Sensitivity analysis of the maximum entropy production method to model evaporation in boreal and temperate forests

The maximum entropy production (MEP) approach has been little used to simulate evaporation in forests and its sensitivity to input variables has yet to be systematically evaluated. This study addresses these shortcomings. First, we show that the MEP model performed well in simulating evaporation during the snow-free period at six sites in temperate and boreal forests (0.68 ≤ NSE ≤ 0.82). Second, we computed a sensitivity coefficient S  representing the proportion of change in the input variable transferred to the latent heat flux (λE). Net radiation (Rn) was the most influential variable (S » 1) at all sites, indicating that an increase in Rn translates into an equivalent increase in λE. The MEP model avoided the issue of oversensitivity to air temperature (S < 0.5 at peak evaporation) and captured limitations to transpiration associated with the atmospheric evaporative demand. Overall, the MEP model offers a promising tool for climate change studies.

Methods
We selected six sites hosting eddy-covariance flux towers from the AmeriFlux and Fluxnet networks. At each site, we modeled evaporation using the maximum entropy production method (Wang and Bras, 2011) for the snow-free period. We performed a sensitivity analysis using the method of Beven (1979) to assess the sensitivity of the MEP model to variations to input variables.

最大熵产生（maximum entropy production, MEP）方法在森林蒸发模拟中的应用尚少，且其对输入变量的敏感性尚未得到系统评估。本研究旨在弥补上述不足。首先，我们证明MEP模型在温带与寒带森林的6个样地无雪期蒸发模拟中表现优异（纳什效率系数NSE介于0.68至0.82之间）。其次，我们计算了敏感性系数S，用以表征输入变量的变化向潜热通量（latent heat flux, λE）传递的比例。净辐射（net radiation, Rn）在所有样地均为影响最大的变量（S远大于1），表明净辐射的增加会对应潜热通量的同等幅度提升。MEP模型规避了对气温的过度敏感问题（蒸发峰值时段S<0.5），并能捕捉与大气蒸发需求相关的蒸腾限制。总体而言，MEP模型为气候变化研究提供了极具潜力的工具。

研究方法
我们从AmeriFlux与Fluxnet通量网络中选取了6个搭载涡度协方差通量塔的观测样地。在每个样地的无雪期，我们采用最大熵产生法（Wang与Bras, 2011）开展蒸发模拟。我们采用Beven（1979）提出的敏感性分析方法，评估MEP模型对输入变量变化的敏感性。