Australian flood data coverage is the hard part

Bushfire mapping in Australia is national and reasonably consistent. Coastal change is measured from satellite imagery with uniform coverage. Heat projections come from the same CMIP6 ensemble everywhere.

Flood coverage is uneven, and flood is the hazard finance teams ask about first.

Why it is uneven

Flood studies are commissioned by local councils, usually catchment by catchment, on their own schedules and budgets. A well-resourced metropolitan council may have probabilistic modelling for every watercourse, updated within the last five years. A neighbouring shire may have a single 1990s study covering one creek, never digitised.

The studies that exist are not directly comparable. Some publish flood planning area boundaries. Some publish design flood extents at specific annual exceedance probabilities. Some publish depth grids. A flood planning area is a planning instrument with freeboard built in. A 1% AEP extent is a modelling output. Treating them as the same layer produces a wrong answer.

How we handle it

Three rules, applied to every run:

Name the study. Every flood result records which study produced it, its adoption year, and what kind of layer it is. If two sites in the same portfolio were assessed against different study types, the appendix says so.

Do not interpolate across gaps. Where a site sits in an LGA with no usable study, it is flagged unassessed. It does not get the state average, and it does not get "low" by default. In a recent 240-site run, 51 sites came back unassessed for flood. That is a fifth of the portfolio, and the reporter needed to know it.

Keep the layer types separate in the output. Planning-instrument boundaries and modelled extents are reported as separate columns rather than merged into one rating.

Why we made this choice

A complete-looking table is easier to put in front of a board than one with a column of blanks. But an unassessed site is a real finding. It shows where a study is needed, and it states the limit of what public data can say.

An assessment that hides its gaps is the same assessment with the uncertainty moved out of view.