A levelling sheet, reduced twice, with the check LEVEL
If the three totals disagree, something is wrong now, while the instrument is still up, rather than in three weeks when a drain runs uphill.
The run
Leave the closing level empty on a run that does not close back onto a known point; the arithmetic check still applies, and it is the half that catches most mistakes.
The readings
A change point carries both a foresight and a backsight: the reading that closes one setup and the reading that opens the next, to the same staff position.
The check
The sheet
The working
How it works
Levelling is the one piece of site arithmetic that proves itself
Reduce the readings by the height of the instrument and you get a column of levels. Reduce the same readings by rises and falls and you get another. If the field work and the arithmetic are both sound, three things are equal:
Σ backsights − Σ foresights
= Σ rises − Σ falls
= last level − first level
That is not a nicety, it is the point. Almost no other calculation on a site tells you it is wrong. This one does, and it tells you now, while the instrument is still set up and the staff is still there, rather than in three weeks when a drain runs the wrong way.
So both reductions are done here and the check is shown. If the three do not agree, stop and find out why before anybody leaves.
What the check catches and what it does not
It catches arithmetic. A transposed digit, a reading entered in the wrong column, a line added twice: any of those breaks the equality and is found in the moment.
It does not catch a reading that was written down correctly and taken wrongly. A staff not held plumb, a staff on soft ground, a bubble not centred, a collimation error in the instrument, the bottom section of the staff not fully extended: all of those produce a sheet that balances perfectly and levels that are wrong. The check is a proof of the arithmetic, not of the survey.
Which is why the field discipline is separate and not optional. Keep backsight and foresight distances roughly equal, so a collimation error cancels. Close the run back to the benchmark you started from, or on to another known one, so the misclosure is a number you can look at. Change point on something solid, and describe it, so the run can be rerun.
Booking it so it can be read afterwards
Write the benchmark down properly: its value, where it is, and where the value came from. A sheet of levels with no stated datum is a sheet of relative numbers, and it cannot be joined to anybody else's work.
Describe every change point and every station in words. “CP1 on the manhole cover, NE corner of the yard” is what lets somebody stand where you stood. “CP1” is not.
Keep the misclosure and say what you did with it. A run that closes within tolerance and is distributed is honest work; a run that closes badly and is quietly adjusted is not, and the difference is only visible if the raw figures are still on the sheet.
And book it as you go rather than reconstructing it in the van. The reason the check exists is that levelling is done by tired people writing numbers in a notebook in the rain, and every part of the procedure is designed around that being true.
The arithmetic runs here in the tab and nothing is uploaded. Nearby: CUT turns a cutting list into which bar each piece comes off, CONCRETE works out what a pour actually needs, RAMP handles gradients and the landings people forget, and FIGURES keeps the rounding honest.
The guide: Rise and fall or height of collimation: which to use, and the check that has to balance
Questions people ask about LEVEL
What is the check?
Three quantities that have to be equal if the sheet is right: the sum of the backsights less the sum of the foresights, the sum of the rises less the sum of the falls, and the last level less the first. It is taught everywhere and skipped almost everywhere, because doing it by hand means adding four columns and each one is another chance to make the mistake you are looking for.
Why reduce it twice?
Because agreeing totals can hide a pair of errors that cancel. The two reductions are independent, so comparing them row by row finds the line the mistake is on rather than telling you that somewhere in forty readings there is one. That is the difference between a check and a useful check.
What is the difference between the two methods?
Height of collimation works out the level of the instrument once per setup and subtracts each reading from it, which is quick and gives no check on the intermediate sights. Rise and fall compares each reading with the one before it, which is slower and checks every line. Doing both gets the speed and the check.
What does the misclosure tell me?
Whether the field work was good enough, which is not the same question as whether the arithmetic was. A run that balances perfectly and closes forty millimetres out has correct arithmetic on bad readings. The allowance grows with the root of the number of setups, because errors accumulate as a random walk rather than in a line.
Does this work without a signal?
Yes, and that is the point. There is no network call on this page at all, so it works in a trench, in a field or at the bottom of a hole, which is where levelling happens.
Doing this often? Foundry, the Engineering Suite's drawing half draws the levels onto a plan to scale, with dimensions, and prints a PDF to scale or saves DXF. On its own, US$2 a month; with everything else, US$5 a month or US$48 a year.