Quantifying Livestock Diet Composition Using Earth Observation (EO) Data for Improved Estimation of Enteric Methane Emissions in Kenya

Abstract

Livestock directly contribute to greenhouse gas emissions, mainly through enteric fermentation and to a lesser extent manure management. Livestock feed composition plays a crucial role in diet quality and the resulting emissions from livestock. However, spatio-temporal variations in diet composition, particularly in tropical environments, remain underexplored. Furthermore, comprehensive livestock diet composition datasets, essential for accurate enteric methane emission estimation, are often lacking in developing countries. Existing datasets from global studies are highly generalised and are based on expert knowledge (Herrero et al., 2013). Other studies undertaken on livestock diet composition have often been done on a local basis (Goopy et al., 2018; Ndung’u et al., 2018; Onyango et al., 2018; Wilkes et al., 2020), with less attention on country-wide studies that can provide detailed assessment and evidence of the impact of livestock diet on greenhouse gas emissions from livestock at national or regional level. For both global and local studies, they often offer incomplete representations of livestock diet composition at best, especially in regions with diverse and poorly documented livestock feeding practices. Where data is available, it has uncertainties resulting from data collection challenges and the assumption of a constant annual distribution of diet composition (MacLeod et al., 2018). This gap in accurate data is particularly problematic in developing countries, where feed composition changes during periods of feed scarcity and plenty (Mutua et al., 2023), and this is seldom captured in existing datasets. In this study, we use freely available Earth Observation (EO) data to generate spatially and temporally explicit livestock diet composition and quality data. Our approach involves three key stages. First, we identify the length of growing period (start and end of wet and dry seasons) at a pixel level using a water balance model that incorporates climate and soil property data to calculate the ratio of actual to potential evapotranspiration (Jones & Thornton, 2009). Second, we determine the feed items available to livestock during these seasons through an extensive literature review and estimate their quantities using EO data including land cover/cover fraction (Buchhorn et al., 2020) and above-ground dry matter productivity from the Copernicus programme (Copernicus, 2024), crop harvest area from FEAST global data repository (ILRI, 2022), length of growing period and crop specific indices and parameters. Third, we calculate the proportion of each feed item in livestock diets at a pixel level to derive diet composition and quality data. Finally, the results are then aggregated by livestock production systems, that are specific to Kenya (Robinson et al., 2018), providing a detailed understanding of livestock diet composition patterns in Kenya. Results indicate that livestock diet composition varied between seasons and livestock production systems. Natural grass was the dominant diet component across all seasons and livestock production systems, with the highest proportion in the diet during the long wet season (41.5%). Concentrates formed the lowest proportion in the diet ranging from 0.9 to 4.7% across seasons and livestock production systems. Diet quality as well varied between seasons and livestock production systems but was in a narrow range (55.4–62.3% dry matter; DM), which was greater than the default digestibility value of 55.0% set by the Intergovernmental Panel on Climate Change (IPCC) for livestock production systems in the region. Notably, mixed rainfed arid systems exhibited lower diet quality across all seasons. Livestock diet quality estimation using EO data demonstrated moderate to high accuracy (R² = 0.50–0.89; RMSE = 1.14–38.44% DM). These findings provide a robust basis for incorporating spatially-and-temporally explicit diet composition and quality data into life cycle assessment (LCA) models to improve GHG emission estimates from the livestock sector. Enhanced accuracy of these models can inform national-level strategies for mitigating emissions and promoting sustainable livestock production.