Wind farms occupy fire-prone terrain but show no evidence of targeting burnt land: a multi-scale, time-ordered analysis for Greece

Stavros Stromatas 

A recurring claim in Greek public discourse holds that wildfires are set to clear land for wind farms, so that turbines end up on previously burnt ground. We test this claim with a design built to avoid the base-rate fallacy of simple overlap statistics, comparing the operational Greek wind fleet against wind-viable land with a spatial case-control model and a time-ordered survival model, on a scale ladder from the parcel to 5 km.
At the parcel scale there is no association: no evidence of site-level targeting. At the landscape scale (2–5 km), when each wind farm is weighted equally, farms sit near burnt land with an odds ratio of about 1.6 compared with the wind-viable base rate; a fire-prone fuel covariate explains roughly half, leaving a residual near 1.4 that survives adjustment for terrain, fuel and protected status. This landscape signal
disappears when individual turbines rather than farms are the unit, because larger developments are less frequently adjacent to recorded burnt areas and dominate the turbine count. The time-ordered test is not robust, and the median fire-to-installation gap is 6–7 years.
The evidence does not support deliberate burning to build; it does show that wind farms occupy fire-prone landscapes to a degree that measured landscape covariates only partly explain.
A secondary result is that the apparent association depends on the scale of development: larger farms are less often adjacent to burnt land than smaller ones, so counting projects and counting turbines give different answers.