Dynamic lag effects of climate and vegetation on biosphere-atmosphere CO2 exchange: a global analysis
Abstract
Climate variability can have both instantaneous and lagged effects on the functioning of terrestrial ecosystems; thereby altering the carbon (C) budgets regionally and globally at different time scales. Yet, the memory effects of past climate & vegetation dynamics on contemporary carbon dioxide (CO2) fluxes are poorly understood. Here we use d a dynamic statistical approach; a Long-Short-Term Memory (LSTM) model able to transmit the memory of past vegetation and climate states (ecological memory, EM) to future predictions of CO2 fluxes. EM effects were quantified by computing a difference metric (Root Mean Square Difference, RMSD) between a reference LSTM model with memory to an LSTM model where memory was destroyed via permutation of the time dimension. The study encompassed CO2 fluxes globally using satellite, climate, and eddy-covariance (EC) data from the period 2001-2018. We f ou nd widespread and pronounced EM effects (i.e. high RMSD between the reference and permutation models) on net ecosystem exchange (NEE) across scales. Strong EM effects were observed from dry and boreal climates, as well as from savannas and both d eciduous and mixed forests, all with RMSD values ranging from 0.3- 0.4 gC m-2 day-1. Globally, limited EM effects were found in water-limited ecosystems , whereas colder and energy-limited regions reveal ed substantial EM effects on CO2 flux responses. At FLUXNET sites, we f ound stronger EM effects on NEE than either on ecosystem respiration (ER) or gross primary production (GPP). By exploring the length of EM effects, we show ed that recent climate and vegetation dynamics (i.e. previous months) dominate the observed EM effects on CO2 fluxes. Finally , we also analyze d the 2018 European heatwave and observe d that capturing EM effects dampen the net CO2 flux responses to this specific extreme event. These findings emphasize the importance of capturing vegetation and climate temporal dynamics to better understand the response of contemporary ecosystem CO2 fluxes.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.B13B..02B
- Keywords:
-
- 0426 Biosphere/atmosphere interactions;
- BIOGEOSCIENCES;
- 0428 Carbon cycling;
- BIOGEOSCIENCES;
- 1631 Land/atmosphere interactions;
- GLOBAL CHANGE;
- 1813 Eco-hydrology;
- HYDROLOGY