Figure 2. Differences between the MANIPULATION (S1-S4) and the HISTORICAL runs (S0), for annual Rh (%; left column), annual Ra (%; middle column), and annual GPP (%; right column), at each site.
3.4.3 Model limitations and potential developments
LPJ-GUESS has difficulties in capturing the physical changes in snow depth induced by WWEs. We identified the lack of surface energy balance in LPJ-GUESS as one of the main limitations to simulating WWE impacts on GT. Heat transfer between air-snow-soil layers is simulated using the Crank-Nicholson finite difference scheme (Crank and Nicolson, 1996). This indirect method of defining heat transfer may significantly affect the computed snow layer and ground temperatures, and the rate and magnitude of snowmelt events. Additionally, the model’s daily timestep may be too coarse to capture the sub-daily freeze-thaw cycles and hydrological processes within the snowpack.
In the model, heat transfer is affected by the changes in snow layer temperature and thermal properties, and the latent heat release upon freezing is not captured (Figure 3). The SA showed that the current model setup is not sensitive to changes in liquid water holding capacity in snow layers (Figure D1-3), potentially due to the simplistic water retention scheme applied. Rainfall infiltration follows a bucket model approach, limited by each snow layer’s maximum water-holding capacity. Precipitation from ROS events and melt water is quickly forwarded to the soil as runoff at the simulated time step (i.e., daily) when the maximum liquid holding capacity of a layer is reached. If the ground is frozen, the excess liquid water will not stay in the bottom snow layer, but drain out as surface runoff. This feature also prevents the formation of ice layers of high thermal conductivity within the snow layers, as many observational studies suggest (e.g., Langlois et al., 2017), which could influence the simulated GT. Observations suggest major ROS events may have durable impacts on GT (e.g. Westermann et al., 2011), but the current model setup cannot capture such persistent effects due to the simple water retention scheme and the lacking processes related to energy balance.
The LPJ-GUESS version used in this study has an intermediate complexity snow scheme, similar to many ecosystem models. These modules are developed and tuned to represent average conditions rather than capturing extreme and smaller-scale phenomena such as WWEs. There is a need for either further snow scheme development in LPJ-GUESS or evaluating a more complex, designated snow model to address the listed shortcomings and capture internal snow dynamics on a finer spatio-temporal scale. A recent extension of the model with detailed land surface processes and surface energy balance LPJ-GUESS LSMv1.0 (Belda et al., 2022) could be used in future studies to assess whether the model-measurements mismatch is reduced when using the LPJ-GUESS with detailed energy balance and sub-daily processes. This land surface model version of LPJ-GUESS needs to be thoroughly evaluated and tested in high latitude environments. Regional modelling studies are required to further understand how WWEs might affect the pan-Arctic carbon budget.