0

How can we identify historical weather data which has changed?

Jessica Steinemann

July 01, 2026 10:19

None

Follow

Hi,

We are designing a pattern for ingesting your weather data repeatedly into one of our data pipelines. We are using the historical data feature and I'm referring to this knowledge base article which states that historic data can sometimes change for some weeks or even a few months after it has originally been received in your systems: https://www.visualcrossing.com/resources/documentation/weather-data/how-historical-weather-data-is-updated/?utm_source=chatgpt.com

We are trying to address the issue of late arriving and corrected data in the context of incremental monthly data loads. 

Are there any watermarks / timestamps available in your data which we could query to identify only data changed since we last queried your API for a certain location (we do so based on UK postcode)? Our current logic simply extracts timestamps since we last extracted (based on event date) but this would miss retrospective corrections of data in your systems, or gaps being filled when more recent data already was extracted.

Thanks for helping me to better understand this,

Kind Regards,

Jess

• Facebook
• Twitter
• LinkedIn

2 comments

• 
  info July 01, 2026 17:21 Official comment

  Thank you for your question.

  There are two main reasons why results from a historical weather query may change after an earlier request:

  1. Additional station data becomes available. Weather-station networks report and finalize observations on different schedules. As delayed data becomes available, it may be incorporated into the historical result. In this case, you can often identify the change by comparing the returned stations list between requests.
  2. An existing station record is revised. A station operator may update previously reported observations following sensor recalibration, quality-control checks, or other data review. These updates can affect values already returned for an existing station and date.

  At present, the API does not provide a per-record update timestamp, watermark, or change feed that would allow you to request only observations revised since a previous extraction. Therefore, an incremental process based solely on the weather event timestamp would not capture retrospective corrections or late-arriving observations for dates already loaded.

  For pipelines that need to accommodate these changes, we generally recommend periodically reloading a rolling historical window in addition to loading new dates. The appropriate lookback period depends on the level of completeness required, but should cover the period in which delayed station reports and quality-control updates are most likely to occur.

  We have forwarded your request for revision timestamps or a change-identification mechanism to our data team for consideration in a future release.

  Regards,
  Visual Crossing Support

• 0
  
  Jessica Steinemann July 02, 2026 07:11

  Thank you for your speedy response. What you propose is the integration pattern we would have chosen in that scenario.