# Integrated region-specific modeling of H5 avian influenza in Asia using ENSO-based forecasts

## TL;DR

This study uses climate data to predict H5 bird flu outbreaks in Asia, showing how El Niño conditions affect different regions and animal types differently.

## Contribution

The novel contribution is an integrative statistical-deep learning framework combining ENSO forecasts with region-specific H5 HPAI modeling for early warning.

## Key findings

- Higher El Niño conditions correlate with fewer poultry H5 outbreaks in East and South Asia.
- Wild bird H5 outbreaks in East and South Asia show complex, phase-specific responses to climate.
- East Asia's wild birds face ongoing H5 epidemic pressure according to climate-informed forecasts.

## Abstract

Highly pathogenic avian influenza (HPAI), particularly of the H5 subtype, remains a persistent threat to poultry, wildlife, and public health across Asia. This study quantifies the influence of the El Niño–Southern Oscillation (ENSO), using the Multivariate ENSO Index (MEI) as the primary predictor, on the climate-driven dynamics of H5 HPAI through region- and host-stratified generalized additive models (GAMs). Seven region–host strata across Asia were modeled separately, revealing pronounced heterogeneity in event frequency. A clear negative correlation with MEI was identified in domestic poultry across East and South Asia, where higher MEI values, corresponding to El Niño conditions, were linked to reduced event frequencies. In contrast, wild bird populations in East and South Asia displayed irregular, multimodal response patterns to MEI, suggesting phase-specific sensitivities to climate variability. A recurrent neural network (RNN) was further employed to forecast MEI trends, which were then incorporated into the GAMs to predict event dynamics. The forecasts highlighted continued epidemic pressure in East Asia's wild birds, in contrast to stable or declining trends elsewhere. Given the zoonotic potential of H5 viruses, these climate-informed risk forecasts could help inform timely interventions to prevent animal-to-human transmission and support integrated One Health preparedness frameworks. This integrative statistical–deep learning framework offers valuable support for short-term early warning and regionally targeted prevention strategies for H5 HPAI preparedness across Asia.

•Region- and host-stratified GAMs revealed heterogeneous effects of ENSO on H5 HPAI event frequency across Asia.•Higher MEI values (El Niño conditions) were negatively correlated with poultry event counts in East and South Asia.•RNN-based MEI forecasts were integrated into the GAM framework for forward-looking riskprediction.•Forecasts indicated continued epidemic pressure in East Asia's wild birds, contrasting with stable or declining trends in other regions.•The integrative framework supports One Health early warning systems and targeted cross-sector prevention strategies.

Region- and host-stratified GAMs revealed heterogeneous effects of ENSO on H5 HPAI event frequency across Asia.

Higher MEI values (El Niño conditions) were negatively correlated with poultry event counts in East and South Asia.

RNN-based MEI forecasts were integrated into the GAM framework for forward-looking riskprediction.

Forecasts indicated continued epidemic pressure in East Asia's wild birds, contrasting with stable or declining trends in other regions.

The integrative framework supports One Health early warning systems and targeted cross-sector prevention strategies.