How deep will water stand in a low spot?

Which is a genuinely useful thing to know, because it is a property of the shape of the land rather than of any particular storm. It does not change when the weather does.

Spill point, in one paragraph

Imagine pouring water slowly into a dip in a lawn. The surface rises evenly. At some moment it touches the lowest notch in the surrounding rim — a gap between two mounds, a dip in a curb, a gap under a fence — and from then on everything you add runs out there.

That moment sets both numbers people care about: the depth at the deepest point, and the extent of the wetted area at the instant it spills.

How it is computed

On a bare-earth elevation surface, every closed depression is raised until it overflows, so that no water is left with nowhere to go. Water spills over a 2 mm lip.

Most drainage software throws away the depth it raised. That is the step we deliberately do not take: the fill depth is kept, because a real hollow beside a house is exactly the thing worth telling somebody about.

Two floors on the answer

• 3 cm. Below that you are reporting measurement noise as ponding, which is printing the error bar as a finding.
• Outside every building. The bare-earth value under a roof is an interpolation across a structure that was removed, so a "hollow" there is an artefact of the data, not a dip in a garden.

And a window: standing water is counted within 20 m of the footprint, because a depression on the far side of the block is not a fact about this building.

Why filling has to happen before anything else

Skip depression filling and millimeter-deep pits eat the entire drainage map. Water routes into an artefact you could not see with a magnifying glass and stops there.

Measured on real ground: 282,572 m² of contributing area in Chicago and 201,312 m² in Pittsburgh drained into interior pits two to four millimeters deep and simply terminated, against largest surviving channels of 1,034 m² and 33,931 m². Almost the whole network was being swallowed by noise.

How we know the depth is right

You cannot validate a fill-to-spill computation against real ground, because nobody knows the true answer on real ground. You can validate it against ground you built yourself.

On synthetic terrain — a bowl of known geometry cut into a 2% slope, where the spill depth can be worked out on paper — the engine fills it to 0.28 m predicted against 0.289 m measured.

That is a check on the arithmetic, and it is worth exactly what it is: proof the routine does what it claims on a surface with no noise in it. It says nothing about how faithfully any particular county's data describes your lawn.

What a modeled depth is not

Four things sit between a hollow in a surface and water in a garden, and we measured none of them:

1. Infiltration. Sandy soil can take the whole event. Heavy clay after a wet fortnight can take almost none.
2. Rainfall. A depth is a capacity. Filling it requires enough water arriving fast enough.
3. Buried drainage. A yard drain, a French drain or a land drain in the middle of that hollow changes everything and is invisible from the air.
4. Recent works. The underlying data is collected on a cycle. A regrade last spring is not in a surface measured three years ago.

So the honest sentence is: the ground here forms a hollow that would hold water to about this depth before spilling. Not: this garden floods.

Where the depth is genuinely useful

• Comparing two corners of one property — same data, same method, so the ordering is trustworthy even where the absolute value is uncertain.
• Deciding where to look first after a heavy autumn, rather than walking the whole boundary.
• Sizing the question, not the solution. A 4 cm hollow and a 40 cm hollow are different conversations.
• Checking a story. If somebody says water sits at the back of the plot, the surface either has a hollow there or it does not.

An independent check that the routing feeding all of this is sane: Pittsburgh Water publishes the location of every storm inlet it maintains, and the 73 published inlets in the frame fall within 10 m of a modeled channel 3.99× as often as the frame's own base rate at a 260 m radius — the control being the share of the frame that sits that close to a channel anyway. Pittsburgh is the only municipality publishing inlets in a form we can consume, so that is one city and not a national result, and inlet positions themselves carry one to two meters of georeferencing error.

Common questions