In this study, a spatiotemporal adaptive fusion method (STAR) that comprehensively considers spatiotemporal background information was used to generate The cloudless Terra Aqua MODIS NDSI dataset in China for the past 20 years. STAR NDSI data generally has the following advantages (1) This dataset has been running for 20 consecutive years, which is the shortest period for long-term hydrological and climate datasets. (2) Cloudless datasets can accurately estimate Calculate the dynamic snow cover, which is consistent with the depth of snow cover in place and the height of the NDSI map of the Earth's remote sensing satellite. Specifically, STAR NDSI The collection eliminates cloud pollution and greatly improves the overall performance of the TAC NDSI dataset. Therefore, this dataset can be used as The basic dataset for hydrological and climate modeling, used to explore various key environmental issues.
The MOD10A1 and MYD10A1 datasets can be accessed through the NASA website (NASA: ttps://search.earthdata.nasa.gov/ ）Obtain
We propose an advanced STAR method that comprehensively utilizes spatiotemporal background information and can completely remove clouds. This method is divided into two steps: spatiotemporal adaptive fusion (STAF) and error correction (EC). The first process includes generating a new NDSI map, including spatial partitioning, adaptive spatiotemporal block determination, and fusion based on Gaussian kernel function (GKF). Considering the spatial heterogeneity of snow cover patterns, the study area was first divided into ten zones. In this way, subsequent processing can be carried out on the basis of partitioning. In addition, the optimal query partition (Q) for each target partition (T) is determined by comprehensively considering the time complexity of snow changes The time distance (t), regional correlation (r), and cloudless fraction (f) of the surface are determined
For regions with extremely fast and fluctuating snow cover, time background reference is likely to introduce incorrect information and amplify errors during the iteration process. This study adopts post-processing methods to reduce the phenomenon of "disorder" in quality assurance maps. Firstly, manually determine the NDSI map with the most consistent snow cover pattern in adjacent times as a reference. Subsequently, the above EC technology is applied to improve spatial consistency between the post-processing region and the original region. Finally, update the quality assurance map.
Optical remote sensing images are severely polluted by clouds, so the MODIS NDSI dataset cannot accurately reflect daily snow cover and melting. Therefore, we propose a two-stage spatiotemporal fusion method called STAR to generate spatiotemporal continuous snow collection. The generation process includes preprocessing TAC and key processing STAR. Provide a quality assessment (QA) method to provide users with data reliability files. On this basis, post processing is utilized to further improve the data quality of individual abnormal areas.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB4016.2023
10.5281/zenodo.5644386
