weather forecast with a barometer: feasible? reliable?

[DiTBho]

How good/bad the weather becomes is the result of how great the difference is between the high pressure and low pressure system.

• high && stable pressure should indicate good weather
• falling pressure should indicate that something is moving in, and you can expect poorer weather
• ...?!? ... something similar - I don't know details, so I suppose - should be used to alert an incoming violent disturbance of the atmosphere with strong winds and usually { rain, thunder, lightning, snow } = (%storm) = ?

A barometer, I think even those integrated inside CASIO's Suunto's and Garmin's wrist watches, is a great thing to have because its meaning should help local weather forecast.

I know the Suuntu Alpha Core watch implements a "storm alert" function. I don't how its CPU calculate the probability.

I'd like to buy some barometer on chip (i2c? SPI? .. mumble), and to write some simple application for the pocket GNU/Linux PDA I have recently integrated into my  90s-titanium bicycle.

It would be extremely good for long wild trips, when on Friday afternoon I leave the lake by bicycle with only an ultralight tent, boots and a sleeping bag in my backpack and arrive in the mountains, to proceed on foot and then descend on Sunday morning.

I mean, knowing if I'm heading towards a probable storm/pouring rain would be great when I still have time to reverse course 

(OK, there are commercial Apps for your smartphone, and even commercial wrist smart watches, or even better professional tools used in modern navigation and meteorology ... but ... I'd like to build my-own portable weather forecast machine

"because science" )



Looking for algorithms, tips and tricks.

[Kleinstein]

It depends a bit on the loacation how good the simple looking at the pressure is. The problem is that the pressure more correlates with the wether right now and less than the next days.
Chances are some free web forcast for the next weak is better than what one gets from just looking at the pressure for the next day.
For reasonable good short term (next few hours) predictions, when there is a chance for local showers the weather radar is great. Just the pressure would not really help.

[rstofer]

“Red sky at night, sailor’s delight. Red sky in morning, sailor’s warning”

I wonder how well that saying actually works..

https://www.loc.gov/everyday-mysteries/meteorology-climatology/item/is-the-old-adage-red-sky-at-night-sailors-delight-red-sky-in-morning-sailors-warning-true-or-is-it-just-an-old-wives-tale/

There is some science behind the adage!

“With a low and falling glass, soundly sleeps a careless ass; only when ’tis high and rising soundly sleeps the careful wise one.”

https://www.sailingtoday.co.uk/cruising/cruising-stories/rain-later-good/

[tggzzz]

How good/bad the weather becomes is the result of how great the difference is between the high pressure and low pressure system.

• high && stable pressure should indicate good weather
• falling pressure should indicate that something is moving in, and you can expect poorer weather
• ...?!? ... something similar - I don't know details, so I suppose - should be used to alert an incoming violent disturbance of the atmosphere with strong winds and usually { rain, thunder, lightning, snow } = (%storm) = ?

Define "good" and "bad" and "location". For example, in the UK high pressure -> hot in summer and cold in winter[1].

"%storm" must include variance. With one barometer the variance will be high; with many distrinbuted barometers and temperature stations the variance will  be lower. For an idea of the variance, do some research on the reliability of weather forecasts from 1940 to 1960.

It would be extremely good for long wild trips, when on Friday afternoon I leave the lake by bicycle with only an ultralight tent, boots and a sleeping bag in my backpack and arrive in the mountains, to proceed on foot and then descend on Sunday morning.

I mean, knowing if I'm heading towards a probable storm/pouring rain would be great when I still have time to reverse course 
...
Looking for algorithms, tips and tricks.

For algorithms, research the fluid dynamics calculations done by the Met Office. Don't expect them to provide a useful answer on a pocket machine.

The best tips are:

• look out the window, upwind
• learn how to interpret SLP diagrams such as this and the others on that site
• for mountains, learn how to interpret skew-t diagrams, and learn how local terrain interacts with the air currents to produce very localised weather conditions that can change very quickly

[1] For the reasons, remember your school geography and physics lessons! If you weren't taught them, find an O-level (now GCSE) geography textbook from around 1970.

[DiTBho]

"%storm" must include variance. With one barometer the variance will be high; with many distrinbuted barometers and temperature stations the variance will  be lower. For an idea of the variance, do some research on the reliability of weather forecasts from 1940 to 1960.

