Search

Deforestation in the tropics causes warming which contributes to regional climate change. Forest loss occurs over a broad range of spatial scales, producing a variety of spatial patterns of cleared and forested land. Whether the spatial attributes of these patterns influence the resulting temperature change remains largely unknown. We adopted a differences-in-differences approach to analyse remotely-sensed forest loss and land surface temperature (LST) data in maritime Southeast Asia. We found that deforestation increased LST, as expected, but that the temperature increases were smaller when forest loss produced more fragmented landscapes in which non-forest and forest edges were heavily interlaced. Temperature increases were greater where the forest loss was more extensive. Warming also extended beyond the location of forest removal, so that forest loss increased temperatures in undisturbed locations up to 6 km away. Different spatial patterns of land clearing, for example, as might be produced by small-holder agriculture as opposed to large-scale deforestation, would therefore have different impacts on the local climate. Conserving forests within 4 km of farmland, urban areas or other sensitive environments may help to avoid temperature increases that reduce land productivity and worsen human health.

Forest cover loss in the tropics is well known to cause warming at deforested sites, with maximum temperatures being particularly sensitive. Forest loss causes warming by altering local energy balance and surface roughness, local changes that can propagate across a wide range of spatial scales. Consequently, temperature increases result from not only changes in forest cover at a site, but also by the aggregate effects of non-local forest loss. We explored such non-local warming within Brazil’s Amazon and Cerrado biomes, the region with the world’s single largest amount of forest loss since 2000. Two datasets, one consisting of in-situ air temperature observations and a second, larger dataset consisting of ATs derived from remotely-sensed observations of land surface temperature, were used to quantify changes in maximum temperature due to forest cover loss at varying length-scales. We considered undisturbed forest locations (1 km ^2 in extent), and forest loss trends in annuli (‘halos’), located 1–2 km, 2–4 km, 4–10 km and 10–50 km from these undisturbed sites. Our research finds significant and substantial non-local warming, suggesting that historical estimates of warming due to forest cover loss under-estimate warming or mis-attribute warming to local change, where non-local changes also influence the pattern of temperature warming.