Landscape fragmentation is statistically correlated with both increases and decreases in wildfire burned area (BA). These different directions-of-impact are not mechanistically understood. Here, road density, a land fragmentation proxy, is implemented in a CMIP6 coupled land-fire model, to represent fragmentation edge effects on fire-relevant environmental variables. Fragmentation caused modelled BA changes of over ±10% in 16% of [0.5°] grid-cells. On average, more fragmentation decreased net BA globally (-1.5%), as estimated empirically. However, in recently-deforested tropical areas, fragmentation drove observationally-consistent BA increases of over 20%. Globally, fragmentation-driven fire BA decreased with increasing population density, but was a hump-shaped function of it in forests. In some areas, fragmentationdriven decreases in BA occurred alongside higher-intensity fires, suggesting the decoupling of fire severity traits. This mechanistic model provides a starting point for quantifying policy-relevant fragmentation-fire impacts, whose results suggest future forest degradation may shift fragmentation from net global fire inhibitor to net fire driver.

Human land use change (LUC) affects a third of the terrestrial surface 1 , and the resulting alteration of land continuity-known as fragmentation 2 -can result in biodiversity loss 3,4 , habitat degradation 5 , changes to the surface energy balance 6-10 and biogeochemical cycling 11 , leading to around one-third of global carbon (C) emissions 12,13 . LUC is forecast to increase substantially by 2100, with expansions in agricultural and settlement areas across all future climate-SSP scenarios of +12-83% 14 and +54-111% 15 , respectively, over a 2015 baseline. Fire is a key component of earth system biogeochemical and ecological dynamics 16-19 , however human-driven perturbations to global atmospheric and hydrologic circulations may alter existing fire regimes, changes that are expected to increase the future frequency and severity of fire events 16,20-24 , the global area prone to frequent fire (+~30%) 25 , and their attendant economic costs 26 , hampering the ability of biological -and implicitly social -systems to respond to broad -scale environmental change 27,28 .

Fire and LUC interact at differing space-time scales via weather and vegetation 29 through fuel structure and landscape fragmentation, "the division of habitat into smaller and more isolated fragments" 30 . Human land fragmentation has existed for at least 10,000 years b.p. 31 , encompassing a spectrum of forms, rationales, and cultural specificities 32 . The increasing yet spatially varying extent of fragmentation may have wide-ranging consequences for the fire regimes that present and future human activity embeds in its surrounding environment, which may differ substantially across ecosystems and timescales. However, the underlying processes linking fragmentation and