Reading path: Coastal water observation →
Find the stated unit and quantity
Start with the dataset’s documentation, column labels, and metadata. Confirm both the unit and the quantity being measured: temperature, water level, salinity, and current speed are not interchangeable just because their values are numeric. A label such as “level” may require a reference datum as well as a unit to be interpretable.
Do not infer a unit from the size of a number alone. A temperature might be expressed on different scales; a water level may be represented relative to a reference surface; a time field might use seconds from a defined starting point. The context determines what the values mean. If documentation is missing, treat the unit as unresolved until you can verify it.
Use ranges as a check, not proof
Plot or summarize the values to find obvious anomalies, such as an unexpected order-of-magnitude shift or a mixture of positive and negative values where the variable’s definition does not allow it. Compare the range with reliable contextual information for the same place, depth, and season. A broad “typical” range is a clue, not a substitute for documentation; unusual conditions can be real.
For example, if two files appear to record the same water temperature but one series is offset by a large constant, a scale or conversion mismatch is one possibility. So are differences in station, depth, or time. Check each explanation rather than correcting the values simply to make the plots agree.
Verify every conversion
Write down the conversion formula and apply it to a few hand-checkable values before converting the full column. Keep the raw values intact and put converted values in a separate field with a clear unit in its name or label. This makes a mistaken formula easier to detect and allows a later reviewer to reproduce the change.
Also check prefixes and time conventions. A value expressed in centimeters differs from one in meters by a factor of one hundred; a timestamp may require a time zone or epoch definition before it can be aligned with another record. Small labeling mistakes can create large apparent differences.
Document what you could confirm
A short data note can identify the original field, stated unit, conversion applied, and the source of the unit definition. If documentation and observed ranges conflict, flag the issue and avoid making a strong physical interpretation until resolved. You can still describe the numeric pattern while being precise that its unit is uncertain.
A plausibility check is a guardrail against preventable errors. It does not certify the measurement, but combining documentation, plots, and a transparent conversion check gives the reader a sound basis for understanding the result.
Have a correction or a source to suggest? Send a note →