Key points are not available for this paper at this time.
High-quality, normalized differential vegetation index (NDVI) time-series data are fundamental for environmental remote sensing applications; however, their quality is often influenced by complicated factors such as atmospheric aerosols and cloud coverage. Hence, in the current study, a robust reconstruction method based on envelope detection and the Savitzky-Golay filter (ED-SG) was developed to reduce noise in the NDVI time-series. To verify the performance of ED-SG, simulation experiments were implemented and NDVI time-series samples were selected for different land cover types derived from MOD09GQ, Sentinel-2 and Landsat 8 OLI of Yangtze River Basin, between December 2018 and December 2019. The experimental results yielded an agreement coefficient and variance of 0.9599 and 0.0006, respectively on simulated time-series, Additionally, the smoothness metrics of evergreen broadleaf forests, evergreen needleleaf forests, deciduous broadleaf forests, herbaceous, and croplands were 0.0019, 0.0017, 0.0012, 0.0012, and 0.0013, respectively. Ultimately, the reconstructed time-series metrics showed significant improvements in robustness and smoothness over conventional methods. Moreover, the simplistic mechanisms of the ED-SG model enabled it to run effectively in the Google Earth Engine over the NDVI time-series of the whole Yangtze River Basin.
Liu et al. (Thu,) studied this question.