# A time-varying geospatial model of habitat suitability for Japanese encephalitis virus vectors and vertebrate hosts in Australia

## TL;DR

This study creates dynamic models to predict where Japanese encephalitis virus spreads in Australia, helping improve surveillance and response.

## Contribution

The study introduces a novel method to combine and rescale habitat suitability models for JEV vectors and hosts.

## Key findings

- JEV habitat suitability was highest in northern and eastern inland Australia during the 2022 outbreak.
- The Great Dividing Range likely limits JEV transmission along the eastern seaboard.
- Despite favorable conditions in 2023, few JEV infections were detected, indicating gaps in understanding its ecology.

## Abstract

In the austral summer of 2021–2022, Australia experienced an unprecedented Japanese encephalitis virus (JEV) outbreak, with detections over 3000 km south of previous occurrences. Given the limited knowledge of JEV transmission ecology in Australia, we developed geospatial models of transmission risk to support the public health response. We created time-varying habitat suitability models for suspected mosquito vectors and ardeid hosts using month-scaled occurrence and covariate data from 2000–2023. Ardeid host presence-absence data were obtained from eBird and BirdLife Australia, with habitat suitability estimated using gradient-boosted regression tree models. A national dataset of Culex occurrences was compiled from mosquito surveillance records, literature, and biodiversity databases. Penalised logistic regression was used to model mosquito vector habitat suitability. Vector and host habitat predictions for the outbreak peak in February 2022 were rescaled using JEV infection locations in the public domain to create a combined habitat suitability surface. Our models aligned with detected JEV infections at the continental scale, highlighting transmission suitability across tropical northern Australia and major inland drainage basins in the East. Unlike existing models, we predicted lower suitability along the eastern seaboard, suggesting a delimiting effect of the Great Dividing Range. Our approach provides the most comprehensive and temporally dynamic models for JEV hosts and vectors in Australia, with a significantly larger vector dataset than previous studies. The novel method of rescaling host and vector outputs into a combined surface offers new insights into JEV transmission risk. Favourable conditions were repeated in 2023 with few detected infections, emphasising that JEV ecology in Australia remains poorly understood. This study’s results can support improvements in arbovirus surveillance systems, promoting earlier detection of circulating viruses. Increased focus on vector ecology and distributions is crucial for better understanding JEV transmission in Australia.

Our work represents a major advance in understanding the distribution of mosquito vectors and vertebrate hosts for Japanese encephalitis virus (JEV) in Australia. Prior to the summer of 2021/22 JEV was known from periodic incursions into Australia’s northernmost extremes, but from December 2021 there was an unprecedented expansion into southern Australia associated with abundant breeding habitat for hosts and vectors associated with a La Niña weather pattern. We brought together mosquito occurrence records, and structured occurrence data for ardeid hosts to inform time-varying, continental scale habitat suitability models for Australian JEV vectors and vertebrate hosts. We present a method of rescaling and combining the joint model outputs to facilitate consideration of a combined model of suitability for pathogen transmission.