MethodologyMéthodologie

ClimateClimat 30 / 100

Five climate criteria are derived from Copernicus E-OBS v31 daily gridded observations (extended to 2025 with ERA5-Land) over a 2015-2025 window, each population-weighted across the department polygon (where people live counts more than empty mountains). Sunshine and rain comfort each blend two complementary signals before ranking, so no single count dominates: sunshine = 0.6 × total sunshine hours + 0.4 × sunny-day frequency, quantile-ranked across the 96 departments ascending (sunnier scores higher); rain comfort = 0.6 × rainy-day frequency + 0.4 × annual rainfall volume, quantile-ranked descending (wetter scores lower — so the dry-but-downpour Mediterranean isn't flattered on day-count alone, nor the gentle drizzly north over-penalised). Winter mildness is the quantile rank of the January mean low (warmer ascending). Heatwaves and mugginess are scored on absolute day-count curves (heatwave days 0→10 and 12→1; muggy days 0→10 and 25→1). All five are then neighbour-blended (70% the department's own value plus 30% a border-length-weighted average of its neighbours) so scores transition smoothly across borders rather than jumping. The two minority blend components are also surfaced on the page in their own right: sunny days (days a year the sunshine fraction — sunshine hours over daylight hours, so night is excluded — exceeds 50%, i.e. the day was more sunlit than not, complementing the total-hours sunshine figure) and annual rainfall (mm/yr, the summed daily-precipitation VOLUME — distinct from frequency: a department can have many light-rain days but a modest total, or few intense days but a high total). Two further intensity counts are display-only (never scored): heavy-rain days (daily precipitation at or above 20 mm) and torrential days (at or above 50 mm, the épisode-cévenol signature), which add intensity to the wet-day frequency that the rain count alone captures — a 2 mm drizzle and a 60 mm deluge are both "rain days" but only the deluge lands in these. Because E-OBS is an areal grid (~11 km cells), a daily cell total understates true point extremes, so the torrential count is a conservative lower bound. The section also shows a display-only day-by-day weather strip — a centered 7-day window (the past 3 days, today, and a 3-day forecast) drawn from the Open-Meteo / Météo-France AROME model. These are recent MODEL values and a short-range forecast, not official station records, and are not part of any score (CC-BY 4.0).

☀ SunshineEnsoleillement Quantile rankRang quantile weight 10poids 10

Quantile rank of annual sunshine hours across the 96 departments (more sun → 10). Neighbour-blended.Rang quantile des heures d'ensoleillement annuelles sur les 96 départements (plus de soleil → 10). Lissé par voisinage.

🌧 Rain comfortConfort pluie Quantile (descending)Quantile (décroissant) weight 6poids 6

Quantile rank of rainy-day count, descending (fewer rainy days → 10). Neighbour-blended.Rang quantile du nombre de jours de pluie, décroissant (moins de jours de pluie → 10). Lissé par voisinage.

🥵 HeatwavesCanicules Absolute curveCourbe absolue weight 6poids 6

Absolute curve on heatwave days (days in ≥3-day runs of Tmax ≥ 30 °C and Tmin ≥ 20 °C): 0 days → 10, 12 days → 1. Neighbour-blended.Courbe absolue sur les jours de canicule (jours en séries ≥ 3 jours de Tmax ≥ 30 °C et Tmin ≥ 20 °C) : 0 jour → 10, 12 jours → 1. Lissé par voisinage.

❄ Winter mildnessDouceur hivernale Quantile rankRang quantile weight 4poids 4

Quantile rank of the January mean daily low across the 96 departments (milder nights → 10). Neighbour-blended.Rang quantile de la température minimale moyenne de janvier sur les 96 départements (nuits plus douces → 10). Lissé par voisinage.

