The short answer
The Salton Sea is California’s largest lake by surface area, and it is there by accident. The state’s water board says today’s sea formed in 1905, when Colorado River floodwater broke through an irrigation canal and flowed uncontrolled into the Salton Basin for 18 months. Water board staff report, April 2026, p. 2; Water board fact sheet, 2024
The break started with a decision. In the fall of 1904 the California Development Company, a private water company short of money, cut a new intake straight from the river to its canal about four miles south of the border, inside Mexico. It put no gate on it. Senate Document 212, p. 3; Cory, Transactions of the ASCE, vol. 76, p. 1287 In 1905 five floods came down the river, tore the cut open, and by that summer the whole Colorado was running downhill into the Salton Sink, a dry basin far below sea level. President Theodore Roosevelt told Congress that making that cut “without controlling devices was criminal negligence.” Senate Document 212, p. 3
It took the Southern Pacific railroad two rock dams and 2,157 carloads of rock to force the river back, on February 10, 1907. Cory, p. 1367 The sea the river left behind never dried up, because the farms around it kept feeding it their drain water. Since a 2003 water deal sent water saved on those farms to Southern California’s cities, the sea has been shrinking.
Watch the episode
The film follows the cut through the government’s own record: the 1907 report President Roosevelt sent to Congress, with its photographs and the telegrams between Roosevelt and the railroad’s boss, E. H. Harriman, and the 1913 paper by H. T. Cory, the engineer who closed the break. It puts their exact words on screen as the narrator reads them. 3D maps on NASA and USGS Landsat imagery show the river turning into the desert, and a side cut through the riverbed shows how the waterfall ate its way upstream. The photographs and footage are real. The 3D diagrams are illustrations built from the record, with heights exaggerated, not an official reconstruction. AI-generated images appear only as plain backgrounds behind the quotations and act titles, and none of them shows the events.
A valley below the river
Southern California’s Colorado Desert has a deep bowl in it. Roosevelt’s message to Congress, drawing on the U.S. Geological Survey and the Reclamation Service, described the farmland at its bottom as lying 200 feet or more below the level of the Colorado River. North of the Mexican border, low hills keep the river out. South of the border, only the river’s own low mud banks stood between it and the dry valley floor. Senate Document 212, p. 1
The river had been in that bowl before. Over thousands of years the Colorado filled it again and again with freshwater lakes that later evaporated. Water board fact sheet, 2024; Carnegie Institution, The Salton Sea, 1914, p. 5
The California Development Company, a New Jersey corporation formed in 1893, set out to farm it. Its engineer, Charles R. Rockwood, brought river water in by a canal that left the Colorado just north of the border, looped through Mexico around the sand hills through a Mexican subsidiary, and came back into California. Senate Document 212, pp. 2, 31-32 The promoters renamed the Colorado Desert the Imperial Valley. Senate Document 212, p. 31 By September 1904, according to the Bureau of Reclamation, nearly 8,000 settlers were irrigating 75,000 acres through 700 miles of canals. Bureau of Reclamation, All-American Canal System
A broke company and a muddy river
Roosevelt told Congress the company had large plans and too little money. Every structure it built to control the water was temporary, made of wood and as small as possible, and the money settlers paid for their water rights did not go into permanent works. Senate Document 212, pp. 2-3
The Colorado also carries a huge load of mud. Silt settled in the company’s canals faster than its dredges could clean it out. By 1903 and 1904 the valley was short of water, and in the spring of 1904, at the start of the hot season, the canal’s intakes silted up. Senate Document 212, pp. 32-33; Cory, p. 1286
The cut with no gate
The company’s lawyer, Edwin A. Meserve, wrote the company’s history for Congress. He recorded that in June 1904 Rockwood told the company’s president, A. H. Heber, that they would have to “take the chance” and cut a new intake below the clogged section, where the canal ran close to the river. The company had no money for structures and no time to get plans approved in Mexico City. In the lawyer’s account, the cut was made to save the valley’s livestock and the crops that weren’t already dead. “Take the chance” is his paraphrase of Rockwood. Senate Document 212, p. 33
