Using a chain of numerical models, we will produce past (hindcasting), present, and future (forecasting) simulations of species distributions and their evolution at the scale of a mountainous landscape. The results of these simulations will allow us to identify potential microrefugia.

The approach will involve (1) high-resolution climate reconstructions/projections and their downscaling to ~1 km2 resolution over the mountain areas where the species are present, (2) the analysis of climatic/microclimatic, topographic and hydrologic factors favouring the existence of microrefugia using niche-based modelling to describe the potential ranges of the species, and (3) time-dependent simulations of the species distribution over time with a dynamic vegetation model (DVM) in which species migration will be implemented.

Climate reconstructions/projections

General circulation model (GCM) climate reconstructions for key periods in the past (e.g., from the Palaeo Modelling Intercomparison Project Phase 3 and projections for the future (simulations from the Coupled Model Intercomparison Project, CMIP5, http://cmip-pcmdi.llnl.gov/cmip5/;Taylor et al., 2012, for the 21st century and beyond) will be downscaled to a 5 km x 5 km resolution over mountain areas to be studied, by using regional climate model (with 2 successive nestings from the original GCM resolution).

Due to the hydrostatic equilibrium approximation, the highest spatial resolution achievable with the model is ~5 km. As a result, further downscaling will be necessary to reach the target resolution of 1 km2 of the niche-based and dynamic vegetation studies. A statistical approach (e.g., Flint and Flint, 2012) will be used, to interpolate MAR results, or station data when available, for meteorological variables weighted by elevation or exposure.

 

Analysis with niche-based approaches

To evaluate the potential occurrence of a microrefugium, we will determine the ecological responses of species using a niche-based model (ie MaxEnt). For each species and each climatic factor identified as significant by the model, we will obtain a probability of presence based on the ecological response (with a threshold above which the species is present). We will test the impact of the landscape topography in the analysis of the niche of each species to assess their response.

Landscape rugosity is the roughness of terrain. High rugosity (deep valleys, ridges, orientation toward or away from sun, etc) create potential microrefugia & hence promote diversity. The topography of mountain areas is different among the selected study areas. Steep gradients and high rugosity may hold many small microrefugia. Broad gradients and low rugosity lead to long migration distances, limited microrefugia and high extinction rates.

 

Simulations with a dynamic vegetation model

We will use the CARAIB dynamic vegetation model developed in Liège. CARAIB is a mechanistic model  simulating the distribution of vegetation and the carbon cycle (i.e., carbon stocks and fluxes) in the terrestrial biosphere. In its standard version, CARAIB does not simulate actual but potential ranges which correspond to areas where climatic conditions theoretically allow the presence of a plant species.As inputs, CARAIB requires mean daily values of air temperature, the amplitude of temperature diurnal variation, relative sunshine duration or surface solar irradiance, relative humidity of the air, horizontal wind speed, precipitation, and soil texture. The hydrological sub-model computes water availability in the soil, actual evapotranspiration, runoff and soil humidity on a daily basis.

Where will the microrefugia be?

The modelling approaches will allow us to simulate past (hindcasting) and future (forecasting) distributions of species, at high spatial resolutions (~1 km2) over the mountain areas studied in the project. Future microrefugial areas where species could persist will be predicted using a range of simulations based on different future climate scenarios. Besides these model simulations, historical and socio-ecological data will be integrated to render future predictions: