Error propagation associated with benefits transfer-based mapping of ecosystem services

Abstract

An increasing number of studies are taking the important first step in global efforts to conserve key ecosystem services by mapping their spatial distributions. However, a lack of primary data for most services in most places has largely forced such mapping exercises to be based on proxies. The common way of producing these proxies is through benefits transfer-based mapping, in which estimates of the values of services are obtained from a small region for particular land cover types, and then extrapolated to a larger area for these same types. However, the errors that may result from such extrapolations are poorly understood. Here, we separate the generalization errors associated with benefits transfer mapping into three constituent components – uniformity, sampling, and regionalization error – and evaluate their effects using primary data for four ecosystem services in England. Variation in ecosystem services within a particular land cover type (uniformity error) alone led to a poor fit to primary data for most services; sampling effects (sampling error) and extrapolating from a small region to a larger area (regionalization error) led to substantial, but highly variable, additional reductions in the fit to primary data. We also show that combining multiple ecosystem services into a single layer is likely to be even more problematic as it contains the errors in each of the constituent layers. These errors are sufficiently large to undermine decisions that might be based on such extrapolated maps. Greatly improved mapping of the actual distributions of ecosystem services is therefore needed to achieve the goal of conserving these vital assets.