Zurich Takes a Methodical Approach to Duplicate Image Replacement — and Other Cities Are Watching

Zurich's city archive, the Stadtarchiv, quietly crossed a threshold this spring: more than 40,000 digitised images in its public-facing urban documentation database had been flagged as duplicates or near-duplicates requiring systematic review and replacement. The figure, drawn from an internal digitisation audit completed in April 2026, underscores a problem that archivists in European cities have been wrestling with since large-scale scanning projects accelerated after 2020.

Duplicate imagery — photographs, maps, and planning documents scanned twice or with slight variation and filed under different reference numbers — may sound like a bureaucratic footnote. It is not. When planners in Zurich's Amt für Städtebau consult the municipal image database to review the visual history of a neighbourhood like Aussersihl or Hürlimann Areal redevelopment, redundant files inflate search results, bury authoritative versions, and in at least some cases have led to contradictory information appearing in public-facing planning documents. The problem scales fast in a city that has been digitising records continuously since 1998.

What Zurich Is Actually Doing

The Stadtarchiv launched a structured deduplication programme in January 2025, partnering with ETH Zurich's Institute for Information Security and Dependability to deploy perceptual hashing algorithms capable of identifying visually similar but not byte-identical images. The collaboration, part of a broader smart-city data governance initiative, processes roughly 3,000 image pairs per week. Flagged duplicates are not automatically deleted — archivists at the Stadtarchiv's Neumarkt 4 offices conduct manual spot-checks before any file is retired or a canonical replacement is designated. That human-in-the-loop approach is slower than fully automated pipelines, but it has so far prevented the accidental loss of images later identified as unique.

The city's housing shortage context adds urgency. Zurich's Wohnungsnot crisis has pushed planning departments to process building permit documentation faster than at any point in the last decade. The Amt für Baubewilligungen processed a record number of applications in 2025, and planners pulling visual reference material from legacy archives need clean, non-redundant data. A single planning consultation that relies on a duplicate — and therefore outdated — aerial photograph of a construction site in Wiedikon or Altstetten can delay a permit decision by days.

How Zurich Compares to Amsterdam, Vienna, and Zürich's Swiss Peers

Amsterdam's Stadsarchief began a comparable deduplication project in 2023, focusing first on its vast collection of canal-zone construction photographs. The Dutch capital opted for a more aggressive automated replacement pipeline, retaining human review only for images pre-dating 1950. The result: Amsterdam cleared its backlog roughly eighteen months ahead of Zurich's projected timeline, but archivists there have acknowledged publicly — in documentation posted on the Stadsarchief's website — that a small percentage of replaced images were later disputed by researchers.

Vienna's Wiener Stadt- und Landesarchiv took a third path. Rather than a dedicated deduplication programme, Vienna embedded duplicate-checking into its ingest workflow from 2022 onward, meaning new material entering the system is screened at the point of upload. That approach does not address the legacy backlog but has effectively stopped the problem from worsening. Basel and Bern, Zurich's nearest Swiss comparators, both have smaller municipal archives and have not yet launched formal deduplication programmes at scale, according to publicly available documentation from their respective city administrations.

The ETH Zurich partnership gives the city an edge that smaller municipalities cannot easily replicate. Perceptual hashing tools developed at ETH's Hönggerberg campus are being refined on Zurich's archive before any consideration of licensing or broader adoption. That gives Zurich's archivists early access to improved detection rates — the current algorithm catches an estimated 94 percent of visually duplicate pairs — but it also means the programme is still in active development rather than running on a proven, stable platform.

For residents and researchers, the practical advice is straightforward: when sourcing historical images of Zurich through the Stadtarchiv's online portal, filter results by verified canonical status — a tag the archive introduced in March 2026 — rather than browsing unfiltered results. The full deduplication review is projected to complete by the end of the third quarter of 2027, at which point the archive expects to retire or consolidate approximately 60,000 files. Until then, the canonical tag is the most reliable way to avoid landing on a redundant image that may no longer reflect the authoritative record.