|Title||Forecasting responses of a northern peatland carbon cycle to elevated CO2 and a gradient of experimental warming|
|Publication Type||Journal Article|
|Year of Publication||2018|
|Authors||Jiang J, Huang Y, Ma S, Stacy M, Zheng S, Ricciuto DM, Hanson PJ, Luo Y|
|Journal||Journal of Geophysical Research: Biogeosciences|
|Date Published||March 2018|
The ability to forecast ecological carbon cycling is imperative to land management in a world where past carbon fluxes are no longer a clear guide in the Anthropocene. However, carbon‐flux forecasting has not been practiced routinely like numerical weather prediction. This study explored (1) the relative contributions of model forcing data and parameters to uncertainty in forecasting flux‐ versus pool‐based carbon cycle variables and (2) the time points when temperature and CO2 treatments may cause statistically detectable differences in those variables. We developed an online forecasting workflow (Ecological Platform for Assimilation of Data (EcoPAD)), which facilitates iterative data‐model integration. EcoPAD automates data transfer from sensor networks, data assimilation, and ecological forecasting. We used the Spruce and Peatland Responses Under Changing Experiments data collected from 2011 to 2014 to constrain the parameters in the Terrestrial Ecosystem Model, forecast carbon cycle responses to elevated CO2 and a gradient of warming from 2015 to 2024, and specify uncertainties in the model output. Our results showed that data assimilation substantially reduces forecasting uncertainties. Interestingly, we found that the stochasticity of future external forcing contributed more to the uncertainty of forecasting future dynamics of C flux‐related variables than model parameters. However, the parameter uncertainty primarily contributes to the uncertainty in forecasting C pool‐related response variables. Given the uncertainties in forecasting carbon fluxes and pools, our analysis showed that statistically different responses of fast‐turnover pools to various CO2and warming treatments were observed sooner than slow‐turnover pools. Our study has identified the sources of uncertainties in model prediction and thus leads to improve ecological carbon cycling forecasts in the future.