# How 'good' is WeatherData data

I am considering using calls to WeatherData for wind speed, precipitation, etc. It seems the data are accurate, and samples compare well with, for example, my local weather reports, however, I am wondering if using WeatherData time-history plots, summary statistics, etc. are defensible in a business context. As an example, when I plot directional wind speed data, for a given period and site, it compares reasonably well with other plots, however, there are differences of the order of +/-15%. Part of the difference may be due to Missing Data and spurious zeros (in my case I simply eliminate these data whereas others may be using schemes to clean the data.)

Do others have a view? Is WeatherData[ ] commercial grade information?

-
I am not sure this is a proper question. It is open-ended, inviting people merely to give opinions rather than definite answers. Also, "commercial grade" is a vague term and no criteria are given for judging WeatherData against this standard. –  m_goldberg Dec 13 '12 at 5:28
Are you sure you getting exact weather station when compare? Just a city can have different weather stations which will have somewhat different results. –  Vitaliy Kaurov Dec 13 '12 at 5:30
I would imagine that the quality of monitoring, and the relative accuracy, presence and frequency of data, would depend strongly on the individual monitoring stations which are run by widely varying organisations. Those from the ICAO network might be of high repeatable quality, those from enthusiastic amateur stations perhaps, but not necessarily, less so. Maybe you could choose the most reliable stations from the areas you are interested in. –  image_doctor Dec 13 '12 at 12:11
Vitaliy, I've been checking the latitude and longitude of the nearest weather stations and pick the station having what seems like the most complete dataset, calls for data at many locations result in lists with Missing[NotAvailable] elements. image_doctor, your right about quality, it's not clear however how you find out if the data are 'good' or not. Hence the post. The simplest way of MMA users checking if the data are sound is to know it's being used in a commercial context by others who have done the QA and got acceptance. Would EPA for example accept the statistics based on the data? –  R Wright Dec 13 '12 at 15:38
To be (only partly) facetious: Try using the Wolfram-curated WeatherData as evidence in a court case and see what happens. –  murray Dec 13 '12 at 16:35
show 2 more comments