QGIS Multi-Criteria Analysis: How I Found a Favourable Site in the Vercors

"Where should a sustainable tourist refuge be built in the Vercors Regional Nature Park?" Ask a hiker, and the answer is a field survey. Ask a geomatics engineer, and the answer starts with the data. In this article I walk through how I tackle this kind of problem — picking the best location for a facility — with a step-by-step multi-criteria analysis in QGIS, without setting foot on site. The case is real: a synthesis assessment from my Master's in Geomatics, but the method is exactly the one I run on assignments for local authorities and consultancies.

I run CODRUM, a geomatics and web development studio based in Morsang-sur-Orge (91), near Paris. The point of a multi-criteria analysis is not a pretty map: it is turning a hunch ("we should build over there") into a list of sites that are ranked, measurable and defensible in front of an elected official, a funder or a public inquiry commissioner. Here is the full funnel, with the real figures.

Why a multi-criteria analysis to choose a site?

Choosing a location means arbitrating between conflicting constraints: slope, sunlight, accessibility, distance to existing facilities, protected areas. A map alone does not decide; a multi-criteria analysis does — by stacking these layers to keep only the zones that satisfy every condition at once.

The principle is a funnel. At the start, the entire Vercors is a "candidate": 83 study municipalities, 51,105 inhabitants. At the end, 406 favourable sites of at least one hectare remain, with one municipality on top — Lans-en-Vercors, with 871 usable hectares. In between: seven successive filters, sixteen QGIS steps, and zero guesswork. The method transfers as-is to a wind turbine, a development zone or a land search — more on that at the end.

Preparing the data: study area, CRS and sources

Before any processing I reproject every layer to EPSG:2154 (RGF93 / Lambert-93). It is a prerequisite, not a detail: one-kilometre buffers and one-hectare thresholds only make sense in a metric projection. The study area is defined by intersecting the Park boundary with the municipalities, using a 99.9% surface-attachment threshold — 83 municipalities retained — then joining INSEE population data.

The sources are open and traceable: the IGN RGE ALTI 5 m DTM for relief (the same data family I use to refine urban relief with HD LiDAR), road and trail networks, and buildings pulled via the OpenStreetMap Overpass API — 425,787 buildings extracted, as the layer was not provided. That reflex (working around missing data with a documented open source) is part of the craft, and it is what makes the study defensible.

Physical criteria: slope and aspect with GDAL

First filter, slope, computed with gdal:slope on the DTM. Above 15°, building becomes costly and risky; below it, 68.4% of the Park stays workable. Second filter, aspect, with gdal:aspect: I keep the southern arc (90°–270°) to maximise sunlight, about 48% of the territory.

I then combine both criteria with the raster calculator (boolean operators, binary masks) and the Park's regulatory mask: only 19.8% of the territory remains, roughly 449 km². At this stage, the order of operations and the DTM resolution already change the result — hence the value of a parameterised, replayable chain.

From raster to usable polygons

The result is still a raster: I vectorise it with polygonize to get measurable objects, after a sieve pass that removes isolated pixel clumps. I then filter on a one-hectare minimum-area threshold: 3,883 polygons are kept (the largest reaching 1,696 ha). That threshold is not arbitrary — below one hectare a site is too small for a real facility. It is a project decision, not a default setting.

Accessibility and exclusions: cascading buffers

A favourable site must also be reachable. I model accessibility by intersecting two buffers: 1 km around paved roads (3,539 km, for logistics and rescue) and 300 m around trails (1,801 km, for tourist appeal). The potential area drops to 728 km².

Then I exclude, by spatial difference, what must be excluded: a 300 m buffer around the 425,787 existing buildings, and a one-kilometre buffer around the tourist facilities already in place (refuges, chalets, cabins, shelters) to ensure an even distribution — another 207 km² removed. Each constraint is a subtraction: the space of possibilities narrows to 294 km² of candidate zones.

Have a siting project to arbitrate? I scope your need for free, over a video call, before any quote.

Result, scoring and synthesis map

Crossing candidate zones with suitability polygons leaves 406 favourable sites of at least one hectare, i.e. 29,421 hectares in total. I rank them by municipality to prioritise prospecting: Lans-en-Vercors comes first (871 ha), ahead of Saint-Julien-en-Vercors (822 ha) and Die (665 ha).

The final map follows Bertin's semiology — Jenks classification, coherent colour progression, full layout (north arrow, scale, locator, legend, sources). Deliverables are an A4 300 dpi map and a reusable sites_favorables_final.gpkg layer. Above all, the whole chain is wrapped in QGIS's graphical modeler: the analysis replays in one click with other thresholds (10° slope, other distances). To go further and weight the criteria rather than combine them yes/no, see AHP weighting and Saaty's method in ArcGIS Pro.

Transferring the method: wind, development zones, local plans and land

Replace "refuge" with a wind turbine, a solar farm, a recycling centre or a development plot: the reasoning does not change. For wind power, you swap in the relevant criteria (distance to dwellings, aeronautical easements, protected areas, grid connection) and check compatibility with the local plan — the favourable-zones map becomes a consultation tool. This is exactly the kind of study I run for local authorities and consultancies, in the continuity of my work on GIS spatial analysis for local authorities in Essonne and the applied multi-criteria spatial analysis in my portfolio.

On pricing I stay transparent: a cartographic audit starts at €800 (a first pass on two or three criteria, to quickly validate a siting idea), a full GIS study at €3,500 (the Vercors-type deliverable: multi-source collection, processing, synthesis map and GIS layer). And because I am also a web developer, I can turn that layer into an interactive Web GIS app where the client adjusts the criteria live. Based in Morsang-sur-Orge, I work across Essonne, Île-de-France and remotely throughout the French-speaking world. Let's talk about your project — free initial scoping, quote within 48 h.

For a full breakdown of every service and price range, see the Services & Pricing page.