Habitat quality and metapopulation e¡ects are the main hypotheses that currently explain the disproportionate decline of insects in cultivated Holarctic landscapes. The former assumes a degradation in habitat quality for insects within surviving ecosystems, the latter that too few, small or isolated islands of ecosystem remain in landscapes for populations to persist. These hypotheses are often treated as alternatives, and this can lead to serious con£ict in the interpretations of conservationists. We present the ¢rst empirical demonstration that habitat quality and site isolation are both important determinants of where populations persist in modern landscapes. We described the precise habitat requirements of Melitaea cinxia, Polyommatus bellargus and Thymelicus acteon, and quanti¢ed the variation in carrying capacity within each butter£y's niche. We then made detailed surveys to compare the distribution and density of every population of each species with the size, distance apart and quality of their speci¢c habitats in all their potential habitat patches in three UK landscapes. In each case, within-site variation in habitat quality explained which patches supported a species' population two to three times better than site isolation. Site area and occupancy were not correlated in any species. Instead of representing alternative paradigms, habitat quality and spatial e¡ects operate at di¡erent hierarchical levels within the same process: habitat quality is the missing third parameter in metapopulation dynamics, contributing more to species persistence, on the basis of these results, than site area or isolation. A reorientation in conservation priorities is recommended.

Summary Understanding the drivers of population abundance across species and sites is crucial for effective conservation management. At present, we lack a framework for predicting which sites are likely to support abundant butterfly communities. We address this problem by exploring the determinants of abundance among 1111 populations of butterflies in the UK, spanning 27 species on 54 sites. Our general hypothesis is that the availability of food resources is a strong predictor of population abundance both within and between species, but that the relationship varies systematically with species’ traits. We found strong positive correlations between butterfly abundance and the availability of food resources. Our indices of host plant and nectar are both significant predictors of butterfly population density, but the relationship is strongest for host plants, which explain up to 36% of the inter‐site variance in abundance for some species. Among species, the host plant–abundance relationship is mediated by butterfly species traits. It is strongest among those species with narrow diet breadths, low mobility and habitat specialists. Abundance for species with generalist diet and habitat associations is uncorrelated with our host plant index. The host plant–abundance relationship is more pronounced on sites with predominantly north‐facing slopes, suggesting a role for microclimate in mediating resource availability. Synthesis and applications. We have shown that simple measures can be used to help understand patterns in abundance at large spatial scales. For some butterfly species, population carrying capacity on occupied sites is predictable from information about the vegetation composition. These results suggest that targeted management to increase host plant availability will translate into higher carrying capacity. Among UK butterflies, the species that would benefit most from such intervention have recently experienced steep declines in both abundance and distribution. The host plant–abundance relationship we have identified is likely to be transferrable to other systems characterized by strong interspecific interactions across trophic levels. This raises the possibility that the quality of habitat patches for specialist species is estimable from rapid assessment of the host plant resource.