What this tool does
Compare target-station accumulation with a matched reference while keeping excluded pairs and the chosen split visible.
When to use it
After retaining original observations and confirming the record basis. Run this decision independently; the other rainfall tools answer different questions and are not required merely to obtain a number.
Calculation or decision method
Use a pair only when both target and reference depths are known. Otherwise omit both from paired accumulation and retain the exclusion. Plot cumulated reference on x and cumulated target on y.
Each segment ratio is sum(target) / sum(reference) over usable pairs on that side of the chosen original-row split. Relative change =100 × (late ratio / early ratio −1). This is a descriptive through-origin increment ratio, not a free-intercept regression or change-point significance test.
A segment needs two usable pairs and positive reference accumulation. Relative change is undefined if either needed ratio is undefined or the early ratio is 0. Dry pairs remain valid observations; undefined is not forced to 0.
Worked example
Target 10,20,30,15,30,45 against reference 10,20,30,10,20,30, split after three rows: paired totals150 and 120 mm; early ratio 1, late ratio 1.5 and relative change+50%. That change does not identify a cause or instruct correction of either record.
Assumptions
Target/reference values cover the same periods and durations in the same supplied order. Reference construction, surrounding climate regime, maintenance history and the reason for the chosen split are external evidence. A reference can be an independently documented station composite; this tool does not build one from an arbitrary selection.
Input and precision rules
Plain decimal, nonnegative depths; blank or NA means missing. Supported nonzero numbers run from 1e-12 to 1e9 in the entered basis. Inputs use ordinary browser floating-point precision, not arbitrary-precision metrology. Reconcile unsupported missing codes, tip counts, timezones, intensity values and changing clocks before entry. A malformed row stops the run rather than disappearing.
Limitations
No automatic correction, break detection, calibration diagnosis, trend significance or missing-data estimation. Paired-only totals omit unpaired rainfall and are not full-record totals. A straight segment alone does not prove data quality or climatic homogeneity. The chart can thin large records to 500 points while calculations and exports retain every row.
Keep the review packet
Compare the descriptive relation with station relocation, exposure, maintenance and reference-network records. Keep original series intact. Use the period-level audit rather than reading a correction factor from the chart. The screen/print preview is limited to 100 rows; audit CSV downloads retain every computed row. Copy result includes the full entered CSV and configuration. Keep sensitive project data within your own handling procedures; the site does not upload it.