Dynamic Inverse Prediction and Sensitivity Analysis With High-Dimensional Responses: Application to Climate-Change Vulnerability of Biodiversity

2016

Ecology

Self Cite

Trait analysis aims to understand relationships between traits, species diversity, and the environment. Current methods could benefit from a model-based probabilistic framework that accommodates covariance between traits and quantifies contributions from inherent trait syndromes, species interactions, and responses to the environment. I develop a model-based approach that separates these effects on trait diversity. Application to USDA Forest Inventory and Analysis (FIA) data in the eastern United States demonstrates an apparent paradox, that the analysis of species better explains and predicts traits than does direct analysis of the traits themselves; trait data contain less, not more, information than species on environmental responses. Whereas variation in some traits is dominated by inherent syndromes (tendency for certain traits to be associated with others within an individual and species), others are strongly controlled by variation in species diversity. There is substantial variation in environmental control on trait patterns, between traits and regionally. In terms of environmental response traits do not aggregate into defined plant functional types, as would be desirable for models.

Section: Species and Environment Contributions To Trait Diversitymentioning

confidence: 99%

Why species tell more about traits than traits about species: predictive analysis

2016

Ecology

Self Cite

“…Because the basic approach has been detailed with diagnostics in both the ecological and statistical literature (Clark et al. , , , Clark ) we focus here on the new innovations associated with seed identification errors. AR(

p

) effects are presented in the application to Pinus , because they are not available from the simulator.…”

Section: Resultsmentioning

confidence: 99%

“…), appear less effective than allowing for heterogeneity in the masting process, between trees and over time (Clark et al. , , Martinez and Gonzalez‐Taboada , Uriarte et al. ).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Foodwebs based on unreliable foundations: spatiotemporal masting merged with consumer movement, storage, and diet

Núñez

Tomasek

2019

Ecological Monographs

Self Cite

114

Mast‐fruiting trees represent a pulsed resource that both supports and destabilizes consumer populations. Whereas a reliable resource is abundant on average and with limited variation in time and space, masting is volatile and localized, and that variability ramifies throughout food‐webs. Theory is developed to evaluate how the space–time structure of masting interacts with consumers who exploit alternative hosts, forage widely in space, and store reserves in time. We derive the space–time–species covariance in resource supply and combine it with the space–time–diet breadth of consumers, or ambit. Direct connection to data is made possible with Mast Inference and Forecasting (MASTIF), a state‐space autoregressive model that fits seed‐trap and canopy observations and predicts resource availability within the canopy and on the forest floor with full uncertainty. A resource score can be assigned to each consumer–habitat combination that integrates the benefits of a high mean supply weighed against the variance cost. As the consumer ambit increases, the volatility of an unreliable resource shifts from a variance cost to a mean benefit. Consumers foraging in the canopy (arboreal arthropods and rodents, song birds) experience space‐time covariance between host trees. Consumers on the forest floor (seed and damping‐off fungi, arthropods, rodents, ground‐nesting birds, mammals) experience instead a redistribution of that covariance by dispersal. For consumers lacking mobility, demographic storage in the form of episodic birth cohorts following mast years is important for population persistence. Consumers additionally compensate volatility with diet breadth. Depending on the dominant masting strategies of host tree species in the diet, habitats differentially limit consumers depending on the misalignment between consumer ambit and spatiotemporal covariance of hosts. The impact of adding or subtracting a diet item can be gauged with the standard error (SE) rule or the benefit of an added diet item balanced against the variance cost, both of which depend on the existing diet, the abundance of the new host, and the consumer's foraging ambit. Results rank habitats by their capacities to support wildlife and other consumers from a resource perspective. Results are connected directly to data, with full uncertainty, by MASTIF.

“…Inverse prediction provides a composite estimate of environmental importance for the entire community (Clark et al. , ). It opens new options for predicting the environment from species, because it combines information from all species in a synthetic prediction with full uncertainty.…”

Section: Introductionmentioning

confidence: 99%

Generalized joint attribute modeling for biodiversity analysis: median‐zero, multivariate, multifarious data

Nemergut

Seyednasrollah

et al. 2017

Ecological Monographs

Self Cite

226

278

Probabilistic forecasts of species distribution and abundance require models that accommodate the range of ecological data, including a joint distribution of multiple species based on combinations of continuous and discrete observations, mostly zeros. We develop a generalized joint attribute model (GJAM), a probabilistic framework that readily applies to data that are combinations of presence‐absence, ordinal, continuous, discrete, composition, zero‐inflated, and censored. It does so as a joint distribution over all species providing inference on sensitivity to input variables, correlations between species on the data scale, prediction, sensitivity analysis, definition of community structure, and missing data imputation. GJAM applications illustrate flexibility to the range of species‐abundance data. Applications to forest inventories demonstrate species relationships responding as a community to environmental variables. It shows that the environment can be inverse predicted from the joint distribution of species. Application to microbiome data demonstrates how inverse prediction in the GJAM framework accelerates variable selection, by isolating effects of each input variable's influence across all species.