Quantification of Greenhouse Gas Emissions from Livestock Using Remote Sensing & Artificial Intelligence

Abstract

Greenhouse Gases (GHGs) from agriculture in Africa are among the fastest-growing emissions in the world with the livestock sector as a major contributor to these GHGs, and is expected to have high emission growth rates. The methods used to quantify livestock GHG emissions require data that is manually collected or outdated because of the low frequency at which it is collected. This research aimed to assess the feasibility of remote sensing and deep learning to quantify grazing cattle GHG emissions in Kisombwa Ranching Scheme in Mubende District. Unmanned Aerial Vehicle (UAV) images were captured and the You Only Look Once (YOLO) v4 and Simple Online Realtime Tracker (SORT) algorithms were applied to create a model to automatically detect and count the number of cattle in the UAV aerial images. The obtained number of cattle was used as an input in the quantification of GHGs from the cattle. Methane (CH4) and Nitrous oxide (N2O) emissions from manure management and enteric fermentation were quantified using Tier 1 guidelines from Intergovernmental Panel on Climate Change (IPCC). The quantified CH4 and N2O emissions were converted into CO2 eq to get the total GHG emissions. The cattle counting approach achieved a high accuracy with an average F1 score of 88.9%, average precision of 97% and average recall of 82.9% on the testing set of images. The total cattle CH4 and N2O GHG emissions were quantified to be 321,121.34 kg CO2 eq yr-1. CH4 and N2O emissions accounted for 282,282.96 kg CO2 eq yr-1 and 38,838.38 kg CO2 eq yr-1 respectively. CH4 was the highest emitted GHG with a percentage of 88% of the total GHG emissions and 12% as N2O. Enteric fermentation contributed the highest CH4 emissions of about 99% of the total CH4 emissions and 87% of the total GHGs. These findings demonstrated that remote sensing and artificial intelligence can be applied to improve the quantification process of livestock GHGs. Therefore, the study recommends the application of this approach on high-resolution satellite images to upscale the reporting of the animal population and livestock GHG emissions in different areas.