Résumé
Natural environments vary in ways that are immensely complex -in both pattern and dimensionality -and so are the population-level responses of living organisms to these environments, especially when they involve interactions between evolution and ecology (1, 2). In light of such complexity, predicting population responses to environmental change may seem out of reach, despite becoming of pressing urgency in the face of the dramatic consequences of ongoing global change for biodiversity (3). While predictive power can be improved to some extent by gathering more (and more useful) data (4), sampling effort and experimental replication cannot be increased without bounds, especially considering the environmental impacts of research itself (5). In any case, detailed analyses will only be available on a small subset of organisms, which will hopefully yield broad insights that can be at least partially transposed to less well-known systems. The success of this endeavor will ultimately rest on our capacity to reduce the complexity of population biological processes in ways that make them accessible to analysis, understanding, and prediction, over useful time windows.
When considering the interplay between changes in the abundance of a species and changes in its phenotypic and genetic composition (alternatively described as eco-evolutionary dynamics (2) or evolutionary demography( 6)), this simplification step involves identifying the key traits and demographic vital rates that underlie variation in population growth, and thereby extinction risk. Remarkably, two such simplifications can be traced back to the same chapter of a book that is almost a century old (7). First, even when a species has a complex life history with age-dependent survivals and fecundities, its population growth rate can still be summarized by a single number -the Malthusian parameter, or intrinsic rate of increase -provided that each age class is weighted by its expected contribution to future generations, quantified by its reproductive value (a result that was later extended to include density-dependent regulation and evolution (8, 9), among other refinements). And second, the rate of adaptation, defined as the amount of change in fitness per generation (or unit time) equals the additive genetic (i.e., heritable) variance in relative fitness (7).
From these premises, it would appear that predicting eco-evolutionary responses of natural populations simply amounts to estimating the heritable variance in fitness in the wild, as recently done over a number of long-term datasets of pedigreed bird and mammal populations worldwide (10). However, this answer would be deceptively simple. Most adaptive evolution takes place in response to environmental change, and the latter may also alter the expression of genetic variation in fitness (11). Indeed, variation in fitness ultimately emerges from variation in traits that determine how well individual phenotypes match the demands set by the ecological context, making it inherently prone to genotype-by-environment interactions. It is therefore crucial to identify key traits that may underlie variation in fitness, and drive eco-evolutionary dynamics in novel and changing environments. With this in mind, the ideal situation for prediction would be one where measuring a few traits at a given time would be sufficient to predict population dynamics far into the future. While no real biological system can fully satisfy these criteria, it is nonetheless important to quantify the timeframe over which genetic and phenotypic variation can explain differences in demographic dynamics. This is what Kumar et al (12) set to test, using an exceptionally long >5-year experiment.