Solving the fourth-corner problem: forecasting ecosystem primary production from spatial multispecies trait-based models

Forecasting productivity and stress across an ecosystem is complicated by the multiple interactions between competing species, the unknown levels of intra- and interspecific trait plasticity, and the dependencies between those traits within individuals. Integrating these features into a trait-based quantitative framework requires a conceptual and quantitative synthesis of how multiple species and their functional traits interact and respond to changing environments, a challenge known in community ecology as the fourth-corner problem. We propose such a novel synthesis, implemented as an integrated Bayesian hierarchical model. This allows us to (1) simultaneously model trait-trait and trait-environment relationships by explicitly accounting for both intra- and interspecific trait variabilities in a single analysis using all available data types, (2) quantify the strength of the trait-environment relationships, (3) identify trade-offs between multiple traits in multiple species, and (4) faithfully propagate our modeling uncertainties when making species-specific and community-wide trait predictions, reducing false confidence in our spatial prediction results. We apply this integrated approach to the world's largest mangrove forest, the Sundarbans, a sentinel ecosystem impacted simultaneously by both climate change and multiple types of human exploitation. The Sundarbans presents extensive variability in environmental variables, such as salinity and siltation, driven by changing seawater levels from the south and freshwater damming from the north. We find that tree species growing under stress have a typical functional response to the environmental drivers with inter-specific variability around this average, and the amount of variability is further contingent upon the nature and magnitude of the environmental drivers. Our model captures the retreat in traits related to resource acquisition and a plastic enhancement of traits related to resource conservation, both clear indications of stress. We predict that, if historical increases in salinity and siltation are maintained, one-third of whole-ecosystem productivity will be lost by 2050. Our integrated modeling approach bridges community and ecosystem ecology through simultaneously modeling trait-environment correlations and trait-trait trade-offs at organismal, community, and ecosystem levels; provides a generalizable foundation for powerful modeling of trait-environment linkages under changing environments to predict their consequences on ecosystem functioning and services; and is readily applicable across the Earth's ecosystems.

Automated summary

The study proposes a novel synthesis by integrating a Bayesian hierarchical model to forecast productivity and stress in an ecosystem, applied in the world's largest mangrove forest. It reveals that tree species under stress show typical functional responses to environmental drivers, with reduced traits related to resource acquisition and enhanced traits related to resource conservation.