Weather forecasts

There has been a discussion under another thread about weather forecasts. I have copied some of the material from that discussion under this more appropriate thread name. I have also updated some details on the recent weather as it shows a good example of 'sunrise over cloud' conditions.

This is how the weather prediction process works:
1. Every six hours observations are made from weather stations, satellites, balloons etc and sent to the Global Telecommunication System (GTS) via satellite. The UN agency WMO has an agreement in place - anyone who contributes data can use the data
2. Global observations from the GTS are downloaded and the 'data assimilation' process converts the observed data into globally complete data for the weather forecast models. The models divide space into boxes 3-dimensionally. If the model has 1/2 degree lat/long boxes, then 259200 boxes cover one layer of the atmosphere. Most models have 40 layers, so use 259200 x 40 to cover the planet and represent the atmosphere. At most of these boxes, there will be no observed data. Some clever schemes are used to invent data. This takes more computing power (by factor of 4) than the forecast itself.
3. The globally complete observations (most of them being fictional) are stuffed into a forecast model. The data are 5-6 hours old already. The model integrates the equations of physics forward in time in short timesteps, writing out to file the newly computed states of the atmosphere - typically every 30 minutes of model time. The models are usually run 10 days into the future.
4. In some modelling centres with big supercomputers, step 2 is rerun with slightly different maths and step 3 rerun with slightly different globally complete observations. Some centres may do up to 10 reruns. Fast computers are needed for this.
5. Maps are drawn from the future weather states which were written to file in step 3.
6. Experts either analyse the data for the region (SAWS) or an automated scheme picks grid boxes of data and presents the data graphically - as with the Norweigian centre, CCN and the mountain weather websites.

So why is Norway better than SAWS?
The answer lies in better assimilation scheme used in step 2 and a much better model in step 3 and a much faster computer. The expert input (step 6) is non-existent in the Norwegian model, which shows that SAWS cannot add skill that wasn't there in their own model.

There are some online forecasts which offer predicted weather for specific peaks. For example:
www.snow-forecast.com/resorts/Mont-aux-Sources/6day/bot
www.mountain-forecast.com/peaks/Sentinel/forecasts/3165

These forecasts can be for peaks which are very close to each other. Why are the forecasts then so different?

Why are the two forecasts for the same place so different?
a) the models are different - one may have a warm bias for the region, for example
b) the inputs of weather at the start will be different
c) the way in which data for the peaks is taken from the model is different - and this is crucial. The models do not have mont-aux-sources or sentinel peak in them. They have a grid box 50kmx50km big (if you're lucky) with average terrain height for that box specified. If in one model the box is half on and half off the escarpment, the surface height for that box would be 2200m or so. If in the other model the box falls squarely over Lesotho, the height could be 3000m. I think this is the case in the two examples (one being 6 or 7 deg warmer).
So although the forecasts are shown to be for peaks, no one at the forecast centre will have spent anytime checking this problem out.

What kind of weather causes sunrise over clouds?
When a high ridges from the South Atlantic to the southern Indian Ocean and pushes cool, moist air onto the escarpment and where the cloud deck which results is below escarpment height. A strong ridging high and an upper air low will cause a deeper layer which causes cloud at and above escarpment height.
31 Oct 2011 is a good example of a layer of cloud caused by a ridging high which is below escarpment height:






These conditions were forecast for Sunday 30 October by forecasts made a few days earlier (on 25 October 2011). But the systems moved more slowly than predicted so that the 'sunrise over cloud' conditions materialised a day later than originally forecast.

Spot on mnt_tiska



This was Sunday morning

Is there historical data available on wind speeds, the wind yesterday on Mashai Pass (near the top) was so strong we had to crawl up the pass!

ghaznavid wrote: Is there historical data available on wind speeds, the wind yesterday on Mashai Pass (near the top) was so strong we had to crawl up the pass!

I don't think there is measured data from the top of the escarpment - only the weather stations at lower altitude.
The observed data which is measured is made available on the GTS
(see step 2 in the weather forecast description posted earlier) and some websites intercept this data. One example is weatherunderground (see www.wunderground.com/cgi-bin/findweather/hdfForecast?query=south+africa for SA stations). Closest station might be Giant's Castle. If you select GC, then you can call up historical data but no strong winds show up for yesterday.GC is too low down.

The other source of data is numerical forecast model data. These take a few days to appear though. One example is
www.esrl.noaa.gov/psd/data/composites/hour/

Meanwhile you can look at the reasons for the strong winds here:
wxmaps.org/pix/af1.00hr.png

but note that this link is not fixed to a date and so will be showing a different situation within a few hours. The reason for the strong winds was that the Berg was in the eastern (divergent) sector of a Rossby wave (the thing that makes the weather). You can see the 'isobars' are v.close together over the Berg. The plot shows conditions at about 5km asl. But at 3000m the basic layout of the chart will have been similar.