Sometimes you come across a weather record that just doesn’t seem right. Perhaps the precipitation is too high based on what you know about the location or perhaps the temperature is too low and you know because you live there. Other times you may find missing records in the data and wonder the cause.
There are various reasons both natural and technical for all of these situations, and while our goal is to have the best weather data in the industry there are sometimes imperfections in the feeds provided by our suppliers.
If you would like to get an explanation for the data that you have, this article is your primary source about how to understand potential data issues and how to report them to us so that we can give you a helpful resolution and make corrections if necessary.
Types of potential data problems
Data quality questions come in several varieties.
- One or more weather metrics out of the expected range based on normal conditions
- Data that is incorrect based on your personal knowledge or observations
- Results that are missing a small number of records chronologically
- Results that are missing one or more weather metrics while other metrics appear as expected
Weather metric out of range based on normal conditions
This situation happens any time one or more weather records appear to be abnormal for the location being queried. Before jumping to the conclusion of a data error, however, consider natural weather extremes that may occur.
For example, a hurricane can pass slowly and drop a large portion of an area’s expected annual rainfall in just a few days. Also, extreme heatwaves, cold snaps, and snow storms do happen.
In northern Virginia, USA, the headquarters of Visual Crossing, we tend to get a relatively modest amount of snow and rarely outside the range of December to February. However, a freak snow storm hit in early October 1979 dropping between 6 and 12 inches of unexpected snow in areas.
If you have a similar example that appears to be outside normal bounds, one good step is to check other sources of historical information and see if there is a natural explanation. Wikipedia is a great source of historical overviews for hurricanes and other notable weather phenomena. You can often search on the date and region name, date and nearby large city name, or date and the type of weather event that could explain your data anomaly (hurricane, heat wave, etc.).
If you can’t find a natural explanation, it is possible that the data itself is wrong based on an historical weather station error or the transmission of the data into our database. In this case, we would like to know about the issue so that we can research it and correct it if appropriate. Please, collect and data outlined in the Reporting a Potential Weather Data Error section below, and send it to our support group. We’ll do our best to find an explanation.
Data that is incorrect based on your personal knowledge
Sometimes you are convinced that specific weather records are incorrect based on your own personal knowledge. Perhaps based on your own observations, you are certain that there was no rain at your location yesterday or you know for a fact that the overnight low in your backyard was significantly cooler than our data reported.
The most common cause for this type of data question is due to weather station placement. Most people don’t have a major weather reporting station in their backyard, so our system must combine data from the closest stations to find the best approximation of of the conditions at a given location.
While we use state-of-the-art interpolations in the process, there is always some variability. This error margin can get worse as your location gets farther from reporting stations. Also, this issue be a more important factor in areas with significant weather differential based on altitude variations, bodies of water, and urban landscapes.
To check what weather stations are being used to supply the data for your query, simply turn on our weather station details feature. This video will show you how to do that in our web query UI, and help you understand the results. In addition, our Historical Weather Dashboard shows a map of the queried location along with the nearby weather reporting stations as points. This will help you understand where stations are located in relation to your query location.
Consider the position of the stations and how differences in elevation, proximity to water, and other factors may make their reports different then your own observations.
If you are convinced that station location differences and other local factors are not the cause of the unexpected results, then let us know, and we’ll research further. Please, collect and data outlined in the Reporting a Potential Weather Data Error section below, and send it to our support group.
Results that are missing records chronologically
Small time gaps in result data are most often caused by the nearby weather station temporarily going offline. This can happen for maintenance or any number of other technical reasons. Usually, our weather engine can find additional nearby stations to fill in the missing data. However, in some cases there are not enough nearby stations within a reasonable search radius. In this case, no data is reported.
The best work around for this issue is to expand the station search area. You can do so easily using our weather station details feature. This video shows you how to see the stations being used and expand the search areas within our web query UI. Of course, these settings are also available in our Weather API.
If you need more information or are seeing a problem that cannot be explains or solved by increase the search radius, please contact our support team with the details in the Reporting a Potential Weather Data Error section below. We’ll be glad to look into the issue further.
Results that are missing specific weather metrics
It is not unusual to see weather records that are missing certain metrics. This can be completely normal and due to the actual weather conditions. For example, if it is not cold and breezy, a Wind Chill metric will not be reported. This is because Wind Chill is only defined in certain conditions.
However, in other cases, metrics that you want may be missing from the results. This is typically because the closest stations do not report the metrics in question. For example, every weather station reports temperature. Temperature is easy to measure with relative cheap instrumentation. However, measuring cloud cover and visibility require more expensive and complex technology. Even a seemingly simple metric such as snow, can be difficult to measure properly. Thus large stations such as those at major airports and government installations tend to report more weather metrics than smaller stations such as those at local airports and schools.
One way to work around this issue is to expand the number of stations considered by the weather engine and also expand the station search radius. If more stations are considered, farther stations can fill in the gaps for closer stations that do not report specific metrics. You can learn how to change the number of stations and the search radius in our web UI by watching this video . You can also change these parameters via our Weather API.
If you need more information on how to get coverage for specific metrics in specific locations, please send a request to our support team. Make sure to include all of the details requested in the Reporting a Potential Weather Data Error section below.
Reporting a Potential Weather Data Error
The more information that we have about the details of your weather query, the more that we can help explain what you are seeing and resolve a potential problem. Since we love to answer weather data questions, our worst support queries are ones that mention a problem but don’t explain how to reproduce it. For this reason, we ask that you please submit the following details when asking our support team about data issues or potentially errant weather results.
- The specific reason that you think the data may be in error
- The specific location(s) showing the problem (Make sure the include the exact address if that is part of your query.)
- A sample of the result table (specific weather records) the shows the problem
- The result date(s) when you see the problem in the data
- The exact start and end date (and time, if appropriate) of your query
- The method used to run the query – web query UI, API, Excel, etc.
- The query URL obtained from the web query UI
- To obtain this log into the web query UI here: https://www.visualcrossing.com/weather/weather-data-services
- Next compose your query via the interface. You should limit the query to the specific location(s) and dates or times show the problem.
- Run the query and check the results. Do they show the error that you expected? If not, carefully compare your original query showing the error to the current one in the UI.
- Click on the Query API button above the query results. This will open the query window shown below.
- Copy the query URL from the Base Query text box.
Please email this entire set of information to our support team at email@example.com. This information will allow us to more quickly and easily reproduce the exact problem that you are seeing. Our weather experts will analyze the raw data and the source data to help explain the issue. Please note that this may take a few days, and we will get back to you as soon as possible with our findings.