Spatio-temporal dataset (2009–2012) of Culicoides spp., vectors of livestock viruses, in France

Abstract

Culicoides biting midges (Diptera: Ceratopogonidae) are vectors of pathogens of veterinary and public health importance, including bluetongue, epizootic hemorrhagic disease, and Oropouche viruses. Following multiple bluetongue virus serotype incursions in France (2000–2008), a nationwide entomological surveillance network (2009–2012) was launched to support livestock health policies and fulfill European Commission requirements. Sampling was conducted using 160 traps deployed across all mainland French départements (French administrative division), excluding three in the Ile-de-France region (Paris, Seine-Saint-Denis and Val-de-Marne). Trapping was synchronized nationally with a weekly frequency during population fluctuations and monthly during known periods of activity/inactivity. Midges were identified to species, sexed, and categorized by physiological status. The network generated the most comprehensive dataset on Culicoides in France, with unprecedented spatiotemporal coverage. Of more than 6.34 million specimens collected, representing at least 83 species, 66% of identified specimens belonged to the Culicoides obsoletus s.l./Culicoides scoticus complex. This unique dataset provides critical insights into species diversity, phenology and ecology, and supports modeling, surveillance, and control of vector-borne diseases at the national scale.

Background

Culicoides biting midges (Diptera: Ceratopogonidae) are small hematophagous insects recognized as vectors of numerous pathogens, including bluetongue virus (BTV), Schmallenberg virus (SBV), epizootic hemorrhagic disease virus (EHDV), African horse sickness (AHSV) and Oropouche virus (OROV)¹. Most Culicoides-borne pathogens pose serious threats to livestock health and agricultural economies, with the exception of OROV which impacts public health^2–6. In the 2000s, a series of unprecedented outbreaks involving several serotypes of Culicoides-borne viruses caused major economic losses and disrupted ruminant production systems across Europe. These outbreaks highlighted critical knowledge gaps on the distribution, seasonal dynamics, and potential role of European Culicoides species in disease transmission. In response, several countries launched national surveillance programs to address these gaps and support improved disease prevention and control strategies.

France experienced its first incursion of BTV in October 2000, when serotype 2 (BTV-2) was detected in Corsica, resulting in 39 outbreaks in 2000 and 335 in 2001^7–9. This marked the beginning of an intensified national focus on vector-borne disease surveillance. In 2002, the French Ministry of Agriculture and Fisheries tasked Cirad (Centre de coopération internationale en recherche agronomique pour le développement) with coordinating a national entomological monitoring network for Culicoides biting midges, the primary vectors of BTV¹⁰. The initial focus was on Culicoides imicola, Kieffer, 1913, a long-known and major vector of BTV and AHSV in Africa, which had been detected in Corsica one month prior to the outbreaks. Sentinel traps were also deployed along the Mediterranean coast to assess its potential establishment on the mainland.

Surveillance activities were progressively expanded in response to the emergence of additional BTV serotypes. BTV-4 was detected in Corsica in 2003 (17 outbreaks), followed by BTV-16 in 2004 (25 outbreaks). In 2006, BTV-8 unexpectedly emerged in northern Europe, including areas where C. imicola was absent. This strain rapidly spread to neighboring countries and reached France, triggering major outbreaks. In 2007, mainland France also experienced a BTV-1 incursion, contributing to over 50,000 outbreaks caused by BTV-1 and BTV-8 between 2006 and 2009^8,11,12. Entomological surveillance was extended to newly affected areas, including the Pyrénées-Atlantiques (2005), northeastern départements (2006), central France (2007), and Brittany (2008). As bluetongue continued to spread across Europe⁸, the European Commission adopted Regulation 2007/1266/EC, mandating harmonized surveillance efforts across member states. These included comprehensive surveillance of bluetongue disease and entomological monitoring to demonstrate the absence of specific serotypes, detect early virus emergence, track disease progression, and define vector-free periods¹³ during which movement restrictions on live ruminants could be lifted.

In line with these regulatory requirements, France established a nationwide entomological surveillance network from 2009 to 2012. A total of 160 light suction traps (Onderstepoort model) were deployed across mainland France and Corsica, with one or two traps per département. Trapping was conducted weekly during spring and autumn, and monthly during the rest of the year. The primary objectives of the network were to document the diversity of Culicoides species, monitor their seasonal dynamics and geographic distribution, and determine the onset and cessation of vector activity (phenology). While directly supporting regulatory decisions regarding vector-free periods, the network also generated an invaluable dataset, comprising millions of specimens collected across a wide range of ecoclimatic zones in all seasons over four successive years. Finally, this network contributed to raise awareness among farmers and veterinarians on Culicoides and Culicoides-borne diseases.

