Highwater

The Environment Agency publishes a list of every planning application it objected to on flood risk grounds. Parsed in full, it contains 23,336 records across 426 planning authorities for 2016-17 to 2024-25. Of those, 67.2% show EA advice was followed, 2.7% (635 applications, 2,949 homes) were granted against its advice — and for 30.0%, or 7,011 applications, the outcome is simply unknown. The EA states plainly that it records the outcome of about 68% of applications and uses a sample.

The absolute numbers rose sharply: permissions granted against EA flood advice went 28 in 2021-22 to 102 in 2024-25. But the honest reading is more interesting than the alarming one. The rate stayed roughly flat at 3–5% across all nine years, which is consistent with the government line that 96% of decisions follow EA advice. More homes were permitted against advice because more applications were objected to — not because councils became more willing to override. Anyone presenting the 3.6-fold rise without the flat rate will be taken apart in the first meeting, and rightly.

The EA reports "over 97% compliance" with its advice. Aviva, joining Ordnance Survey AddressBase to the national flood risk assessment, found one in nine new English homes built 2022–2024 is in a medium or high flood risk area — 43,937 homes. Both statements are true. The gap between them exists because the EA counts objections it made, not homes that were built, and because surface water flooding sits largely outside its objection trigger despite being the fastest-growing risk.

Two structural defects make the official dataset nearly unusable. It carries no spatial reference of any kind — no coordinates, no UPRN, no postcode, no site address — so it cannot be joined to flood maps without scraping 426 council portals. And each annual release silently restates history with no versioned archive, so the trend above is not reconstructible from official sources by anyone who did not save previous years’ files.

Highwater leads with the outcome gap, not with mapping. The existing volunteer planning aggregator already carries decision status and decision date alongside coordinates, retrievable free and anonymously — so the 7,011 unknown cases can be closed without scraping a single council portal. If the granted-against-advice rate in the unknown set even approximates the known set, the true national compliance figure is materially worse than the published one, and that is the finding government needs before it changes the policy.

Geocoding follows from the same source rather than driving the work. Where portal retrieval is unavoidable, the Agency’s own dataset supplies a current directory of every authority’s planning portal, and its guidance explicitly tells users to look applications up that way. One trap is load-bearing: authority reference formats are not nationally unique — the same reference exists in four different councils — so any lookup must be constrained by authority or it will silently geocode records to the wrong county while appearing perfectly healthy.

It maintains a versioned archive of every annual release, so trends survive the EA’s restatements. This matters urgently: the statutory consultee reform proposes removing the use of objections where flood directions do not apply, which would break the only national time series that exists, at the moment it is most needed.

And it addresses the blind spot directly. The Environment Agency’s statutory trigger is, as a matter of law, fluvial and tidal only — the flood zones that trigger consultation are defined purely by river and sea probability. Surface water consultation falls to lead local flood authorities, whose responses sit on roughly three hundred separate portals with no central register, no dataset and no reporting requirement. Crucially, the ministerial referral that applies when a council overrides the Agency does not apply to surface water objections. No escalation trigger, no referral, no record.

The policy history has three steps and must be told in full. Planning guidance brought other flood sources in during 2022; the December 2024 framework codified it for individual applications; then September 2025 guidance created an escape hatch available to no other flood source, allowing the sequential test to be disapplied where a site-specific assessment shows occupiers would remain safe from surface water for the lifetime of the development. Surface water was brought in, hardened, and then singled out for an exemption. That is a stronger finding than a simple extension — and stating it as a simple extension invites immediate correction.