Avian Influenza Warning Map

About the Project

The Bird Flu Radar (GUI) has been developed for EFSA as part of the project named 'Implementation of an Early Warning System for Highly Pathogenic Avian Influenza' (NP/EFSA/BIOHAW/2022/01).

The purpose of this project was to develop a spatiotemporal risk assessment to estimate the probability of HPAI outbreaks in the wild bird population by spatiotemporal unit across Europe.

The spatiotemporal risk assessment model is now updated in real time with new results being displayed every Monday morning based on the latest HPAI outbreaks (H5 and H7) reported in wild birds in Europe. Results are displayed for the current week and predictions for the forthcoming three weeks. These predictions should be treated with the appropriate amount of caution.

Please note that the model and this GUI will be updated and improved during the course of the project. In particular the key next steps are:

December 2025 - Final model validation

Important notes: Delay in outbreak reporting has a major impact on the quality of the model prediction. To take into account the outbreaks occurring several weeks ago but only reported within the last week, the model is re-run every week for the last 16 weeks. In practice, this means that model outputs are updated every week not only for the week to come but also for the last 16 weeks. In addition, at this stage, Cyprus and Türkiye are not included yet in the model calculations because of lack of some input data used in the model.

Project members:

Sovon, the Dutch Centre for Field Ornithology coordinates the project.
The Catalan Ornithological Institute (ICO) used citizen data on abundance and distribution of wild birds available in the EuroBirdPortal (EBP) to estimate the expected relative abundance of 12 wild bird species across Europe.
The British Trust for Ornithology (BTO) used data from EURING DataBank to provide estimates for long-distance movements of migratory populations and short-distance dispersal models of the same 12 wild bird species in Europe.
EpiMundi developed the spatiotemporal risk assessment model using as input the data provided by ICO and BTO.

Spatiotemporal risk assessment model description
In this model, the probability of an HPAI outbreak (Pintro) is defined as the probability that HPAI virus enters a given area via wild bird movements (probability of entry) and is able to be transmitted to at least one other wild bird in the area under consideration (probability of establishment).

The risk factors considered in the model include:

Wild bird order and species
Number of migratory birds
Migration patterns
Notification of avian influenza outbreaks at origin

The 12 wild bird species included in the model are: Common teal (Anas crecca), Eurasian wigeon (Mareca penelope), Mallard (Anas platyrhynchos), Greater White-fronted Goose (Anser albifons), Greylag Goose (Anser anser), Pink-footed Goose (Anser brachyrhynchus), Bean Goose (Anser fabalis), Common Pochard (Aythya ferina), Tufted Duck (Aythya fuligula), Canada Goose (Branta canadensis), Whooper Swan (Cygnus cygnus), Mute Swan (Cygnus olor).

These 12 wild bird species account for 89% of European ring recovery data for waterfowl.

The spatial scale of the final model was defined based on the EEA 50x50km grid cell. The temporal scale of the final model was defined as a week of a year using as reference the weeks used by EBP.

The model sensitivity and specificity were assessed by comparing the value of the probability of an HPAI outbreak in a given spatiotemporal unit with the actual HPAI outbreaks reported in ADIS.

A full description of the model is available in the EFSA External Scientific Report published in 2022 (see reference below), but changes made to the model compared to this initial model description are briefly summarised below.

Model update January 2022 - the probability of HPAI introduction in the grid cell i at time t takes now better into account what happened in that same grid cell at time t-1, t-2 and t-3. The new optimal cutoff value is Pintro = 0.04 and leads to a sensitivity and specificity of 81%. Model update August 2023 - The EU Animal Disease Information System (ADIS) is now used as a source for information on HPAI outbreaks for all countries but the United Kingdom (UK)

Model update September 2024: the performance of the model has been reassessed after 1.5 year working on real time. The optimal cutoff value remains stable (Pintro =0.04) but leads to lower sensitivity and specificity, 73% respectively.

Model update September 2025: predictions of Pintro are now available not only for the current but also for the following three weeks.

Input Data
Four main sources of data have been used as input for the model:

Expected abundance of the 12 wild bird species across Europe (output of ICO's model)
Long-distance movement of migratory populations across Europe (output of BTO's model)
Short-distance movement of migratory populations across Europe (output of BTO's model)
Reported outbreaks of HPAI (H5 and H7) in wild bird populations (ADIS for all countries but UK, while UK HPAI outbreaks are extracted from EMPRES-i)

References:

EFSA External Scientific Report 2022 - Development of a prototype early warning system for avian influenza in the EU based on risk-mapping
Short training video explaining how the model works

The information on this page was produced and adopted by Sovon, ICO, BTO, EuroBirdPortal and EpiMundi in the context of a contract with The European Food Safety Authority (EFSA), awarded following a public procurement procedure. It aims at informing stakeholders on the probability of highly pathogenic avian influenza outbreaks occurring in the EU based on spatiotemporal risk models. The information reflects the state of knowlege available on the date of dispatch. EFSA cannot be held liable for any errors, inaccuracies or inconsistencies with regard to text and/or data contained therein. Therefore, EFSA accepts no responsibility or liability arising out of, or in connection with the information provided. The present document is not to be considered an output adopted by EFSA.

Loading map...

Key

Category SPECIAL:

Probability of Introduction (Pintro):

Category 10:

Category 9:

Category 8:

Category 7:

Category 6:

Category 5:

Category 4:

Category 3:

Category 2:

Category 1:

Category 0:

- - - - Optimal Threshold

Sensitivity:

Category sens 1:

Category sens 0:

Pintro (Poultry):

Category sens 10:

Category sens 9:

Category sens 8:

Category sens 7:

Category sens 6:

Category sens 5:

Category sens 4:

Category sens 3:

Category sens 2:

Category sens 1:

About the Project

Loading map...

Loading key...

Bird Flu Radar Subscription Service

Impact Survey

Filtering tabular data

Movements by Month/Season:

Species/Group selection:

Initial focal area:

Destination focal area (movements by month only):

Season of ringing (movements by month only):

Filter data

Species/Group selection:

EBP Data:

Season of ringing:

Focal area:

Select Layers

EBP data:

Ringing data:

Focal area:

Error occurred: