Estimating rates of butterfly decline from distribution maps: The effect of scale Original Research Article

Analysis of rates of decline for butterflies in the British county of Hertfordshire, from presence/absence data in grid squares before and after 1970, showed that complete extinction took place in 66·9% of 2 km squares occupied before 1970 (average for 18 species). For 12 species of intermediate rarity, a 2 km grid resulted in estimates of decline that were on average 35% higher than estimates based on a 10 km grid, the scale at which butterflies have been mapped nationally. Even estimates of decline based on a 2 km grid are likely to be underestimates because pre-1970 records are incomplete and because 2 km grid squares still conceal declines within squares. In Plebejus argus, for which the exact location of every local population is known in North Wales, a 2 km grid would seriously underestimate declines, for example giving only a 56% loss of grid squares if 90% of local populations were to become extinct. Our results and analysis of simulated distributions indicate that: for a few of the very rarest species, declines on grid maps may closely reflect population losses; for species of intermediate rarity, grid maps identify but underestimate population losses; for common species, population losses fail to be detected on grid maps. Per-population extinction rates for butterflies of intermediate rarity, and even for some relatively or very common species, may have been as high as extinction rates for some of the rarest. Because most of the commoner species initially had many populations per grid square, their declines have been underestimated or have not been detected by existing mapping schemes. We propose a scheme for monitoring changes in the status of common as well as rare butterflies in a network of intensively mapped grid squares at different scales.