💧 MugginessLourdeur Absolute curveCourbe absolue weight 4poids 4

Absolute curve on muggy days (mean temp ≥ 22 °C and humidity ≥ 70 %): 0 days → 10, 25 days → 1. Neighbour-blended.Courbe absolue sur les jours lourds (temp. moyenne ≥ 22 °C et humidité ≥ 70 %) : 0 jour → 10, 25 jours → 1. Lissé par voisinage.

LifestyleCadre de vie 30 / 100

The affordability score is built from the median residential price per square metre, computed from DGFiP DVF (Demandes de Valeurs Foncières) sale records over 2023-2025; the three Alsace-Moselle departments (57/67/68), which the Livre-Foncier regime excludes from DVF, carry notary-derived estimates instead. It is scored on fixed absolute price thresholds — about €4,000/m² maps to 1 and about €800/m² maps to 10 — so cheaper departments score higher and a department's score does not shift just because prices move elsewhere. The taxe foncière rate (DGFiP local-tax dataset) and the low-to-high price band (the 10th and 90th DVF percentiles) are shown for context and are not part of the score.

The landscape score is 30% a curated editorial judgement of a department's scenic and natural quality + 70% a real geographic composite. The data 70% blends three measured components: land cover (CORINE Land Cover 5-poste per-commune statistics — share of natural land minus an artificialisation penalty, weight 0.45), relief (per-commune IGN altitudes — the department's peak elevation and its relief amplitude, weight 0.30), and protected areas (the share of the department covered by IGN BDTOPO parcs & réserves — national/regional parks and nature reserves, unioned in Lambert-93 so overlapping designations are not double-counted — plus a small heritage bonus for UNESCO World Heritage and Grands Sites de France, weight 0.25). Mountainous, heavily-protected departments (Hautes-Alpes, Savoie, Lozère, Corse) score high; dense lowland departments (Hauts-de-Seine, Val-de-Marne) score low.

Population counts and the annualised growth trend come from the INSEE census (inter-census annualised growth rate); density combines those counts with the IGN ADMIN EXPRESS department geometry (and GHSL where finer population weighting is needed); median living standard is from INSEE FiLoSoFi 2021. The socio-professional make-up — the share of the occupied active population (actifs occupés 15-64) in each of the six INSEE CS groups (farmers, artisans/shopkeepers, cadres & higher intellectual professions, intermediate professions, employees, manual workers) — comes from the INSEE Recensement de la population 2022 (table DS_RP_EMPLOI_LR_COMP, lieu de résidence), pop-weighted to the department. The headline is the cadres share ('office professionals' — the office/knowledge, remote-capable class), shown beside the télétravail share (the share working without commuting, from INSEE RP 2022 mobilités professionnelles, table DS_RP_NAVETTES_PRINC). The two are paired on purpose: télétravail conflates genuine office remote work with at-home trades (farmers and artisans whose workplace is their home), and across the 96 departments it correlates with farming (about +0.84) and negatively with the cadres/knowledge share — so the cadres % is the cleaner 'remote-capable' signal (a high télétravail share in a farming department mostly reflects farmers and artisans working on-site at home, not office remote work). The on-screen labels stay neutral; this interpretation lives here in the methodology. The six shares regroup into office / at-home-trades / on-site cohorts. INSEE publishes no télétravail-by-occupation cross-tab at department or commune level, so the panel never decomposes the télétravail figure by occupation — it reports the marginals and the cross-department regularity only. Two of these feed the scorecard: the population-trend score is set on fixed absolute thresholds (about −1.0%/yr maps to 1 and about +1.2%/yr maps to 10), and the tourism score is a quantile rank across the 96 departments of the department's total tourist bed-capacity (INSEE DS_TOUR_CAP, 2024 — a raw bed count, not overnight stays). Neither population-trend nor tourism is neighbour-blended. The socio-professional shares, the télétravail share and the median living standard are display-only context, never scored, weighted or filtered. Two deprivation stats sit beside the median living standard: the poverty rate (the share of people living below 60% of the national median standard of living, INSEE FiLoSoFi 2021, indicator TP60) and the localised unemployment rate (the share of the 15-64 active population out of work, INSEE Recensement de la population 2022). They move together (about +0.77 across the 96 departments) and are both display-only — except that the poverty rate is the single source for one scored criterion. social_cohesion (in the scorecard's Risk & Safety group) is an inverted quantile rank of the poverty rate across the 96 departments: more deprivation maps to a lower score, the same higher-is-better polarity as flood and fire risk. It was added in the 2026-06-15 scoring review as the one new axis that is statistically distinct from the criteria already scored — it correlates only about −0.03 with real-estate prices, so it carries information the affluence cluster (price, income, cadres share, which all move together around +0.7 to +0.85) misses, rather than re-counting wealth. social_cohesion ships at a default weight of 0: it is opt-in and changes no ranking unless a user raises its slider; the 100-point weight budget is unchanged. Unemployment itself is never scored — only the poverty rate feeds social_cohesion, and only as that inverted rank, never as the raw percentage.

