Estimating Carbon Stock using field data, Satellite Imagery, and Cloud-Based Machine Learning Algorithms: case study of Mubende District

Abstract

Quantifying carbon stock is a good step in pursuit to mitigate GreenHouse Gases (GHGs). The purpose of this research was to estimate carbon stock changes in Mubende District as a consequence of Land Cover Change (LCC) through the integration of traditional approaches, Machine Learning (ML) algorithms and Remote Sensing (RS). Random Forest (RF) ML Algorithms for Land Cover classification using Landsat 8 and 9 images were used and LCC was assessed. To determine biomass and soil carbon, tree measurements, grass, and soil samples were collected from Mubende District and later taken for laboratory analysis. The carbon values obtained were used in carbon modelling. Upon carbon stock modelling using the InVEST model, total carbon stock values for the years 2013, 2016, 2018, 2020, and 2022 were as follows: 6,953,193.02 Mg C/ha, 7,316,239.83 Mg C/ha, 7,429,244.21 Mg C/ha, 7,849,321.58 Mg C/ha, and 6,899,078.59 Mg C/ha respectively. Results show that carbon stock change values between 2013 and 2016, between 2016 and 2018, between 2018 and 2020, and then between 2020 and 2022 were observed to be: 363,046.80 Mg C/ha, 113,004.37 Mg C/ha, 420,863.59 Mg C/ha, and -950,242.96 Mg C/ha respectively. The negative value indicates that there was carbon loss in form of carbon dioxide emissions into the atmosphere while a positive value indicates gain in carbon stock. Ultimately, forests, trees, vegetation, and agricultural land acted more as a sink between the years 2013 and 2020 than a source of GHGs. Subsequently, the aforementioned land cover classes were converted into sources of CO2 between 2020 and 2022. It is therefore recommended that methods used in this study are applied in estimation of carbon stock for other districts in the whole of Uganda for policymaking purposes. Additionally, further research should be performed to ascertain how carbon stock changes and land cover changes affect climate.