Mapping and Evaluation of NDVI Trends from Synthetic Time Series Obtained by Blending Landsat and MODIS Data around a Coalfield on the Loess Plateau

The increasingly intensive and extensive coal mining activities on the Loess Plateau pose a threat to the fragile local ecosystems. Quantifying the effects of coal mining activities on environmental conditions is of great interest for restoring and managing the local ecosystems and resources. This paper generates dense NDVI (Normalized Difference Vegetation Index) time series between 2000 and 2011 at a spatial resolution of 30 m by blending Landsat and MODIS (Moderate Resolution Imaging Spectroradiometer) data using the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) and further evaluates its capability for mapping vegetation trends around a typical coalfield on the Loss Plateau. Synthetic NDVI images were generated using (1) STARFM-generated NIR (near infrared) and red band reflectance data (scheme 1) and (2) Landsat and MODIS NDVI images directly as inputs for STARFM (scheme 2). By comparing the synthetic NDVI images with the corresponding Landsat NDVI, we found that scheme 2 consistently generated better results (0.70 < R2 < 0.76) than scheme 1 (0.56 < R2 < 0.70) in this study area. Trend analysis was then performed with the synthetic dense NDVI time series and the annual maximum NDVI (NDVImax) time series. The accuracy of these trends was evaluated by comparing to those from the corresponding MODIS time series, and it was concluded that both the trends from synthetic/MODIS NDVI dense time series and synthetic/MODIS NDVImax time series (2000–2011) were highly consistent. Compared to trends from MODIS time series, trends from synthetic time series are better able to capture fine scale vegetation changes. STARFM-generated synthetic NDVI time series could be used to quantify the effects of mining activities on vegetation, but the test areas should be selected with caution, as the trends derived from synthetic and MODIS time series may be significantly different in some areas.

Abstract licence: CC BY

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.