https://doi.org/10.1080/22797254.2024.2438638

The importance of trees in non-forest landscapes has been the focus of only a few studies.
However, these trees provide many important ecosystem services. In this study, we mapped
and quantified these trees using Sentinel-2 (S2) and very high-resolution (VHR) Google satellite
imagery without any field campaigns. We performed a Random Forest (RF) classification to map
the spatial distribution of native trees in different scenarios. The optimal model showed an
overall accuracy and kappa of 0.99 and 0.98, respectively. We mapped 40,500 km2 of tree cover,
including native tree cover (approximately 29,565 km2 ≈10.5%), excluding plantations, regional
and provincial parks, and water bodies in the Canadian prairie region of Saskatchewan.
According to our results, the highest numbers of native trees were found in the eastern and
northwestern parts of the study area – cluster “BLK_1” and the “Black” soil zone, with total cover
of 5,388 and 13,233 km2, respectively. The lowest numbers of native trees were found in the
southwest side of the study area – cluster “BRN_6” and the “Brown” soil zone, with total cover
of 2.38 and 979.5 km2, respectively. This research is important as detecting and quantifying
native trees is an integral part of studies on carbon sequestration, economics, and effective
management strategies.

https://doi.org/10.1080/22797254.2024.2438638