The GLASS Fractional Vegetation Cover (FVC) product is based on machine learning methods to train a relationship model from preprocessed reflectance to FVC value, which is used to produce global land surface vegetation cover products. The GLASS-FVC product has a spatial range of the global land surface, with a time resolution of 8 days, and is monitored a total of 46 times throughout the year. Among them, the vegetation coverage remote sensing dataset products produced based on AVHRR data have a time range of 1981-2020, using latitude and longitude projection with a spatial resolution of 5 km × 5 km; the vegetation coverage remote sensing dataset products produced based on MODIS data have a time range of 2000-2021, using SIN projection with a spatial resolution of 0.5 km × 0.5 km. The GLASS-FVC product output format is HDF-EOS standard format, including a vegetation coverage dataset.
This dataset collected data from three bands, H25v04, H25v05, and H26v05, in Gansu Province from 2000 to 2021.
University of Maryland GLASS Products http://www.glass.umd.edu/Download.html
The GLASS FVC product algorithm is based on machine learning methods and uses training samples generated from globally distributed high-resolution satellite data. Initially, the GLASS FVC product algorithm used for MODIS data was generated using the Generalized Regression Neural Network (GRNN) method, training sample data from the Poetic Mapper (TM) and Enhanced Poetic Mapper plus (ETM+) datasets. However, in the process of generating long-term global GLASS FVC products, it was found that the computational efficiency of the GRNNs method was not satisfactory. Therefore, four machine learning methods were evaluated, including backpropagation neural network (BPNN), GRNN, support vector regression (SVR), and multivariate adaptive regression spline (MARS).
We have also developed the GLASS FVC algorithm for AVHRR data to be used in conjunction with the GLASS MODIS FVC product. It is based on the GLASS MODIS FVC product and can achieve continuity in FVC estimation from AVHRR and MODIS data.
Extensive validation experiments were conducted using high-resolution satellite data and estimated values from ground measurements.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.GSEER.DB4144.2023
10.12072/ncdc.gseer.db4144.2023
