State of Play

Flood risk is a development constraint like bushfire: zones from low to high, numbers, and measures that follow from them.

• Land is constrained. Prime, unconstrained land is largely developed or unaffordable, so what remains is constrained ground. Flooding can hold up an approval entirely, or push a project past feasible design and construction cost.
• The information is improving. Catchments are modelled more accurately, and the estimated likelihood of severe floods rises as the historical rainfall record lengthens. The latest flood information is not new risk: the risk was always there, and what has changed is how well it is understood.
• The terminology has not kept up. Flood events do not take turns, and a one per cent chance in one year does not reduce the chance the following year: across an average 80 year lifespan, the 1% AEP event carries about a 55% chance of occurring. That is why the old once-a-century framing is retired, here and on the map.

• There is no statewide visual mapping system for flood control lots, no public section 10.7 system, and no single place to see what mapping exists.
• Councils are responsible for flood mapping, and each presents it differently: different viewers, terminology and depth of detail. Some publish full model grids, others legacy PDF reports and nothing else, and application fees vary.
• The NSW Planning Portal’s statewide flood layer is a powerful tool for town planning but excludes current flooding data: counted in August 2026, it showed 622 remnant polygons across roughly ten local government areas, dated May 2019, with whole council areas and floodplains blank.
• The 2022 NSW Flood Inquiry made twenty eight recommendations, including better community flood data provision and education. The NSW Government supported all twenty eight, and the SES states that public flood information remains each council’s responsibility.
• So the gap is not recognition. The need is acknowledged and supported. There is simply no funded program to deliver it, and the data to build it is already published.

Every council chooses its own colours, its own catchment names and its own platform, and none of them agree. A resident who moves thirty kilometres meets a different colour language for the same risk, and flood information is often buried deep in a council website under a catchment name the public has never heard of. Hosting of maps on council specific platforms, and inconsistency in how layers are shown, is part of the ongoing confusion.

Premiums have separated from the average. Average annual home insurance, October 2025: outer western Sydney and the Blue Mountains $5,350, against a national average of $2,938.

148,546 NSW properties are projected to be uninsurable by 2030. Riverine flooding drives about 80% of the national high-risk total.

In March 2021 about 65,000 people across Greater Sydney were under evacuation warnings and orders, the majority in the Hawkesbury-Nepean valley. The NSW Floodplain Management Program funded 53 projects totalling $10.4 million in 2024-25, against that exposure.

An owner wanting to know about their own land is asking a narrow question: how deep, how often, and what it means for what they intend to build. Three things get in the way, and only one of them is cost.

Fragmented

No consistent format, fee or data standard. Flood information fees run from $250 to $375, with model data at $1,190. Most sit buried in fees-and-charges PDFs.

Out of date

An estimated 250 to 400 adopted studies. The majority are older than three years and likely more than half older than five. Of everything on the state portal, 76.8% was published before ARR 2019, which multiplied the modelling effort roughly tenfold, so bringing a pre-2019 study current is not a light revision.

Locked

Of everything published on the state portal only 15.6% carries an open licence, and of the datasets holding real spatial data, 85% are restricted. Even open files need a registered account to download.

Every NSW flood dataset on the state portal, counted. A study being published is commonly taken to mean the information is available. In practice four things sit between a study appearing on the portal and a map being able to draw it.

By dataset. A third-party licence means the consultant or council retained rights: readable on the portal, not republishable.

Counting datasets says nothing about whether the particular thing a practitioner needs exists. These count datasets whose own description names the product, then narrow to the openly licensed ones, then to the ones that also hold a data file.

Velocity is the clearest case. 271 datasets across the state describe holding it, 54 are openly licensed, and 8 both carry an open licence and hold a data file. For the AR&R hazard classification the figure is 1: not per cent and not per council, but one dataset in the state. Every other published hazard classification is either behind a licence that forbids republication or exists only as a figure in a report.

These are claims made by the publisher in the dataset description, not verified contents. An archive described as holding velocity has not been opened to confirm it, and a package that holds velocity without saying so is not counted.

The distinction that explains most of this page: a right of access is not a right of reuse. The Government Information (Public Access) Act gives anyone the right to obtain information a NSW council holds, flood studies included. It grants no right to republish what is obtained. A study can therefore be public in the sense that any person can read it, and locked in the sense that no one may lawfully put it on a map for the next person.

That is why the licence column above matters more than the published column. LEVEE republishes mapping, so it can only carry material whose licence permits republication, which today means CC BY. A restricted study is recorded, counted, and left where it is, however openly it can be read in place.

While almost every council publishes its flood study information publicly, fewer than half provide it for free and open reproduction by engineers.

Councils are counted, not listed. This research does not rank councils and will not publish a list of the ones publishing least. The constraint is the licence attached to each study, and a council is often bound by terms an earlier consultancy contract set years before anyone asked the question.

Every council catchment the platform considers is asked the same five questions, and the answers become a public record rather than an internal note:

1. Does adopted flood mapping exist?
2. Is it published, and where?
3. What format: model grids or legacy PDF reports?
4. Under what licence?
5. How current, and how much resistance to reuse?

The output is a council-by-council audit of NSW flood data accessibility, format and currency, published as a findings report and as the operating platform itself. Gaps in map coverage directly demonstrate gaps in accessibility, data or cooperation, and the intent is that the findings inform policy improvement at a state level, not just a new mapping system.

Every study is catalogued, including where licensing prevents display. LEVEE records that the study exists and links to the source, so gaps in visual coverage speak for themselves.

The naming rule is fixed: name the good, aggregate the rest. Councils performing well may be named positively. Councils that are not appear only in aggregate figures. No league table, no ranking, descriptive language only. Once a council participates, naming becomes a reward rather than a penalty, which is the right incentive and keeps the research collaborative rather than adversarial.

Openly licensed spatial datasets, after report archives are excluded. Between them these publishers hold most of the state’s total, and every one of them decided to license openly when nothing obliged them to.

Wollondilly Shire Council also publish openly, as reports rather than as data files.

This list grows as councils license data openly, and it is the only list on this page. If your council publishes openly and is not named here, that is an error worth telling us about.

Share of each year’s published material carrying an open licence. Counted by publication year rather than by the date a record was loaded onto the portal, so the bulk historical load spreads back across the years it belongs to instead of showing as two meaningless spikes.

Open licensing has sat between a fifth and a third of new material for six years running, with one better year. On the current trend the gap does not close on its own.

LEVEE is built in four parts. Each gauge shows how far that part has risen, what is counted behind it, and what is not there yet. River stage equals project stage, which is why these are named stages rather than a percentage: a percentage of an unfinished thing is a guess, and a stage is a fact somebody decided.

Counted from the published map on 28 August 2026. This is the part of the state’s flood information that is both openly licensed and actually on a map, against the audit on the other tab.

Figures are counted from the published map manifest and the licence register on the date shown, and are republished as they change. A layer only appears on the map, and only counts here, once its source licence is recorded and permits publication. Nothing on this platform replaces a section 10.7 planning certificate.