Climate change is one of the greatest threats recently, of which developing countries are facing most of the brunt. In the fight against climate change, forests can play an important role, since they hold a substantial amount of terrestrial carbon and can therefore affect the global carbon cycle. Deforestation, however, is a significant challenge. There are financial incentives that can help in halting deforestation by compensating developing countries for their efforts. They require however assessments which makes it essential for developing countries to regularly monitor their stocking. Based on the aforementioned, forest carbon stock assessment was conducted in Margalla Hills National Park i.e., Sub-tropical Chir Pine Forest (SCPF) and Sub-tropical Broadleaved Evergreen Forest (SBEF), in Pakistan combining field inventory with a remote-sensing-based approach using machine learning algorithms. Circular plots of a 20 m radius were used for recording the data and Sentinel-2 (S2) and Sentinel-1 (S1) satellite data were used for estimating the Aboveground Biomass (AGB). The performances of Random Forests (RF) and Support Vector Machine (SVM) were explored. The AGB was higher for the SCPF. The RF performed better for SCPF, but SVM was better for SBEF. The free available satellite data in the form of S2 and S1 data offers an advantage for AGB estimations. The combination of S2 and S1 for future AGB studies in Pakistan is also recommended.
Keywords: AGB; Forests; Machine Learning; Remote Sensing.
© 2024. The Author(s), under exclusive licence to Springer Nature Switzerland AG.