Relative influence of environmental variables in boosted regression tree models including spatial variables.

<p>Relative influence of each environmental variable from four categories (physical, land use, chemical, and spatial) on the boosted regression tree models predicting the main axis of regional faunal composition (Metacom 1) in 168 low-mountain streams and rivers in central Germany categorised into all sites (N = 168), headwaters (N = 48), mid-sized (N = 69) and large rivers (N = 51). The top plot shows the relative influence of each individual variable and the bottom the combined influence of each category of variables. Note catchment size was removed from all size-based models (i.e. headwaters, mid-sized and large rivers). See <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135450#pone.0135450.t001" target="_blank">Table 1</a> for a description of variables and <a href="http://www.plosone.org/article/info:doi/10.1371/journal.pone.0135450#pone.0135450.t003" target="_blank">Table 3</a> for the results of BRT analyses.</p>