🏡 Real estateImmobilier Absolute thresholdSeuil absolu weight 12poids 12

Absolute thresholds on the median €/m² (DVF): ~€4,000/m² → 1, ~€800/m² → 10 (cheaper → higher). Not neighbour-blended.Seuils absolus sur le prix médian €/m² (DVF) : ~4 000 €/m² → 1, ~800 €/m² → 10 (moins cher → plus haut). Non lissé.

🏔 LandscapePaysage 70% data + 30% editorial70 % données + 30 % éditorial weight 9poids 9

70% a measured geographic composite (CLC land cover 0.45 / IGN relief 0.30 / IGN protected areas 0.25, with a heritage bonus) + 30% an editorial judgement; rescaled to 1-10. Not neighbour-blended.70 % un composite géographique mesuré (occupation des sols CLC 0,45 / relief IGN 0,30 / aires protégées IGN 0,25, avec un bonus patrimonial) + 30 % un jugement éditorial ; ramené sur 1-10. Non lissé.

🧳 TourismTourisme Quantile rankRang quantile weight 6poids 6

Quantile rank of total tourist bed-capacity across the 96 departments (more beds → 10). Not neighbour-blended.Rang quantile de la capacité d'accueil touristique (lits) sur les 96 départements (plus de lits → 10). Non lissé.

📈 Population trendTendance démographique Absolute thresholdSeuil absolu weight 3poids 3

Absolute thresholds on the annualised growth rate: about −1.0%/yr → 1, about +1.2%/yr → 10. Not neighbour-blended.Seuils absolus sur le taux de croissance annualisé : environ −1,0 %/an → 1, environ +1,2 %/an → 10. Non lissé.

InfrastructureInfrastructure 13 / 100

Three infrastructure criteria, each scored on fixed absolute thresholds (not a rank against other departments). Healthcare uses the DREES APL accessibility index (2023; GP availability per inhabitant), where APL 2.0 maps to 1 and APL 5.0 to 10. Schools use the DEPP brevet pass rate (2025 session), where 78% maps to 1 and 95% to 10. Fiber uses the ARCEP FTTH coverage percentage (Q1 2026 observatory), where 70% maps to 1 and 100% to 10. Healthcare and fiber are then neighbour-blended (70% own value + 30% a border-length-weighted neighbour average); schools is NOT blended. The Base Permanente des Équipements (INSEE BPE, 2024) backs the amenity counts shown for context.

🏥 HealthcareSanté Absolute thresholdSeuil absolu weight 7poids 7

Absolute thresholds on the DREES APL accessibility index: APL 2.0 → 1, APL 5.0 → 10. Neighbour-blended.Seuils absolus sur l'indice d'accessibilité APL (DREES) : APL 2,0 → 1, APL 5,0 → 10. Lissé par voisinage.

🏫 SchoolsÉcoles Absolute thresholdSeuil absolu weight 4poids 4

Absolute thresholds on the DEPP brevet pass rate: 78% → 1, 95% → 10. Not neighbour-blended.Seuils absolus sur le taux de réussite au brevet (DEPP) : 78 % → 1, 95 % → 10. Non lissé.

