September 30

The Accuracy Experiment

A big thank you to everyone who has agreed to take part in the accuracy experiment for the rain predictor (I’m going to need a name soon).  You should find some buttons in the app itself to vote on whether you found the prediction accurate or not.  Here is a link to the app:

Rain Predictor

If you are unable to leave a comment from within the app, here is a link to survey:

Accuracy Survey

My mobile is  0408 553 175 if you are stuck at any point.  Many thanks again.

Jason Jeffers


Category: Uncategorized
August 14

New Version out Soon

I’m just about to release a new version of the weather app that is ready for me test more systematically with users.  In the mean time here is a link to the current version.

I now have quite a bit of experience using it myself and do use it regularly, mainly for working out whether to drive my kids to school or walk.  I have a reasonable feeling for how I trust it (something that doesn’t just automatically happen even for me).  From what I’ve experienced:

  •  If it reports more than a 25% chance of rain, that is high and means there is rain around and you’re in danger of getting wet (that’s my cut off for driving my kids to school).
  • There is a major bug in this version that means you should not trust the very first answer it gives back!!  You need to wait wait a minute for the second answer once it has full updated your position.  It has taken me a while to work out why this happens but it is fixed for the new version.
  • I have found it is quite accurate (accurate enough to be useful is for you to decide) but it is literally predicting whether there is ANY rain.  I have lost count of the number of time I have seen it predict rain, I look out and think it is mistaken, then find when I actually walk outside to double check that it technically is actually raining but so lightly that it wouldn’t bother anyone.  There is a rain level prediction that is difficult to interpret in this version and it has been changed for the new version.
  • I know people want rain alerts and I have a way of sending email alerts in the new version but not this one.
Category: Uncategorized
March 5

Weather Beta Version Released (Almost)

I’ve been beavering away on a major update to the weather app.  You can see the results at:

At the moment I can’t trust the output on the new version so I’ve also kept the old version with predictions that I trust.  They both should produce the same results because the use the same model but for some reason don’t.  I’ll get the bottom of it soon.  Here is where you find the old version:

For the old version you can produce predictions for a location by clicking on the map at the location where you want to see the predictions (The screen needs to be scrolled to the top though :(, sorry).

Category: Uncategorized
January 21

New Version of Website

I’ve updated the web version of the rain predictor so that it now updates automatically when new predictions are created.  The graph has also been updated to display rain probabilities in a comprehensible way. It also updates automatically and show the probability of rain at the selected location by the number of minutes into the future.

Category: Uncategorized
December 24

The Weather App

This year I have been working on a smart phone app that uses GPS location to predict how likely it is to rain in the next few minutes at that location. I had to put it on hold for a few months but I’ve been able to work on it again over the last few weeks and now have an alpha web version that you can check out (alpha version is tech speak for ‘I’m too embarrassed by the number of bugs to show anyone but you can probably see what I’m trying to do’).

At the moment I have predictions for Melbourne and Mt Gambier which you can access from the link below:

To interpret the predictions you read them just like the Bureau of Meteorology radar images except that they predict what will happen in the future.  The page cycles through 12 images: the first two images show the actual rain as indicated by the BOM radar and the remaining images are the predictions in 6 minute intervals up to an hour ahead.  The time for the prediction is displayed at the top.  Unfortunately the predictions won’t update automatically when new data comes in you will need to manually refresh the web page (one of those alpha version things). The darker the display, the higher probability of rain: it doesn’t predict the heaviness of the rain at the moment though.   You can click on the pause button to stop the animation and fix on a single prediction.

There is also a difficult to read graph at the bottom which predicts the probability of rain at an exact location (not your GPS location just yet).  To set the location click on the location of interest in the main display.  At this precise time of writing this feature is not working but that will probably change in a few hours because it is working on my development site.  Another alpha version warning: you will need to scroll to the top of the web page for the location to correctly register.

I’ll add more locations over the next few days and I’m expecting to make quite a few changes over the next week or so.  Check out the link and let me know any suggestions or problems you have (  Always access the link from this blog ( because I will be making a lot of changes and could even completely change the web location, but I’ll publish any changes here.  The predictions update every 6 minutes when new data comes in from the Bureau but don’t be surprised if you get old predictions because my web server is not happy doing the number crunching.  I’ll relieve it of it’s duties when I get a chance (alpha version again, ah the list of bugs is never ending).





Category: Uncategorized