Supplementary material
Fig. 1. The response curves based on three different regularization values (1, 3 and 5), defined in the Maxent software, of the four most important predictors in Model 1 for the common eider. The first row of graphs displays the response curves for distance to vegetation, the second row exposure, the third row elevation and the fourth row distance to forest. Higher regularization values result in more smooth curves and thus more general models. The probability of presence (logit output) is shown on the y-axis and the range of the environmental variable is shown on the x-axis.
Fig. 2. The response curves based on three different regularization values (1, 3 and 5), defined in the Maxent software, of the four most important predictors in Model 2 for the common eider. The first row of graphs displays the response curves for distance to vegetation, the second row exposure, the third row elevation and the fourth row distance to forest. Higher regularization values result in more smooth curves and thus more general models. The probability of presence (logit output) is shown on the y-axis and the range of the environmental variable is shown on the x-axis.
Fig. 3. The response curves based on three different regularization values (1, 3 and 5), defined in the Maxent software, of the three most important predictors in Model 1 for the herring gull. The first row of graphs displays the response curves for elevation, the second row distance to forest and the third row exposure. Higher regularization values result in more smooth curves and thus more general models. The probability of presence (logit output) is shown on the y-axis and the range of the environmental variable is shown on the x-axis.
Fig. 4. The response curves based on three different regularization values (1, 3 and 5), defined in the Maxent software, of the three most important predictors in Model 2 for the herring gull. The first row of graphs displays the response curves for elevation, the second row distance to forest and the third row exposure. Higher regularization values result in more smooth curves and thus more general models. The probability of presence (logit output) is shown on the y-axis and the range of the environmental variable is shown on the x-axis.