Carbon sequestration in smallholder agricultural landscapes in the Tropics has the potential to be a win-win scenario. Trees in agricultural landscapes provide a global environmental benefit, global warming mitigation, along with local environmental benefits such as, soil erosion protection, while simultaneously enhancing human well-being through livelihood diversification and provisioning services. Accurately estimating carbon stocks and flows in these dynamic and heterogeneous landscapes remains problematic. Remote sensing methods are relatively accurate for carbon stocks in forests at different scales (i.e. regional, continental, global). Woody vegetation in agricultural landscapes often does not meet the definitions of forest applied in these estimates. These trees are frequently managed distinctly and more intensively than trees in forests, with farmers selectively maintaining or harvesting whole trees or components of them as needed as part of their management of multifunctional agricultural landscapes. We evaluate two methods of estimating carbon stocks based on local field data and high-resolution remote sensing data, 1) direct estimation, and 2) stratification by land cover classes and application of average values. We compare these estimates with available carbon maps and land cover classifications, and global carbon storage values to understand sources of variability and uncertainty. We use a densely populated agroforestry landscape in western Kenya with ten years of high-resolution satellite and field data as a case study.
There is wide variability in existing carbon stock maps and very few products that capture the dynamic nature of trees in agricultural landscapes. Unlike in most forests, we do not find a strong relationship between canopy cover and aboveground biomass. Many estimations that use remote sensing data rely on this relationship. Detailed and accurate biomass reference data is expensive to collect; however, given the potential benefits of increasing trees on farms, it is necessary to have estimates that are sufficiently accurate at the scale on which many management decisions are made.