... thus, the storm-alarm feature implemented by Suunto, Casio and Garmin in their wristwatches and promoted by their affiliated sports retailers sounds more more "marketing" than science 

I had exactly this doubt the last time I saw their wristwatch in a climbing supply store and was tempted to buy it.

It was only 120 Euro instead of 190 Euro for a Suunto Alpha Core with an embedded barometer and thermometer, which CASIO calls "twin sensors", but it's nothing new even with Garmin, since an embedded thermometer is always a problem in a wristwatch unless you to take off your wristwatch and place it somewhere on the ground or on a tree to correctly measure the ambient temperature otherwise it will measures something close to the temperature at your wrist.

I'm starting to think that, even using fuzzy logic, the correct slice of pressure can be obtained from a mini cluster of barometers (say, a sensor on your bike's front fork, one on your rear fork, one on your wrist and one on your hat ... are four barometers enough for a local reading? probably ... no, and) depends on the temperature of each reading, and you better have more sensors around you.

Ummm, so ... I'm going to re-think, with more radio link capability (wifi 2.4-5Ghz dual band, + hackRF kit and all the GNU/Linux SDR(1) support). This way I can take more serious reading from weather stations around me and the software on my PDA can provide better forecasts.


( it won't work  )


(1) SDR = software defined radio

[DiTBho]

The storm alarm notifies you that a pressure drop of 4 hPa / 0.12 inHg or more has occurred during a 3-hour period. Suunto Core will activate an alarm

(Suunto Core Alpha Stealth Watch - Weather Trend Indicator & Storm Alarm)

[DiTBho]

Just to give you an idea of Km and time I usually run on my bicycle, I can run 160 Km in about 10 hours (with peaks of 45Km/h but only in the plain), not great performance, but 3 hours of warning before a storm are enough to find shelter.


So, my specs are weather forecast for the next 3 hours (should be enough) 

[CatalinaWOW]

My commercial weather station has this kind of barometric prediction built in.  In my location it is better than flipping a coin by a small margin.  For your purposes false positives and false negatives are the critical issues.  False positives because of the cost of missing a perfectly fine trip, and false negatives for the obvious reason.  Only you can assign the value to these outcomes.  A cautious person will not mind the opportunities missed, and deplore miserable or dangerous trips.  Adrenaline junkies the opposite. 

I have personally found the predictions of such low value as to be worthless.  Not as good as the ten day out predictions by the national weather service.  That is the barometric prediction for tomorrow is not as good as the ten day prediction from nine days ago.  But if you are going to be out of communication for more than ten days there might be some value.

[tggzzz]

"%storm" must include variance. With one barometer the variance will be high; with many distrinbuted barometers and temperature stations the variance will  be lower. For an idea of the variance, do some research on the reliability of weather forecasts from 1940 to 1960.

... thus, the storm-alarm feature implemented by Suunto, Casio and Garmin in their wristwatches and promoted by their affiliated sports retailers sounds more more "marketing" than science 

You do remember these, don't you? Is there any reason to belive those storm alarm algorithms are any better?

I'm starting to think that, even using fuzzy logic,

Why is "fuzzy logic" relevant? In what way isn't it "magic"?

the correct slice of pressure can be obtained from a mini cluster of barometers (say, a sensor on your bike's front fork, one on your rear fork, one on your wrist and one on your hat ... are four barometers enough for a local reading? probably ... no, and) depends on the temperature of each reading, and you better have more sensors around you.

With 4 such sensors you would have more data.

Please outline how they would give more information, and what the importance of a "slice of pressure" might be.

Ummm, so ... I'm going to re-think, with more radio link capability (wifi 2.4-5Ghz dual band, + hackRF kit and all the GNU/Linux SDR(1) support). This way I can take more serious reading from weather stations around me and the software on my PDA can provide better forecasts.

You need to expand the concept of "around you" by over 3 orders of magnitude (or 6 if you are thinking in terms of area).

[iMo]

High pressure means there is less water vapor in the air (the dry air is heavier), the low pressure means there is more water vapors in the air (the air with a lot of water vapor weights less). It is generally an indicator but not extremely reliable one. The sudden large changes towards the low pressure usually show the wet air is coming at you..

[TheCalligrapher]

"The weather is a thing that is beyond me altogether. I never can understand it. The barometer is useless: it is as misleading as the newspaper forecast.

