How to simulate canopy temperatures in a global, process-based model?

The frequency and intensity of high temperature events will very likely increase in the future, which could have significant effects on agricultural production. A suitable tool to assess potential heat stress damages in crops is the climate-driven simulation of crop growth and development processes with computer models. While studies show that process-based models reproduce observed yield variabilities, the temperature sensitivities of underlying growth and development processes are often not in accordance with observational data, which can be a significant source of uncertainty especially in future projections. We intend to reduce these uncertainties by improving process responses to high temperatures in the dynamic global vegetation model LPJmL.

A common weakness of models, including LPJmL, is the use of air temperatures in crop related processes. Depending on climatic factors and water status, these can deviate strongly from canopy temperatures, which can have significant effects on the triggering of temperature-related process responses. As a first step, we thus implemented a combination of energy balance and empirical model in LPJmL that computes canopy temperatures based on equations of Penman and Monteith and empirical findings from Idso and Jackson. First preliminary results of future scenarios (SSP585) show that projected wheat yields are substantially higher or lower in some regions when using canopy temperatures compared to solely air temperature-driven LPJmL simulations. However, while the implemented approach assumes neutral atmospheric stability and thus requires little computing capacity, a comparison study showed that more complex methods that include stability correction factors better reproduce observed canopy temperatures. The difficulty with these complex canopy temperature computations is that the high computing costs can be a limitation for already computationally expensive global models. To solve this problem, we built a complex stand-alone model based on the Monin-Obukhov Similarity Theory for computing canopy temperatures with consideration of the stability conditions and from this derived two emulators that reproduce the results of the complex model with significantly less computing power. The two emulators describe upper and lower canopy temperature bounds under two extreme states of water stress as a function of air temperature, radiation, wind, vapor pressure deficit and leaf area index. For this, we chose parametric models with a third-order polynomial basis function that also include interaction terms of the different variables. To train the emulators, we used a global dataset that covers a broad range of combinations of different weather variables. These two emulators will be implemented in LPJmL to simulate canopy temperatures by first calculating upper and lower canopy temperature bounds and from this deriving final canopy temperatures through scaling with actual water stress. We will then compare the results to those of the approach that assumes neutral atmospheric conditions.

The computation of canopy temperatures is a first step towards better crop yield projections accounting for responses to high temperatures. The first preliminary results highlight the importance of improving the representation of canopy temperatures in global models to better estimate future agricultural yields and to identify potential risks to food security.