Figure 3. Multivariate partial dependence plots offer a chance to better grasp driver interactions. These plots show the neural networks’ approximation of driver richness relationships when two inputs co-vary. In this example, we see how species richness is expected to respond across different temperatures (a), seasonality in primary productivity (b) and limiting biogeochemical variables (c) in response to changing depth and elevation. This is an example of the power of ANNs to approximate functions and the integration of interactions. This can be explored for any possible interaction.

Figure 3. Multivariate partial dependence plots offer a chance to better grasp driver interactions. These plots show the neural networks’ approximation of driver richness relationships when two inputs co-vary. In this example, we see how species richness is expected to respond across different temperatures (a), seasonality in primary productivity (b) and limiting biogeochemical variables (c) in response to changing depth and elevation. This is an example of the power of ANNs to approximate functions and the integration of interactions. This can be explored for any possible interaction.