3. Modelling Distributions after Climate Change
Having completed and run three different GAMs using 1, 3 and 5 variables, this process is basically the same as modelling for current distributions. Once you have identified the model and variables that best describe the current distribution of the study species within your country of choice, you will use the future climate data to identify the distribution of suitable bioclimatic envelopes for your study species in 2050 under the A2 climate change scenario.
- Identify the best model. You should consider the visual match between the observed and modelled species distributions in ArcView, but the ultimate decision should be based on the receiver operating characteristic values you obtained (check the R console, and look at the AUC for each of the models' predictions). Remember, ROC values above 0.8 are good models.
Future distribution modelling.
The following should be carried out in R, and each step is explained in the script's text
- In R, using this model's script (remember, you saved each of the three models' scripts with a different name), load up the future environmental data (A2_2050.dbf).
- Project the model using the new data.
- Add latitude and longitude values to the projection.
- Estimate the optimal threshold for cutoff values.
- Transform the probabilistic values into presence/absence values for the chosen values.
- Output the projection as a txt file.
Examining the outputs.
- Load the txt file into Arcview to examine the projected climate change-driven distribution in 2050.
- Using this DBF file, you can then output the data as an ESRI Shapefile, which can be loaded onto the web or shared amongst other members of the course.