The flood floor

Why a street, and why the council owns it

AnBanjir's shape traces to a response architecture still in its first decade, a street-resolution gap no national alert closes, sensors that stop above the council, connectivity that dies during the flood, an unverified crowd already reporting it — and the privacy regime that decides which pattern survives.

Honest status · designed

The findings below are real and cited. The product is designed, not yet running — the evidence floor is what a council tender would be built on, not a claim of traction.

What the record actually says

One young agency, one landmark disaster, one upstream-only sensor layer, one alert-fatigue pattern, one connectivity failure — and the crowd and privacy regimes that decide which design survives.

2016

The national disaster architecture is barely a decade old

Malaysia centralised disaster response only recently — NADMA was stood up after years of fragmented, agency-by-agency handling. That is not a criticism; it is the reason the operating layer is still missing. Command was built top-down for the catastrophic national event. The recurring, hyper-local flash flood — the one that shuts a single underpass every monsoon — was never the design case, and no PBT was handed a console for it.

NADMA Act 2015 (Act 774) · agency operational 2016

Dec 2021

The largest peacetime evacuation in recent Malaysian history moved tens of thousands

The late-2021 monsoon produced the biggest evacuations the country had seen, exposed fatal gaps — people trapped in underpasses and car parks as water rose faster than any alert predicted — and made 'which street, how high, how soon' a question every council suddenly had to answer live. The science existed at the federal level; the street-level operating picture did not. That deficit is the exact product AnBanjir is.

December 2021 floods · multiple-state evacuation, Shah Alam & Selangor loss of life

Federal

JPS gauges and MetMalaysia radar are real — and stop above the council

The federal agencies already run the sensors and models: water-level gauges on major rivers, nowcast radar on the weather. What they publish is a basin- and state-scale product. A gauge on the main river says nothing about the tributary nullah backing up under a specific junction, and no federal feed arrives as an actionable, thresholded alert inside a council's own control room. Integrate downstream; never duplicate upstream.

JPS (DSI) river gauges & flash-flood guidance · MetMalaysia Godang radar

State-wide

Generic alerts trained the public to ignore them

When the warning covers an entire state, the resident whose road is bone-dry learns to discount it — and the one whose kerb is rising learns nothing actionable. The alert that saves the underpass is scoped to a street and a risk band, in language a person can act on tonight. Resolution is not a nice-to-have; it is the difference between a working bell and one people have tuned out.

Siren / broadcast alert fatigue · observed behaviour

During the flood

Connectivity fails in precisely the hours the system is needed

Towers go dark, power is cut for safety, and the map the responders were given stops updating at the moment the situation is worst. A flood product that requires four bars is a flood product that fails during the flood. The state — gates closed, routes forbidden, shelters full — has to mutate locally and reconcile legally on reconnect. This is inherited engineering, not a roadmap promise.

Observed tower/power outages during severe flooding events

#banjir

The crowd already reports floods — into an unstructured, unverified void

During every event thousands post waterline photos to social platforms: enormously valuable, entirely unusable as an operating signal — no location certainty, no corroboration, no way to separate a genuine rise from a panic post or a recycled image. Crowd-sensing research is consistent: the failure mode of every abandoned pilot is a single unverified observation surfaced as fact. The answer is resolution before the map moves, not a ban on the crowd.

Social-media flood reporting, 2021 onward · crowd-sensing literature

Finding → rule

So AnBanjir is built this way

from: the fiefdom doctrine

One council, one instance

AnBanjir signs with MPJ, or MBSA, or DBKL — alone. Nothing crosses between councils by contract clause or by schema. A state or NADMA view is a read-only overlay, licensed separately and never writing back. That isolation is what makes each tender pass legal review.

from: the 2021 outage record

Offline is the design centre

The state engine mutates on the council's own LAN and resolves on reconnect — the machinery inherited from AnKampung's gatehouses. Continuity is not a resilience checkbox bolted on later; it is the shape the whole console is built around.

from: #banjir's autopsies

Crowd reports carry proof or weight-zero

A waterline photo is matched against gauge telemetry, hazard layers and neighbouring sightings before it moves a map state. Reporters hold a pseudonymous confidence history (AnVerified rails); an uncorroborated sighting shows as unconfirmed — visible, honest, harmless.

from: the federal-sensor reality

Integrate-downstream, never-duplicate

MetMalaysia radar and JPS hazard are licensed inputs the council already has rights to use. AnBanjir adds the operating layer on top of them — it does not build a competing weather service, which is the only posture the federal agencies will partner with.

from: accountability's location

Thresholds belong to the engineer

Where the danger line sits, which band triggers which alert, is set by the council engineer who answers for that underpass — never defaulted by a vendor. The product measures and warns; the authority to declare and act stays human and local.

from: crowd-sensing research

Confidence is a published column

Every street-state cell carries its sample depth and corroboration score, visible to the control room and the resident alike. A map that hides uncertainty gets used as a weapon after the water goes down; this one stays an honest instrument.

Why here, why now

The Malaysian flood reality

2016
the year national response was centralised — the local layer is still young

Command exists for the catastrophic national event. The recurring hyper-local flash flood, owned by a PBT with no console, is the gap AnBanjir is designed to close.

Basin
the scale federal gauges and radar publish at

The science is real and upstream. What no feed delivers is a thresholded, street-scoped alert inside the council's own room, on the council's own thresholds.

Street
the resolution a trustworthy alert must reach

Anything coarser and the public tunes out the bell. AnBanjir's whole promise is answering 'which street, this hour, which direction' — not 'danger, somewhere'.

0
cross-council joins, and crowd reports that leave without consent

Instances are air-gapped by construction; a resident's flood report crosses to their council only if they choose to bring it. The isolation is the design, not a policy on the roadmap.

What is not yet true

The honest gaps

Designed, not yet running

AnBanjir is a designed product, deliberately last in the pilot order — it needs a council champion and one monsoon of sensor density before the alert bell can be trusted. No live council instance exists today; every reading here is a shape, not a measurement.

No real screenshots

Every mockup here is labelled 'design reference · not a capture'. There is no control room in the field to photograph, and we will not fake one — the console figures illustrate the format, not a flood that happened.

The opening offer

What is credible now is a single reference deployment — one PBT, one flood-prone catchment, one honest season — won with a council champion. The console, the offline engine and the crowd resolver are the product; the tender is scoped to one instance.

Warn the street. Route the rescue. Keep the map alive when the tower dies.

AnBanjir is designed, and it is built to be adopted one council at a time — a single PBT owns its catchment, its thresholds, and its console; nothing joins across councils. The opening offer is a reference deployment: one flood-prone catchment, one control room, one offline-capable state engine. Council enquiries are welcome today.