Erratic birth rates challenges ecologists to explain the transitions in forests that are happening now and to anticipate the consequences for food webs. Uncertain seed limitation remains perhaps the greatest obstacle to predicting forest response to climate change. Relatively few, site-specific studies explain small fractions of the variation in seed production, dispersal, germination, and seedling survival. These studies do not provide a basis for extrapolation to continents or even landscapes due to poorly parameterized effects of species composition, tree life history, habitat variables, and quasi-periodic variation (‘masting’). From the perspective of seed predators this volatile seed supply constitutes an unreliable resource. As climates change, poorly-understood fecundity responses will ramify throughout food webs. Consumers will experience change each in its own way, depending on its capacities to cope with covariance structure in time, in space, and across a diverse diet of alternative host trees that vary in quality and are dispersed throughout the canopy and forest floor. The questions are many. What are the changes in tree fecundity that are happening now? To what extent are these changes controlled by direct climate responses versus the changing structure of forests? How does the space-time structure of masting interact with the capacities of consumers to exploit alternative hosts, to forage widely in space, and to store reserves in time?
I discuss a new synthetic collaboration to answer these fundamental questions of forest response and its impact on food webs, MASTIF. We combine efforts of dozens of collaborators from across the globe, but focus here on North America. The overview includes i) new theory to evaluate, quantify, and predict changes in recruitment that are underway now and ii) and how we quantify these effects. I discuss two theoretical innovations. First is a shift in emphasis from ‘states to rates’. This is the notion that future states (e.g., community composition of multiple species) are never observed, but those we care about most (the decade scale) follow from what is observable now, the current rates of change. Second is theory needed to evaluate resource supply at the scales of space, time, and diet breadth experienced by consumers. From the first continental scale, long-term network of tree fecundity data we determine where seed production is limiting tree recruitment, and we attribute variation to climate, species, and the structure of forest stands, which varies regionally and continues to change over time. Unlike meta-analysis of aggregated statistics from individual studies, MASTIF accomplishes Bayesian inference (a single posterior distribution) on the full network. With the goal of understanding the types of consumers that forests can support, we incorporate changes in fecundity and predict spatio-temporal covariance in the diverse diets of mast consumers with their capacities to average over variation through movement in spaces, storage in time, and substitute resources that vary asynchronously.