There was one hanging up in a hotel at Oxford at which I was staying last spring, and, when I got there, it was pointing to “set fair.” It was simply pouring with rain outside, and had been all day; and I couldn’t quite make matters out. I tapped the barometer, and it jumped up and pointed to “very dry.” The Boots stopped as he was passing, and said he expected it meant to-morrow. I fancied that maybe it was thinking of the week before last, but Boots said, No, he thought not.

I tapped it again the next morning, and it went up still higher, and the rain came down faster than ever. On Wednesday I went and hit it again, and the pointer went round towards “set fair,” “very dry,” and “much heat,” until it was stopped by the peg, and couldn’t go any further. It tried its best, but the instrument was built so that it couldn’t prophesy fine weather any harder than it did without breaking itself. It evidently wanted to go on, and prognosticate drought, and water famine, and sunstroke, and simooms, and such things, but the peg prevented it, and it had to be content with pointing to the mere commonplace “very dry.”

Jerome K. Jerome "Three Men in a Boat"

[james_s]

Do you have a smartphone? The weather app on mine provides a far more detailed and accurate forecast than a barometer. Certainly a barometer can provide valuable information, but meteorology is a complex topic and there is much more to predicting the weather than watching the pressure.

[IanB]

High pressure means there is less water vapor in the air (the dry air is heavier), the low pressure means there is more water vapors in the air (the air with a lot of water vapor weights less). It is generally an indicator but not extremely reliable one. The sudden large changes towards the low pressure usually show the wet air is coming at you..

I'm not in any way a meteorologist, but isn't there more to it than that? High pressure usually indicates a stationary, stable air mass, while low pressure usually indicates winds and swirling air masses that are usually associated with storms?

[iMo]

You may have stable weather with both high and low pressures. The changes make the storms and windy weather (and usually the abrupt changes from high to low, like from 1035hPa to 910hPa within 1 hour - that would indicate large water masses in the clouds or fog are approaching you, and that generates storms, rain and the pressure front the strong winds).

[tggzzz]

High pressure means there is less water vapor in the air (the dry air is heavier), the low pressure means there is more water vapors in the air (the air with a lot of water vapor weights less). It is generally an indicator but not extremely reliable one. The sudden large changes towards the low pressure usually show the wet air is coming at you..

Er. No. The statements in your first sentence are not even wrong.

High pressure is caused when air masses descend. Low pressure is the reverse.

Bonus questions: how much does a fully developed cumulo nimbus cloud weigh? What is the humidity inside the cloud?

[iMo]

..High pressure is caused when air masses descend. Low pressure is the reverse..

I doubt it is related to the movement of the air masses. It is simply because the dry air is heavier..

[DiTBho]

Why is "fuzzy logic" relevant? In what way isn't it "magic"?

Some algorithms are entrusted with drawing inferences from the data.

      { barometer, thermometer } reading -> { stormy, rainy, change, fair, very dry }

AI-based algorithms are very good at this, but consume too much power, and my little PDA is a little battery powered ARM@400Mhz with only 64Mbyte of ram.

I hacked the battery, tripled its capacity (from old NiHm to modern LiPo), installed a modern GPS module; now I have about 40 hours of battery life with a new frequency scaling kernel hack that forces CPU @ 50Mhz, the wifi completely off and screen back-lit at 0% of la its brightness when not in use and 40% when you need to take a quick look.

Not bad for a survival travel computer! And I'm working on my front wheel (700x24,2U) to install a hub dynamo capable of generating up to 6-Watt so that I can charge the PDA battery while pedaling.

Fuzzy-logic consume less resources, it's not as great as AI-based (tensor-flow) algorithms, but if well-done and well-tuned it is less sensitive to small variations in values than every direct approach, and if well integrated with statistics, it allows discrete inferences.

Hence, even if it's a sub-sub-optimal approach, it should be more robust in classifying discrete events from noisy data-set from digital barometers (which are very noisy).

(d'oh, the ARM chip is an intel PXA, there is no FPU ... no DSP ... kernel and userspace emulation)

[tggzzz]

..High pressure is caused when air masses descend. Low pressure is the reverse..

I doubt it is related to the movement of the air masses. It is simply because the dry air is heavier..

Bonus questions: how much does a fully developed cumulo nimbus cloud weigh? What is the humidity inside the cloud? What is the humidity immediately outside the cloud?

