Kang Sinkyu, Hong Suk Young
Department of Environmental Science, Kangwon National University, Chuncheon 200-701, Republic of Korea.
Department of Agricultural Environment, National Academy of Agricultural Science, Wanju 565-851, Republic of Korea.
PLoS One. 2016 Mar 23;11(3):e0151395. doi: 10.1371/journal.pone.0151395. eCollection 2016.
A minimum composite method was applied to produce a 15-day interval normalized difference vegetation index (NDVI) dataset from Moderate Resolution Imaging Spectroradiometer (MODIS) daily 250 m reflectance in the red and near-infrared bands. This dataset was applied to determine lake surface areas in Mongolia. A total of 73 lakes greater than 6.25 km2in area were selected, and 28 of these lakes were used to evaluate detection errors. The minimum composite NDVI showed a better detection performance on lake water pixels than did the official MODIS 16-day 250 m NDVI based on a maximum composite method. The overall lake area detection performance based on the 15-day minimum composite NDVI showed -2.5% error relative to the Landsat-derived lake area for the 28 evaluated lakes. The errors increased with increases in the perimeter-to-area ratio but decreased with lake size over 10 km(2). The lake area decreased by -9.3% at an annual rate of -53.7 km(2) yr(-1) during 2000 to 2011 for the 73 lakes. However, considerable spatial variations, such as slight-to-moderate lake area reductions in semi-arid regions and rapid lake area reductions in arid regions, were also detected. This study demonstrated applicability of MODIS 250 m reflectance data for biweekly monitoring of lake area change and diagnosed considerable lake area reduction and its spatial variability in arid and semi-arid regions of Mongolia. Future studies are required for explaining reasons of lake area changes and their spatial variability.
采用一种最小合成方法，根据中分辨率成像光谱仪（MODIS）每日250米分辨率的红波段和近红外波段反射率数据，生成了一个15天间隔的归一化植被指数（NDVI）数据集。该数据集被用于确定蒙古的湖泊面积。总共选取了73个面积大于6.25平方千米的湖泊，其中28个湖泊用于评估检测误差。与基于最大合成方法的官方MODIS 16天250米分辨率NDVI相比，最小合成NDVI在湖泊水体像素的检测性能上表现更佳。基于15天最小合成NDVI的湖泊面积总体检测性能显示，相对于28个被评估湖泊通过陆地卫星数据得出的湖泊面积，误差为-2.5%。误差随着周长面积比的增加而增大，但对于面积超过10平方千米的湖泊，误差随着湖泊面积的增大而减小。2000年至2011年期间，这73个湖泊的面积以每年-53.7平方千米（-1）的速率减少了-9.3%。然而，也检测到了相当大的空间差异，比如半干旱地区湖泊面积有轻微到中等程度的减少，干旱地区湖泊面积则迅速减少。本研究证明了MODIS 250米分辨率反射率数据在双周监测湖泊面积变化方面的适用性，并诊断出蒙古干旱和半干旱地区湖泊面积有相当大的减少及其空间变异性。未来的研究需要解释湖泊面积变化及其空间变异性的原因。
