Analysis of a phylogenetic epidemiology approach to predicting the establishment of multi-host plant pests

Forecasting emergent pest spread is paramount to mitigating their impacts. For host-specialized pests, epidemiological models of spread through a single host population are well developed. However, most pests attack multiple host species; the challenge is predicting which communities are most vulnerable to infestation. Here, we develop a phylogenetically-informed approach to predict establishment of emergent multi-host pests across heterogeneous landscapes. We model a beetle-pathogen symbiotic complex on trees, introduced from Southeast Asia to California. The phyloEpi model for likelihood of establishment was predicted from the phylogenetic composition of woody species in the invaded community and the influence of temperature on beetle reproduction. Plant communities dominated by close relatives of known epidemiologically critical hosts were four times more likely to become infested than communities with more distantly related species. Where microclimate favored beetle reproduction, pest establishment was greater than expected based only on species composition. We applied this phyloEpi model to predict infestation risk in California using weather data and complete tree inventories from 9,262 1-km² grids in 170 cities. Regions in the state predicted with low likelihood of infestation were confirmed by independent monitoring. Analysts can adapt these phylogenetic ecology tools to predict spread of any multi-host pest in novel habitats.

Supplementary Data 1. Vegetation group, plot size, tree basal area (BA) and density, attacked tree density in 2017 and 2018, species richness, density- and basal area-based phylogenetically weighted susceptibility estimates, and ISHB generation estimates for each plot in Ventura, Orange, and San Diego Counties. Plots are presented in order from northwest to southeast within each county and vegetation group (California Broadleaf Forest & Woodland [BFW]; Semi-natural [SN]; and Southwestern North American Riparian Woodland, Forest, or Wash Scrub [RW, RF, or WS]).

Supplementary Data 4. ISHB generation estimates using degree-day models and wPS estimates using individual tree data from the California Urban Forest Inventory, aggregated to 1-km² grids in 170 cities across the extent of California.

Supplementary Data 5. Subset of 832 grids that were independently monitored for FD–ISHB across its known infested and non-infested geographic range and used to test the model.

Supplementary Data 6. Phylogenetic distance matrix of all potential species ISHB could encounter calculated using a dated ultrametric phylogenetic tree developed by Lynch et al. (2021).

Supplementary Data 7. Fusarium dieback-invasive shot hole borer host traits as described in Lynch et al. (2021).

Supplementary Data 8. Species density by plot matrix used to calculate wpS.

Supplementary Data 9. Species basal area by plot matrix used to calculate wpS.

cdd_calcs2.jl: Julia script used to compute the output raster bioclimatic map of beetle generation estimates (Fig. 2). The input data consists of a 20 year-long raster stack of daily temperature minimums and maximums obtained from Daymet (https://daac.ornl.gov/DAYMET/guides/Daymet_Annual_V4R1.html).

phyloEpi.R: R script used to develop the phyloEpi model, perform the data analyses, and produce graphs.