Saturday, December 1, 2012

Can Smartphone Observations Revolutionize Weather Forecasting?

I believe that one of the biggest revolutions in weather forecasting is literally in your hands:  your smartphone.
The large Galaxy S3 has a decent atmospheric pressure sensor.
It turns out that a number of new smartphones, such as the Galaxy S3, have relatively good atmospheric pressure sensors in them.  Intended to provide elevation information for a variety of apps, these pressure sensors--if collected and quality controlled--could provide thousands, if not millions, of pressure observations each day across the nation.  The implications are huge, particularly since surface pressure is a uniquely valuable meteorological observation.  As described below, a private-sector company has already created an app to collect and display these pressures, and meteorological research (including myself and others at the UW) are doing research on getting maximum benefits from these pressure observations.

First, which smartphones have pressure sensors in them?  Right now they are mainly Android 3 phones, including: the Galaxy Nexus, Galaxy S3, Galaxy Note, Galaxy Note II, Nexus 4, Nexus 10 and Xoom.  Tens of millions of these phones have been sold worldwide, so there are undoubtedly millions in the U.S. alone.  Millions of weather observatories, if only if they were collected in real time.

One company has stepped up to the plate:  a small Canadian firm called, appropriately enough, Cumulonimbus Software.  They have developed a nice app for the Android 3 phones called PressureNet2.1  that can be downloaded for free.  This app accesses your

pressure observation and your location (smartphones use GPS and/or cell phone tower triangulation) to a central database, and allows you to view the observations taken by others.  Here is a screenshot of the graphics provided by the app:


I have been talking to the developers of PressureNet2 and they said that thousands of folks signed up for their app before and during Hurricane Sandy hit the East Coast.  Here is a plot of the smartphone pressure observations for an area of New York City during landfall.  Amazing....you can see the pressure minimum and the calibrations were really quite good.

Pressure is a uniquely valuable observation...in many ways better than temperature or wind.  Surface pressure expresses the state of the deep atmosphere, since it depends on the weight of the air above.  Pressure doesn't have the same exposure "issues" as temperature (is the thermometer in the sun?) or wind (is the anemometer behind a tree or house?).  Pressure sensor can be inside a house or outside...doesn't really matter since the pressure is essential the same.   Pressure errors are generally easy to remove (e.g., elevation differences or errors in a particular barometer).  And you can also use the  pressure change information.

Even better a number of research groups include those at the UW are studying the use of new data assimilation methods to take advantage of pressure observations. (data assimilation is the combination of models and observations to produce the best possible three dimensional descriptions of the atmosphere.  Data assimilation provides the starting point of weather forecasts done by computers).   With Professor Greg Hakim and Graduate Student Luke Madaus we are testing the use of advanced ensemble-based data assimilation using surface pressure, with sophisticated quality control of the pressure observations.  Here is an example of a Puget Sound convergence zone.  You can see the radar an area of heavy precipitation over Seattle.

These figures show the analysis of our data assimilation system with (left) and without (right) additional pressure observations.  No contest...pressure really helps.


So imagine the possibilities.

Millions of surface pressure observations across the country EVERY HOUR, allowing us to define local scale features, such as the cold pools and dry lines associated with the initiation of severe convection.    Using these observations we could radically improve our definition and description of fine-scale atmospheric features and short-term forecasts.

This could be a major advance in weather prediction technology.  And then we can use the smartphones to provide folks with highly detailed and accurate forecasts at their location.  so the benefits flow both ways.

Exciting times ahead!


18 comments:

LatigoLiz said...

Requires Android... Boo! :(

Rob W. said...

I've been waiting for the car GPS systems that give traffic information to also start displaying weather information - rain/snow/fog. The first two shouldn't be that difficult in getting the information from current, accurate sources.

Polistra said...

Extremely interesting!

Might finally get back to the data density we had in the 1890s when every little town and whistlestop had a telegrapher who reported twice daily to the Weather Bureau.

Or at least we might finally get weather data from actual cities, instead of airports.

Westside guy said...

I'll be curious to see whether this turns out to actually be as useful as you hope, Cliff. I suspect you'll get 99+% of your readings clustered in major cities, where I imagine you guys already have networks of permanent pressure sensors (which would be more accurate and calibrated, to boot).

