This differential leveling calculator reduces a run of staff readings to reduced levels, compares the result with the benchmark you closed on, and hands the misclosure back out to every point in proportion to the work done — by distance if you measured it, by number of sights if you did not. Start from a known level, enter back and fore sights down the line, and read the misclosure, the allowable tolerance and the adjusted levels. Free, in your browser, no login.
One line per sight, in the order it was taken. Fill Back for a back sight, Fore for a fore sight, or both when the instrument moved at that point (the change point). Add a distance for each sight if you paced or taped it — leave it at 0 and the distance method falls back on sights. Blank last rows are ignored.
Check the readings — every sight needs a positive back or fore reading, and a line that closes on a benchmark or on the start needs the last reading to be a fore sight.
| Point | Back | Fore | Rise | Fall | RL | Dist | Corr | Adj. RL |
|---|
Arithmetic check: Σ back − Σ fore must equal Σ rise − Σ fall, and that value must equal the last RL minus the first. The calculator tests all three and says so below the table.
A level run gives pairs of staff readings. At each change point the instrument is read on the
staff held at the previous point — the back sight — and then on the staff at the new point —
the fore sight. If the fore sight is smaller than the back sight the ground went up, and the
rise is back − fore; if it is larger the ground fell, and the fall is
fore − back. Each reduced level is the previous one plus the rise, or minus the fall.
The two columns have to balance. Summing gives Σ back − Σ fore = Σ rise − Σ fall = RL_last − RL_first,
which is an arithmetic check that catches a transposed digit in the reduction but not a wrong
reading in the field — that only shows up as a misclosure.
The same numbers come out of the other bookkeeping method: after each back sight the instrument
height is HI = RL_back + back, and the fore sight gives
RL_fore = HI − fore. It is the same arithmetic written once per set-up instead of once
per sight. This calculator runs rise and fall, because the two column checks above fall out of it
for free, but the pairing is what matters: a fore sight is only ever reduced against the instrument
height its own back sight produced.
If the run ends on a known level — a benchmark, or the starting point again — the computed level
will not agree with it exactly. The difference is the misclosure:
e = computed − known. A positive misclosure means the run came out high, and every
adjusted level has to be brought down.
With the distance rule, the correction at a point is
c_i = −e × (d_i ÷ Σd), where d_i is the cumulative distance to that point.
It is proportional and it closes exactly: at the last point the sum of the fractions is 1. The
sights rule does the same thing with the number of sights in place of distance, which is the honest
choice when you have no distances and the set-ups were of similar length. Neither rule improves the
survey; they both just spread a number you cannot remove.
The usual standard is proportional to the square root of the length of the run,
e_allow = k × √L, with L in km and k in mm. The square root
reflects the way random errors accumulate: a run twice as long has about 1.4 times the error, not
twice. The order/class system in the table below is the long-standing FGCS scheme (Federal Geodetic
Control Subcommittee, 1984); it is still what most specifications quote, even where the current
positional-accuracy standard has replaced orders with uncertainty values.
| Order / class | Max loop misclosure | Typical use |
|---|---|---|
| First order, class I | 4 mm × √K | National framework; deformation and research networks |
| Second order, class I | 6 mm × √K | Metropolitan control, large engineering projects |
| Second order, class II | 8 mm × √K | Local control networks, topographic mapping |
| Third order | 12 mm × √K | Construction, small projects, local surveys |
K is the length of the loop in kilometres. For a one-way line between two benchmarks (a section rather than a loop) first order class I tightens to 3 mm × √K; the other classes keep the same constant. China's GB/T 12898 applies a similar √L rule with different constants — ±12√L mm for third order and ±20√L mm for fourth order on flat ground, relaxed to ±15√L and ±25√L mm in mountainous terrain. Always take the figure from the project specification.
Where the specification gives a fixed figure per set-up instead of a distance rule, divide it by the square root of the number of set-ups. And if the misclosure is more than about twice the allowance, adjusting is pointless: re-observe the run.
Almost every large misclosure comes from four habits, and none of them is arithmetic.
Equalise the sights. Keep the back sight and the fore sight the same length at each set-up. The curvature and refraction error and the residual error of the instrument's line of sight both grow with the difference between the two distances; if the sights are equal they cancel. It is the single most effective thing you can do, and it costs nothing but walking.
Change points properly. A change point must sit on solid, stable ground, be read on a straight staff held vertical, and be the same point for the closing and the opening sight. Use a turning plate on soft ground. Two hand spans from where you placed the staff to where you thought you placed it is a real error, and it enters the run twice.
