Parsimony vs predictive and functional performance of three stomatal optimization principles in a big‐leaf framework

Abstract

 Stomatal optimization models can improve estimates of water and carbon fluxes with relatively low complexity, yet there is no consensus on which formulations are most appropriate for ecosystemscale applications. We implemented three existing analytical equations for stomatal conductance, based on different water penalty functions, in a big-leaf comparison framework, and determined which optimization principles were most consistent with flux tower observations from different biomes.  We used information theory to dissect controls of soil water supply and atmospheric demand on evapotranspiration in wet to dry conditions and to quantify missing or inadequate information in model variants. We ranked stomatal optimization principles based on parameter uncertainty, parsimony, predictive accuracy, and functional accuracy of the interactions between soil moisture, vapor pressure deficit, and evapotranspiration.  Performance was high for all model variants. Water penalty functions with explicit representation of plant hydraulics did not substantially improve predictive or functional accuracy of ecosystemscale evapotranspiration estimates and parameterizations were more uncertain, despite having physiological underpinnings at the plant level.  Stomatal optimization based on water use efficiency thus provided more information about ecosystem-scale evapotranspiration compared to those based on xylem vulnerability and proved more useful in improving ecosystem-scale models with less complexity.

One Stomatal Model to Rule Them All? Toward Improved Representation of Carbon and Water Exchange in Global Models

“…We found three models (CMax, CAP, and MES) to be difficult to constrain (Figure 1), and one model (CMax) to contain a low influence parameter that could be fixed or omitted (cf., Zenes et al., 2020). At the ecosystem scale, Bassiouni and Vico (2021) confirmed that the less physiologically meaningful parameters of CMax were the most uncertain among those of three g s optimization models. From an operational perspective, constraining multi‐parameter models can require over three times as many function evaluations than for one‐parameter models.…”

“…(2010) for a review of >30 empirical g s formulations—at least 10 new stomatal optimization schemes have been proposed in the past 5 years (Anderegg et al., 2018; Dewar et al., 2017; Eller et al., 2018; Huang et al., 2018; Lu et al., 2020; Mrad et al., 2019; Novick et al., 2016; Sperry et al., 2017; Wolf et al., 2016). Some g s optimization models have been compared (Anderegg et al., 2018; Bassiouni & Vico, 2021; Dewar et al., 2017; Mrad et al., 2019; Novick et al., 2016; Wang et al., 2020) or evaluated against empirical analogs (Eller et al., 2018, 2020; Sabot et al., 2020; Venturas et al., 2018; Wang et al., 2019). Yet, comparisons rarely control for both univariate model sensitivity (e.g., isolating soil moisture impacts from vapor pressure deficit impacts) and agreement with observations (but see Novick et al., 2016).…”

Leaf Shedding and Non-Stomatal Limitations of Photosynthesis Mitigate Hydraulic Conductance Losses in Scots Pine Saplings During Severe Drought Stress

“…The approaches range from empirical relationships (e.g., Leuning, 1995), to mechanistical descriptions based on optimality theory such as maximizing C gain per unit of transpiration (e.g., Medlyn et al, 2011), maximizing transpiration while reducing conductivity loss (Sperry and Love, 2015), or maximizing C gain while minimizing loss in hydraulic conductivity (e.g., Sperry et al, 2017;Eller et al, 2018;2020). The success of these models has been mixed, leading to a good representation of broad monthly and annual transpiration and productivity patterns but often failing to capture subtler responses arising when drought stress intensifies (e.g., Drake et al, 2017;Yang et al, 2019;De Kauwe et al, 2020;Bassiouni and Vico, 2021;Mu et al, 2021;Nadal-Sala et al, 2021). Hence, challenges to model tree drought responses and mortality persists albeit increasing developments of optimization-based tree hydraulic models over the recent years.…”

Source Relationships and Model Structures Determine Information Flow Paths in Ecohydrologic Models

“…We build upon previous metrics for model predictive and functional performance based on information theory (Bassiouni & Vico, 2021; Ruddell et al., 2019) by investigating the variability of these metrics along a source dependency axis (Equation ). We define the predictive performance for a given model as follows, where Z is the observed target variable, Z mod is the modeled output, and I ( Z ; Z mod ) is their mutual information: $$${A}_{p}=1-\frac{I\left(Z;{Z}_{mod}\right)}{H(Z)}$$$ A p represents the information about the observed target variable that is missing from the model output and ranges from 0, for a perfectly accurate model, to 1, if Z mod does not reduce any uncertainty in Z .…”