Mexico’s 1904 concession let the company take water from the river inside Mexico, but no intake could be built there until Mexico’s engineers had approved the plans for its structures. Senate Document 212, pp. 32-33 Rockwood later wrote that the company hesitated for that reason, not because it feared the river would break through, until its attorney in Mexico City telegraphed it to go ahead. Cory, pp. 1286-1287
According to Rockwood, the cut was begun late in September 1904 and finished in about three weeks. It was 3,300 feet long, about two-thirds of a mile, straight through soft river silt. He sent gate plans to Mexico City in early November 1904; they were not approved until December 1905. Cory, p. 1287 In the lawyer’s account, Rockwood asked for lumber to build a controlling gate a month before the floods were expected. Heber promised it, but the company never had the money and the lumber never came. Senate Document 212, p. 33 By December 1904, water from the new cut was already reaching the Salton Sink. Senate Document 212, p. 33
Five floods
Rockwood’s plan was to close the cut before the summer flood with a dam of brush and earth, the method the company had used to close its cuts for years. He was betting on the river records. The gauge at Yuma, he said, had been read for 27 years, and in that time there had been only three winter floods, and never two in one winter. Cory, pp. 1287-1288
In 1905 there were five. In March, crews threw a barrier of piles, brush and about 10,000 sandbags across the cut, and the river undermined it. A fourth and a fifth flood swept away the dams that followed. Cory, pp. 1288-1289 The floods widened the cut from about 60 feet to about 150. Its banks caved in, and by the summer of 1905 the whole Colorado River was running through the canal toward the sink. Roosevelt put it in the spring floods; Rockwood wrote that the entire river was running through the canal by August. Senate Document 212, p. 3; Cory, pp. 1289-1290
The waterfall that walked upstream
The river had two ways to go from the break. Toward the Gulf of California it dropped about 100 feet in 80 miles, about a foot and a quarter a mile. Toward the bottom of the Salton Sink it dropped about 387 feet in 95 miles, about four feet a mile. That route was more than three times as steep, and the river took it. Cory, pp. 1292-1293
A new sea spread across the floor of the sink. It drowned the New Liverpool Salt Company’s works and forced the Southern Pacific to move its main line to higher ground; the geologist W. P. Blake recorded 67 miles of track moved. Cory, p. 1293; Carnegie Institution, p. 5
Racing downhill, the water dug the beds of the New and Alamo rivers deeper, and the deepening worked its way back upstream as a moving waterfall. Water pouring over a drop digs out the soft ground at its foot; the edge breaks off, and the falls stand a little further upstream. The government’s photographs, printed for Congress in January 1907, show falls about 1,500 feet wide, roughly a quarter of a mile, and 40 to 60 feet high, “cutting back upstream at the rate of about one-third mile per day.” Senate Document 212, photograph plate; Cory, p. 1323
When the falls reached the border towns, they carried away part of Mexicali, including its railroad station and a brick hotel, and threatened Calexico across the line. Cory, p. 1323; Senate Document 212, photograph plate No source we read records any deaths.
Roosevelt warned Congress what would follow if the break wasn’t closed before the spring flood of 1907. The channel would cut so deep that water could no longer flow by gravity into the valley’s canals, which would wipe out its farms and towns, and the cutting could work back up the river past Yuma to the government’s Laguna Dam. Senate Document 212, pp. 3-4
The railroad’s first rock dam
By then the railroad was running the water company. On June 20, 1905, the Southern Pacific lent it $200,000. In return, stockholders pledged 6,300 shares, the railroad named three directors, including the company’s new president, Epes Randolph of the Harriman lines, and it took over management. Senate Document 212, pp. 20-21; Cory, pp. 1291-1292 Cory estimated the railroad stood to lose $1 million a month if its southern transcontinental line was cut. Cory, p. 1345
In April 1906 Rockwood gave up his post as chief engineer, and H. T. Cory became the company’s general manager and chief engineer. Cory, pp. 1314-1315 On October 11, 1906, the company’s new headgate, a structure the company’s lawyer said cost about $150,000, buckled and went out under the river. Cory, pp. 1341-1342; Senate Document 212, p. 35
