time period with no vegetation activity was then defined as when LST is less than 278 °K (ca. 5 °C) for each pixel. The maximum NDVI value in this period was then selected from each valid year from 1982 to 2013. The mean of these maximum values was calculated as the background. Finally, values less than the background were identified as snow contaminated and were replaced by the background NDVI. We further applied a three-point median-value filter to the NDVI time series to remove these sudden drops.
The research area is limited to the north hemisphere (north to 30 °N). Besides, pixels identified as the bare area or evergreen forests according to the MODIS land cover type product in 2010 were excluded. Considering the potential error in the MODIS products, we also excluded pixels with multiyear mean annual NDVI less than 0.2 or mean winter (December to February) NDVI larger than 0.6. In addition, we further excluded pixels with annual maximum NDVI occurs in non-summer seasons and pixels with bi-peak NDVI time series in more than thirty percent of the years.
Daily NDVI was firstly generated to temporal composited data via MVC with a bimonthly time step, from the 1st day of the month to the 15th day and from 16th day to the end. This newly generated NDVI data was then used in the following analysis. Snow contamination is considered to severely influence the LSP detection. Here, we remove the influence from snow using a method similar to (Zhang et al 2015). Firstly, LST climatology was calculated with MOD11C2 data of 2000 to 2015. Secondly, the time period with no vegetation activity was then defined as when LST is less than 278 °K (ca. 5 °C) for each pixel. The maximum NDVI value in this period was then selected from each valid year from 1982 to 2013. The mean of these maximum values was calculated as the background. Finally, values less than the background were identified as snow contaminated and were replaced by the background NDVI. We further applied a three-point median-value filter to the NDVI time series to remove these sudden drops.
NDVI le).