The dataset described in this paper originates from this 2009–2012 entomological surveillance network. It consists of three interconnected datasets documenting (i) sampling events, (ii) Culicoides biting midge occurrences and abundance per species and per physiological state, and (iii) environmental variables associated with each trapping event. The spatial and temporal resolution of this dataset is unmatched at the national scale, offering a comprehensive resource for the study of Culicoides ecology and phenology in France.

Since its establishment, this dataset has been instrumental in addressing a range of scientific questions, from the dispersal capacity and cryptic diversity of vector species^14–16, to the spatial and temporal modeling of Culicoides midges abundance using genetic data¹⁷, statistical methods and machine learning approaches^18–20. The dataset has also supported investigations into vector-virus associations and the epidemiological dynamics of disease outbreaks^21,22, notably during the emergence of SBV. In addition, it has informed risk assessments concerning the introduction of exotic diseases such as AHSV and EHDV via vector dispersal or animal trade²³. Beyond its applications in disease ecology, the dataset has contributed to methodological advances in species identification and vector surveillance tools²⁴.

The synchronized, nationwide surveillance system from which this dataset originates provided a unique opportunity to investigate Culicoides distribution and phenology across diverse ecoclimatic regions of France. With its exceptional spatial and temporal resolution, the dataset captures detailed information on species composition, seasonal activity, and environmental conditions. This comprehensive coverage allows for robust analyses of population trends, community dynamics, and ecological drivers of vector activity. As such, the dataset is a valuable resource for entomologists, epidemiologists, and policymakers involved in vector-borne disease surveillance. It has already advanced our understanding of major livestock disease vectors such as BTV, EHDV, and SBV, and holds considerable promise for enhancing predictive models in the face of climate change and emerging disease threats.

This dataset played an important role in refining control measures, including vaccination campaigns, natural immunity, and movement restrictions. As a result, BTV transmission gradually declined in France, with only 83 outbreaks reported in 2009 and just one in 2010. By December 2012, BTV circulation had ceased, and mainland France regained its bluetongue-free status. The national entomological surveillance network was subsequently discontinued, though the data it produced continue to support research and inform policy well beyond its operational timeframe.

Methods

Between 2009 and 2012, Culicoides midge population monitoring in mainland France and Corsica was conducted by national authorities and research groups in compliance with European requirements. The surveillance network deployed 160 light suction traps (Onderstepoort model) with one or two traps allocated per administrative division (french départements). We used the Petites Régions Agricoles classification²⁵ [results of the intersection of agricultural regions (historically defined zones of homogeneous agricultural productions) and départements] to rationalize the location of collection sites, which were selected to sample the agricultural regions of interest for animal productions. Overall, this approach ensured the coverage of the whole territory with the constraint of at least one trap per département. Final choice of the collection site was done, for each département, with the Directions départementales de la cohésion sociale et de la protection des populations (DDcsPP), i.e. deconcentrated veterinary services.

The monitoring program targeted farms housing mostly cattle (82%), commonly sheep (25%), rarely goats (11%) and exceptionally horses (<1%). It should be noted that some sampling sites left the monitoring network during the program, and were replaced by a similar farm nearby in the same département. As a result, over the four years of monitoring the 160 traps were distributed across a total of 210 farms.

Sampling frequency and duration

The sampling schedule varied by season. From mid-February to April and November to mid-December, traps were set one night per week, whereas a monthly schedule was adopted for the remainder of the year (In January, and between May to October). In total, 14,895 collections were performed under the supervision of the DDcsPP.

Sample collection and preservation

The traps used for the monitoring were black-light suction traps, also known as OVI traps (Onderstepoort Veterinary Institute, Pretoria, South Africa). These traps are known to capture higher numbers of Culicoides midges compared to other trap types^26,27. Light tubes were replaced every year.

