About the project
Over the past 20 years, landscape fire activity has decreased in Africa, particularly in the northern hemisphere. Questions remain with regards to potential anthropogenic and environmental drivers of this change and its magnitude. This project aims to utilise recent Earth Observation datasets and machine learning techniques to address this challenge.
Global landscape fires burn on average 4.6 million km2 annually, around 4% of Earth’s vegetated surface. Africa has the greatest burned area where 13% of its non-desert land surface is fire affected annually. Analysis of satellite burned area datasets reveals that the area affected by fire has decreased over the past 20 years.
Suggested drivers of this change include land cover change, principally agricultural expansion and changes in climate. However, there are uncertainties with regards to the extent to which the former drives landscape fire activity due to limitations of agricultural land cover datasets. A further uncertainty stems from the use of moderate spatial resolution (500 m) burned area data which underestimate burned area by up to 80% in Africa. This poses a question regarding the extent of the burned area decline and our capacity to detect it since moderate resolution datasets omit small fires and these are the most prevalent, particularly in agricultural areas.
This study proposes investigating the extent to which agricultural expansion and land cover change have influenced the reduction in African burned area using recent, high spatial resolution (30 m) land cover change datasets that describe changes in agricultural and woody cover (Potapov et al., 2022). Along with climate data, these will be combined with machine learning methods to identify the key drivers. This will be supported by the development of a high spatial resolution burned area dataset to map fire activity in agricultural landscapes and help determine the trends of the area burned in Africa.
