Aa
Contrast
Text size

Blog / Data In

The Map Nobody Can Read

10 July 2026

Twelve years of spatial science reduced to one static screenshot and two weeks of coordinate system emails.

The science is twelve years deep, and the spatial data is a mess. Geospatial management is where the earlier failures in collection, checking and integration finally become visible to the outside world. They are not visible to your team, who learned to navigate the chaos. They are visible to the person across the table who asks a simple question about where you sampled and watches you struggle to answer. This post is about GIS environmental analysis when the files themselves are the problem.

What it looks like

A coastal research programme had twelve years of spatial monitoring data: sampling stations, habitat boundaries, species records, bathymetric surveys and protected area overlays. It was scientifically irreplaceable and spatially a disaster. The GIS files lived on four shared drives, two personal laptops and an external drive that a former coordinator had taken home "temporarily" in 2019. Folders were called "GIS_final", "GIS_final_USE_THIS", "GIS_2021_updated_v2" and "OLD", and the last one held files newer than half of those outside it.

The coordinate systems were a guess. Some files were in WGS84, others in a regional projection from a government partner that nobody had documented. The study area boundaries had changed three times in twelve years, but the old ones persisted in every analysis, because nobody was sure which file belonged to which year.

For a major EU programme review the request sounded simple: show us where you work and what you have found. The most experienced GIS analyst took ten days to prepare it. The result was a static screenshot from QGIS, placed in a presentation. A panel member asked whether the stations could be filtered by year to show how the programme had grown, and the answer was no. Another asked for the boundary file to overlay on their own data. The file that was sent used a coordinate system that matched nothing on their side, and the email thread to sort it out lasted two weeks. Twelve years of spatial science became one screenshot and two weeks of emails.

Four places where it goes wrong

Files are scattered with no order. Spatial data fragments faster than any other kind. A shapefile is a set of four to six files that must stay together, the coordinate definition sits in a sidecar file that gets lost when files are copied, and rasters need particular software versions. The result is what every GIS analyst recognises at once: hundreds of files in nested folders with names that were logical when created and opaque three years later. Nobody knows which boundary is current or which coordinate system a file uses without opening it. And broken spatial data often looks fine until you plot it and your stations appear in the middle of the sea.

Coordinate systems get confused. They are invisible until they are costly. Data in WGS84 and data in a local projection look identical in a spreadsheet, and plotted together they can be hundreds of metres apart. A partner's boundaries in one system laid over your stations in another produce a map that looks plausible and is scientifically wrong. Conversions make it worse: someone converts a file, does not record it and saves it next to the original under a slightly different name. I watched a marine monitoring team spend three weeks working out why their species models gave implausible results, only to find that the habitat layer and the occurrence records had been silently misaligned for two years.

Maps cannot be shared with non specialists. Your analyst makes accurate, defensible maps, but they exist as QGIS projects or ArcGIS documents that need particular software, data paths and knowledge to open. A funder, a ministry contact or a community partner cannot open them. So the maps are exported as static images, stripped of the ability to zoom, filter or click a station to see what was found there. Every follow up question means a new export, which takes days and raises three more questions.

Time is not managed. Monitoring is temporal by nature. Where you sampled in 2015 is not where you sampled in 2023, boundaries change as programmes grow, classifications are refined and protected areas change with policy. Without temporal management, history is either overwritten, which makes past analyses impossible to reproduce, or it piles up without structure, so that knowing which boundary applies to which period takes institutional memory. A biodiversity programme could not answer a basic question for a climate impact assessment: how had its sampling coverage changed over fifteen years? Rebuilding the answer from archived files took six weeks, and they were not fully confident in it, because they could not be sure they had found every file.

What fixes it

Spatial data should be as accessible, searchable and shareable as any other data in your system, for anyone who needs it, with or without GIS software.

Start with a spatial database as the single source of truth. PostGIS, the spatial extension of PostgreSQL, does what folders of files cannot. Geometries live in a database instead of loose files, every layer has a defined coordinate system enforced at import, and queries combine spatial and tabular data. History can be versioned and access controlled, and nothing fragments when someone copies a file to a laptop. Moving from folders of shapefiles is a one off effort, and afterwards spatial data behaves like all your other data and is backed up with it.

Enforce one coordinate system at intake. Choose WGS84, or whatever your main funder or regulator requires, and build the check into intake in the same way as species name validation. Data in other systems is transformed on the way in, the transformation is documented and the original is archived. From then on everything in your system speaks the same spatial language.

Publish web maps for people without GIS. GeoServer, Mapbox or QGIS Cloud can publish your layers as interactive maps that open in a browser. A funder clicks a link and sees the monitoring network, zooms, filters by year, clicks a station for details and passes the link to a colleague. A static screenshot answers the question that was asked. A live map lets people ask the questions you did not expect and answers them at once, which is what good environmental data visualization is for.

Manage time in the database. Every boundary change is recorded with a start and an end date, and every analysis refers to a specific version, not just the current one. Asking where you sampled in 2018 returns the 2018 boundaries, and comparing 2018 with 2023 coverage is a database query, not a six week reconstruction.

The programme rebuilt its spatial infrastructure along these lines: shapefiles migrated to PostGIS, coordinate systems standardised and a web map published with filtering by year. At the next review they sent a link. The panel filtered by year, clicked stations and downloaded boundary files in the format they needed, and the follow up questions were answered in the meeting instead of in a two week thread. The director told me: "We had the best spatial dataset in the room. For the first time, people could actually see that."

The pattern

Geospatial chaos looks like a technical problem, something for the GIS specialist to sort out when there is time. It is a visibility problem in technical clothes. Your spatial data is the most readable form of your work for non scientists. A stakeholder understands a map before a statistics table, a community partner connects with a map of their coastline before a regression, and a ministry contact grasps the scope of a programme from a spatial overview before they read the methods.

If the map is wrong, out of reach or impossible to share, the quality of the science behind it does not matter. The first impression is a flat screenshot that cannot answer questions, and that is not the impression of scientific excellence. Fix the infrastructure and people see your organisation differently, not because the science improved but because it finally became visible in the form that non scientists understand. Your monitoring network deserves to be seen.

Escaping Data Chaos

Also on LinkedIn

This article was first published in the Escaping Data Chaos series. The stories in the series are composites of real projects, anonymised.

Read it on LinkedIn

Recognise this problem?

One conversation to find out if we can help.

Book a free 15 minute call