Figure 5. Main predictors of soil food web dissimilarity within forest and grassland habitats. The relative importance of each predictor was estimated using Generalised Dissimilarity Models (a). It is measured as the sum of the coefficients of the three I-splines of the focus predictor and represents the total amount of change along the predictor gradient. Partial curves for three of the most important predictors of food web dissimilarity in forest (b) and grassland (c). Each panel shows the food web dissimilarity as a function of an environmental predictor when holding all other variables to their mean. The slope at any point on the curve indicates the rate of food web dissimilarity at that position along the environmental gradient (x-axis), while the total height reached by the function indicates the total amount of food web dissimilarity due to that environmental predictor. SOM: Soil Organic Matter, FDD: Frost Degree Days, GDD: Growing Degree Days, AP: Annual Precipitation, NDVI: Normalized Difference Vegetation Index.