Rivrdog said...

collection of data on large scale is not new, and predated smartphones.About ten years ago, NOAA commissioned a Colorado company to develop equipment to observe multiple data indicators in the mid-atmosphere, by installing automated observation on regional aircraft, which fly well below the altitude profiles of long-haul aircraft. The issue with this program was the down time to install the equipment on the aircraft, and the additional maintenance involved, but models showed a huge dearth of data from that layer of the atmosphere, so it had to be done.

The lesson from that program is that NOAA needs to be prepared to spend more money to speed the programs along, and money seems to be the sticking point.Lots of money for theoretical research, but the purse strings are tightened when it comes to actual hardware or data bases.

To be fair, buyers of these smartphones ought to be compensated in some way for being data points. Tax relief against some of the many cell phone taxes might be enough incentive.

David Burch said...

This is a very exciting development. One more reason to be glad I follow this blog. Thanks.

A good way for users to test these barometers is at www.starpath.com/barometers

The free service finds the closest ten stations to user's lat-lon that have accurate pressure online, then the program allows users to automatically fit a 3D pressure surface to the the stations of their selection for two times, and the service will interpolate for their location at a common time. It then runs a self consistency check to confirm that the stations are all equally valid.

The elevation correction for user's device is also computed, as is an accurate elevation if that is not known.

JewelyaZ said...

Downloaded and installed! 995.09 mb here now. :-)

Max Powers said...
This comment has been removed by the author.
Gerry Barnett said...

or, build your own app... Open Data Kit at UW... http://opendatakit.org/

Rivrdog said...

@ David Burch: just look at the NOAA observations for the airport nearest you and read the METAR, which will have the altimeter setting. You'll have to look up the elevation of the airpatch and figure your elevation from your GPS app, but if they are within a couple hundred feet, you have a decent pressure reading.

KE7DKG said...

Google already does this with their Traffic Data on Google Maps. As long as you have not opted out, your Android phone sends an anonymous ping to Google with your location and speed every few seconds, they then use that data for traffic information is your location matches a roadway.

Hassan Yate said...

I am a relatively new reader to this blog and find its content fantastic, thank you!

By the way you are actually referring to Android 4.0 phones and up (Ice Scream Sandwich & Jelly Bean etc) and not Android 3.

Rivrdog said...

I was doubtful that it would work on my older Xoom tablet, but I downloaded it and it fired right up!

I have an elevation here of 133.3 meters, including my office being on the second floor, and that jives fairly well with a 10 mbar difference from the average of the Portland cluster to my west at 45-75 meters lower elevation less elevation.

We had a decent FROPA here last evening, but I couldn't define it well across the cluster, the westernmost extent of which is 36K wide, east-to-west. I'm on the far east end of that cluster currently.

Anthony Crespo said...

Currently better than anything else out there!

David Burch said...

ref to rivrdog note: The database we use includes all metars, but for the best results it is not enough to just look at the closest station, regardless of what it is. First, the nearest station could be in error, second it does not interpolate for time or location. The free service we have online takes all of this into account... and it accesses the national elevation data set for an accurate elevation as well. All you need to know is the so called removal correction, ie how high off the ground is your instrument, then feed that back in along with the elevation to get the slp offset.

To learn about barometer errors at metar reports, look up the station in question in the Gladstone Family reports. They track these errors, much the same way we can spot them with our 3d-fitting program. We do a self consistency check amongst all data selected for the fit.

Users can get their lat-lon from Google Earth.

If there is any interest in it, i would be happy to explain why as mariners we care about accurate pressure.... or it can be found in our book called The Barometer Handbook.

Daniel said...

Would that be considered "cloud sourcing"?

Dan said...

Nice, downloading now.

Tyler Tanton said...

Perhaps I'm missing something here, but if the pressure sensor was included with the hardware as a means of better determining elevation, then without it, the elevation reading has a larger amount of error. Now if the phone has a large amount of elevation error, and reports the pressure to a weather server, then the pressure itself has a large amount of error synoptically. The pressure reported by the phone cannot be used for both elevation and atmospheric pressure since that would negate each one's integrity.

What am I missing here...?