Sunday, 16 December 2012

A weather update

When I acquired my new weather station Frances commented that I will be able to spend many hours analysing the data.  As usual she is spot-on, but it has taken me a while to sort things out with a database.

Before getting to that a quality statement may be helpful.  The temperature recording has continued to track very well against our other sources (various thermometers around the house; BoM reports on TV news).  Wind direction seems reasonable but needs more study.  We have only had one day with and it appears that the new system has under-recorded that compared to the old technology.  I will reassess this and if needs be, relocate the device if a better spot can be found.

What follows is based on 25 days of observations which means the weather isn't affected greatly by day length but the averages could be affected by outlier observations.

My first effort was to look at some temperature data to see how it varied through the day.

I am surprised at the flatness of the average in the early morning (ie up to 7am) and pleased to see the drop off from 6pm onwards.  In local mythology that reflects the Easterly winds kicking in, bring the more moderate temperatures from the coast.  So the next step was to see which way the wind blows through the day.

Trying to sort this out in an easy-to-present form has been a little challenging.  A first effort was to look at the proportion of observations with wind from various directions.  That is summarised in the following pie chart.
This is not strictly speaking a rosette since the points don't quite match the directions (eg the centre of the 'W' slice is closer to NW than W per se).  However it isn't bad in showing that there have been few breezes recorded from the Northern quadrant (from NW to NE).  Possibly this reflects trees getting in the way in that direction?  A key learning experience in this is the difficulty in dealing with text descriptions of wind directions.  It became far easier when converting them to angular degrees (E=90, S=180 etc)

The next effort was to contemplate how the direction varied through the day.  My default hypothesis was from the NW during the day swinging round to the East in the evening.  Trying to summarise the data to investigate this proved particularly challenging.  The eventual approach I have adopted was to

  • count the number of observations for each direction for each hour and then take the direction with the maximum count.  
  • As always with matters statistical tied scores were difficult and I decided that I would simply average the angular degrees between the ties .  (For example if W [270 degrees] and South [180 degrees] got the same score, the most common direction was averaged as 225 [ie SW].)  
  • A philosophical point arose when at one point the scores for NW and SE were the same, and the question arose whether to take the average assuming a veer through N (equating to NE) or a veer through S (equating to SW).  Since the 'surrounding values' were SW or S I assumed the veer to be Southerly.
  • I then plotted these values on a graph and smoothed them a little with a 4 point moving average.

The smoothed data show the wind coming from the West in the morning, moving towards South in the afternoon and returning to NW in the evening.  A little different to expectations, but more in the order of "interesting" rather than causing thoughts about "errors" to arise.


No comments: