Plotting a Cold Change

Saturday 23rd January 2010 saw a classic heatwave/cold front event occurring up the eastern coast of New South Wales Australia, and I observed from Sydney, watching things progress during the day via the internet, as well as from my own home, where I have a view across parts of Sydney.

Oz Weather v2.1 introduced graphing of weather history as a new feature, and the graphs from that day show the change very clearly indeed. The following graph is a composite of the different ones available in Oz Weather, although I have overlaid a transparent bar indicating the time when the main changed occurred.

A summary of the changes:

  • Temperature dropped from about 41°C to 22°C.
  • Humidity jumped from 10% to 85%
  • Wind jumped from 30km/h to 65km/h with gusts to over 95km/h just as the change came through, and the direction shifted from NW to S.
  • Interestingly, the pressure started to rise an hour or so before the main change, and there was a little rain from some thunder cells that developed following the change.

The Doppler (wind) radar also showed the approach of the wind change very clearly. Unfortunately I didn’t save a graphic from when the change was passing right through Sydney, but an earlier shot shows the change passing through Stanwell Park, to the south of Sydney.

The key point here is to note that blue indicates wind towards the radar location (centre of crosshairs) and yellow indicates wind away from the radar location. So this is showing strong NorthWest winds (blowing offshore) over the Sydney region, but from the South at Stanwell Park and below. This picture was a lot more striking as the change passed through Sydney itself, but I’ll have to wait for another event to show that off better!


Oz Weather v1.4 Released

Sales of Australian weather apps have been quiet recently (more details to come in a future post), and right now Oz Weather is at #17 overall ranking of Australian paid apps. This is a very respectable position by almost any standard, although much less impressive than it has been historically.

I am attributing this mainly to the fact that the weather is now inter-seasonal and hence rather bland, especially when compared to the exceptional heatwave and tragic bush-fires in Victoria two months ago, during which sales peaked quite dramatically. In a previous article I revealed the relationship between interest in weather and Oz Weather sales levels, and this apparently continues to hold true, as weather query rates have been below 1 per app per day recently, indicating an historically low interest level in the weather from app users.

However, Oz Weather development continues never-the-less! Version 1.4 has just been approved by Apple. This approval took 4 working days (6 days elapsed), which seems to be about par for the course.

The main new feature is a much-improved city selection dialog. Although it has been possible since v1.0 to use GPS location via the “Locate Me” crosshairs button on the add city screen, it seems that many users never even noticed it was there, and thus were scrolling through the list of 274 cities in alphabetical order in the hope of finding somewhere nearby by finding familiar place names.

The new screen design has the following improvements

  • Divides cities into separate lists for each state
  • Makes the “Near Me” button much bigger and easier to find
  • Also allows users to view a list of  locations near to any specified location

Add City Dialog Add City Dialog 2

The other changes are a little more subtle. For example I have added grip bars to the current weather views, thus making it more obvious that you can scroll sideways to find observations for other nearby locations within your vicinity. I’ve almost lost count of the number of emails I’ve received from users requesting additional observation locations, and who seemed to be very pleasantly surprised when I told them they could already just scroll sideways to get exactly those locations.

Side Grips

And the least visible change has been a very satisfying one for me, as I believe that I may finally have conquered an obscure problem that could cause a crash. It was a fairly rare crash, but not rare enough to prevent some understandably annoyed users from leaving some bad reviews. Sadly there is no mechanism in place allowing developers to respond to people who report their problem via a review. But here is a suggestion for Apple. When someone submits a review with, say, 2 stars or less, then prompt the user with a message like the following.

If you have experienced problems with this app, then please contact the developers via their supplied contact email address for support. Developers are unable to contact you or respond to you regarding any problem that you report only via a review.

Apple already does prompt users with messages when they leave reviews – why not just add an extra one  like this? It could greatly help users and developers alike.

What Pushed Oz Weather To #1?

Some time ago, Tom Adams (CTO, MoGeneration) asked me whether there was any link between Oz Weather sales and the unusually stormy weather we’d been having in Queensland. And at that time I couldn’t think of any obvious way to find out.

However, a few days ago inspiration finally struck. It occurred to me that one of the server stats I’d been recording was producing a pretty good indication of how unusual or interesting the weather was on any given day. That statistic is the number of weather queries per installed app.

Typically this seems to be 1.0 to 1.5 weather queries per day per app, but this peaked at about 4.3 weather queries per app during the recent Victorian extreme heatwave and bushfires. In fact there were more than 100,000 queries on that day alone, from an installed base of about 24,000 units.  So the next step was to visualise the data graphically to see if there was any relationship in general over time, as shown below.

Sales Vs Weather

There does indeed appear to be a relationship here! The blue curve is a logarithmic best fit, although a linear fit would probably have been as good.

The conclusion I am drawing from this is that there is “base” level of sales of around 150 to 200 per day, and that rate increases by 100 or more sales per day as the weather becomes more extreme, interesting or newsworthy.

You might question whether I’ve got the relationship the wrong way around here ie. whether in fact the new daily sales are causing the higher query rate than usual as people first tried out the app. No doubt this is indeed a factor. However, given that new daily sales are typically only about 1% of total apps already in use, it seems very unlikely that this could actually account for the large changes observed.

In terms of app rankings, it interesting to note that Oz Weather finally jumped into #1 position in the paid app rankings on a day when there were 4.1 weather queries per app. It remained there for 12 days, while the rate of weather queries averaged 2.5 per app (ranging from 1.5 to 4.3). It finally fell back to #2 position again on the day that the weather query rate had dropped back to 1.3 per app, and since then it has dropped further to #6 position, with an average weather query rate of just 1.2 queries per app over the last 20 days.