A hybrid change-detection approach for monitoring tropical lowland rainforest degradation, using remote sensing and machine learning algorithms

Abstract

The lowland rainforest of Akure forest reserve is highly rich in biological diversity (both flora and fauna), the forest has declined due to an increase in human pressure, owing to illegal logging activities, agricultural expansion, and an increase in urbanization. Despite the importance of forests and the forestry sector to Nigerians and its economy, a reliable database that gives accurate information about the forest extent in the country is not available due to a lack of resources such as financial and expertise. However, there is a need for a robust monitoring system that will contribute to the Reduction of Emissions from Deforestation and forest Degradation (REDD+) program to take quick action against forest degradation and management for sustainable use. This study is to monitor forest degradation in the Akure forest reserve using a hybrid change detection approach. Two methods were used. The first one includes multidate comparison methods to analyse the extent of forest change for temporary snapshots (2001, 2006, 2011, 2018, 2021) using random forests algorithm. These indicated that forest cover reduced from 94% to 88.3% in 2006, 71.4% in 2011, 54.1%, and 45.9% in the years 2018 and 2021 respectively while degraded forests increased from 5.8% (2001) to 11.7% (2006) to 22.4 % (2011) to 42.4 (2018) and 50.8% (2021). The second method used was spectral trajectory-based methods to analyse positive and negative changes in the Akure forest reserve using the Landtrendr algorithm. This algorithm uses pixels spectral from Landsat datasets to extract information on how the forest has changed due to disturbance and the process of its recovery over 20 years (2001-2021). In comparison, the first method showed a stepper forest decline than the second method.