Film 01 · Old River, Louisiana31°04′36″N 91°35′51″W

1 November 2026 · 8:34

The Mississippi Is 20 Feet Above Its Escape Route

Watch on YouTube ↗

1 · The short answer

Water wants the shorter, steeper way

At Old River, in Louisiana, the Mississippi runs beside a channel that leads to another river, the Atchafalaya. Down the Mississippi it is 315 river miles to the Gulf of Mexico; down the Atchafalaya, 142 2. On 1 April 2019 the Mississippi at Knox Landing stood 6.1 m (20 ft) above that channel 1. Water wants the shorter, steeper way.

Deltas do this all the time: over the last seven thousand years the Mississippi has switched course every one to two thousand years 4. By 1950 the Atchafalaya was taking close to a third of the flow, and in 1951 the Mississippi River Commission predicted it would capture the main river “between 1968 and 1985, probably around 1975” 6. Congress answered in 1954: a control structure at Old River, to hold the split at about 30% 8. Measured in 2008–2010, it was 29.5% 7.

The structure nearly failed in the flood of April 1973 9. We re-ran the 2019 flood on the real ground, checked against river gauges and satellite flood maps 15. With the same flows and the structure removed, the Atchafalaya’s share went from 28% to about 45% within six days. The model holds the riverbed fixed, so it leaves out the scouring that made 1973 dangerous: the real number would likely be higher.

2 · Key figures

The step, the shortcut, the test

1 April 2019: the Mississippi at Knox Landing stood 6.1 m (20 ft) above the outflow channel.Fig. 01Data-based reconstruction31°04′36″N 91°35′51″W
1 April 2019: the Mississippi at Knox Landing stood 6.1 m (20 ft) above the outflow channel. [1]
Two ways to the Gulf: 315 river miles down the Mississippi, 142 down the Atchafalaya.Fig. 02Measured data31°04′36″N 91°35′51″W
Two ways to the Gulf: 315 river miles down the Mississippi, 142 down the Atchafalaya. [2]
Same flows, structure removed: the Atchafalaya’s share goes from 28% to about 45% by day 6. Simulation.Fig. 03Simulation
Same flows, structure removed: the Atchafalaya’s share goes from 28% to about 45% by day 6. Simulation. [15]

3 · Evidence

Legend · how we checked the model

Measured dataData-based reconstructionSimulation
0.15–0.84 mmodel vs. four river gauges (measured levels)
0.66flooded area vs. Sentinel-1/2 (critical success index; 1 = perfect)
0.64vs. the independent Copernicus Global Flood Monitoring map
28→45%Atchafalaya’s share by day 6 with the structure removed: simulation, not a forecast
What was modelled
The flood of 1 April 2019, with the measured river flows (Mississippi 51,300 m³/s, Red 3,850 m³/s), on USGS lidar ground, with river beds from USACE surveys where they exist.
Against the gauges
Water levels within 0.15–0.84 m of four gauges (Knox Landing, Red River Landing, Old River outflow channel, Simmesport). Atchafalaya flow 15,000 m³/s against 15,600 measured.
Against the satellites
Flooded area compared with Sentinel-2 and Sentinel-1 radar: critical success index 0.66, and 0.64 against the independent Copernicus Global Flood Monitoring map. 1 is perfect; above 0.7 is considered good and below 0.5 poor.
Where it misses
It floods some farmland west of the Red River that the satellite saw dry (small levees and roads), and leaves dry a flooded forest that the radar saw underwater: the model gets water only from the rivers, not rain.
What “without the structure” means
Same model, same 2019 flows, structures removed and channels open, riverbed held fixed. It shows where the water would go in the first days, not how the river would dig a new channel over years.

Corrections. None so far. If you spot an error, write to hello@reasonofplace.com. Corrections are published here with the date.

More figures5 figures
Turnbull’s Bend and Shreve’s Cut, 1831, over the 1839 survey.Fig. 04Data-based reconstruction
Turnbull’s Bend and Shreve’s Cut, 1831, over the 1839 survey. [5]
The split the structure holds: about 70/30. Measured 2008–2010: 29.5%.Fig. 05Measured data31°04′36″N 91°35′51″W
The split the structure holds: about 70/30. Measured 2008–2010: 29.5%. [7] [8]
April 1973: the scour hole at the Low Sill structure, about 90 m across and 18 m deep.Fig. 06Data-based reconstruction
April 1973: the scour hole at the Low Sill structure, about 90 m across and 18 m deep. [9]
18 May 2011: 17 gates open at the Morganza Spillway.Fig. 07Data-based reconstruction
18 May 2011: 17 gates open at the Morganza Spillway. [10]
Our model: the April 2019 flood on lidar terrain, model hour 144.Fig. 08Simulation31°04′36″N 91°35′51″W
Our model: the April 2019 flood on lidar terrain, model hour 144. [15]
All sources15 sources
  1. USACE CWMS (Knox Landing) and USGS gauge 07381482, 1 April 2019 (our calculation; ~0.3 m datum uncertainty).USGS 07381482 ↗ CWMS data ↗
  2. USACE New Orleans District, Old River Control (2009): route lengths.PDF (copy) ↗
  3. Edmonds 2012, Geophysical Research Letters 39, L08402.DOI ↗
  4. Roberts 1997; Frazier 1967; Coleman, Roberts & Stone 1998, Journal of Coastal Research 14(3).Roberts 1997, PDF ↗ Coleman et al. 1998, PDF ↗
  5. Fisk 1952 (plate 2, survey by Geo. T. Dunbar, 1839); Mississippi River Commission 1881.Fisk 1952, PDF ↗
  6. Mississippi River Commission, The Atchafalaya River Study (1951), Vol. 1, pp. 36 and 43.ERDC Library ↗
  7. Allison et al. 2012 (29.5%, 2008–2010).PDF ↗
  8. Flood Control Act of 1954 (P.L. 83-780); House Document 478 (1954); USACE, Old River Control (2009).USACE 2009, PDF (copy) ↗
  9. USACE ERDC SR-23-2, TR-22-22 and MP-25-3; USACE photographs (public domain).SR-23-2 ↗ TR-22-22 ↗ MP-25-3 ↗
  10. USACE Mississippi Valley Division, post-flood report (2011); National Weather Service.USACE release ↗ NWS, May 2011, PDF ↗
  11. NWS and USGS gauges (2019: 211 days above flood stage at Baton Rouge).NWS gauge BTRL1 ↗ USGS 07374000 ↗
  12. Andrus & Bentley 2023, Journal of Coastal Research 39(6).DOI ↗
  13. Couvillion et al. 2017, USGS Scientific Investigations Map 3381.USGS ↗
  14. H. A. Einstein, USDA Technical Bulletin 1026 (1950).AgEcon Search ↗
  15. Our simulation, checked against USGS gauges, Sentinel-1/2 (Copernicus) and the Copernicus Global Flood Monitoring map; Bernhofen et al. 2018 for the CSI scale.How we checked the model Bernhofen et al. 2018, DOI ↗

Imagery: EOxCloudless 2016 by EOX (CC BY 4.0, contains modified Copernicus Sentinel data 2016) · Contains modified Copernicus Sentinel data 2019 · © OpenStreetMap contributors (ODbL) · USGS 3DEP lidar · USDA NAIP · USACE · NASA Black Marble and Blue Marble · EC JRC/Google Global Surface Water (Pekel et al. 2016).