Modelling mobile agent‐based ecosystem services using kernel‐weighted predictors. Issue 5 (13th February 2018)

• Record Type: Journal Article
• Authors: Goedhart, Paul W.; Lof, Marjolein E.; Bianchi, Felix J. J. A.; Baveco, Hans (J.) M.; van der Werf, Wopke
• Editors: Carvalheiro, Luísa
• Is Part Of: Methods in ecology and evolution. Volume 9:Issue 5(2018)
• Page Start: 1241
• Page End: 1249
• Publication Date: 2018-02-13
• Subjects: biocontrol -- conservation planning -- natural pest control -- pollination -- source–sink dynamics -- spatial distribution -- spatial extent -- statistical model; Ecology -- Periodicals; Evolution -- Periodicals; 577
• DOI: 10.1111/2041-210X.12972
• Languages: English
• ISSNs: 2041-210X

Abstract

Abstract: Agriculture benefits from ecosystem services provided by mobile agents, such as biological pest control by natural enemies and pollination by bees. However, methods that can generate spatially explicit predictions and maps of these ecosystem services based on empirical data are still scarce. Here we propose a generic statistical model to derive kernel functions to characterize the spatial distribution of ecosystem services provided by mobile agents. The model is similar in spirit to a generalized linear model, and uses data of landscape composition and ecosystem services assessed at target sites to estimate parameters of the kernel. The approach is tested in a simulation study and illustrated by an empirical case study on parasitism rates of the diamondback moth Plutella xylostella . The simulation study shows that the scale parameter of the exponential power kernel can be estimated with limited bias, whereas estimation of the shape parameter is difficult. For the case study the model provides biologically relevant estimates for the kernel associated with parasitism of P. xylostella . These estimates can be used to generate ecosystem service maps for existing or planned landscapes. The case study reveals that predictions can be sensitive to the parameter values for the width and shape of the kernel, and to the link function used in the statistical model. In the last two decades numerous empirical studies assessed ecosystem services at target sites and related these to the surrounding landscape. Our method can take advantage of these data by estimating underlying kernels that can be used to map the spatial distribution of ecosystem services. However, empirical data that can discriminate between alternative kernel shapes remain critical.