Cory turned to rock. His crews built railroad trestles across the break and dumped rock from trains faster than the river could carry it away. After fifteen months of other attempts, the rock dam stopped the river in 24 days, and on November 4, 1906, the Colorado was back in its old channel. Cory, pp. 1345-1346, 1349; Senate Document 212, p. 3
Harriman and Roosevelt
It held for about a month. On December 7, 1906, a flood got under the new levees just south of the dam, levees Roosevelt called poorly constructed, and within days the entire river was flowing into the valley again. Senate Document 212, pp. 3, 12
On December 13, E. H. Harriman, president of the Southern Pacific, telegraphed Roosevelt that his railroad had already spent about $1 million on the river and about as much again on its tracks, and: “it does not seem fair that we should be called upon to do more than join in to help the settlers.” Senate Document 212, p. 9 A few days later he asked for the government to do the work, under the Reclamation Service, with the railroad lending trains, tracks and quarries. Senate Document 212, p. 10
Roosevelt refused on December 20. The government could not act inside Mexico without Congress and a treaty, he answered, and Congress was adjourning that day. “It is incumbent upon you to close break again,” he wired, and “there is not the slightest excuse for the California Development Company waiting an hour for the action of the Government.” Senate Document 212, pp. 10-11
Harriman replied the same day: “you seem to be under the impression that the California Development Company is a Southern Pacific enterprise. This is erroneous.” In the same telegram, he ordered his officers back to close the break, trusting the government to help with the cost later. Senate Document 212, p. 11
On January 2, 1907, Roosevelt wired that he had been told the railroad’s people had taken a majority of the company’s stock and elected its officers, and asked whether that was so. Harriman answered that the railroad held the 6,300 shares only as collateral for its loan, and that the contract let it choose three directors and the president. Senate Document 212, pp. 17-19 The company’s own lawyer called the pledged shares “a majority of the stock.” Senate Document 212, p. 34
2,157 carloads of rock
The second closing was harder. Crews drove two railroad trestles across the new break, with piles up to 90 feet long, in a strong current, and floods tore out the first trestle three times. On January 14, 1907, a railroad superintendent wired that they had used up every available pile in San Diego and Southern California. Cory, p. 1366
Rock dumping from the finished trestle began at 5 p.m. on January 27, 1907, from big side-dump cars the crews called “battleships.” By the next daylight, 145 of them had been unloaded. Cory, p. 1367 At 11 p.m. on February 10, 1907, the second break was closed and all the water was going down the river’s old channel to the Gulf. Before the water was shut off, 2,157 carloads of rock had gone into the gap. Cory, pp. 1367-1368
That same day the Salton Sea reached its highest level, about 84 feet deep at its deepest. Blake gave it as about 45 miles long and up to 17 miles wide that month. Carnegie Institution, pp. 5, 117 The sources disagree on its area at the time, so we don’t give one.
Criminal negligence?
Roosevelt told Congress: “The making of this cut in a bank composed of light alluvial soil above a depression such as this without controlling devices was criminal negligence.” Senate Document 212, p. 3 That was his judgment in a message to Congress. No one was charged in the record we read.
Rockwood never accepted it. “We have since been accused of gross negligence and criminal carelessness in making this cut,” he wrote in 1909, “but I doubt as to whether any one should be accused of negligence or carelessness in failing to foresee that which had never happened before.” Cory, p. 1287
Cory worked for the railroad that had taken over Rockwood’s company. He wrote that when he first saw the break in August 1905, he thought the cut “practically criminal.” Four years later he was convinced it had been “absolutely imperative,” because the settlers needed the water. A headgate should have been provided, he wrote, and “The real mistake was not in ‘putting all the eggs in one basket,’ but in not ‘then watching that basket.’” Cory, pp. 1290-1291 He also argued that the 1905 floods were so great they would probably have flooded the sink anyway, as a flood in 1891 had put six feet of water over the salt works. Cory, p. 1293 That is his opinion. The record shows that the water which made today’s sea ran through the company’s cut for about 18 months.
