The Grand River crested at Comstock Park on April 8, 2026, about half a foot beneath major flood stage — close enough that the next forecast (it kept raining) put Grand Rapids on track for one of the highest readings on record. The city took 5.79 inches (147 mm) of rain in the first two weeks of April, already two inches above the full-month average, on top of a March that delivered roughly twice normal precipitation. NASA Earth Observatory’s April 11 composite — Landsat-derived, false-coloured to separate water from vegetation — shows the floodplain footprint against the same scene a year earlier.

Read it as a system, not a weather report. Two inputs meet a saturated buffer: March rain pushed soil moisture toward field capacity, the northern Lower Peninsula carried an above-normal snowpack into April, and the April rain arrived with the runoff coefficient already near one. Snowmelt adds the second forcing. Downstream, the queue builds at the dams: reservoirs designed to time-shift inflows hit storage faster than they can release, and NOAA’s National Water Prediction Service gauges started flashing the topology — Newaygo evacuations, Cheboygan Dam waters rising, a scenic drive and an airport runway gone.

←TODAY: Grand River crested April 8, 2026 at ~0.5 ft below major flood; aging Michigan dams stressed by rain-on-snow.
→3012: Compound rain-on-snow becomes the design baseline; dependency-graph literacy is licensure-grade work.
Fulcrum: The single point of failure is always the one you didn’t draw.

The remote-sensing layer is the part PAZ readers can borrow today. Landsat 8 and 9 deliver 30 m surface reflectance on a 16-day revisit and the tiles are free — USGS EarthExplorer, or the Microsoft Planetary Computer STAC catalog if you want them programmatically. Copernicus Sentinel-1 SAR is the European analogue and cuts through cloud, which matters when the storm system is still over your site. The EU Floods Directive (2007/60/EC) already obliges member states to publish flood hazard and risk maps; the satellite tile is how you check whether reality is staying inside them.

The harder lesson is the dam one. Michigan’s 2020 Edenville–Sanford failure in Midland County displaced roughly 11,000 people and ran a nine-figure damage tally, and the 2026 stress test arrives with most of that aging-dam fleet still in place. The architecture diagram in the asset register is not the dependency graph. The dependency graph is what fails when the operator can’t reach the spillway because the access road is underwater, or the SCADA link goes down because the substation is flooded — the second-order coupling between the asset and the people, vehicles, and infrastructure required to operate it.

Atelier: On a Swiss Plateau river-edge brief (Aare, Limmat, Reuss), the Michigan event is a calibration point: rain-on-snow compound forcing is no longer a 2005-flood outlier. The PAZ Atelier-Code move is to overlay Sentinel-1 SAR flood extents from the last five DACH events on the cadastral, the 2007/60/EC hazard maps, and the project’s own utility plan — then draw the as-fails graph. Which substation, which pump, which two access roads. Two days of work; the value runs the life of the project.

Hack: This Hack teaches you to compute NDWI — Normalized Difference Water Index, the same band-math that makes flood water glow on NASA’s composite — from a free Landsat 9 tile, so you can put yesterday’s floodplain on your site model this afternoon. Domain: Workflow. Tools: rasterio, numpy, one scene from USGS EarthExplorer.

import rasterio, numpy as np
with rasterio.open('LC09_B3.TIF') as g, rasterio.open('LC09_B5.TIF') as n:
    green, nir = g.read(1).astype('f4'), n.read(1).astype('f4')
ndwi = (green - nir) / (green + nir + 1e-9)   # > 0 ≈ open water
np.save('ndwi.npy', ndwi)

Mask ndwi > 0 for a binary water layer, polygonise, drop into Rhino or QGIS over your site. Thirty-metre resolution, decades of archive — enough to ground-truth a flood-risk overlay against actual events, not just modelled return periods.

Pick a river your office has work on this year. Pull two Landsat tiles — a dry-season one and the worst flood you can find — compute NDWI on both, and draw the dependency graph for the riskier scenario before the next brief lands on the desk.

Sources & Further Reading

• Primary: NASA Earth Observatory — Spring Rains Saturate Michigan