The relative value of field survey and remote sensing for biodiversity assessment

1. The importance of habitat for biodiversity is well established, but the two most commonly used methods to measure habitat (field survey and remote sensing) have seldom been explicitly compared. 2. We compare high-resolution sample-based field survey (Countryside Survey) with medium-resolution remotely sensed habitat data (the highest resolution of Land Cover Map available) for Great Britain. Variation in abundance of 60 bird species from 335 1 km squares was modelled using habitat predictors from the two methods. Model comparisons assessed the explanatory power of (i) field survey vs. remotely sensed data and (ii) coarse information on habitat areas (Broad Habitats) vs. fine-grained information on Landscape Features. 3. Field survey data (combining Broad Habitat and Landscape Feature predictors) explained more variation in bird abundance than remotely sensed data (comprising Broad Habitat predictors only) for 57 species and had significantly higher mean explanatory power, averaged across 60 species models. The relative explanatory power of remote sensing, as a proportion of that provided by field data, was measured at 73%, aver aged across 60 species models. Predictions from field survey Broad Habitat data were more accurate than those from either remotely sensed Broad Habitat data or field survey Landscape Feature data, averaged across 60 species models. 4. High-resolution data generate more reliable models of predicted local population responses to land use change than lower resolution remotely sensed data. Collection of field data is typically costly in time, labour and resources, making use of remote sensing more feasible for assessment at larger spatial extents if data of equivalent value are produced, but the cost–benefit threshold between the two is likely to be context specific. However, integration of field survey with remotely sensed data provides accurate predictions of bird distributions, which suggests that both forms of data should be considered for future biodiversity surveys.