Aiming at the problem that the existing snow cover products in the "the Belt and Road" area are underestimated in mountain areas and woodlands, based on multi-source remote sensing data, a set of snow cover data in the "the Belt and Road" area was generated by using the MARS model combined with the method of land type characteristics to automatically identify snow cover. By leveraging the advantages of machine learning in solving nonlinear fitting problems and avoiding the misjudgment of traditional snow remote sensing recognition in complex terrain and terrain, the accuracy of this product in mountainous and forested areas has significantly improved compared to existing MODIS snow products.
MOD09GA is sourced from the website of the United States Geological Survey.
Extract detailed land cover information using FSC images from Landsat-8 as ground truth data, combined with the CGLS-LC100 dataset. A MARS model based on MOD09GA reflectance was constructed by combining land features to achieve efficient snow recognition. The entire process includes data preprocessing, feature optimization, model training, and validation, ultimately forming an efficient automatic recognition algorithm. This algorithm demonstrates speed and high accuracy in snow area recognition, effectively addressing the challenges faced by existing products and providing important technical support for regional environmental monitoring and climate change research.
| # | number | name | type |
| 1 | 2022YFF0711702-05 | National key R & D plan |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.NIEER.DB6622.2024
10.12072/ncdc.nieer.db6622.2024
