Predicting plant responses to novel drought regimes is challenging for models that rely on empirical coefficients due to great uncertainty when such models are extrapolated out of the range for which they are calibrated. A recently developed model, which predicts plant responses to environmental variables using an optimization of carbon-gain versus hydraulic-risk trade-off, may help improve predictions, as it relies on plant traits and biophysical processes. The theoretical basis of this model is that plants will open or close their stomata as to maximize instantaneous carbon gain while avoiding excessive damage to xylem due to cavitation. To test this model we applied four drought treatments (control, drought recovery, moderate drought, and severe drought) to aspen (Populus tremuloides Michx.) saplings in a research garden. We measured all traits and variables required to run the model as well as the main model output variables. First, we tested the optimization algorithm by using predawn xylem pressure as an input, and compared seven measured and modeled outputs. Second, we tested the full model which calculates root-zone water budget and xylem pressure hourly throughout the growing season. We also used the model to explore the relationship between hydraulic conductance and plant mortality.
Results/Conclusions
We succeeded at imposing different levels of drought stress for each treatment, leading to the death of two moderate-drought and three severe-drought saplings. We were able to measure all inputs except for rhizosphere resistance, which was estimated to be 50-55%. The optimization algorithm performed well when run from measured predawn xylem pressures. We ran the model both with and without xylem refilling, and obtained very similar results because the measurements exhibited an intermediate response between both modeled outputs. The percent mean absolute error (MAE) without refilling averaged 27.7% for midday xylem pressure, transpiration, net assimilation, leaf temperature, sap-flow, diffusive conductance, and soil-canopy hydraulic conductance, whereas with refilling MAE averaged 28.1%. When the full model was run without refilling from irrigation and rain the average MAE was 31.2% for the same observations. A logistic probabilistic regression indicated that all plants that were projected to reach or exceed 85% loss in soil-to-canopy hydraulic conductance died, whereas surviving plants never reached this threshold. Summarizing, the model provided robust predictions of gas exchange, physiological performance, and hydraulic mortality thresholds without recourse to empirical coefficients. Thus, the model provides a useful tool for predicting responses to novel climatic stresses.