A key question in the management of pastoral systems in semi-arid grasslands is how grazing and other management affects soil carbon. Soil carbon may be a key mediator of soil fertility and the capture of available rainfall, but the influence of management on soil carbon is not well understood. In this study, the authors conducted a ground survey of vegetation and soils at 86 sites across 8 different conservancies within the Northern Rangelands Trust in Samburu and Isiolo Districts in northern Kenya. This phase of sampling was designed to accomplish two objectives: 1) to establish a baseline for comparison of soil carbon over time, and 2) to test a predictive model that estimates the accrual rate of carbon based on a few soil and vegetation characteristics. To accomplish these objectives, sites were non-randomly selected to encompass a wide range of soil types across as many conservancies as possible, and to compare areas of different livestock grazing management: no to light livestock grazing (Core Areas), moderate livestock grazing (Buffer Areas), and heavy continuous livestock grazing (Village Areas). While all areas have the potential for wildlife grazing, this grazing pressure is not managed and therefore is not factored into our analysis at this stage.

At each site, soil and vegetation were sampled together. Soil was sampled to 20 cm depth, and analyzed for total organic carbon, texture (percent sand, silt, and clay), and bulk density. Vegetation was sampled by clipping live, aboveground plant biomass, and analyzed for lignin and cellulose content. In addition, current and past grazing intensity were estimated for each site. These soil and vegetation parameters (i.e., soil texture, lignin and cellulose content, historic average grazing intensity) along with interpolated average annual rainfall, were entered into a predictive soil carbon dynamic model called SNAP. The SNAP model was used to predict current soil organic carbon (SOC) stocks based on the estimated history of grazing. The mean predicted SOC stocks were then compared with mean observed SOC stocks for each type of management, and predicted SOC stock at a particular site was compared against observed SOC at the same site.

The results suggested that within the NRT Conservancies, the model predicted mean and individual site SOC values with more than 90% accuracy. The SNAP model results suggest that prolonged, heavy, continuous grazing in the NRT Conservancies over the past 30 years has greatly depleted SOC stocks, but that reduction in grazing intensity will lead to recovery of SOC at a potential rate of 0.3-0.5 tons C/ha/yr across a variety of soil types. Because of past degradation, there is a large capacity for recovering SOC stocks in the Conservancies. These results suggest that planned grazing management beginning in the NRT Conservancies should help restore SOC and productivity in these semi-arid grasslands, and could result in an economically viable carbon offset project. Further sampling planned in the coming months will help to validate the accuracy of these assessments across all Conservancies potentially participating in the grazing management program, and more precisely assess the progress of specific grazing management actions within a few selected conservancies.