The traps were positioned as near as possible to the animal resting areas at night (outside or inside stables - in the latter case, the degree of building opening was recorded). The position of each trap was permanent during the whole survey and GPS coordinates were recorded at the installation. Traps were operated from dusk to dawn. In these traps, insects are attracted by the black light of the trap and collected using a fan, which directs them into a beaker containing soapy water. The soap allows the insects to sink and prevents them from drying out. A thin net, with mesh size of 1.2 millimeters, was placed around the trap light to prevent the collection of large insects. Once collected, the specimens were transferred to 70% ethanol at room temperature for storage and transport to identification centers.

Specimen identification and processing

The samples were processed at three specialized sorting centers: Cirad (Centre de coopération internationale en recherche agronomique pour le développement) and EID-Med (Entente interdépartementale pour la démoustication du littoral méditerranéen) in Montpellier, and IPPTS (Institut de parasitologie et de pathologie tropicale) in Strasbourg. Expert entomologists performed species identification using morphological keys and the IIKC database^28,29. Morphological identification relied on key features such as sensory pit shape and size, wing spot patterns, and the number and morphology of spermathecae in females or the aedeagus and parameres in males. Damaged or incomplete specimens were excluded. Non-Culicoides specimens were not identified.

Identification was conducted to the species level or, where differentiation was not possible, to the complex level. Indeed, due to the difficulties to reliably separate some species or the existence of cryptic diversity, some species were grouped together at a complex level: i) C. cataneii Clastrier, 1957 and C. gejgelensis Dzhafarov, 1964, ii) C. obsoletus (Meigen), 1818 and C. scoticus Downes and Kettle, 1952, and iii) C. sejfadinei Dzhafarov, 1958 and C. tauricus Gutsevich, 1959. Note that the authors of a species are indicated in brackets when this species has previously been placed in another genus.

In cases of excessively large samples, subsampling was performed to streamline the identification process, following a protocol adapted from Van Ark and Meiswinkel (1992)^21,30. Specifically, if the total insect volume exceeded 3 mL, a subsample was taken, offering significant time savings. In some cases (in particular when the trap screen was damaged or misplaced allowing large insects to enter), before the subsampling, a preliminary sorting step was conducted under a stereomicroscope to isolate Culicoides midges from other insect species based on morphological traits. The subsampling details are recorded in the measurement or facts extension (Table 1). In total 912 samples were subsampled. Certain specimens were slide-mounted for quality control and identification training purposes.

Table 1.

Descriptive summary of variables included in the measurement or facts extension.

measurementID	Variable description	Range of values	Source
ALT	Global 30 Arc-Second Elevation Data in meters (GTOPO30)	[1; 1395]	E-OBS40,41
BETAIL	Density of horse, cattle, sheep and goat in the canton (ind/ha)	[1; 2.76]	2010 agricultural census42
BOV	Density of cattle in the canton (ind/ha)	[0; 2.67]
CAP	Density of goat in the canton (ind/ha)	[0; 0.31]
OVI	Density of sheep in the canton (ind/ha)	[0; 1.47]
EQU	Density of horse in the canton (ind/ha)	[0; 0.07]
SURF_CANT	Surface of the canton (ha)	[725; 97440]
CLC	Thematic classes of land cover and land use	15 categories	CORINE Land Cover 2018 (vector/raster 100 m), Europe, 6-yearly43
ECO_CLI	Biogeographical regions	4 categories: Continental, Atlantic, Mediterranean, and Alpine	European Environment Agency44
EVI	The enhanced vegetation index	[−0.10; 0.96]	MOD13Q1 v061 from MODIS32,45
		−3000 values are to be considered as missing values
NDVI	Normalized Difference Vegetation Index	[−0.07; 0.88]
		−3000 values are to be considered as missing values
PP	Photoperiod / daylength (hours)	[7.94; 16.52]	Meteor package46,47
FG_ERA5	Daily mean wind speed (m s−1) calculated from 10m u-component of wind (windu) and 10m v-component of wind (windv) $\sqrt{{\mathit{wind}}_u^2+{\mathit{wind}}_v^2}$	[0.01; 13.44]	ERA5-Land hourly data from 1950 to present31
HU_mean_ERA5	Daily averaged 2 m dewpoint temperature. A measure of the humidity of the air (°C)	[−20.39; 21.56]
QQ_ERA5	Daily average surface net solar radiation (J m-2)	[197473; 28013636]
RR_ERA5	Daily precipitation sum (mm)	[0; 68]
TP_mean_ERA5	Daily mean2 m temperatures (°C)	[−15; 29]
TP_min_ERA5	Daily min 2 m temperatures (°C)	[−23; 25]
TP_max_ERA5	Daily max 2 m temperatures (°C)	[−10; 37]
swvl1_mean	Daily average volumetric water in the soil in layer 1 (0–7 cm) of the ECMWF Integrated Forecasting System (m3 m-3)	[0.08; 0.69]
TPsoil_mean	Daily average temperature of the soil in layer 1 (0–7 cm) of the ECMWF Integrated Forecasting System (°C)	[−3.81; 30.24]
BUILD_OP	Building opening rate (%)	4 categories: 0–5, 5–25, 25–75, and 75–100	Communication for operators
SITUATIONPIEGE	Trap position	2 categories: inside, and outside
VENTDEBUT / VENTFIN	Wind at trap installation / collection	4 categories: no wind, slight breeze to light wind, medium wind, strong wind
NUAGEDEBUT / NUAGEFIN	Cloud at trap installation / collection	4 categories: no cloud cover, low cloud cover, medium cloud cover, high cloud cover
PLUIEDEBUT / PLUIEFIN	Rain at trap installation / collection	4 categories: no rain, fog / drizzle, light rain, heavy rain
SOUSECH	Subsampling protocol	2 categories: no, yes	Communication from the entomological expert in charge of identification
VTOT	In case of subsampling, volume total before subsampling (mL)	[3.4; 322]
		0 must be considered as NA
VTRAIT	In case of subsampling, treated volume during subsampling (mL)	[1; 6]
		0 must be considered as NA
PRETRI	Pre-sorting of Culicoides genus individuals before sub-sampling	2 categories: no, yes

