Strong and consistent associations of waterbird community composition with HPAI H5 occurrence in European wild birds

Since 2014, highly pathogenic avian influenza (HPAI) H5 viruses of clade 2.3.4.4 have been dominating the outbreaks across Europe, causing massive deaths among poultry and wild birds. However, the factors shaping its broad-scale outbreak patterns remain unclear. With extensive waterbird survey datasets of about 7,000 sites across Europe, we here demonstrated that H5N8 occurrence in wild birds in the 2016/17 and 2020/21 epidemics as well as H5N1 occurrence in 2005/06 epidemic were strongly associated with very similar waterbird community attributes, pointing to the possibility of similar interspecific transmission processes between different epidemics. A simple extrapolation of the model constructed from the 2016/17 epidemic can well predict the H5N8 pattern in wild birds in 2020/21 epidemic. We also found a dilution effect of phylogenetic diversity that was always negatively correlated with H5 occurrence in wild birds. In contrast, H5N8 occurrence in poultry was subject to different risk factors between the two epidemics. In general, waterbird community composition play a much more important role in determining the spatial pattern of H5N8 in wild birds than in poultry. Our work contributes to reveal the factors driving H5N8 patterns, and highlights the value of waterbird community factors in future HPAI surveillance and prediction.


Introduction
Emerging infectious diseases are one of the most critical challenges to humanities. 47 The pandemic of COVID-19 has attracted the world's great attention and medical 48 resources, but shadows over many others in recent years. Over the last two decades, 49 outbreaks of highly pathogenic avian influenza (HPAI), mostly caused by H5 viruses, 50 have occurred frequently in European countries (Verhagen et al. 2021). Europe has   In this paper, we investigate the spatial patterns of H5N8 outbreaks during the 82 2016/17 and 2020/21 epidemics in Europe. We applied waterbird census data from 83 about 7,000 sites across Europe to 1) explore and compare the risk factors associated 84 with H5N8 occurrence in wild birds and poultry, and between the 2016/17 and 85 2020/21 epidemics; and 2) determine the importance of waterbird community factors 86 in predicting H5N8 spatial patterns.

88
The H5N8 data collected from FAO (see Methods) included 1,630 outbreak cases 89 from the 2016/17 epidemic with 778 cases in wild birds and 852 in poultry ( Figure   90 1A), and included 1,581 outbreak cases reported from the 2020/21 epidemic with 743 91 cases in wild birds and 838 cases in poultry ( Figure 1B).

92
Risk factors for H5N8 occurrence 93 We conducted univariate analyses to explore risk factors for H5N8 occurrence in the  Relative importance of community predictors in explaining H5N8 patterns 113 We then constructed variance partitioning analyses to compare relative importance of

141
Our analyses demonstrate that waterbird community composition was strongly 142 associated with the spatial patterns of HPAI H5N8 occurrence in wild birds.     while the maximum number for any waterbird species was used for the sites with 251 more than one years' data. We also excluded rare species-those occurring in < 10 252 sites or with counts < 20 birds. These approaches have been widely used in many 253 studies on waterbird community analyses.

254
Matching waterbird community data to H5N8 outbreak sites 255 As bird surveillance sites usually are not matched exactly with the H5N8 outbreak 256 sites, we here applied a cross-selection procedure to solve this spatial mismatch and 257 13 generate disease "presences" and "pseudo absences" for further analyses. To achieve 258 this, we first searched all surveillance sites around each outbreak site within a 10-km 259 radius, which is related to surveillance zones for HPAI outbreaks in EU countries Agency CCI 300-m annual global land cover products) as the H5N8 outbreak site. We 272 used all matched presences to construct models but also tested the robustness by using 273 data excluding the poor matches. 274 We then assigned surveillance sites sufficiently distant (> 60 km, three times the 275 diameter for HPAI surveillance zones) from any outbreak site as potential H5N8 276 absence sites. We tested the sensitivity of our results by changing this distance 277 threshold to 40 km and 80 km. We assumed the absences could represent a site 278 unsuitable for infection, i.e., a true absence. However, there were, for spreading 279 14 pathogens, also false absences resulting from a lack of surveillance and/or reporting, 280 or because the pathogen had not been introduced into the region (Phillips et al. 2009).

281
To limit the effect of these false absences, we also set a maximum distance to the  Statistical analyses 332 We used a logistic regression framework to link H5N8 presence and absence to the 333 predictors. As our data had low prevalence values (< 10%), we applied a 334 bootstrapping procedure with 1,000 repetitions, selecting all HPAI H5N8 presence 335 sites with an equivalent number of absence sites, to minimize the problem of data 336 imbalance. The absence sites were randomly selected under the condition that each 337 absence site was more than 20 km away from others, in order to avoid duplicated 338 absences. 339 We first performed univariate analyses for both the 2016/17 and 2020/21 epidemics to