Brent in Edmonton

Here’s the plot:

And the code:

 1 library(RPostgreSQL)
 2 library(ggplot2)
 3 library(plyr)
 4 drv <- dbDriver("PostgreSQL")
 5 con <- dbConnect(drv, dbname="power", host="HOST", user="USER", password="PASSWD", port=5432)
 6 all <- dbGetQuery(con, "
 7     SELECT extract(year from obs_dt) AS year, 
 8            trim(to_char(obs_dt, 'Month')) AS month, 
 9            CASE WHEN extract(month from obs_dt) between 5 and 9 THEN 'Summer' ELSE 'Winter' END AS season, 
10            obs_dt, power 
11     FROM observations
12     WHERE obs_dt > now() - interval '1 year' 
13     ORDER BY obs_dt;")
14 all$month <- factor(all$month, c("January", "February", "March",
15     "April", "May", "June", "July", "August", "September",
16     "October", "November", "December"))
17 all$season <- factor(all$season, c("Winter", "Summer"))
18 cum.seasons <- ddply(all, .(season), summarize, power = unique(power), ecdf = ecdf(power)(unique(power)))
19 f <- function(x) {
20     r <- quantile(x, probs = c(0.01, 0.25, 0.5, 0.75, 0.99))
21     names(r) <- c("ymin", "lower", "middle", "upper", "ymax")
22     r
23 }
24 o <- function(x) {
25     subset(x, x < quantile(x, probs = c(0.01, 0.99))[1] | quantile(x, probs = c(0.01, 0.99))[2] < x)
26 }
27 png(file = "ninety_nine_boxplot_w_jitter.png", width = 800, height = 400)
28 ggplot(data = all, aes(x = month, y = power))
29     + stat_summary(fun.data = f, geom = "boxplot", color = "blue", fill = "slateblue")
30     + stat_summary(fun.y = o, geom = "point", size = 1, color = "red", position = position_jitter(width = 0.1))
31     + scale_y_continuous(name = "Power (Watts)")
32     + scale_x_discrete(name = "Month")
33     + opts(title = "Power consumption by month")
34 dev.off()
35 
36 # vim:ft=r

The trick is to build the boxplot elements (box and whisker aesthetics) using a function that uses the quartile function, then subset the points to outside the range of the whiskers using another function. The plot is generated by using these functions inside stat_summary. Thanks to kohske on stackoverflow for showing me this technique.

Thanks much for your comment and explanation of what could be going on with this winter. I wish I knew (or understood) more about the larger scale weather patterns and how they relate to what we see on the ground. And I look forward to the possibility of an average snowfall for once.

Cheers,

Chris

I ran across your pages while searching temperature data bases for Fairbanks. I am currently a City Manager for a North Slope village, but live here in Fairbanks now (resided on the North Slope Sep 2009-Aug 2010). In a former life I was a university professor with specialization in climate dynamics focusing on inter- and intra-annual seasonal variability in North American temperature and precipitation.

I read with interest your brief analysis regarding what we might expect for Winter 2010-11 here in Fairbanks as a function of the La Nina SSTA. I have done a little work in the past on this subject myself, and began to look at it again recently. In general I’ve found two different La Nina modes for North America as they relate to the general atmospheric circulation pattern, and this year we’ve found ourselves facing the lower frequency version characterized by a frequent “over the top block” that connects the Bering Sea ridge to the North Atlantic ridge. This results in the negative Arctic Oscillation (AO) we have observed much of the winter which has resulted in frequent cold air dumps into the eastern CONUS (not a typical La Nina response). When this occurs we tend to see winter snowfall in Fairbanks somewhere between 50-75% of average, which is where we appear to be heading this year. Of course if the small-scale details of the “ice storm” last November had been different, we could actually be much closer to our 30-year average now.

At any rate, since La Ninas tend to persist for about 26 months, and our most recent event began this past summer, we should have another shot at a near-average snowfall winter for 2011-12 with below “normal” temperatures for the period. It’s unlikely that this second La Nina winter would include an over the top block, so perhaps this is something positive to anticipate.

Take care,

Dave

Very cool!!!, I would very much like to see the code Do you write to a database?