Frost Prediction

[wd40]

After a bit of digging, I came up with this equation for predicting the low temperature for the night and early morning. It can be more accurate than local forecast since you are using your stations data.

Min Temp= *Dew-point - ((*Relative humidity - 40)/4)
*This uses reading taken at 6 pm standard time.

It assumes that most frost conditions occur on clear nights with few clouds and little air movement therefore becoming less accurate on cloudy evenings at the time of the sensor observations. It also assumes humidity reading are not too low.

This formula has been around since the early 1900s.

Randy

[steve]

What temperature units does that formula use?

[wd40]

Fahrenheit.

I think there is a Swedish study done about the same time but could not get that document.

[nking]

This formula has been around since the early 1900s.

Hi Randy - could you post the link for this - many thanks

[wd40]

Neil, here it is.

http://docs.lib.noaa.gov/rescue/mwr/Suppl/no16-1920.pdf

It is a document from 1920. There are several methods shown but the posted equation is considered the basic approach.

If Steve sees merit in the idea of frost prediction with Cumulus, i think this formula would offer a start.

I have reviewed the value of adding wind and pressure to formula. Wind at the time of the prediction might be high, then by morning be calm. High Pressure would suggest clear skies but again from my reading, it could be misleading.

I am no authority so anyone says I am full of whuey, I want argue.

[steve]

If Steve sees merit in the idea of frost prediction with Cumulus, i think this formula would offer a start.

It's interesting and potentially useful; I'm just a bit worried about adding things like this now because of the the Forecast and Cloud base syndromes: "Cumulus said it was going to be frosty and it rained" etc etc

[wd40]

Yes, I see the problem from the past post. It might be better for folks just to make the calculation themselves when the conditions look questionable.

Best I can tell is this would be about 80% right when the station is located in an average topography area. Mountains, valleys, and coastal weather requires specialized correction factors. Like the forecast, and cloud base, frost prediction could be a problematic as well. You have a good point.

No problem.

Randy

[steve]

I still think it's interesting enough to give it a try...

[Synewave]

Maybe include it as an optional webtag, so we can choose to use it or not. I'd certainly be interested to give it a try.

I use Javascript to show current conditions, and based on the cloudbase calculation I can say it is cloudy if below 100m, partly cloudy above 1200m and most likely sunny if above 1500m.

Regarding the forecast, mine really has been pretty accurate over the past 6 months.

So all in all, I'm pleased with everything Steve, and I would be keen on trying out any frost prediction features with the winter coming here very soon in the UK.

[nking]

Thanks for the link Randy

I still think it's interesting enough to give it a try...

The UK Met office has got round the problem of the public loosing confidence in weather forecasts; they introduced probability of an event. This action in itself has made the public aware that nothing is for certain at a local level. The Zambretti forecast is reputed to be 90% accurate, not sure if this is based upon empirical research or not. Perhaps we all need to be reminded that Cumulus forecasts are not 100% and neither would a frost prediction but that doesn’t mean the information isn’t of value, a talking point and generally interesting fun. So Steve, I hope you do give it a try

I came across this site when I was nosing around about frost predictions http://biomet.ucdavis.edu/frostprotecti ... FTrend.htm
might be of interest to someone.
edit - take a look at this http://www.fao.org/docrep/008/y7223e/y7223e0b.htm

[wd40]

Here is a different way to look at it. What about a graph for "select a graph" so when the dew point, the expected low based on local station conditions(The formula) and the freeze line meet frost should occur. Let the observer figure if a frost event is coming. No projections, just information. Hard to say if it would be any better than a forecast.

Would that be a lot of code?

Randy

[beteljuice]

As the beteljuice has mentioned before for various formulae involving RH / DewPoint - Most of our kit is inaccurate or more like won't give a reading of 100% RH.

Makes everything just that bit more ...............

[wd40]

I ran the calculation using the 6 pm reading and it calculated out a morning temp of 57.25F Wunderground projecting 58F. Clear tonight.

[wd40]

Here is another calculation method from New Zealand.

Smith (1973) produced and tested a simple forecast scheme, for use in Canterbury, to estimate the next morning’s grass minimum temperature, using an understanding of the maximum rate of cooling prescribed by radiation physics and measuring the weather conditions at 15.00 hrs NZ standard time of the current day:
(3) Tg = 1/3 ( T + Td /2 ) - c
where Tg is the forecast grass minimum temperature (oC)
T is the 15.00 dry-bulb temperature (oC)
Td is the 15.00 dew-point temperature (oC)
c is a constant c=8 for May. Sept. and Oct.
c=9 for Jun. and Aug.
provided that:
1. the cloud amount is <1/4 of low or medium clouds and <3/4 high clouds
2. the wind speed is < 5 kt from E or NE (wind off the sea) or < 7 kt from any other direction.
For example, a psychrometer reading at 15.00 in September gives a temperature of 6 oC and a dew-point temperature of 3 oC.
Tg = 1/3 (6 + 3/2) - 8 = -5.5
An

[PeterP]

Cumulus does show "air frost".

air frosts.JPG

We get many more ground frosts that dont seem to register as "air frosts". Is there a difference?