Based on the developed random forest method, the first accurate high-resolution time-series tree cover product was generated for Southeast Asia (SEA), which integrates Planet Norway's International Climate and Forest Initiative (NICFI) images and Sentinel-1 synthetic aperture radar data on the Google Earth Engine platform. The Planet NICFI tree cover map was surveyed with good accuracy and high consistency over six years (2016-2021). The baseline tree cover map of 4.77 meters can be converted into forest cover maps of various resolutions in SEA to meet the needs of different users. In addition, this tree cover product can help address rounding errors in forest cover mapping by calculating the number of isolated trees and monitoring the removal of narrow forest cover.
Planet-NICFI
Sentinel-1
Integrate Planet NICFI and Sentinel-1 SAR images to generate high-resolution (4.77 m) annual tree cover map products for SEA, covering the period from 2015 to 2021.
The data processing process is as follows: (1) Mapping the definition of tree coverage; (2) Image preprocessing; (3) Generate time series of tree coverage products; (4) Statistical accuracy assessment.
The tree coverage map is evaluated in terms of user accuracy, producer's accuracy, and overall accuracy. The accuracy of the user and the overall accuracy of the tree coverage map exceed 0.083. The performance of ESA WorldCover in 2020 and 2021 is similar to that of the Planet NICFI tree cover map. In particular, the accuracy of users, manufacturers, and the overall accuracy of ESA WorldCover 2020 decreased by 0.020, 0.008, and 0.017, respectively (Figure 4). This may be because SAR images are used as input and RF based machine learning methods are applied to classify the tree cover.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
https://cstr.cn/31253.11.sciencedb.07173
10.57760/sciencedb.07173
