Application of the Maximum Entropy Production Model of Evapotranspiration over Partially Vegetated Water-Limited Land Surfaces

Abstract

The maximum entropy production (MEP) model based on nonequilibrium thermodynamics and the theory of Bayesian probabilities was recently developed to model land surface fluxes, including soil evaporation and vegetation transpiration. This model requires few input data and ensures the closure of the surface energy balance. This study aims to test the capability of such a model to realistically simulate evapotranspiration (ET) over a wide range of climates and vegetation covers. A weighting coefficient is introduced to calculate total ET from soil evaporation and vegetation transpiration over partially vegetated land surfaces, resulting in the MEP-ET model. Using this coefficient, the model outputs are compared with in situ observations of ET at eight FLUXNET sites across the continental United States. Results confirm the close agreement between the MEP-ET predicted daily ET and the corresponding observations at sites characterized by moderately limited water availability. Poor ET results were obtained under high water stress conditions. A regulation parameter was therefore introduced in the MEP-ET model to properly take into account the effects of soil water stress on stomata, yielding the generalized MEP-ET model. This parameter considerably reduced model biases under water stress conditions for various heterogeneous land surface sites. The generalized MEP-ET model outperforms several popular ET models, including Penman–Monteith (PM), modified Priestley–Taylor–Jet Propulsion Laboratory (PT-JPL), and air-relative-humidity-based two-source model (ARTS) at all test sites.

Comparing the Performance of the Maximum Entropy Production Model With a Land Surface Scheme in Simulating Surface Energy Fluxes

“…On the other hand, MEP performance is slightly better compared to CLASS results at the wettest site, as previously discussed in Figure e. The LE overestimation in MEP at sites with low water availability was previously reported by Hajji et al (), with the original version of the MEP model (Wang & Bras, ). To amend the poor performances, the authors improved the model with a water stress factor, also used in this study (see more in Appendix ), showing that better simulation of evapotranspiration was achieved regardless of the water stress condition.…”

“…To amend the poor performances, the authors improved the model with a water stress factor, also used in this study (see more in Appendix ), showing that better simulation of evapotranspiration was achieved regardless of the water stress condition. However, our results at the driest site, using the same improvements offered by Hajji et al (), still exhibit, to a lesser extent, the deficiency of the model in estimating LE under limited water availability. More detailed analyses are given in subsection using a drought index (scPDSI).…”

Estimation of Evapotranspiration of Amazon Rainforest Using the Maximum Entropy Production Method

“…It is reasonable to assume that ground heat flux is nonnegligible for these land surface types. Surface humidity data are, however, unavailable but can be estimated from soil moisture and temperature (Hajji et al, 2018) or replaced by near-surface air humidity. Hence, the MEP ET at the basin scale can be further improved by addressing ground heat flux in sparsely vegetated areas.…”

A comparison of methods to estimate groundwater recharge from bare soil based on data observed by a large‐scale lysimeter

“…The targeted value reaches 1.0, demonstrating that the variance of residuals is much smaller than the variance of observations, and the model performance is relatively good (Hajji, Nadeau, Music, Anctil, & Wang, 2018).…”