Vegetation phenology is a comprehensive indictor for the responses of terrestrial ecosystem to climatic and environmental changes. Remote sensing spectrum has been widely used in the extraction of vegetation phenology information. However, there are many differences between phenology extracted by remote sensing and site observations, with their physical meaning remaining unclear. We selected one tile of MODIS data in northeastern China (2000-2014) to examine the SOS and EOS differences derived from the normalized difference vegetation index (NDVI) and the simple ratio vegetation index (SR) based on both the red and near-infrared bands. The results showed that there were significant differences between NDVI-phenology and SR-phenology. SOS derived from NDVI averaged 18.9 days earlier than that from SR. EOS derived from NDVI averaged 19.0 days later than from SR. NDVI-phenology had a longer growing season. There were significant differences in the inter-annual variation of phenology from NDVI and SR. More than 20% of the pixel SOS and EOS derived from NDVI and SR showed the opposite temporal trend. These results caused by the seasonal curve characteristics and noise resistance differences of NDVI and SR. The observed data source of NDVI and SR were completely consistent, only the mathematical expressions were different, but phenology results were significantly different. Our results indicated that vegetation phenology monitoring by remote sensing is highly dependent on the mathematical expression of vegetation index. How to establish a reliable method for extracting vegetation phenology by remote sensing needs further research.
植被物候是陆地生态系统响应气候和环境变化的一项综合性指标.遥感光谱已经被广泛用于提取植被物候期,但遥感提取的物候期与站点观测差别很大,其物理意义尚不明确.本文选取中国东北部的一景MODIS数据(2000—2014年),分析了基于红波段和近红外波段的归一化差值植被指数(NDVI)和简单比植被指数(SR)提取的植被生长季起始期(SOS)和结束期(EOS)的差异.结果表明: 两者的物候期存在显著差别,基于NDVI提取的SOS比SR提取的SOS平均早18.9 d,基于NDVI提取的EOS比SR提取的EOS平均晚19.0 d,NDVI得到的生长季长度更长.基于NDVI和SR提取的物候期的年际变化也存在显著差别,超过20%的像元SOS和EOS甚至表现出相反的年际变化趋势.上述差异与两种植被指数自身的季节曲线特征和抗噪性差异有关.NDVI与SR观测数据来源完全一致,仅数学表达形式不同,提取的物候期结果却存在显著差异.说明遥感监测的植被物候期高度依赖于植被指数的数学表达形式,如何建立可靠的植被物候期遥感提取方法仍需进一步研究.
Keywords: mathematical expression; remote sensing; temporal trend; vegetation index; vegetation phenology.