Complementary measurements and facts

For each sampling campaign, additional information was recorded, either by on-site measurements or by further processing. On-site measurements included qualitative variables, i.e. trap position (inside or outside the stable) and building opening rate (recorded at the time of the first observation on the farm), and semi-quantitative ones, i.e. wind/cloud/rain at trap installation and collection, as well as quantitative ones: date and time of trap installation and collection. The operator also had the opportunity to leave textual/non-formalized comments for each event. A posteriori measurement aimed at describing the environment of the trap and most of them were retrieved using GIS approaches intersecting exact GPS coordinates with spatial dataset collected from various sources such as ERA-5 land³¹ or MODIS³². Those data include elevation, livestock densities in the canton, land cover classes, biogeographical region, vegetation indexes, daily wind speed, daily temperatures (air and soil), daylight duration, surface net short-wave radiation flux, daily precipitation sum, and volume of water in soil layer 1 (0–7 cm). The data were restructured using R software to meet GBIF publication criteria.

Data Records

The database is openly accessible through the Global Biodiversity Information Facility (GBIF)³³, where it can be downloaded as a Darwin Core Archive (DwC-A)³⁴. It is available under a CC BY 4.0 license. The dataset was standardized to the Darwin Core structure as sampling-event data, and fit to the FAIR principles^35,36, ensuring its findability, accessibility, interoperability, and reusability. It is presented as a set of files in tab-delimited txt format.

This sampling-event dataset comprises three interrelated tables based on the DarwinCore (DwC) standard³⁷: the event core (‘event.txt’), the occurrence extension (‘occurence.txt’), and the measurement or facts extension (‘measurementorfacts.txt’), respectively designed to capture data related to trapping events, Culicoides midges identification, and complementary measurements or facts for each event, such as environmental data. These tables are linked through the primary key (‘eventID’) defining the relational structure of the database and facilitating analyses of insect populations across temporal, spatial, and ecological dimensions by relating them to sampling conditions.

The event core contains 14,895 sampling events carried out between 2009 and 2012 across 210 farms. The occurrence extension contains 8,683,785 records—8,520,893 of which report no individuals (zero counts), while 162,892 report at least one individual. In total, these records describe 6,340,177 Culicoides specimens. The measurement or facts extension contains 491,535 event-associated measurements from 33 parameters (Table 1), with value distributions illustrated in Fig. 1. All occurrence records are arthropod insects belonging to the Diptera order, the Ceratopogonidae family, and the Culicoides genus.

Fig. 1.

Distribution of values of quantitative variables among observations/events, for more information about variables, see Table 1.

