Parameter sensitivity and transferability for simulating ET and GPP of Dryland ecosystems across a climate gradient

Ecohydrology models are essential tools for projecting ecosystem shifts and quantifying water and carbon fluxes across ecosystems. Parameter uncertainty limits our ability to simulate ecosystem processes. We evaluated sensitivity of parameters in the widely used Farquhar and Ball-Barry algorithms for simulating evapotranspiration (ET) and gross primary productivity (GPP) for three sagebrush ecosystems across an elevation/climate gradient within the Reynolds Creek Critical Zone Observatory in southwestern Idaho, USA. This gradient spanned annual precipitation rates of 292 to 800 mm and GPP from 420 to 849 gC m−2. We used a Monte-Carlo approach of 10,000 to 20,000 model runs to assess the distribution of optimal parameter values for each site. Distributions of best-case values for sagebrush parameters controlling carbon uptake were similar for the Wyoming big sagebrush and mountain big sagebrush sites but differed for the low sagebrush site. Differences in sagebrush transpiration parameters were consistent with soil water regimes for these different species and subspecies. Parameter values for the herbaceous understory showed less similarity between sites and years, which was attributed to the varying composition of grasses and forbs. Peak seasonal GPP was replicated by the model, with understory vegetation contributing up to 80 % of peak GPP. R2 values for simulated GPP ranged from 0.78 to 0.90 for the optimization period and 0.61 to 0.89 for the validation period. A composite parameter set for each site obtained by combining parameter values whose best-case distributions did not significantly differ, resulted in no meaningful difference in simulations. This study reveals the extent to which parameters can be transferred across ecosystems and years, thereby reducing parameter uncertainty which is a critical step forward in assessing future trajectories of ecosystems and carbon fluxes.