Read to the full station. Read all three hairs and check that the middle hair is the mean of the other two; a staff read to one millimetre on one sight and guessed on the next shows up in that check, not in the reduction.
Close the job, not just the loop. A run that returns to its own start proves only that no gross error was made; it does not detect a reference mark that has moved. Close onto an independent benchmark whenever one exists, and note the difference — sometimes the benchmark is what has moved, and that is a finding, not a failure.
The default table is a closed loop of four set-ups returning to a benchmark at 50.000 m.
Each line is one staff position, so the distance column is the length of that sight and the
cumulative distance is the running total from the start.
| Point | Back | Fore | Rise | Fall | RL | Cum. dist (m) |
|---|---|---|---|---|---|---|
| BM (start) | — | — | — | — | 50.000 | 0 |
| P1 | 1.428 | 1.536 | — | 0.108 | 49.892 | 300 |
| P2 | 1.502 | 1.494 | 0.008 | — | 49.900 | 600 |
| P3 | 1.455 | 1.512 | — | 0.057 | 49.843 | 900 |
| BM (close) | 1.507 | 1.483 | 0.024 | — | 49.867 | 1200 |
Both arithmetic checks agree: Σ back = 5.892, Σ fore = 6.025,
Σ rise = 0.032, Σ fall = 0.165, and in each case the net is
−0.133 m, which is also 49.867 − 50.000. The loop closes
133 mm low.
Over 1.2 km the allowance at k = 4 mm/√km is
4 × √1.2 = 4.38 mm, and at k = 8 it is 8.76 mm. The
misclosure is 30 times the tighter figure, so this run fails every order in the table — which is
what a survey with a blunder looks like. The miss is not spread evenly through the four set-ups
either: 108 mm of it arrives on the very first one, which is where you would look first.
Adjusting anyway, the distance rule gives −e × (cum ÷ 1200):
+33.3 mm at P1, +66.5 at P2, +99.8 at P3 and
+133.0 at the benchmark, taking the adjusted levels to
49.925, 49.967, 49.943 and exactly 50.000.
Switch the method to number of sights and each of the four gets +33.3 mm
instead. Neither method makes the loop better than it is — both only spread a number you cannot
remove, which is why the ratio to the allowance is on the result panel and the misclosure is not.
Reduce the readings rise and fall, take the difference between the computed and the known closing level as the misclosure, then give every point a correction proportional to the distance (or the number of sights) up to it. The corrections sum to minus the misclosure, so the adjusted run closes exactly on the benchmark.
It is the difference between the level the run gives for the closing point and the level that point is known to have. A positive misclosure means the run came out above the truth. It is not an error you can find and remove after the fact — the adjustment only spreads it, and the size of the misclosure against the allowance is the real measure of the work.
By distance if you have the distances, because it is what the tolerance is expressed against and it weights a long sight more than a short one. If you did not measure, by number of sights is the honest approximation, provided the set-ups were of similar length. Never adjust by distance that you made up so the two methods agree.
Typically a constant in millimetres times the square root of the run in kilometres — 4 mm for precise control, 12 mm for site leveling, 20 mm for rough work. Take the figure from the project specification. If the misclosure is more than about twice the allowance, adjust nothing and re-observe.
Because the earth's curvature, atmospheric refraction and any residual tilt in the instrument's line of sight all act in proportion to distance. At a set-up with equal sights those effects are identical on both readings, so they cancel in the difference. Unequal sights leave the difference of the two errors in the run.
A point that is read twice: as a fore sight from the set-up you are leaving and as a back sight from the next one. It is the only point that carries the level from one instrument position to the next, so it must be stable and it must be the same physical mark both times.
Because it only proves the reduction is internally consistent. If the field readings are wrong — a staff misread, a change point moved, a bubble not centred — the arithmetic still balances perfectly. Only closing on a known level, or repeating the run, finds that class of error.
Only with a geoid model in between. Leveling gives height above the geoid (orthometric height); GNSS gives height above the ellipsoid. The separation is tens of metres and varies along the line, so the two are not interchangeable without a conversion. The geoid versus ellipsoid guide covers the reasoning.
Yes — free, no login, no account, no usage limit. Every calculation runs locally in your browser, so no survey data is uploaded.
Calculations run locally in your browser. Results are for reference — confirm against the governing design standard before construction or procurement. About · Privacy & cookies · Terms of use · Developers