Reading path: Coastal water observation →
Identify exactly what you used
A comparison is easier to reproduce when another reader can locate the same inputs. Record the provider, dataset title or identifier, station or region, retrieval date, and version if one is supplied. A general note such as “coastal temperatures online” does not identify which file or record supported the result.
Save a copy of the downloaded data when permitted, or record a stable identifier and the way to retrieve it. Online records can be updated, corrected, or presented through changing interfaces. The retrieval date helps distinguish the version you analyzed from a later download.
Write down the processing path
List each transformation that affects the values: selecting dates or stations, converting units, changing time zones, resampling, averaging, removing flagged readings, or handling missing values. State the order when it matters. For example, converting every observation before calculating a daily mean may differ from averaging values that were rounded first.
Keep raw and processed data separate, with clear field names. A short script or an ordered notebook of steps can make the work repeatable, but even a compact manual workflow should be specific enough that another person can follow it. Record software settings or formula choices if they influence the output.
Explain comparison rules
When comparing stations, define how observations were matched in time and space. Did you pair exact timestamps, choose the nearest reading within a set interval, or compare daily summaries? Did you require matching depths? These rules can change which pairs are included and therefore change the result.
For example, two records with different sampling intervals might be compared using hourly means. Document the time window, minimum coverage, and treatment of partial hours. If filtering removes many points from one station but few from another, report that imbalance rather than presenting the final lines as equally complete.
Make the result independently checkable
A caption or methods note should state the variables, units, period, stations, and key processing decisions. Include exclusions and limitations that could alter the pattern. If an uncertainty interval or a smoothed line is shown, explain how it was generated and what it represents.
Reproducibility is not paperwork separate from analysis. It lets readers test whether a pattern survives reasonable choices and helps you find mistakes when you revisit the work. A clear record of source, date, version, and transformations turns a one-off comparison into evidence that can be checked and improved.
Have a correction or a source to suggest? Send a note →