Optimizing Ecosystem Parameterization Using Genetic Algorithm: Addressing Uncertainties and Equifinality in Modeling Plant Physiological Processes and Canopy Radiative Transfer

Canopy structural and leaf photosynthesis parameterizations such as maximum carboxylation capacity, slope of the Ball–Berry stomatal conductance model, leaf area index, leaf chlorophyll content and canopy height are crucial for modelling plant physiological processes and canopy radiative transfer. These parameterizations also represent the largest sources of uncertainty in predictions of mass and energy exchange across ecosystems. While gradient-based inversion methods are commonly used, they often lack accuracy due to their susceptibility to becoming trapped in local minima. Stochastic approaches on the other hand alleviate this problem but they suffer the disadvantage of being computationally intensive, requiring substantial computing power to explore the full parameter space. Additionally, many process based models exhibit high nonlinearity and discontinuities, making gradient computation challenging. We propose an optimal moving window inversion framework based on genetic algorithms, using the Soil Canopy Observation Photochemistry and Energy Fluxes (SCOPE 2.0) model to constrain key ecosystem parameters. This inversion framework incorporates constraints from observed turbulent and energy fluxes, as well as net outgoing radiation and spectral reflectance, to narrow the parameter search space. Results from several ecosystems showing the advantage of this method featuring both C3 and C4 photosynthetic pathways under stressed and unstressed conditions will be presented. Further, the potential of this approach to address parameter equifinality issues commonly encountered when optimizing multiple parameters will also be discussed.