Frost warnings

Hi Duane, It would be great to get frost predictions.

This is really good for horticulture types.



I’ll second that. I grew up on a Cranberry farm - and frost played a very big role in our lives.

I would also ask for an ‘Offset’ where a user can custom dial in an adjustment value pertaining to their area in hopes of being more accurate - for frost protecting…

Does anyone know how the frost predictions are done? I haven’t seen them in the data downloaded from data sources. Maybe it’s a formulae, so if you can help find this, I can look into how to include this.

1 Like

Do you mean a bias correction? I think if we can find the way frost is calculated then I can look at this offset.

Cheers, Duane.


From what I understand it goes A frost/freeze warning is made when winds are below 10 mph and the air temperature is below 32 degrees.

not sure if that helps any

On the Cranberry farm we had thermostats set at roughly 32* or 33* to kick on the pumps to help frost protect the plants. It varied at each location and we adjusted those setpoints as necessary.

1 Like

Sounds like a black art. Since I’m not a meteorologist, I’m reluctant to give any predictions without a well regarded equation. I don’t want to be responsible for the loss of a cranberry crop :slight_smile:

But, I think there is a better alternative. We need a way to define new data based on other data. For example, if temp < x and wind_speed < y then highlight region with color z and opacity 10%.

This approach is very similar to what I would like to do with notifications with the extra output of a highlighted field on a map. It’s a very interesting concept and I’ll put it on the to-do list.

Cheers, Duane.


Notes from a previous post and link -

“A frost/freeze warning is made when winds are below 10 mph and the air temperature is below 32 degrees.
A freeze warning is made when winds are above 10 mph and the temperature is below 32 degrees.”

When the temperature is at freezing levels etc. - the plants own moisture moves to the surface of the plant - the water then becomes frozen - as the sun warms up the plant - the frozen moisture is evaporated - and the plant suffers from basic dehydration.

To prevent this from happening - we irrigated - and covered the plant with extraneous water. The water on the outside of the plant still freezes, but the plant does not suffer from dehydration and recovers nicely.

It is really a pretty basic principal. The hard part comes from location specific attributes (Windward side of a hill or mountain, in a secluded cove or in a wide open area with no cover). That is the reason why I asked for a bias or an offset. I honestly don’t know how much wind affects this - we strictly went by a thermometer and adjusted the trigger or setpoints as necessary. By the way - If we were successful at this, I don’t think it could be that hard of a thing to do. I would assume that windchill is a factor - but I personally can’t add additional information on that.

I’m no longer farming - but I am responsible for my wife’s Dahlias… So any help in protecting those plants would be greatly appreciated.


Yes, but if Flowx says there will be no frost and the next morning your wife finds the Dahlias frosted, will she blame me or you?? :wink:


She would see that there was a bias to use - kiss me and slap me on the back if the head for not using it correctly… That is repeatable - problem is - I have lost some hair over the years… :astonished: Truth.
Give us the tools to adjust it as necessary - and we will cover your back. :+1:

1 Like

As a past Hobbyist Developer - if you feel uncomfortable with possible litigation - you can always put in a disclaimer - I did. I personally wouldn’t hold you responsible - but there are questionable people out there…
Keep up the good work.

1 Like

Hello Duane,

Metservice has a reasonable looking page for a starting point. It’s got a few formulas there as well.
I’ve only skim read it however.


1 Like

Not all husbands are like you. There will be some that will throw Duane under the bus instead of facing their wife.

Disclaimers is a solution but I don’t like those since no one reads them and it adds cruff to the app.

Remember, I’m not a meteorologist, so I don’t make predictions - it’s outside my field. I’ve only computed derived data once and that was relative humidity from dew point, pressure and absolute humidity. But this was a well defined equation in the meteorological field. I don’t think I calculate relative humidity any more since it has since been released by the data source.

Unless there is a standard, validated and well accepted equation for frost, I am reluctant to make predictions.

The solution is to build a tool for users to build (and share) their own predictions.

Currently, Flowx is the messenger and a tool. This makes it very easy to answer issues about inaccuracies. Making weather predictions is not an area I’m comfortable going into. I don’t understand enough about it.


Thanks @Chris. This suggests 4 methods to estimate frost with the expectation of fine tuning. Looks like frost prediction is location dependant.


There are a few ways to get frost prediction into Flowx:

  1. create the prediction outside of Flowx and it’s brought in as a data source. This might be a Python script to take data from all the data sources and create one frost prediction.
  2. create a tool to add (and share) predictions to Flowx.
  3. find a data source predicting frost.

I like these approach because it keeps Flowx neutral and just a tool. I’m happy to help with any of these.

The other thing I like about the approaches above is that it can be applied to other predictions, e.g., lightning… Flowx just pulls the data or prediction in.

Cheers, Duane.


Tasty! I would likely pay for premium for this. Weather is an extremely local experience. Urban microclimates and regional topology largely make weather info gross estimations. Over time, having access to Flowx data, the brain picks up on relevant assumptions on the consistency of the gross estimations; as relevant for a given extremely local experience (home). Gives us the tools to see the data that makes Flowx graphs and a tool to make simple equations, which result in our own graphs. Wow! If that can be overlayed on the map as well. Kudos! Brownie points! Amazeballs!

Just need more time. Actually, I have the same amount of time as any other, I just need less other things to do.


@Windsailor, I was just talking to Ohan on another post. Ohan wants sunset predictions. I was thinking maybe we can create third party predictions. In other words, you come up with the maths and, if possible the python code, and I can run it on the workers to generate the prediction which we pull in as a data source. There are some extra things we need to iron out but I’m sure we can find a way.

My main goal is to keep Flowx separate from predictions, i.e., just the messenger.

Cheers, Duane.


My turn to whelp.
You do realize your asking someone who was raised on a dirt farm with limited programming ability.?
We used something that was very similar to the Grass Minimum Temperature method in the link above, an electronic temperature probe with an activation setpoint set ‘XYZ’ inches above the plant with a mercury probe as visual verification. And we adjusted each location as necessary. I would imagine the height of the plant would also be a factor on what method to use… And / or have layered methods based on priority and conditions with bias options…??? That seems a bit convoluted - usually simple works best.
Give me some time to think about it and I’ll email you.