Computationally efficient joint species distribution modeling of big spatial data

Abstract

The ongoing global change and the increased interest in macroecological processes call for the analysis of spatially extensive data on species communities to understand and forecast distributional changes of biodiversity. Recently developed joint species distribution models can deal with numerous species efficiently, while explicitly accounting for spatial structure in the data. However, their applicability is generally limited to relatively small spatial data sets because of their severe computational scaling as the number of spatial locations increases. In this work, we propose a practical alleviation of this scalability constraint for joint species modeling by exploiting two spatial-statistics techniques that facilitate the analysis of large spatial data sets: Gaussian predictive process and nearest-neighbor Gaussian process. We devised an efficient Gibbs posterior sampling algorithm for Bayesian model fitting that allows us to analyze community data sets consisting of hundreds of species sampled from up to hundreds of thousands of spatial units. The performance of these methods is demonstrated using an extensive plant data set of 30,955 spatial units as a case study. We provide an implementation of the presented methods as an extension to the hierarchical modeling of species communities framework.

Joint species distribution modelling with ther‐package Hmsc

“…Compared to H msc 2.0 (see Ovaskainen, Tikhonov, Norberg, et al, ), H msc 3.0 includes several extensions, enabling one to ask how environmental conditions influence species‐to‐species association matrices (Tikhonov, Abrego, Dunson, & Ovaskainen, ), to infer species‐to‐species associations from time‐series data of species‐rich communities (Ovaskainen, Tikhonov, Dunson, et al, ), and to apply H msc to large spatial data (Tikhonov, Duan, et al, ). Furthermore, H msc 3.0 offers much improved flexibility with respect to the random error structures, model fitting efficiency and greater functionality for post‐processing the results and for making predictions.…”

Forecasting community reassembly using climate‐linked spatio‐temporal ecosystem models

“…Each panel also indicates the last year of biological data (dotted vertical line), representing the year when the model starts forecasting biological variables based on a projection of ocean physics. (Thorson et al 2015, Tikhonov et al 2020. Applying EOF to community ecology identifies correlations between annual variation in density (within or among species) at any two locations, where these locations can be geographically distant from one another and can be more tightly correlated than locations between the two.…”

Neural hierarchical models of ecological populations

“…The current state of the art multi‐species occupancy models use approximately one‐tenth of the number of species, at about half the number of sites, and are static in the sense that colonisation and extinction dynamics through time are not represented (Tobler et al ). Species distribution models are increasingly scalable due to advances in approximate Gaussian process models (Tikhonov et al ), but multi‐species dynamic occupancy models have not previously been reported at this scale. This is particularly relevant for extensions of the BBS case study, given the volume of (imperfect) bird data accumulating through citizen science programs (Sullivan et al ).…”