Penzance - Unbroken Sunshine & Forecast Verification

[Graham Easterling]

Following on from Len's verification 2 or 3 weeks ago.

So there is a consistent discrepancy. 24hrs a bit more accurate than 48 hours. 72 hours would be further out again, almost always in the same direction. So the model forecast normally starts way too low, than gradually homes in a bit. This is largely a spring/summer phenomena in Penzance

Today had the right elements for a big discrepancy, very light offshore breeze. NE backing NW, now more W, and unbroken sunshine. I do wonder if too much reliance is put on reports from Camborne which would be (& was today) much cooler in the light onshore wind which tends to funnel up the valley.

Anyway, today was a glorious day for a walk from Lands End top  Sennen. Unbroken sunshine, E visibility (Scilly very clear) and almost calm, even on the cliff tops.

Graham

Penzance

[Freddie]

On Thursday 18 April 2024 at 16:03:16 UTC+1 geaste...@gmail.com wrote:

I do wonder if too much reliance is put on reports from Camborne which would be (& was today) much cooler in the light onshore wind which tends to funnel up the valley.

I would be inclined to agree.  As part of observed data assimilation the model values would be "nudged" towards the observed values.  I should imagine the observations from both Camborne and Culdrose would have the largest impact - anything else observed is too far away, so would have a very small impact.  In the general case this works well - but in the case of western Cornwall with its irregular coastline and orography, maybe not so well.  It likely works well for the majority of inland Cornwall but not the immediate coastal margins.

-- 
Freddie
Alcaston
Shropshire
148m AMSL
http://www.hosiene.co.uk/weather/

Stats for the month so far: https://www.hosiene.co.uk/weather/statistics/alcaston/latest.xlsx 

[Graham Easterling]

I would be inclined to agree.  As part of observed data assimilation the model values would be "nudged" towards the observed values.

I think today may be a good example of that. It's been unbroken sunshine in Penzance (so far - it's now 13:00) This was correctly forecast. At 09:00 the temperature prediction for Penzance at 13:00 was 14C. At noon the forecast was lowered to 12C, whilst when I checked a moment ago it was 15.2C with a light F2 NE breeze.

However it is a fair bit cooler along the north coast with a directly onshore N sea breeze.

In addition there is now some cloud build up near the north coast, where the northerly sea breeze hits the ENE gradient wind.

A big advantage with being on the south coast, as opposed to the north, is that any cloud build up down the spine of the peninsula, or along the sea breeze front, is the opposite side of the sky to the sun!

I would be interested in any recommended reading on how the models work for actual weather at a location, what they include and what they don't (like salt spray in visibility forecasts). I think the models are remarkably good now at forecasting the synoptic situation 3 or 4 days ahead, but actual weather in a rather complicated area like coastal Cornwall can still be fairly poor. Certainly understanding of the sea breeze is still very limited. Long Rock, 1 mile east of Penzance and Marazion have some of the strongest sea breezes in Cornwall, with the marsh behind offering no resistance. Penzance sees only limited sea breeze development with the hills behind, and a different aspect. Something widely known (hence all the windsurfing at Long Rock) but not really appreciated by the models. Mid afternoon a fresh onshore sea breeze can become a N offshore in just a few minutes, due to the afternoon heating of the NW facing N coast. A quick look at forecast winds & actual on sunny days under very slack pressure, quickly reveals the limitations of the forecast winds.

Graham

Penzance

[Graham Easterling]

Situation at 1:40. The sea breeze front cloud being swept WSW towards Scilly

First signs of cloud developing more generally now

Graham

Penzance

[Freddie]

On Sunday 21 April 2024 at 13:22:20 UTC+1 geaste...@gmail.com wrote:

I would be interested in any recommended reading on how the models work for actual weather at a location, what they include and what they don't (like salt spray in visibility forecasts). I think the models are remarkably good now at forecasting the synoptic situation 3 or 4 days ahead, but actual weather in a rather complicated area like coastal Cornwall can still be fairly poor. Certainly understanding of the sea breeze is still very limited. Long Rock, 1 mile east of Penzance and Marazion have some of the strongest sea breezes in Cornwall, with the marsh behind offering no resistance. Penzance sees only limited sea breeze development with the hills behind, and a different aspect. Something widely known (hence all the windsurfing at Long Rock) but not really appreciated by the models. Mid afternoon a fresh onshore sea breeze can become a N offshore in just a few minutes, due to the afternoon heating of the NW facing N coast. A quick look at forecast winds & actual on sunny days under very slack pressure, quickly reveals the limitations of the forecast winds.

In the case of western Cornwall the limiting factor is the distance between the model grid points.  Any feature (such as the variation in orography influencing sea breeze development and strength) needs to be large enough to be modelled over a number of grid points, otherwise the model does not resolve it.  Until the model grid length gets to < 500 metres I think that Penzance sea breezes will continue to be poorly modelled.

You say that "understanding of the sea breeze is still very limited", but it's not a case of understanding (the models can and do handle detail like sea breezes perfectly well), but the inability of the models to resolve the local detail in order to correctly forecast events, because that detail is "sub-gridscale".

For the case of salt spray affecting visibility - you are correct.  It is not included in the visibility forecast.  I don't think observed horizontal visibility is an element that is included in data assimilation (i.e. it is ignored), but it is something that can be emulated given a good enough algorithm.  I'm not sure if there are any plans to do this.