Evaluating the potential for species to adapt to changing climates relies on understanding current patterns of adaptive variation and selection, which might vary in intensity across a species' niche, hence affecting our inference of where adaptation might be most important in the future. Here, we investigate the genetic basis of adaptation in Lactuca serriola along a steep precipitation gradient in Israel approaching the species' arid niche limit and use candidate loci to inform predictions of its past and future adaptive evolution. Environmental association analyses combined with generalised dissimilarity models revealed 108 candidate genes showing nonlinear shifts in allele frequencies across the gradient, with 66% of these genes under strong selection near the dry niche edge. We detected selection acting on genes with separate suites of biological functions, specifically related to phenology and responses to environmental stressors, including osmotic stress, at the dry niche edge, and related to biotic interactions and defence closer to the niche centre. The adaptive genetic composition of populations, as inferred through polygenic risk scores, point to intensified selection operating towards the dry niche edge. However, inference of past and future evolutionary change predicts larger adaptive shifts occurring in the mesic part of the range, which is most affected by climate change. Our study reveals that adaptive shifts in response to climate change can be heterogeneous across a species' range and not necessarily strongest near its niche edge.