Revegetation projects face the major challenge of sourcing the optimal plant material. This is often done with limited information about plant performance and increasingly requires to factor resilience to climate change. Functional traits can be used as quantitative indices of plant performance and guide provenancing, but trait values expected under novel conditions are often unkown. To support climate-resilient provenancing efforts, we develop a trait prediction model that integrates the effect of genetic variation with fine-scale temperature variation. We train our model on multiple field plantings of Arabidopsis thaliana and predict two relevant fitness traits -- days-to-bolting and fecundity -- across the species' European range. Prediction accuracies were high for days-to-bolting and moderate for fecundity, with the majority of trait variation explained by temperature differences between plantings. Projection under future climate predicted a decline in fecundity, although this response was heterogeneous across the range. In response, we identified novel genotypes that could be introduced to genetically offset the fitness decay. Our study highlights the value of predictive models to aid seed provenancing and improve the success of revegetation projects.