The database structure is illustrated in Fig. 2. Field names were also based on DwC terminology and their meaning (i.e. data dictionary) is available in the List of Darwin Core terms³⁶. It includes: ‘eventID’, ‘eventType’, ‘eventDate’, ‘startDayOfYear’, ‘endDayOfYear’, ‘year’, ‘month’, ‘samplingProtocol’, ‘samplingEffort’, ‘eventRemarks’, ‘decimalLatitude’, ‘decimalLongitude’, ‘geodeticDatum’, ‘coordinatePrecision’, ‘coordinateUncertaintyInMeters’, ‘country’, ‘countryCode’, ‘locationID’, ‘municipality’, ‘county’, ‘stateProvince’, ‘habitat’, ‘type’, ‘modified’, ‘language’, ‘license’, ‘ownerInstitutionCode’, ‘week’, ‘occurrenceID’, ‘basisOfRecord’, ‘individualCount’, ‘organismQuantity’, ‘organismQuantityType’, ‘sex’, ‘reproductiveCondition’, ‘occurrenceStatus’, ‘scientificName’, ‘kingdom’, ‘phylum’, ‘class’, ‘order’, ‘family’, ‘taxonRank’, ‘institutionID’, ‘superfamily’, ‘tribe’, ‘scientificNameAuthorship’, ‘measurementID’, ‘measurementType’, ‘measurementValue’, ‘measurementUnit’, ‘measurementDeterminedBy’.

Fig. 2.

Entity-relationship diagram of the dataset. PK denotes the primary key. For more details about the field names, see the Darwin Core Quick Reference Guide³⁶.

The geographical coverage of the database spans mainland France (including Corsica Island), extending from 41.5°N to 50.8°N in latitude, from 4.51°O to 9.5°E in longitude. Over the 210 sampled farms, the majority were cattle farms (65.2%, n = 137), followed by mixed livestock farms (21.0%, n = 44). During the 2009–2012 study period, a total of 168, 171, 166, and 162 traps were deployed for each year respectively, with weekly trapping nights during spring and autumn, and a monthly for the remainder of the year (Fig. 3). The light trap locations covered a range of biogeographic regions, most sites falling predominantly within the Atlantic zone (Fig. 4).

Fig. 3.

Trap collection schedule for the 2009–2012 study period.

Fig. 4.

Geographic distribution of Culicoides trapping sites across mainland France from 2009 to 2012. Trap locations are overlaid on biogeographic regions, with French département boundaries outlined for reference.

Data Overview

In total, more than 80 Culicoides species were recorded, with 92.9% of collected specimens belonging to only seven species/species complexes: C. pulicaris (Linnaeus), 1758 (1.6%), C. punctatus (Meigen), 1804 (2.2%), C. newsteadi Austen, 1921 (2.6%), C. chiopterus (Meigen), 1830 (2.8%), C. imicola (8.2%), C. dewulfi Goetghebuer, 1936 (9.4%) and the C. obsoletus s.l./C. scoticus complex, which alone represented an overarching 66% of all captured individuals.

Notably, C. obsoletus s.l./C. scoticus complex remained dominant across the south-to-north gradient in the majority of mainland France areas except for the island of Corsica, where C. imicola accounted for most individuals captured (Fig. 5).

Fig. 5.

Spatial Culicoides population structure with seven most abundant Culicoides species across different administrative regions in France.

In addition, C. dewulfi displayed greater relative abundance in the north-western regions of France compared to other areas, while C. newsteadi was more prominent near the Mediterranean basin. Overall, despite the year-to-year variability, shifts in less common species’ relative abundance were marginal indicating consistent habitat preference.

Long-term observations reveal species-specific temporal trends with certain years displaying several abundance peaks within a single season. This can be seen for the C. obsoletus s.l./C. scoticus complex, as well as C. punctatus and C. chiopterus species, which presented multimodal abundance peaks during the corresponding 2011–2013 activity periods, whereas 2009–2011 showcased unimodal patterns (Fig. 6). Despite a general trend of higher trap capture counts during the warmer months, the timing of abundance peaks varied across years in the four biogeographic regions (not shown), highlighting distinct inter- and intra-species population dynamics.

Fig. 6.

Intra-annual variability of seven most abundant Culicoides species found in France for the 2009–2012 period.

