Met Office Datapoint archive
Here is a minimal data distribution in two .zip
[data updated past end of 2015, now contains a complete year, 2015]
- Verbatim XML received from Met Office servers
- CSV of decoded into time series 22 July 2014 onwards to date
- Decode of XML is non-trivial, beware.
- Data files may overlap in time
- A few gaps
- Hourly data for ~170 land stations
- Time series have been missing data filled so the series are contiguous in time.
- Missing data marked nk
- Only some stations have all weather parameters
- Additional column with spreadsheet format time
- >12,000 data rows.
- All supplied data parameters
- Additional column, computed clear weather insolation for the station latitude / longitude / altitude / temperature / pressure, if parameter known or sane default. As of date figure is guessed to be about 5% higher than typical simple site instruments.
Are known data defects, open Met Office support ticket from January 2015 and assumed will not fix. Data fails audit. Some wind speeds are never present. Problems with all parameters. (some details are on my own blog, daedalearth)
Data supplied by Met Office under OGL
This cannot be computed conventionally, data is far too short. However, the author’s Synth software can make a good estimate of the annual cycle. Where checked these fit published data well, see eg. http://www.yr.no
Example for above. Error is probably mostly the result of recent weather, just the last 18 months whereas the formal result used a decade.
Plotted above is average of hourly estimate for each month, no adjustments or party trick, simply as it came out for a random station.
How to do this? (only works if the data is long enough, a few new stations on have a few months of data, probably too little)
Extract the parameter of interest, eg. temperature into text file as a single data column.
Incantation at command line or console
synth fs=hour annual=1 offset=1 decay=0.2 gens=2 interp=nk meritmode=0 resume=0 <filename>
Note carefully: this is trickery, detector mode has been overridden from the normal annual mode of Fourier to least squares mode so a best shape fit is done. (modes run 0…5)
This will take some time to run, sorry. (maybe 10 seconds)
Output will be filename with extensions
- .met which is the annual cycle on 12 and 6 months
- .rem which is the data less annual cycele
Command line item offset=1 or offset=0 to taste. Could also use the .asc file as live spreadsheet time series generator and do maths on data. This way you get access to the deduced offset.