Evaluating a new hydraulic implementation in LPJ-GUESS for three sites in north Europe

Abstract: Drought is projected to increase in frequency and intensity and impacts trees with increased water stress and increased mortality rate. Water stresses can cause hydraulic failure-related mortality or carbon starvation due to tree species having different strategies to deal with water stresses. LPJ-GUESS-HYD (Hydraulic implementation of a new plant hydraulics scheme in LPJ-GUESS) was developed to include strategies plants are taking to deal with drought. The new model is an enhanced version of LPJ-GUESS, introducing parametrizations of hydraulic mechanisms in plants. LPJ-GUESS (Lund-Potsdam-Jena general ecosystem simulator) is a state-of-the-art ecosystem model, which combines a Dynamic Global Vegetation Model (DGVM) with a more detailed representation of vegetation dynamics, to simulate vegetation at a regional scale. The aim of this study was therefore to evaluate a new version of the dynamic global vegetation model LPJ-GUESS-HYD with an upgraded hydraulic implementation by testing the model results’ accuracy to observed data and the behaviour of mostly hydraulic parameters, never tested in Sweden. A parameter calibration was done to improve the hydraulic model representing GPP (gross primary production) and ET (evapotranspiration). After two calibrations the model was improved with a final mean RRMSE of around 40% aggregated using all sites and model outputs (ET and GPP). The hydraulic model underestimated ET while the standard version represented ET fluxes better, with a lower site mean RRMSE compared to observations. The hydraulic model predicted GPP better than ET, even if the model tends to overestimate the carbon fluxes for HTM and NOR. A One At a Time (OAT) sensitivity test of LPJ-GUESS-HYD was done to evaluate which parameters cause the highest variability in model outputs representing GPP and ET fluxes. The yearly mean changes during the simulation period (2010-2019) of carbon and water fluxes showed high sensitivity to isohydricity (λ), optimal forcing pressure to maintain canopy conductance (ΔΨmax), maximum sapwood conductivity (Ksmax), water potential representing 50% loss of conductance (Ψ50) and the ratio of intercellular to ambient CO2 (λmax). Cavitation slope (d) (how fast air bubbles form in the tree xylem and prevent water from being pulled upward) and the proportion of resistance located below ground and above ground (b) showed negligible sensitivity for all sites. Recommendations for further studies are varied and range from testing parameter interactions to including additional parameters in an improved sensitivity test and calibration. It is crucial to further evaluate the hydraulic model to improve the water fluxes predictions and to go beyond the scope of the current observations to strengthen climate change prediction, including how ecosystems react to increase drought conditions in future scenarios.