📡 FiberFibre Absolute thresholdSeuil absolu weight 2poids 2

Absolute thresholds on ARCEP FTTH coverage: 70% → 1, 100% → 10. Neighbour-blended.Seuils absolus sur la couverture FTTH (ARCEP) : 70 % → 1, 100 % → 10. Lissé par voisinage.

TransportTransport 12 / 100

The transport score is infrastructure-primary rather than blended: a department's own high-speed-rail access dominates, with closeness to Paris as a lighter modifier. The formula is (7 if the department has a TGV/Intercités station else 2) + 3 × (1 − min(parisHours, 6.5) / 6.5), clamped to 1-10 — so a department with a TGV station scores 7-10 regardless of distance, one without scores 2-5. Inputs are a hand-maintained TGV-station flag and a manually-sampled shortest SNCF itinerary to Paris (2025 timetables); OpenStreetMap motorway references and the nearest commercial airport (OurAirports) are shown for context. Unlike the climate and most infrastructure scores, transport is NOT neighbour-blended.

🚆 Transport & connectivityTransport et connectivité FormulaFormule weight 12poids 12

(7 if the department has a TGV/Intercités station else 2) + 3 × (1 − min(parisHours, 6.5) / 6.5), clamped to 1-10 — a TGV department scores 7-10 regardless of distance. Not neighbour-blended.(7 si le département a une gare TGV/Intercités, sinon 2) + 3 × (1 − min(heuresParis, 6,5) / 6,5), borné à 1-10 — un département avec TGV obtient 7-10 quelle que soit la distance. Non lissé.

Risk & safetyRisque et sécurité 15 / 100

Flood risk measures the BREADTH of flood exposure, not its severity: the share of a department's communes flagged for flooding in the BRGM Géorisques GASPAR registry — the union of the risq risk register, PPRN flood-prevention plans, and AZI flood atlases. Fire risk is the fire-season (Jun-Sep) mean of the Copernicus CEMS Fire Weather Index (FWI, ERA5 reanalysis), area-weighted over the department — meteorological fire-WEATHER danger (how favourable the weather is to fire start and spread), NOT observed burns or fuel load, covering all of France uniformly. Both are scored by inverted quantile across the 96 departments, so the most-exposed departments score lowest. The 'Major risks present' breakdown, the seismic-exposure share, and the nuclear-nearby flag are all read per-commune from the same GASPAR risk register (lib_risque labels: Séisme, Feu de forêt, Nucléaire, Mouvement de terrain, Risque industriel, …), with the department's full commune count as the denominator. The 50 granular GASPAR labels are collapsed into 14 plain categories; each is shown as an absolute very-low→very-high exposure band (share of the department's communes listing it) plus the department's tie-aware rank versus the 96.

⚠ Natural riskRisque naturel Inverted quantile (avg)Quantile inversé (moy.) weight 8poids 8

round(avg(flood_risk, fire_risk)). Flood risk is the inverted quantile of the share of communes flood-exposed (BRGM GASPAR union); fire risk is the inverted quantile of the fire-season mean Fire Weather Index (Copernicus CEMS). Most-exposed → lowest. Not neighbour-blended.round(moy(risque_inondation, risque_incendie)). Le risque inondation est le quantile inversé de la part des communes exposées (union BRGM GASPAR) ; le risque incendie est le quantile inversé de la moyenne saisonnière de l'indice Fire Weather Index (Copernicus CEMS). Les plus exposés → les plus bas. Non lissé.

🛡 SafetySécurité Absolute thresholdSeuil absolu weight 7poids 7

Absolute thresholds on the recorded-crime rate (SSMSI, offences per 1,000 inhabitants), inverted so less crime → 10. Not neighbour-blended.Seuils absolus sur le taux de délinquance enregistrée (SSMSI, infractions pour 1 000 habitants), inversé pour que moins de crimes → 10. Non lissé.