Abstract
The prediction of species geographic redistribution under climate change (i.e. range shifts) has been addressed by both experimental and modelling approaches and can be used to inform efficient policy measures on the functioning and services of future ecosystems. Dynamic Global Vegetation Models (DGVMs) are considered state-of-the art tools to understand and quantify the spatio-temporal dynamics of ecosystems at large scales and their response to changing environments. They can explicitly include local vegetation dynamics relevant to migration (establishment, growth, seed production), species-specific dispersal abilities and the competitive interactions with other species in the new environment. However, the inclusion of more detailed mechanistic formulations of range shift processes may also widen the overall uncertainty of the model. Thus, a quantification of these uncertainties is needed to evaluate and improve our confidence in the model predictions. In this study, we present an efficient assessment of parameter and model uncertainties combining low-cost analyses in successive steps: local sensitivity analysis, exploration of the performance landscape at extreme parameter values, and inclusion of relevant ecological processes in the model structure. This approach was tested on the newly-implemented migration module of the state-of-the-art DGVM, LPJ-GM 1.0. Estimates of post-glacial migration rates obtained from pollen and macrofossil records of dominant European tree taxa were used to test the model performance. The results indicate higher sensitivity of migration rates to parameters associated with the dispersal kernel (dispersal distances and kernel shape) compared to plant traits (germination rate and maximum fecundity) and highlight the importance of representing rare long-distance dispersal events via fat-tailed kernels. Overall, the successful parametrization and model selection of LPJ-GM will allow simulating plant migration with a more mechanistic approach at larger spatial and temporal scales, thus improving our efforts to understand past vegetation dynamics and predict future range shifts in a context of global change.
Highlights
It is widely accepted that climate change is affecting the geographic distribution of species worldwide (Pecl et al, 2017; Lenoir et al, 2020)
According to the four summary statistics of the local sensitivity analysis (LSA), we identified the two most influential parameters as the mean (SDDd) and maximum (LDDd) dispersal distances for local and long-distance seed dispersal (LDD), respectively (Fig. 1 and Fig. S2)
All parameters related to the dispersal kernel (SDDd and LDDd) showed an overall consistency across species regarding the positive sign of their relationship with migration rate, though of different magnitude relative to the species (Fig. 1, Fig. S2 and 410 Table S1)
Summary
It is widely accepted that climate change is affecting the geographic distribution of species worldwide (Pecl et al, 2017; Lenoir et al, 2020). 45 Though the inclusion of more detailed mechanistic representations of the migration process can potentially improve the predictive power of a model, the larger number of simulated equations and parameters, each with its inherent uncertainty, may increase the overall uncertainty of the model predictions (Snowling and Kramer, 2001). An assessment of these errors is needed to increase our confidence in the model predictions, and/or to identify parameters and representations of ecological processes that require further improvement.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
