Assimilating disturbance: Toward real-time carbon monitoring and forecasting

Abstract:

Current approaches to bottom-up carbon monitoring rely heavily on the detection of land-use, land-use change, and forestry (LULUCF) through remote sensing, but often account for the carbon impacts of disturbance using simple look-up tables. By contrast, process models are frequently used to analyze and predict disturbance and carbon dynamics in greater detail, but once observations are available we need to update the model’s predictions, especially for stochastic processes such as disturbance. State data assimilation (SDA) is designed specifically to fuse model predictions and observations, nudging modeled states back toward reality in proportion to the uncertainties in the model and the data. However, current SDAs are designed to update continuous states, rather than discrete changes like disturbance, which causes the “nudge” to be spread over many years, carbon to be accounted incorrectly, and productivity to decrease at undisturbed sites because of the way SDA borrows strength across sites. Here we develop a new Bayesian SDA algorithm that combines a discrete Multinomial state-and-transition framework with conventional ensemble filtering SDA approaches. In the Forecast step of the Forecast-Analysis cycle, different ensemble members are assigned to simulate either an undisturbed status-quo or different alternative disturbances. In the Analysis step new observations update both the probability that disturbance occurred, and the carbon pools conditional on disturbance. We demonstrate the efficacy of this algorithm using simulated data with prescribed disturbances. Next, we apply the system to a set of Ameriflux tower sites with known disturbances, assimilating only remotely sensed data and using tower data to validate our ability to improve our estimates of carbon pools and fluxes. This approach has the potential to improve carbon-cycle monitoring, reporting, and verification, real-time disturbance detection, and carbon and LULUCF forecasting capabilities.