Quantifying the sensitivity of net ecosystem exchange of CO2 (NEE) and of evapotranspiration (ET) to environmental drivers, their interactions, and their timescales of influence has implications for improving carbon and water cycle models. We developed a stochastic antecedent model (SAM) in a Bayesian framework to evaluate the sensitivity of NEE and ET to environmental conditions (e.g., vapor pressure deficit [VPD], temperature [Tair], precipitation, PAR, and soil moisture) and their interactions, using daily flux data that span the growing seasons of 2008 to 2014 from two semiarid AmeriFlux sites from the New Mexico Elevation Gradient (NMEG): a piñon-juniper woodland (US-Mpj) and a ponderosa pine forest (US-Vcp). The direct (main) effects of current conditions explained 66-71% (ET) and 58-60% (NEE) of the variability in these fluxes at the US-Vcp and US-Mpj sites, respectively. Incorporating antecedent conditions and interactions among environmental drivers improved model fit by 6% to 13%. Although model fit did not improve substantially, these details provide insights into the importance of the timing of each covariate for driving NEE and ET and provide a means to evaluate combinations of factors that control changing sensitivity. Although atmospheric covariates (VPD, Tair) 1-2 days prior to the flux measurement were most important for driving fluxes, deep soil moisture 5-6 days before the measurement was significant for driving ET at US-Vcp and for driving NEE at both sites. At both sites, ET and NEE were most sensitive to precipitation received 2 to 4 months prior to the flux measurement, so winter precipitation was most important in driving early growing season fluxes, and growing season precipitation controlled late-growing season fluxes. At both sites, the sensitivity of NEE to VPD decreased as deep soil moisture increased, but this effect was less pronounced at high Tair. ET sensitivity to VPD was higher at the US-Vcp site and increased as deep and shallow soil moisture increased. We hypothesize that variations in the sensitivity of NEE and ET to combinations of environmental drivers reflect the dominance of different processes (e.g., E versus T) at different times of the growing season.
Samuels-Crow, K.E., Ogle, K., Litvak, M.E. (2018): Drivers, Interactions, and Timescales for Carbon and Water Fluxes in semiarid ecosystems. Abstract B13C-06 presented at 2018 AGU Fall Meeting, Washington, D.C., 10-14 Dec.