How much information is there in arthropod data about the landscape around sampling sites? Exploring a theory based on site-specificity of organisms

Abstract

The importance of local versus landscape drivers of biodiversity is presently intensely discussed, which raises the question what information ecological sampling can provide about the relative importance of these factors and how the amount of information is distributed over different spatial scales. Here, we have tried to assess the amount of the information in sets of arthropod samples on four landscape characteristics, i.e., the percentage arable land, semi-natural area, urban area, and edge density, at spatial scales varying from 100 m to 3000 m around sample sites. A large, existing dataset of different studies from all over Europe was used for that. Random Forests were used for predicting the characteristic classes of the surrounding area. The accuracy of the predictions, calculated as the reversed Normalized Brier score, was used as measure of the amount of information. The results showed that, at least in Europe, the amount of information is different between edge density on the one hand, and arable land, semi-natural area, and urban area on the other hand. In case of edge density, the information decreased from 100 m to 250 m around the sample site, then increased to get a hump-shape between 250 and 3000 m, with the maximum amount at 1750 m. In case of the other three landscape characteristics, the information decreased from 100 m to 1000 m, and then stayed equal or slightly increased. These results could be explained by assuming that organisms present at a sample site are either site-specific, or non-site-specific. Site-specific organisms are thought to enable predictions of characteristics at the small scales, while non-site-specific organisms are thought to indicate characteristics of larger scales. The results implied that, for study designs, it is important to be aware of the type of processes that result in the presence of species at sample sites. For effective conservation measures for arthropods, the results showed that landscapes at a spatial scale of at least 9.6 km2 should be taken in consideration in Europe.