[floobydust]

OP it's very limited by looking at one variable only. I watch barometric pressure and the changes can tell you when a system is moving in/out, but not what effect it will have. We had a recent large jump up/down in two days from 100 to 104kPa (graph) as a cold front moved in and down to -30°C and then back up to 0°C. Will it snow? Stare at the graph and it's too difficult. You need a few long-term averages and compare to the trend. I find you can forecast a weather change only when there are extreme excursions in barometric pressure, over X hours.

For a long time, weather forecasting based on a barometer was life/death for a seamen. Despite the finest brass and oak, the gauges were not enough.
Royal Navy Admiral FitzRoy Storm_glass (he did barometers too) at Lee Valley discontinued.
I did try the Kikkerland Storm Glass but it did nothing, so I did the gentle reheat and recrystallize and... it did nothing. The formula/Camphor or construction might be wrong nowadays. It is using both temperature and pressure.

Tempest_prognosticator or leech barometer probably more accurate than some meteorologists lol.

[golden_labels]

With enormous amount of global data and detailed models we have now, predicting weather events for a few days ahead is unreliable. In large urbanized areas it can hardly reach a few hours, still with huge uncertainity: predictions are breaking due to air currents and very fine scale required to model them. Sunny weather is generally easy to predict, but anything related to clouds, thunderstorms or strong winds… nope.

That should give you the upper bound, of what you may expect from a barometer. A barometer, that gives you single value from a single point on Earth.

My guess, unsupported by evidence, is: if you were a farmer and had no access to better information, waking up in the morning, looking and the barometer and expecting either rain or fine weather was still very valuable even if it failed 3/4 of times.

But nowadays your national weather instututions are likely publishing forecasts from their models, with 10 km and 15 min resolution, very reliable for 24h and still pretty good for 48h. Outside cities, that is.

Unfortunately Google Scholar does not yield any interesting results. I suspect that’s because barometer use started to drop, replaced with national scale forecasts, before methods permitting testing their effectiveness became widely known and used in science (second half of 20th century).

[SiliconWizard]

Well sure, we were using barometers along with others cues to predict weather for centuries before computers and more elaborate models appeared.

Now you need to define what "forecasting" means to you.

What you can predict with just atmospheric pressure and its changes is limited. Sure a relatively fast drop in pressure is usually not a good sign. But you don't know what's really going to happen. Some heavy wind, or rain, or a storm, or... a combination thereof. A slow drop may not mean really bad weather OTOH. It's probably going to be cloudy, but you can go with a low pressure for days on end without a drop of rain.

[tggzzz]

My guess, unsupported by evidence, is: if you were a farmer and had no access to better information, waking up in the morning, looking and the barometer and expecting either rain or fine weather was still very valuable even if it failed 3/4 of times.

In the UK the rule of thumb is that tomorrow's weather will be like today's weather. That is probably 50% correct.

A barometer will increase that slightly.
Looking upwind will increase it more.
In hilly areas, the predictability is worse.

Nowhere is the predictability sufficient for the OP's requirements. That's why weather (cf climate) is a traditional conversation topic.

[RoGeorge]

TI has had a wrist watch with a MSP430 as a devboard.
https://www.eevblog.com/forum/microcontrollers/msp430-wrist-watch-project/

It's programmable, has an atmospheric pressure sensor and a low power RF link with its own USB dongle for PC.  The pressure sensor is used there as an altimeter, but should be possible to use it for weather forecast if you tinker a little with the code.

[DiTBho]

TI has had a wrist watch with a MSP430 as a devboard.
https://www.eevblog.com/forum/microcontrollers/msp430-wrist-watch-project/

It's programmable, has an atmospheric pressure sensor and a low power RF link with its own USB dongle for PC.  The pressure sensor is used there as an altimeter, but should be possible to use it for weather forecast if you tinker a little with the code.

so, let's go with triplex { barometer, thermometer,  altimeter }.
I can integrate all of these sensors inside my Japanese PDA

[tggzzz]

TI has had a wrist watch with a MSP430 as a devboard.
https://www.eevblog.com/forum/microcontrollers/msp430-wrist-watch-project/

It's programmable, has an atmospheric pressure sensor and a low power RF link with its own USB dongle for PC.  The pressure sensor is used there as an altimeter, but should be possible to use it for weather forecast if you tinker a little with the code.

so, let's go with triplex { barometer, thermometer,  altimeter }.
I can integrate all of these sensors inside my Japanese PDA

A humidity meter would be more useful. I doubt the results would be much use for you, and certainly less useful that a decent online forecast - or even using your eyes.