Create MOGREPS Mean/Std Archive

[devsjc]

The MetOffice provides ensemble data which is used by National Grid in their internal forecasts. In order to achieve parity with their current solution, we want to incorporate this same data into our testing.

This will require finding and then downloading an archive of MOGREPS data. The number of years of data, as well as the maximum horizon, parameters, and init times are to be determined. I suspect they will match that of the UKV larger horizon dataset (#97) also needed for the same work package, which is to say:

• init_times: [00, 03, 06, 09, 12, 15, 18, 21]
• steps: TBD (Increase UKV Archive Horizon 55 Hours #97)
• parameters: [t, r, vis, si10, wdir10, prate, lcc, mcc, hcc, sde, dswrf, dlwrf]

[For NIA Work Package 6]

[devsjc]

@peterdudfield CEDA does not have a MOGREPS archive (that I can find). The only other source I can seem to find is this one: https://registry.opendata.aws/mogreps/ which stops at 2019. This would leave us a gap that would cover most of our other datasets! Not sure how best to proceed.

[peterdudfield]

And just to double check, Is it easy to get live?

[devsjc]

Easy enough yes, I would just need to create a new order with MetOffice and add some capability to the consumer.

[peterdudfield]

Easy enough yes, I would just need to create a new order with MetOffice and add some capability to the consumer.

If we cant pull a Archive, maybe we have to pull it witht he MetOffice live, and use that for testing it with

[devsjc]

From the live service, we can't exactly match the parameters of the uk standard order. MOGREPS is missing

• Visibility at 1.5m
• Relative humidity at 1.5m
• Downward shortwave radiation flux at surface - ground

[devsjc]

Pricing

The cost is very dependant on the data we request. Assuming we want to be as close to the normal UKV data set, that is to say

• 9 parameters (best we can do)
• 48 time steps
• 24 init times

Then this is ~480GB (this is because the MOGREPS model has 34 members). We are currently using about 115GB, so this would just fit in the highest tier monthly subscription band, which is 600GB at £612.50 per month, or £445 more per month than what we currently pay. However there is a danger here: we're probably about to go out to 55 hours and timesteps for the UK archive and then I assume also this archive:

• 9 parameters
• 54 time steps
• 24 init times

which makes it ~535GB and as such pushes us beyond the highest tier subscription band and into the invoice bands, which would mean looking at ~£700 per month, or £550 more than current costs.

If we didn't need that much data, and did e.g. double the init times of ECMWF, so

• 9 parameters
• 54 time steps
• 4 init times (6 hourly)

That is ~90GB, so we would be covered by the next subscription tier up, the 250GB, £272.50 per month tier, which is an extra 100 quid a month.

Matching ECMWF, so

• 9 parameters
• 48 time steps
• 2 init times (12 hourly)

is ~40GB, and wouldn't cost any more money than what we're already paying (just!)

MetOffice Pricing: https://datahub.metoffice.gov.uk/pricing/atmospheric

[devsjc]

Aside from the above, I am in conversations to try to get an existing archive.