Improving Yasso15 soil carbon model estimates with ensemble adjustment Kalman filter state data assimilation

Abstract

Model-calculated forecasts of soil organic carbon (SOC) are important for approximating global terrestrial carbon pools and assessing their change. However, the lack of detailed observations limits the reliability and applicability of these SOC projections. Here, we studied whether state data assimilation (SDA) can be used to continuously update the modeled state with available total carbon measurements in order to improve future SOC estimations. We chose six fallow test sites with measurement time series spanning 30 to 80 years for this initial test. In all cases, SDA improved future projections but to varying degrees. Furthermore, already including the first few measurements impacted the state enough to reduce the error in decades-long projections by at least 1 tCha(-1). Our results show the benefits of implementing SDA methods for forecasting SOC as well as highlight implementation aspects that need consideration and further research.