In January 1907 the director of the U.S. Geological Survey, Charles D. Walcott, advised that no artificial cut in the river’s west bank should ever be allowed again, and that water should be taken from the river only where the headworks could stand on solid rock. Senate Document 212, pp. 7-8 That took more than 30 years. After Hoover Dam, construction of the All-American Canal began in 1934. Imperial Dam, which feeds it, was built from January 1936 to July 1938, and the first irrigation water came through in 1940. The canal runs about 80 miles, entirely inside the United States, so the valley’s water no longer has to pass through Mexico. Bureau of Reclamation, All-American Canal System
Why the sea never dried up
Cut off from the river, the new sea was expected to disappear like the lakes before it. In 1923 the Geological Survey reported it shrinking by evaporation at five or six feet a year, and predicted that waste water from the valley’s irrigation would eventually balance the evaporation and hold the sea at about a constant size. USGS Water-Supply Paper 497, pp. 68-69
That is what happened. For decades the sea was kept up mainly by drain water from the farms, about 80 percent of it from the Imperial Valley. Water board staff report, April 2026, p. 2; Water board fact sheet, 2024
Why it is drying now
In 2002 the state approved a long-term transfer of water conserved on Imperial Irrigation District farms to San Diego and other Southern California water agencies, part of the 2003 Quantification Settlement Agreement. To soften the blow to the sea, the district had to send it 15 years of “mitigation water.” That ended on December 31, 2017. Water board staff report, April 2026, p. 4; Water board fact sheet, 2024; Order WR 2017-0134, Exhibit A, finding 1
As of March 10, 2026, the water board’s staff report, citing the Pacific Institute, put the sea’s surface at 243 feet below sea level, about 14.6 feet lower than in 2003. About 24,490 acres of lakebed were exposed, roughly 38 square miles, and the sea was shrinking by about 2,400 acres a year, close to four square miles. Water board staff report, April 2026, p. 3 The water is now more than twice as salty as the Pacific Ocean, and algae fed by farm runoff are depleting its oxygen, which makes it hard for the fish and birds that depend on it. Water board staff report, April 2026, p. 2
The exposed lakebed holds decades of farm and waste-water residue. The water board says wind lifts dust from it that adds to poor air quality in the communities around the sea and is a public health risk for the people who live there. Water board staff report, April 2026, p. 2; Water board fact sheet, 2024 We have not read health studies of the dust.
The 2028 deadline
The state’s 2017 order requires habitat and dust-control projects on 29,800 acres of the exposed lakebed, about 47 square miles, to be finished by the end of 2028. Order WR 2017-0134, condition 24; Water board staff report, April 2026, p. 5, Table 1 We have not checked how far along those projects are; the program’s 2026 annual report was not reachable when we did our research.
So why is there a sea in the California desert? Because in the fall of 1904 a water company with no money cut an open trench into the Colorado River’s bank, without Mexico’s approval and with no gate, and bet on the weather. Five floods came. It took a railroad, more than two thousand carloads of rock and a standoff with the president to push the river back, and the sea it left is still there, drying into a dust problem of its own. When the end-of-2028 deadline arrives, we will check whether California delivered.
Primary sources
- U.S. Senate, Imperial Valley or Salton Sink Region: Message from the President of the United States, S. Doc. No. 212, 59th Congress, 2d Session, January 12, 1907 (Library of Congress copy scanned by the Internet Archive). Roosevelt’s message (printed pp. 1-7), the Geological Survey director’s statement (pp. 7-8), the Roosevelt-Harriman telegrams and the loan contract (pp. 9-26), the company attorney’s history (pp. 31-37) and the government’s photographs. Printed page numbers are cited above.
- H. T. Cory, “Irrigation and River Control in the Colorado River Delta,” Transactions of the American Society of Civil Engineers, vol. 76, December 1913, pp. 1204-1453, with C. R. Rockwood’s own account quoted. Printed page numbers are cited above.
- D. T. MacDougal and collaborators, The Salton Sea, Carnegie Institution of Washington Publication No. 193, 1914 (W. P. Blake’s chapter, p. 5; MacDougal’s, p. 117).
- J. S. Brown, The Salton Sea Region, California, U.S. Geological Survey Water-Supply Paper 497, 1923, pp. 68-69.
- U.S. Bureau of Reclamation, Boulder Canyon Project: All-American Canal System, project history.
- California State Water Resources Control Board, Order WR 2017-0134 with Exhibit A, November 7, 2017.
- California State Water Resources Control Board, Salton Sea fact sheet, 2024.
- California State Water Resources Control Board, Division of Water Rights, Staff report: Annual Public Workshop on the Status of Phase 1 of the Salton Sea Management Program, April 2026 (PDF pages cited above).
