How we work
How KEI turns a logger download into a discharge record, and why each step is taken. Every picture is drawn from a real processing session in the KEI Hydro Processor.
Each sheet is one picture, then why the step is needed, what was done and the rule we follow.
Five steps, always in this order. Each one changes the record a little, and each change has a reason.
A logger file holds pressure, not discharge. Raw level sat 247 mm under the staff gauge reading, and discharge first came out at 76.5 L/s against 62.6 L/s measured.
Imported two files, made 8 deletes on the Pressure graph, lifted the data set onto the staff gauge (+247 mm), set a drift on Stage Final (0 to +12 mm) and a drift on the rating (+6 to +3 mm).
Work down the chain in order: pressure, stage offset, stage final, discharge.
A minute spent on the file name and both ends of the file saves an hour of untangling a bad import.
The name promises 5 August to 18 September, downloaded at 16:15. If the file's first row, step or last row disagree, the dates are wrong or readings are missing.
Opened the file, read the first row (2026-08-05 14:35), saw the times step by five minutes, jumped to the last row (2026-09-18 16:15) and matched it to 1615-7 in the name.
Before importing, check the file's first row, step and last row against its name.
The file's original date column read 2008-05-26 14:35 for a reading taken at 14:35 on 5 August 2026. Worked out afterwards: 5,585 of its 12,693 times had been misread.
Checked the corrected date column at the first and last rows, removed the misread column, and imported the corrected one as the Timestamp.
If a logger's dates look wrong, check how the file was opened before blaming the logger.
The creek's logger and a barometer. The level cannot be worked out until both are in.
The Harrop file holds Harrop Creek's own in-water sensor as well as the barometer. That sensor reads 3.7 kPa above Procter's: 0.38 m of water.
In Column mapping, set the barometer file's "Abs. kPa" and "Water Temperature" to "Ignore", kept "Baro pressure", and kept the logger's own time zone and every reading.
Read every MAP TO line before importing; bring in only the column you need.
The sensor feels air and water together: of 95.419 kPa, 94.396 kPa was air. With only the level logger file in, the new period had no stage at all.
Saw the sensor's pressure carry on alone past 5 August with no water level, imported the barometer file for the same period, and the raw level and stage appeared.
Import the barometer for the same period before judging the water level.
Everything downstream is worked out from the two pressures. Readings that are wrong because the sensor misbehaved, was out of the water or was being downloaded are deleted here, in the Absolute or the Barometric pressure, never in the stage: bad readings that are not real stage must never reach the stage.
The download's first and last sensor readings sat 2.44 and 2.34 kPa below the hour of readings beside them: about 0.25 m of water level each.
Zoomed in on the Pressure graph until single readings showed, selected the one bad reading, and deleted it from the sensor's pressure only. Twice: 5 August 14:35 and 18 September 16:15.
Delete faulty, out-of-water and download readings in the pressures, never the stage.
The logger gives the shape of the record. The staff gauge and the flow measurement tie it to the creek.
The logger gives the shape of the record; the visit's readings tie it down: staff gauge 0.356 m, twice, and a 43.5 L/s flow measurement on 18 September 2026.
Dragged the visit's digital field sheet onto Field Readings, checked the preview and added its three records: two staff gauge readings and one flow measurement.
Add the visit's field readings before you set an offset, a drift or a rating shift.
One constant per data set, matched to the first staff gauge reading. Nothing else happens in this step.
The logger only reads the water above itself: 0.1217 m on 5 August, when the staff gauge read 0.369 m. The rating curve works on the staff gauge's scale.
Selected the new data set in the Stage Offset graph and dragged it up until its start met the staff gauge reading: +247.4 mm.
In Stage Offset, apply one offset to the data set and nothing else.
Raw level sits about a quarter of a metre under the staff gauge, and each data set needs its own offset: +244.4 mm for one, +247.4 mm for the next.
Selected each data set edge to edge (1,508 readings, then 12,694) and lifted it until its start met its first staff gauge reading.
Offset to the first staff gauge reading and leave the end for drift.
If the start fits the staff gauge and the end does not, the sensor has drifted. The correction is drawn so anyone can see it.
The offset made the start fit the staff gauge, but the last hour of record averaged 0.344 m against a staff gauge reading of 0.356 m: 12 mm low.
On the Stage Final graph, with the data set selected, left the start at 0 and moved the end up 12 mm onto the staff gauge reading.
Fit the start with the offset; fix a miss at the end with Drift.
The rating curve turns stage into discharge. The flow measurements say whether it still fits.
Discharge before drift (hour means) read 76.5 L/s against 62.6 L/s measured on 5 August, and 50.6 against 43.5 L/s on 18 September: the curve gave too much flow.
Moved the rating's point of zero flow up 6 mm at 5 August, easing to 3 mm at 18 September; and -3 mm to +5 mm for the data set before.
To bring the discharge down, move the point of zero flow up.
One line from each sheet. They will be added to as more sites are recorded.