Is Niagara Falls Turned Off at Night? The Treaty and the Erosion

No. The 1950 Niagara Treaty never switches the falls off. It sets minimum flows over the Falls, and the International Joint Commission (IJC) gives them as plain numbers: not less than 2,832 m³/s (100,000 ft³/s) between 8 a.m. and 10 p.m. Eastern Standard Time from April 1 to September 15, the same figure between 8 a.m. and 8 p.m. from September 16 to October 31, and not less than 1,416 m³/s (50,000 ft³/s) "at any other time". That is the whole clause, read on 2026-10-27. The IJC adds that the flow in question is the American and Horseshoe flows added together.

A floor is not a flow. The sources we read do not give the actual hourly flow, so none of them can tell you how much water goes over the crest at 11 p.m. in July. The IJC directive that created the Niagara Board of Control in 1953 does ask the Board to report the amounts passing over each fall (IJC Board FAQ), and we found no public hourly series. One more wrinkle is ours, not the IJC's: the treaty says Eastern Standard Time, and United States daylight saving time runs from the second Sunday of March to the first Sunday of November (NIST), which covers every day of the tourist-hour period. Read literally, the daytime window is then 9 a.m. to 11 p.m. on a wall clock, and 9 a.m. to 9 p.m. in the autumn period. We found no source saying how the operators apply it.

The treaty tells the power plants what they must leave in the river. It does not tell a visitor what the falls will look like at ten o'clock.

If you want to see the falls after dark, the Niagara Falls at night page lists the tours that run then, and this site earns a commission if you book through it, at no extra charge to you. Our guide to choosing a waterfall tour is worth reading first.

Three pieces do the work. The treaty calls for one representative from each country to determine how much water is available and record how much the power plants take; the two governments designated them the International Niagara Committee in 1955. Their on-site representatives inspect the flows weekly, according to the IJC, and a shortfall against the minimum is investigated and entered in an annual report to both governments. The structure that makes the minimum possible is the International Niagara Control Works, 18 sluice gates about 0.8 km upstream of the Horseshoe Falls, operated by Ontario Power Generation and the New York Power Authority under the oversight of the IJC's Niagara Board. Raising or lowering the pool behind it changes both the flow over the falls and the water available for the power tunnels. Everything above the minimum, once other priority uses are met, is split equally between Canada and the United States, except 141.6 m³/s (5,000 ft³/s) of Long Lac and Ogoki diversions that belong to Canada alone. Sources: the IJC pages on the Niagara Committee and the Control Works, read 2026-10-27.

Twice, in a sense, and neither time was the treaty. In March 1848 the river itself stopped. Niagara Parks writes that the flow ceased for 30 hours after wind pushed the ice fields of Lake Erie into the head of the river and wedged them there; people who were used to the sound heard a strange silence, and some walked out on the exposed riverbed. The same page says the ice boom put in place each winter since 1964, between Fort Erie and Buffalo, has kept it from happening again. The wider story of ice and frozen falls has its own page, snowmelt, ice and frozen waterfalls.

The second stop was built. In June 1969 the U.S. Army Corps of Engineers closed a channel between Goat Island and the New York shore with a rock cofferdam, which sent the river to the Horseshoe and left the American Falls nearly dry for more than five months while engineers and geologists studied the rock at its foot; the dam was taken out on November 25, according to the Niagara Falls USA tourism site. Wikipedia gives the same outline and adds that a plan to clear the pile of fallen rock was dropped on grounds of expense. These are secondary sources, so hold the details loosely: we did not find the Corps' own report. Stopping a Niagara fall took a dam and a plan, and the Horseshoe carried the extra water.

Fast, once. The measured figure for the Horseshoe is 5 feet a year, give or take 1 foot, between 1842 and 1905, from G. K. Gilbert's comparison of crest surveys in USGS Bulletin 306, published in 1907. That is about 1.5 meters. For the American Falls he found less than 3 inches a year from 1827 to 1905. Spread over the whole retreat, the IJC's Niagara Board gives an average of about 1 meter a year, which fits 11 km in about 12,500 years (our division: 0.88 m). Then the number drops. The Board puts the current rate at about 0.1 meter (4 inches) a year, and a 2009 paper in Geomorphology by Yuichi Hayakawa and Yukinori Matsukura describes a historic decline from about 1 meter a year a century ago to about 0.1 now (DOI).

Then the sources split. The New York State Museum says the falls are migrating upstream on the order of one foot a year, about 0.3 meter, three times the IJC figure. The 2009 paper is blunter than either. It says the current rate is supposed to be less than 0.1 meter (4 inches) a year, and that no certain record of the recent recession is available, citing Tinkler (1993, 1994). So the number repeated most often today rests on estimates, not on a recent survey.

Treat the present rate as an estimate: the 2009 paper that cites it says no certain record of recent retreat exists.

Why it slowed matters more than the number. The usual explanation is the diversion of water to power plants, which began in the late 19th century. Hayakawa and Matsukura tested it against a model built on flow volume, crest length and rock strength. In their calculation the Horseshoe would retreat at about 0.4 meter a year if natural flow came back with today's crest shape, and at about 0.3 if flow had never been cut and only the crest had changed. Diversion is part of it. The rest is geometry: the Horseshoe crest lengthened from 420 m in 1842 to 762 m in 2000, and the IJC says erosion tends to slow when the crest is horseshoe-shaped and speed up when it narrows to a V-notch. The paper concludes that diversion is not the only reason for the slowdown. Its flow table, built from IJC and Tinkler figures, shows how much water there was to begin with.

One caution on that last row. The paper presents 1,770 m³/s as the annual mean under the treaty floors and does not show its working. Weighting the floors by the hours they apply gives about 1,890 m³/s by our arithmetic, so treat both as stand-ins for a flow that nobody published. The paper also dates the agreement to 1953; the IJC says the treaty was signed on February 27, 1950, and dates its directive creating the Niagara Board of Control to 1953, so we follow the IJC.

Less than people hope. Do not pick an hour for the water. Neither the treaty nor any source we read gives you a flow for 9 a.m. against 11 p.m., and the daytime window is wider than most sightseeing days, so an evening visit in July can still fall inside it on the literal reading above. If a listing says the falls are turned down or at full power at some hour, it repeats a claim we could not find in the treaty text, the IJC pages or Niagara Parks. If you are choosing between the two shores, the comparison of the border falls sets out what each operator says, and the Niagara Falls tours page sorts the ways of seeing both.

The erosion rate gives you no deadline either. At about 0.1 meter (4 inches) a year the crest would move roughly a meter in a decade, if the estimate holds, and the New York State Museum says the falls may reach softer rock upstream in about 15,000 years at current rates. What you cannot get from any source is tonight's flow or this year's retreat. What you can plan around is ice. Niagara Parks says a crust of ice can form over the spray and make the falls look stopped, while the water keeps running underneath.