Combining climatic and genomic data improves range-wide tree height growth prediction in a forest tree

Predicting adaptive-trait variation across species ranges is essential to assess the potential of populations to survive under future environmental conditions. In forest trees, multi-site common gardens have long been the gold standard for separating the genetic and plastic components of trait variation and predicting population responses to new environments. However, relying on common gardens alone limits our ability to extrapolate predictions to populations or sites not included in these logistically expensive and time-consuming experiments. In this study, we aimed to determine whether models that integrate large-scale climatic and genomic data could capture the underlying drivers of tree adaptive-trait variation, and thus improve predictions at large geographical scales. Using a clonal common garden consisting of 34 provenances of maritime pine (523 genotypes and 12,841 trees) planted in five sites under contrasted environments, we compared twelve statistical models to: (i) separate the genetic and plastic components of height growth, a key adaptive trait in forest trees, (ii) identify the relative importance of factors underlying height-growth variation across individuals and populations, and (iii) improve height-growth prediction of unknown observations and provenances. We found that the height-growth plastic component exceeded more than twice the genetic component. The plastic component was likely due to multiple environmental factors, including annual climatic variables, while the genetic component was driven by the confounded effects of past demographic history and provenance adaptation to the climate-of-origin. Distinct gene pools were characterized by different total genetic variance, with broad-sense heritability ranging from 0.104 (95% CIs: 0.065-0.146) to 0.223 (95% CIs: 0.093-0.363), suggesting different potential response to selection along the geographical distribution of maritime pine. When predicting height-growth of new observations, models combining population demographic history, provenance climate-of-origin, and positive-effect height-associated alleles (PEAs; previously identified by GWAs) explained as much variance as models relying directly on the common garden design. Noteworthy, these models explained substantially more variance when predicting height-growth in new provenances, particularly in harsh environments. Predicting quantitative traits of ecological and/or economic importance across species ranges would therefore benefit from integrating ecological and genomic information.