Soil thermal dynamics, snow cover, and frozen depth under five temperature treatments in an ombrotrophic bog: Constrained forecast with data assimilation

Abstract

Accurate simulation of soil thermal dynamics is essential for realistic prediction of soil biogeochemical responses to climate change. To facilitate ecological forecasting at the Spruce and Peatland Responses Under Climatic and Environmental change site, we incorporated a soil temperature module into a Terrestrial ECOsystem (TECO) model by accounting for surface energy budget, snow dynamics, and heat transfer among soil layers and during freeze‐thaw events. We conditioned TECO with detailed soil temperature and snow depth observations through data assimilation before the model was used for forecasting. The constrained model reproduced variations in observed temperature from different soil layers, the magnitude of snow depth, the timing of snowfall and snowmelt, and the range of frozen depth. The conditioned TECO forecasted probabilistic distributions of soil temperature dynamics in six soil layers, snow, and frozen depths under temperature treatments of +0.0, +2.25, +4.5, +6.75, and +9.0°C. Air warming caused stronger elevation in soil temperature during summer than winter due to winter snow and ice. And soil temperature increased more in shallow soil layers in summer in response to air warming. Whole ecosystem warming (peat + air warmings) generally reduced snow and frozen depths. The accuracy of forecasted snow and frozen depths relied on the precision of weather forcing. Uncertainty is smaller for forecasting soil temperature but large for snow and frozen depths. Timely and effective soil thermal forecast, constrained through data assimilation that combines process‐based understanding and detailed observations, provides boundary conditions for better predictions of future biogeochemical cycles.

Realized ecological forecast through an interactive Ecological Platform for Assimilating Data (EcoPAD, v1.0) into models

“…Datasets are pulled from SPRUCE archives and stored in the EcoPAD (v1.0) metadata catalog for running the TECO model and conducting data-model fusion or forecasting. The TECO model has been applied to simulate and forecast carbon dynamics with productions of CO 2 and CH 4 from different carbon pools, soil temperature response, snow depth, and freeze-thaw cycles at the SPRUCE site (Jiang et al, 2018;Huang et al, 2017;Ma et al, 2017).…”

“…The attribution of uncertainties in EcoPAD (v1.0) will rely on an ensemble of ecosystem models, the data assimilation system, and climate forcing with quantified uncertainty. Jiang et al (2018) focused specifically on the relative contribution of parameter uncertainty vs. climate forcing uncertainty in forecasting carbon dynamics at the SPRUCE site. By assimilating pretreat-ment measurements (2011)(2012)(2013)(2014) from the SPRUCE experiment, Jiang et al (2018) estimated uncertainties in key parameters that regulate peatland carbon dynamics.…”

“…Jiang et al (2018) focused specifically on the relative contribution of parameter uncertainty vs. climate forcing uncertainty in forecasting carbon dynamics at the SPRUCE site. By assimilating pretreat-ment measurements (2011)(2012)(2013)(2014) from the SPRUCE experiment, Jiang et al (2018) estimated uncertainties in key parameters that regulate peatland carbon dynamics. Combined with the stochastically generated climate forcing (e.g., precipitation and temperature), Jiang et al (2018) found that external forcing resulted in higher uncertainty than parameters in forecasting carbon fluxes, but caused lower uncertainty than parameters in forecasting carbon pools.…”

“…By assimilating pretreat-ment measurements (2011)(2012)(2013)(2014) from the SPRUCE experiment, Jiang et al (2018) estimated uncertainties in key parameters that regulate peatland carbon dynamics. Combined with the stochastically generated climate forcing (e.g., precipitation and temperature), Jiang et al (2018) found that external forcing resulted in higher uncertainty than parameters in forecasting carbon fluxes, but caused lower uncertainty than parameters in forecasting carbon pools. Therefore, more efforts are required to improve forcing measurements for studies that focus on carbon fluxes (e.g., GPP), while reductions in parameter uncertainties are more important for studies in carbon pool dynamics.…”

Forecasting Responses of a Northern Peatland Carbon Cycle to Elevated CO2 and a Gradient of Experimental Warming

“…Thus, the expectation on how long it would take for experimental effects to be observed may be different from real observations. Our EcoPAD platform can update parameters regularly as new data sets become available and reveal when model simulations depart from experimental observations (Huang et al, ; Ma et al, ).…”