 cm,snow classification accuracy（Sa） of AE_10D is only 48.44%,and overall accuracy（Oa） is 55.43%;2） Snow classification accuracy of MOYD_10D is seriously influenced by snow depth.When snow depth is between 1 and 3 cm,the classification accuracy is low,with Sa of 20.63% merely.While snow depth is more than 4 cm,Sa is 53.56%,and Oa is 93.54%.When snow depth is more than 11cm,Sa is 88.57%;and 3） For MDAE_10D image,Sa is 20.86% when snow depth is less than 3 cm.Sa increases with increase of snow depth.When the depth is over 4 cm,Sa will reaches 59.28%,and Oa will reaches 93.66%.When snow depth is more than 11 cm,Sa reaches to the maximum value of 91.43%.For MDAE_10D,which has the characteristics of AE_10D and MOYD_10D,the Sa and Oa are improved.Accordingly,MDAE_10D has an evident role in snow disaster monitoring and evaluating in pastoral areas.