University of Alaska Fairbanks
Geophysical Institute

Beyond the Mouse 2010 - The geoscientist's computational chest.

Lab 6: Matlab I/O 2

"Programming is legitimate and necessary academic endeavor."
Donald E. Knuth

Lab slides

none.

Exercise 1:

Download these two files and store them in the same directory:

Execute the script such that the figure appears on your screen. It looks a little weird with spikes; like a random fence. We'll take care of that later. For now, I want you to discover the properties of the figure/axes you can change with the property editor: in the figure window menu go to 'edit'->'axes properties' and/or 'figure properties'. Try to find a button that says 'More Properties'; change things and see what happens. Take note of the respective names of the properties; they are similar to what you would use as parameters in get/set. Also try adding a title, and labels for x and y axis. There's nothing to turn in, just keep in mind that basically every figure property you can change in the Figure Editor can be changed from the command line/ your scripts.

Exercise 2:

Now we're getting a little more serious; the display of the data as shown in the figure wrong, i.e. does not correspond to the actual timeseries in the file! We've got to correct for this and make things look nicer. If you check line 5 in fai_temp.m you will see that I plot temps over dates where dates is a vector that holds the dates of temperature measurements in the format 'YYYYMMDD' (check FAI_temps.txt if you don't believe me). Matlab does not know anything about the semantics of the dates vector for the x-axis and assumes dates contains integer values. This is where the weird spaces in your figure come from: we have big gaps from 1980-12-31 to 1981-01-01, i.e. 19801231...19810101, and 19811231 - 19820101, and so on. Matlab assumes those gaps are values and plots a continuous x-axis which results in it reserving space on the x-axis. That's not wrong and also not too bad, but by default plot connects all data points with a line. Now everybody thinks there's data where you have none.

This becomes clear when you replace line 5 in fai_temp.m with: plot(dates, temps, '.'). Understood? OK. Change it back!

Let's fix this now! Matlab's internal representation of times is using serial date numbers. Instead of using dates as given in the original file, let's do this:

You might notice that the tick labels are in strange places. Let's try having them every 5th year. There are many ways of doing this. For the most generic one we only need to know that the date format is 'yyyymmdd.' To get tick marks in the style '1985 - 1990 - 1995 ... ' without actuall knowing upper and lower bounds of the time series, we can do this (we treat the yyyymmdd dates as actual numbers here):

Alright. Now YOU know that your x-axis tick spacing is 5 years, but it's not really obvious in the figure as the datenumber labels haven't been converted to a reasonable string yet. We should change the tick labels!

Great, now that we know what we're dealing with we can add title, x, and y labels. I recommend you to use '\circF' as part of your ylabel argument to clarify that temperatures are in degrees Fahrenheit (we know that from experience - the numbers in the file certainly aren't degrees Celcius, they remain undocumented though).

If you want to get a better idea about how the temperature evolved over this long time, you could smooth it using a moving average approach over a period of 365 days:

Your final figure should look similar to this:
Fairbanks Intl Airport Temperature data, example

Exercise 3: Setting axes

Now, with very few steps you can add a lot of extra information to that figure. Let's add a spectrogram. Here's what I want you to do: Your final figure should look similar to this:
Fairbanks Intl Airport Temperature data, spectrogrma, example

Helpful Matlab Functions:

Here is a list that might be of help.

ronni <at> gi <dot> alaska <dot> edu | Last modified: February 01 2013 21:41.