← Back to the map

How the Broadmoor low spot map was made

This is terrain only. It is not a flood forecast. New Orleans drains through a city-wide system of catch basins, underground pipes, canals and pump stations. This map ignores all of it on purpose. It answers a narrower question: if water could only move over the ground, where would it go and how much would each spot hold? Real flooding depends on how much of the drainage system is working, how fast the rain falls, and how full the canals are.

What the map shows

Broadmoor covers about 362 acres. All of it sits below sea level: the ground ranges from -7.3 ft to 1.6 ft, with a typical (median) elevation of -4.2 ft. Elevations are feet relative to NAVD88, the standard vertical datum, which is close to sea level.

Low spots: places where water would pond because the ground around them is higher, each filled to the point where it would overflow into the next. Colored by water depth.
Rainfall: what happens when 1 to 10 inches of rain falls at once and nothing drains away.
Flagged streets: street blocks that cross a low spot at least 1.5 ft deep.
Catch basin check: whether each catch basin sits at the lowest ground nearby.
Street profiles: click any street to see the ground along it, with water drawn in.

Data

Layer	Source	Notes
Bare-earth elevation	USGS 3DEP 1 m DEM, project LA_2021GreaterNewOrleans_C22	Built by USGS from the 2021 Greater New Orleans LiDAR. 1 m cells, NAD83 / UTM zone 15N, NAVD88 meters. Read straight from the USGS cloud files, only the window needed.
Broadmoor boundary	City of New Orleans, Neighborhood Statistical Areas
Street centerlines	City of New Orleans, Centerline	Split into blocks as published by the City.
Catch basins	data.nola.gov, Catch Basins (Department of Public Works)	The City's current catch basin layer.
Drain pipes, major drains and canals, pump stations	Sewerage and Water Board asset layers, hosted by City of New Orleans GIS	A 2018 snapshot released after the August 5, 2017 flood. The City notes it is not being updated. Shown for context only.
Rainfall reference marks	NOAA Atlas 14 at Broadmoor	1-hour: 2.3 in (2-yr), 3.4 in (10-yr). 24-hour: 5.5 in (2-yr), 8.4 in (10-yr).

How it was made

Cut the elevation model. Read the 1 m DEM for Broadmoor plus a 1,968 ft buffer so water flowing in from nearby blocks is counted. Elevations were converted to US survey feet. All measuring was done on the original 1 m grid in NAD83 / UTM zone 15N, so the LiDAR was never resampled. Web layers were reprojected to Web Mercator at the end.
Why a plain "fill sinks" does not work here. The usual GIS approach fills every pit until water can flow off the edge of the map, then subtracts the original ground. In New Orleans the whole city is a bowl ringed by levees and ridges, so that approach marks most% of Broadmoor as sink, up to 4 ft deep. That is technically true and not very useful: it describes the bowl, not the spots where water gathers first.
Build a depression hierarchy instead. The ground was split into basins, one for every low point, using a priority-flood watershed. Neighboring basins merge at the lowest point on the rim they share, and this repeats until everything reaches the edge of the analysis area. The result is a tree of low spots nested inside bigger low spots.
Low spots layer. Every basin is filled to the level where it first spills into a neighbor. Dips shallower than 1 ft are folded into the larger low spot around them, and low spots shallower than 0.5 ft or smaller than 300 sq ft are left off. For each one the script records the deepest point, area, volume in gallons, bottom and overflow elevations, the ground that drains to it, and how much rain on that ground would fill it.
Rainfall layer (fill, spill, merge). Rain lands evenly on every 1 m cell and runs to its basin (here using a finer 0.1 ft noise rule, so small dips count). Each basin fills; when it is full the extra water spills over the rim into the neighbor, and when two neighbors are both full they become one pond with one water level. Water that reaches the edge of the analysis area leaves. This method follows Barnes, Callaghan and Wickert (2020), "Computing water flow through complex landscapes, Part 3: Fill-Spill-Merge". Every pond keeps a perfectly flat surface, and the water stored never exceeds the rain that fell.
Streets. Each street block was sampled every 1 m along its centerline. A block is flagged when it crosses a low spot at least 1.5 ft deep. In the rainfall view, a block turns red when water over the center of the street reaches 0.5 ft (6 inches), which is enough to stall many cars, and orange from 3 to 6 inches.
Catch basins. A catch basin counts as "at a low point" when the ground there is within 0.25 ft of the lowest ground within 30 ft. Because catch basin points are mapped near the curb rather than on the exact grate, the check uses the lowest ground within about 5 ft of each point. In Broadmoor, 859 of 1,147 catch basins pass. Separately, 8 low spots at least 6 inches deep have no catch basin within 15 ft.

Rainfall results for Broadmoor

Rain (in)	Acres under water	Deepest water	Street blocks with 6 in+	Gallons in Broadmoor

Broadmoor holds more water than falls on it because the higher ground around it drains in.

Limits

No drainage system. Catch basins, pipes, canals and pumps are left out. With them working, far less water would stand, and it would clear faster.
No soaking in and no timing. All rain arrives at once and none soaks into lawns. That makes the rainfall view a worst case for the ground alone.
Bare earth only. Buildings and trees were removed by USGS. Water would really flow around buildings, and raised slabs keep many homes above street level.
LiDAR accuracy. The elevation model is accurate to roughly a few inches. Depths under about 3 inches are within that noise.
Snapshot in time. The LiDAR was flown in 2021. Street work since then, including Broadmoor drainage projects, may have changed the ground.
Edges. Water that reaches the edge of the analysis area is treated as gone. Low spots touching the edge are marked as less reliable.
Drainage layers are old. The pipe, canal and pump station layers are a 2018 snapshot and are shown only for reference.

Reproduce it

Everything on the map comes from one Python script, process.py, using free and open-source tools (rasterio, NumPy, SciPy, scikit-image, GeoPandas, Shapely). The map uses MapLibre GL JS with an OpenFreeMap basemap. Run python process.py to rebuild all data from the original sources. Data generated .