Net ecosystem exchange of carbon (NEE) represents the overall “breathing” of the biosphere and is a key indicator of past, present, and future global carbon cycling. However, ecosystem models poorly predict NEE across a wide range of time-scales, including daily. NEE responses can be complex because they are simultaneously modified by extrinsic forcings (e.g., climate, environmental drivers) and intrinsic dynamics (e.g., physiological acclimation, changes in vegetation density, composition, and structure). In particular, two challenges exist: 1) understanding how concurrent and antecedent environmental conditions (exogenous “memory”) simultaneously influence ecosystem behavior, and 2) quantifying how such extrinsic forcings trigger intrinsic changes that feed back to influence ecosystem behavior (endogenous memory). To address these challenges, we implemented a Bayesian model to quantify exogenous and endogenous memory of daily NEE by synthesizing data from 49 FLUXNET sites representing 9 IGBP-biome types across the globe. Site-level daily NEE data spanned 5-21 years, yielding 534 site-years of daily NEE and meteorological records. Using these data, we ask: How important are antecedent conditions (memory) for NEE dynamics? Over what time-scales do antecedent conditions influence NEE? How does NEE memory and sensitivity to exogenous drivers (e.g., moisture availability) vary across ecosystems?
Results/Conclusions
Compared to a model lacking exogenous memory (no antecedent effects), we found that a model with exogenous memory explained an additional 3-30% of NEE variation across the sites. An autoregressive model of the NEE residuals, representing the intrinsic evolution of the ecosystem (endogenous memory), explained an additional 4-35%. After accounting for exogenous and endogenous memory, model fits (R2) ranged from 0.48 (a temperate Eucalyptus forest) to 0.95 (deciduous forest sites). The time-scales of influence associated with soil moisture (over past two weeks) and precipitation (over past year) varied greatly across sites, while those of vapor-pressure deficit (VPD; ~1 day lag) and short-wave radiation (no lag) were more consistent across site. Sites share similar patterns in many of the environmental sensitivities, but differ primarily in their baseline NEE, precipitation, and VPD effects. Our findings indicate: 1) evaluations of NEE responses need to consider ecological memory, and 2) contrary to current thought, sites within a biome do not always exhibit similar NEE responses (in both exogenous memories and sensitivities). Overall, this synthesis reveals new insights into the time-scales of exogenous influences on NEE and variation within and among biomes; we discuss potential mechanisms underlying variations in memory and environmental sensitivities.