Activity Number:
|
420
- Highlights from "Bayesian Analysis": Bayesian Methods for the Public Good
|
Type:
|
Invited
|
Date/Time:
|
Thursday, August 6, 2020 : 10:00 AM to 11:50 AM
|
Sponsor:
|
Section on Bayesian Statistical Science
|
Abstract #309385
|
|
Title:
|
Additive Multivariate Gaussian Processes for Joint Species Distribution Modeling with Heterogeneous Data
|
Author(s):
|
Jarno Vanhatalo* and Marcelo Hartmann and Lari Veneranta
|
Companies:
|
University of Helsinki and University of Helsinki and Natural Resources Institute Finland
|
Keywords:
|
linear model of coregionalization;
hierarchical model;
heterogeneous data;
spatial prediction;
model comparison;
covariance transformation
|
Abstract:
|
Species distribution models are key tools in ecology. Their key components are description for species distribution response along environmental covariates and spatial random effects that capture distribution patterns not explained by environmental covariates. Joint species distribution models (JSDMs) additionally include interspecific correlations in these components which improve their descriptive and predictive performance compared to single species models. We developed JSDMs where the environmental responses are modeled with additive multivariate Gaussian processes coded as linear models of coregionalization. These can be seen as an extension of the current state-of-the-art JSDMs to semi-parametric models which allow wide range of functional forms and interspecific correlations between the responses. We propose efficient approach for inference with Laplace approximation and parameterization of the interspecific covariance matrices on the Euclidean space. We compare the proposed model to analogous single species models and parametric single and joint species models in interpolation and extrapolation tasks. The proposed model outperforms the alternative models in all cases.
|
Authors who are presenting talks have a * after their name.