Two methods, and the one you use depends on the size
Flow is area times velocity. Everything else is bookkeeping about how you got those two numbers.
For anything you can catch in a container — a spring box, a pipe outlet, a small chute — skip velocity entirely. Time how long a known volume takes to fill and divide. It is exact, it takes thirty seconds, and the only way to get it wrong is to miss part of the flow or trust a bucket that is not the size printed on it. Five gallons in 12.5 seconds is 24 GPM.
For an open channel you have to measure both halves. A tape across the water gives the width, a stick or a rod gives depths at equal spacing across it, and a float timed over a measured run gives velocity. The calculator uses the trapezoidal rule on the depths, which treats the two edge readings as half-weight because they sit on the water line, and the interior readings as full weight.
Why the float number is always too high before correction
A float sits on the surface and drifts with the fastest thread of water it can find. Velocity in a channel is not uniform: it is near zero at the bed and at both banks, and highest just below the surface in mid-channel. So a float measures roughly the maximum, and the mean over the whole wetted section is meaningfully lower.
The correction factor is the ratio between the two, and it is an input on this page rather than a constant because it depends on your channel. A smooth concrete ditch has a flatter velocity profile and a factor closer to 0.9. A rocky bed with weed and irregular banks drags far more, and the factor drops. If you use 1.0 you are asserting that the water at the bed moves as fast as the water at the surface, which it does not.
This is the single largest source of error in the method, and it is worth being honest that a float gauging is a good working number rather than a precise one. Anyone measuring for a permit application uses a current meter that reads velocity at points through the depth, because the correction disappears when you measure the profile instead of guessing it.
Picking the reach
The section you measure decides how good the answer is. What you want is a straight uniform stretch with the flow all going one way: no pool, no riffle, no eddy behind a rock, no back-current along a bank. If the float wanders across the channel or hangs up, move.
| What you see | What it does to the reading |
|---|---|
| Pool with slow deep water | Big area, low velocity, large error either way |
| Riffle over cobbles | Surface velocity far above mean, correction badly wrong |
| Straight run, even bed | What the method assumes |
| Weed or debris in the section | Area overstated, real flow lower |
| Wind straight up or down the channel | Pushes the float, adds error with no warning |
More depth readings are better on an uneven bed, and the marginal value of extra readings is high early and low later. Eight across a six foot channel is reasonable; two is a guess with arithmetic on top.
What the number is for
Flow is the input to almost everything else on a property that involves moving water. Combined with a head measurement it becomes power — the micro hydro power calculator takes exactly the GPM or cfs this page produces. On its own it tells you what a spring can supply, which sets the storage on the cistern sizing calculator. For a well rather than a surface source, the equivalent measurement is a drawdown test, which the well yield and drawdown calculator covers.
What flow is not is permission. A measured number tells you what the channel is carrying and nothing about what you may take out of it.
Taking water out of a stream, spring or ditch is regulated almost everywhere in the United States, and the rules are not the same in any two states. Depending on where you are it can involve a water right, a diversion permit, a fish passage or screening requirement, a minimum flow you must leave in the channel, and a separate review if the stream is navigable or carries listed species. Some of this is federal, most of it is state, and a little is local. Start with your state water resources or environmental quality agency before you move a shovel or a length of pipe. This page does arithmetic on numbers you measured and states no rule of any kind.
Questions people ask
How many depth readings do I need?
Enough that the profile of the bed shows up. On a smooth even channel six to eight readings across is plenty. On a bed with a deep slot on one side and a shelf on the other you want more, because the trapezoidal rule can only interpolate between the points you give it and a slot between two readings simply does not exist as far as the arithmetic is concerned. Take the first and last readings at the water edge, where the depth is zero or close to it, and space the rest evenly. If two adjacent readings differ by more than about half the deeper one, put another between them.
What float should I use?
Something that sits low and catches little wind. A partly filled bottle, an orange, or a short piece of waterlogged wood all work; a leaf or an empty bottle does not, because the wind moves it more than the water does. Start it well upstream of the measured run so it is up to speed before it crosses the start line, and keep it in the same thread of current on every run. Time from a fixed mark to a fixed mark rather than from a shout, and run it at least three times. If your times spread widely, the reach is telling you it is not suitable.
Can I use the bucket method on a stream?
Only if you can genuinely catch all of it, which in practice means a pipe, a spring box outlet or a chute you have built. The failure mode is silent: water that splashes past the container, or that keeps flowing while you are swapping containers, simply never appears in the answer, and you get a confident number that is too low. If the flow overshoots or sprays, it is not a bucket measurement. For anything in an open channel, use the cross-section method instead and accept the correction factor as the price.
Why does my measurement matter more than the average annual flow figure?
Because averages hide the shape. A stream that averages 100 GPM across the year might run 400 in the snowmelt and 15 in September, and a system built on the average is oversized for nine months and starved for three. Any scheme that has to work continuously is sized on the low flow, not the mean. That is also why this page asks for your own judgement about the driest month rather than applying a factor: you know your watershed and nobody at a distance does. If you can only take one measurement, take it in the dry season.
Do I need a permit to measure the flow?
Measuring is not diverting, but the moment you propose to take water out, or to build anything in or near the channel, you are in regulated territory, and the rules vary by state and sometimes by watershed. That can include water rights, diversion permits, screening and fish passage, minimum instream flows, and separate review for navigable water or listed species. Start with your state water resources or environmental quality agency. Getting the measurement first is useful, because the first thing they will ask is how much water is there and how you know.