Before attempting to model daily temperature between the Southwest Nova Scotia (SWNS) weather stations, an examination of the temperature data alone can help us answer several important questions.

 
1. Where are the SWNS weather stations? What does the temperature dataframe look like?
2. How many active stations were recording temperatures each day throughout of the study period?
3. How does daily temperature mean, maximum and minimum vary from station to station?
4. How much does daily temperature variation between stations change over the seasons?
5. How do station temperatures rank amongst eachother, from hottest to coldest, over the seasons?

 

 

1. Where are the SWNS weather stations? What does the temperature dataframe look like?

There are a total of 98 SWNS stations involved in collecting the data for this project. The map, below, shows their placement over the region:

The temperature data from the stations is stored as a dataframe that includes: information about the weather stations (unique ID and coordinates), a date field, and three daily temperatures (minimum, mean and maximum); for the years 2012 - 2017.

stationid date_time temp_min temp_mean temp_max EASTING NORTHING
27141 2012-01-01 -0.8 2.9 6.5 383226 4991426
27141 2012-01-02 1.0 6.4 11.7 383226 4991426
27141 2012-01-03 -9.8 -3.9 2.1 383226 4991426
27141 2012-01-04 -12.2 -10.0 -7.8 383226 4991426
27141 2012-01-05 -8.0 -3.9 0.3 383226 4991426
27141 2012-01-06 -10.5 -7.8 -5.0 383226 4991426

2. How many active stations were recording temperatures through each day of the study period?

A rough calculation would estimate the numbers of dataframe records as follows:
 
# of data frame records = 365 days x 6 years x 98 stations ~ 21.5k records

 
This is close to, but larger than the actual number of records, 188163. Over the 6 years, stations were gradually introduced, then retired, and down for periods of time. The number of active stations on each day of the study period can be visualized to quickly understand of the coverage of temperature recordings over the six years.

 
The amount of active daily stations in 2012 began in the low 80’s and slowly increased over 150 days to the low 90’s. At approximately 250 days in there was a drop-off back to daily counts in the low 80’s. The year finished off with a slight increase of stations in the last few weeks, to approximately 85 active stations per day.
The active station count in 2013 and 2014 remained mostly constant at approximately 95. However there are several days that sporadically drop stations, mostly in winter periods (within the first and last 50 days of the year). For the years 2015 and 2016, the active station count remains steady in the high 90’s. However, in 2017 the active station count immediately starts to drop and continues throughout the year, finishing off at approximately 50 stations.

3. How does daily temperature mean, maximum and minimum vary from station to station?

An important assumption to challenge is: the stations’ daily temperature recordings are different from one another. Based on local knowledge of SWNS, we assume that if the stations are reasonably spread over the region, that they will record varying daily temperatures. If the stations are not recording varying temperatures, we can only conclude from this data that temperature is constant throughout SWNS, and modelling efforts would be redundant.  

A plot of all daily temperatures (daily mean, maximum and minimum) at all stations was generated for the last day of each month in 2015. The year 2015 was selected, as this year has the most consistently high daily active station count. Also this year towards the middle of the 2012-2017 study period, giving a median-like representation. From the individual daily plots, we can see how much temperatures vary between stations on a single day. To ease legibility, stations are arranged on the x-axis by increasing daily temperature mean, and station ID’s are not shown. How daily temperature ranks at each particular station is examined later on.
 

 

The plots show that daily temperature means range sizes are approximately 5 degrees celcius, at a minimum, for each day. The daily temperature maximums and minimums have even larger ranges. The plots increase confidence in the assumption that stations are recording a range of temperatures, and that this assumption remains constant over all seasons. The range sizes are summarized in a table:  

Date Range of Daily Temp. Means Range of Daily Temp. Minimums Range of Daily Temp. Maximums
2015-01-31 5.50 8.40 9.31
2015-02-28 10.21 19.92 7.73
2015-03-31 6.88 9.50 7.36
2015-04-30 6.41 4.30 9.26
2015-05-31 10.32 11.11 13.71
2015-06-30 8.13 5.16 13.74
2015-07-31 10.03 7.80 13.91
2015-08-31 5.52 9.25 10.49
2015-09-30 5.23 5.01 5.83
2015-10-31 14.03 9.30 20.54
2015-11-30 6.76 11.08 7.20
2015-12-31 6.55 7.22 7.95

The daily temperature maximums are the most variable, with range sizes hovering around 10, and even reaching 20. The daily temperature minimums have similar ranges to the daily maximums, although are more often smaller on a per-month basis. The daily means have lower ranges that tend to fall around 5 or 6, although 4 months had ranges of 10 or larger. The comparatively lower ranges of daily temperature means are expected, since they are derived from daily temperature minimums and maximums. For example, if the minimum and maximum are 0 and 10, the mean will be 5. The same is true if the minimum and maximum are 4 and 6.

4. How much does daily temperature variation between stations change over a year?

The plots in the above section demonstrate that we can expect to find variation daily temperatures among stations, on potentially any day of the year. By zooming out, and looking at these variations for every single day of 2015, we can see how temporal scale affects the distribution of temperature throughout SWNS.

The smooth lines of the plot show a seasonal trend in the daily temperature variables. The daily maximum temperatures vary most in the middle 2015, decreasing towards the beginning and end. The opposite is true of daily minimum temperatures. They have the largest ranges at the beginning and end of the year, with a dip in the middle. The daily mean temperatures are mostly balanced by the opposing effects of the minimmum and maximums, although they seem to dip in the middle of the year. By looking at the same plots for the other years in the study, we can confirm or deny that these trends may be real.

4. Does the season affect variation between stations?

In the previous section, the smooth-line plot showed a seasonal trend in the range sizes of daily temperature variable between weather stations in 2015. The daily temperature maximums in 2015 have a wide range of values over the stations in the summer, and a small range in the winter. The opposite is true of daily temperature minimums in 2015. By generating the same plot for all years, we can determine if this trend is consistent:

A similar seasonal trend is visible for all the years in the study period, 2012 - 2017. The trend in daily temperature maximums remains very consistent in shape across the years. The daily temperature minimum has the same basic idea, but looks to be shifted between the years. In 2015, the trend in ranges for daily temperature minimums mirrors that of daily temperature maximums. However, in the other years, the trend is not mirrored, but lagging by approximately 50 days.

5. How do station temperatures rank amongst eachother, from hottest to coldest?

We have confirmed that stations are recording daily temperatures that vary from eachother. By ranking the stations from hottest (Rank = 1) to coldest (Rank = 98), and plotting them, we can quickly visualize how the temperature is distributed over SWNS. One plot from the first day of each season will give us idea of how the distribution changes over the seasons.

Spring:
Summer:
Fall:
Winter:

The plots reveal a seasonal change in the ranking among stations. This information is important for our modelling, as it suggests that environmental variables that impact temperature, have different strengths of influence through time.