The finer time-scale seasonal dynamics of the most abundant Culicoides species reveal alternating periods of increased abundance (Fig. 7). Each species showcased distinct phenology, with varying activity period length throughout the year. Species within the C. obsoletus s.l. and C. scoticus complex activity period were mostly confined to the warmer months, whereas C. imicola and C. dewulfi indicated delayed seasonal onset in relation to other species, with activity extending further into the autumn. A distinct pattern can be seen in C. newsteadi species, which present an almost continuous presence but highly variable abundance throughout the year. An overall decrease in abundance across all species can be seen during December/February months, which correspond to the cold season.

Fig. 7.

General trends of the most abundant Culicoides species seasonal dynamics observed during the 2009–2012 period. Horizontal black bars represent the median, the boxes represent the quartiles 1–3 of the monthly abundance data. Extreme values were excluded to better visualize the central trends across species and months.

The trapping sessions exhibited high variability, ranging from very low to exceptionally high numbers of specimens per event. In general, most traps collected few Culicoides spp., although a small number of sessions yielded extremely large catches (Fig. 8). Since one of the principal objectives of the year-round surveillance campaign included studying the species-specific phenology, trapping was maintained even during the winter months, when activity is low. This resulted in many trapping sessions with little or no captures.

Fig. 8.

Distribution of positive Culicoides spp. counts per trap from 2009 to 2012 in France. The values are represented on a logarithmic scale to reduce overdispersion.

The 2009–2012 Culicoides monitoring campaign uncovered a rich species diversity and heterogeneous population dynamics across the French mainland and Corsica. Long-term surveillance captured interannual changes in both population dynamics and abundance, highlighted by the varying timing and duration of Culicoides activity periods from one year to the next. The wide geographical range of investigated sites, combined with a weekly-to-monthly trap collection schedule, allowed to capture smaller-scale seasonal Culicoides spp. dynamics and species-specific patterns. Despite the large variability in specimen counts across different trapping sessions, global spatio-temporal trends can be observed due to the sheer number of data collected.

Technical Validation

All identifications included in the database were performed by entomological experts. The database was revised, and all authors carefully checked the complete dataset for potential technical errors or inconsistencies.

Regarding data management, several quality control procedures were implemented: consistency checks were carried out for dates, verification of geographic coordinates against the municipality of the trap location, and validation of the sum of individuals (ensuring that the total number matched the sum of males and females). Whenever errors were detected, they were corrected accordingly.

For species identification, all the technicians in charge of morphological identification were trained by an expert. Regular morphological identification workshops were organized to ensure standardization among the technicians and maintain a high level of expertise. A verification procedure was also put in place: a sample of specimens from each identification center was sent to the lead expert, who confirmed the identifications. This verification was carried out punctually and did not result in significant changes to the identifications initially made. Importantly, among the team of identifiers, there was a recognized expert in Culicoides taxonomy (Dr. Delécolle JC) who trained all personnel involved in identification and provided continuous training throughout the project.

Usage Notes

These data have already been used in several studies that could be investigated further, including (i) characterization of (1) the ecological niche of specific Culicoides species, (2) species communities and (3) associated vector activity in mainland France; (ii) assessment of the environmental factors influencing midge population distribution, dynamics, biting activity and epidemic risk; and (iii) spatial and temporal projection of these dynamics in various environmental and climatic contexts.

Author contributions

Hammami P., Baltusyte I. and Hélène Guis wrote the paper. Hammami P. collected the post-hoc data and formatted the dataset. Baltusyte I. and Taconet P. performed the descriptive presentation of the dataset. Delécolle J.C., Setier-Rio M.L., Mathieu B., Venail R., Balenghien T. and Garros C. supervised and/or carried out the entomological data collection. Balenghien T. and Garros C. managed the entomological surveillance network. All authors reviewed the paper.

Data availability

The data that support the findings of this study are openly available through the Global Biodiversity Information Facility (GBIF) at 10.15468/8vepgt.

Code availability

All analyses were conducted using R version 4.5.0³⁸. The code used to generate the figures presented in this study is publicly available on Zenodo: https://zenodo.org/records/15528313³⁹. Entomological data can be accessed via the GBIF repository (10.15468/8vepgt)³⁴. The spatial structure data are available in the “france” folder within the same Zenodo repository. The analyses relied on the following R packages: tidyverse (version 2.0.0), sf (version 1.0.19), scatterpie (version 0.2.4), and scales (version 1.3.0).

Competing interests

The authors declare no competing interests.