# Community density patterns estimated by species distribution modeling: The case study of an insect virus interaction

## TL;DR

This study examines how an insect virus interacts with its host in the Andes using species distribution models and finds limited correlation between host density and virus prevalence.

## Contribution

The paper introduces a method using species distribution models to analyze virus-host interactions with non-sympatric data.

## Key findings

- Locality and storage bag explained 8% and 26% of the total variance in infection status.
- Predicted virus prevalence was not significantly correlated with predicted host density (R=-0.053).
- 26% of the total insect population in the study range was expected to be infected based on the model.

## Abstract

We studied the interaction between the invasive potato moth T. solanivora and its granulovirus PhopGV in the northern Andes. Host density was analyzed based on 1206 pheromone trap data from 106 sampled sites in Ecuador, Colombia and Venezuela. The prevalence of the virus was assessed at 15 sites in 3 regions in Ecuador and Colombia. Infection status was analyzed for spatial structure at different scales: storage bag, storage room, field, locality, country. Locality and storage bag explained 8% and 26%, respectively of the total variance in infection status in glm analysis. The field versus storeroom effect differed between localities. GLM species distribution models were optimized for bioclimatic variables for both insects and viruses. Predicted virus prevalence was not significantly correlated with predicted host density at sampled virus sites. Over the entire climatic range covered by the study, the correlation was R=-0.053. Of the total population insect in this range, 26% were expected to be infected based on the model. This basic method of using species distribution models to analyze average correlations between species densities can help investigate statistical relationships across a range of trophic models using existing non-sympatric data, with little or no additional sampling effort. It removes confounding time-lag effects and allows the use of data collected separately in the different species. The approach is correlative, and cannot be interpreted in terms of causality or outside the study area.