West Nile virus (WNV), a mosquito-borne pathogen, poses a growing public health threat in Europe. While climate change is suspected as a driver of its spread, a formal evaluation of this causal relationship is lacking. Anthropogenic climate change significantly impacts ecosystems globally, affecting biodiversity distribution and impacting human well-being. Changes in climate influence the distribution and phenology of disease vectors like mosquitoes and ticks, impacting pathogen transmission. Climatic variables (temperature, precipitation, humidity) affect mosquito habitat suitability, distribution, abundance, and biting behavior, influencing transmission rates and the duration of the extrinsic incubation period. However, the extent to which climate change drives observed changes in mosquito-borne disease distribution remains poorly understood. WNV, maintained in a bird-mosquito cycle primarily involving *Culex pipiens*, has spread across Europe since the 1950s, with major outbreaks occurring since 1996. The virus exhibits high genetic diversity, with WNV-2 being the most widespread lineage in Europe. While WNV circulation is linked to temperature, drought, and winter conditions, the impact of long-term climate change on human infections hasn't been quantified. Other factors, like land use changes (irrigated croplands, fragmented forests), the presence of standing water, and biodiversity loss, could also influence WNV transmission. This study addresses this gap by using the IPCC framework to investigate the contribution of climate change to WNV spatial expansion in Europe. It leverages observational reanalysis climate datasets and counterfactuals from ISIMIP, which removes long-term climate trends, allowing for the isolation of climate change's effect.

Existing literature establishes a link between WNV occurrence and high temperatures in spring and summer, summer droughts, and warm winters. Studies have shown that high spring and summer temperatures, lower water availability, and drier winter conditions are key determinants of WNV occurrence in Europe. The impact of climate change on WNV transmission has been explored, with studies highlighting the role of rising temperatures and altered precipitation patterns. However, a comprehensive quantitative assessment of the contribution of long-term climate change, accounting for other factors like land use and population changes, has been lacking. Previous research has also highlighted the roles of irrigated croplands and fragmented forests in favoring WNV outbreaks, as well as the importance of standing water bodies and the complex interplay between avian biodiversity and mosquito infection rates.

This study employs ecological niche modeling using a boosted regression tree (BRT) approach to estimate WNV ecological suitability (risk of WNV circulation). The analysis is conducted at the NUTS 3 level (third administrative level in European countries). Data included 13 years (2007-2019) of confirmed WNV human infections from the ECDC database, along with climatic, land-use, and human population data (2000-2019) from four ISIMIP3a reanalysis datasets (GSWP3-W5E5, 20CRv3, 20CRv3-ERA5, and 20CRv3-W5E5). Models were trained on present-day data and projected onto historical (1901-2019) conditions, using both factual and counterfactual climate data. The counterfactual data removed long-term climate trends, creating a 'no-climate-change' scenario. Model performance was assessed using AUC and a prevalence-pseudoabsence-calibrated Sørensen's index (Slppc). The relative influence (RI) of each environmental factor was also determined. Finally, the study compared the evolution of the estimated population at risk of exposure to WNV under factual and counterfactual climates, using two ecological suitability thresholds (0.1 and 0.5). An optimized NUTS map was used to homogenize polygon size across different European countries. The response variable used was presence/absence of at least one confirmed, non-imported human case, rather than incidence data, to account for variations in surveillance effort.

Ecological niche models showed good predictive performance (AUC and Slppc > 0.8). Summer near-surface air temperature had the highest relative influence on WNV ecological suitability in the GSWP3-W5E5 model, followed by winter air temperature, winter relative humidity, summer precipitation, and fall relative humidity. Among land-use factors, managed pastures and rangeland had the highest influence. Simulations using the GSWP3-W5E5 dataset showed a clear increase in WNV ecological suitability in several European regions (northern Italy, Carpathian mountains, lowlands in Hungary and eastern Romania, Aegean region) since the 1980s under factual climate conditions. This increase was absent in the counterfactual simulations. Similar trends were observed in 20CRv3-W5E5 and 20CRv3-ERA5 datasets, but not in 20CRv3. The population at risk of exposure almost doubled (threshold 0.1) or increased sixfold (threshold 0.5) in the factual GSWP3-W5E5 scenario compared to the counterfactual, demonstrating the significant contribution of climate change.

The study's findings demonstrate a strong link between climate change and the increased risk of WNV circulation in Europe. The observed increase in ecologically suitable areas for WNV since the 1980s coincides with rapid warming in Europe and the establishment of WNV hotspots in various countries. While previous studies highlighted the importance of spring temperatures, this study found that summer and winter temperatures had a greater influence on WNV ecological suitability in the models. This discrepancy might be attributed to the use of presence/absence data as the response variable compared to other studies that used incidence. The study also found that cropland density positively correlated with WNV ecological suitability, while managed pastures and rangelands might also play a role. The study acknowledges limitations such as the exclusion of biotic factors like mosquito and bird diversity/abundance due to data limitations, the use of a simplified set of land-use variables, and the use of presence/absence data instead of incidence data to account for surveillance heterogeneity. However, the study's approach using counterfactual climate data from ISIMIP provides strong evidence for climate change's significant contribution to WNV spatial expansion.

This study provides compelling evidence for the substantial contribution of climate change to the increased risk and spatial expansion of West Nile virus in Europe. The findings highlight the importance of incorporating climate change projections into future surveillance and intervention strategies. Future research should investigate the impact of future climate change scenarios on WNV distribution and transmission dynamics, incorporating additional factors like vector and bird population data and more detailed land-use information, for refined risk assessments. The methodology presented offers a valuable framework for integrating climate data in epidemiological studies, furthering the understanding of climate change impacts on infectious disease.

This study has several limitations. The inability to incorporate biotic factors like mosquito and bird diversity/abundance into the models is a significant constraint. The simplified set of land-use variables may have limited the accuracy of the ecological niche model. The use of presence/absence data instead of incidence data, chosen to account for surveillance heterogeneity, could influence results. Attribution is to climate change in general (irrespective of cause) using detrended reanalyses, rather than specifically to anthropogenic climate forcings, although this is strongly implied. Finally, using WNV occurrence data up to 2019 may not fully capture recent expansion into areas not highlighted as high-risk by the models.