Canada – 2: GHCN/GISS Comparisons with Environment Canada

I’m a bit behind this week due to volcano-related travel disruption. It has also taken me longer than anticipated to download and plot the data I wanted.

While looking at Canada in Part 1 I knew there was station data on the Environment Canada Climate Data website. My problems started when I decided to take a look at it.  You see I was hoping for some recent data for stations that were no longer actively reporting into GHCN, and otherwise expecting broad agreement with the GHCN/GISS data. Well, I did fulfil the first but not the second expectation, so am left a bit puzzled by it. 

On the Environment Canada climate site there is a nice interactive map that allows you to zoom in by region (e.g. map right) and then click on stations to get data. I downloaded the data as CSV files and used the pivot table function in Excel to aid conversion to annual data (D-N). Up-to-date GHCN/GISS data was downloaded from the GISS station data site (Unadjusted i.e.”after combining sources at the same location” and Adjusted “after homogeneity adjustment”).

On the quality of data, Environment Canada’s statement is here:

“Data collection, processing, quality control checks and procedures have evolved and changed over the years since the earliest data was observed in 1840. Changes have been particular rapid in recent years. For example, the advent of automatic weather observing stations established a new set of challenges for quality control.

The vast majority of observational data is accurate but the database contains some incorrect values, which show up from time to time. Environment Canada continues to review quality control procedures, both as current data is observed and incorporated into the database, and retrospectively for historical data. Be aware that data can be erroneous and that some values may change over time as quality control procedures identify and deal with doubtful data.”

I took this to mean that this is “raw” but quality controlled data; I am unable to determine what, if any adjustments (e.g. TOBS, SHAP) have been made in the data.  Anyway, here is the first set of station data I looked at and compared with GISS data – Prince Albert, Saskatchewan:

So, an immediate surprise – the data from Environment Canada matched neither versions of the GISS data.  Not only that, but the GISS unadjusted data was warmer in the past than the EC data but was adjusted to be cooler.  This is an example of “wrong way” correction for urban growth and urban heat island effects within the GIStemp program and seems to happen when surrounding rural stations have a greater rate of warming than the station being adjusted (Prince Albert has ‘semi-urban’ status: population 31,000).
Hmm, there is a 0.5-1 deg C difference in some of those temperatures – why? Which set is right? Who do you believe? And is this repeated for other stations, which I’ll come to in a moment?

I was inspired to look at high latitude stations by a post (Infilling the True Hypothetical Cow) a week or so back in which Boballab downloaded data for Alert in the Canadian Arctic and looked at the effect at grid square level of adding back in data that is no longer being reported into the GHCN datafile. He did this at grid square level.  Now he found good agreement between Environment Canada and GISS adjusted data when the data was anomalized with a 1951-1980 baseline.

I wanted to to look at additional high latitude stations and I listed (from GHCN v2.mean) those that had at least 20 years of data up to 1988 or later and were above 60N in latitude: 37 stations in total. Of these I found 18 on the EC site, although 3 (Hall Beach, Inuvik and Fort Simpson) had so little data that it was not worth using. The 15 that were in common with the GHCN/GISS data are shown and named on the map below; the GHCN/GISS Stations not on the EC site are marked but not named.

Four: Alert, Eureka Resolute and Tuktoyaktuk are above the Arctic Circle; together all the stations cover the three Canadian Provinces Territories of Nunavit, Northwest Territories and Yukon Territory.

On comparison the stations fall roughly into three categories (for which only the latter two are mutually exclusive):

• Those in which there is a historical record from GHCN/GISS, which ceases typically around 1989-90, but for which Environment Canada has more recent data.  The GISS series may restart in the last few years:  Alert , Burwash, Dawson, Hay River, Mayo, Tuktoyaktuk, Watson Lake, Yellowknife

Alert (Nunavut)

Hay River (NWT)

• Those for which the GHCN/GISS data typically runs “warmer” than the Environment Canada data: Alert, Baker Lake, Burwash, Coral Harbour, Eureka, Tuktoyaktuk, Whitehorse, Yellowknife

Coral Harbour (Nunavut)

Eureka (Nunavut)

• Those for which the GHCN/GISS data typically runs “cooler” than the Environment Canada data: Fort Smith, Hay River, Mayo, Norman Wells, Resolute, Watson Lake

Fort Smith (NWT)

Resolute (Nunavut)

Hmm, no overall pattern, in that we can’t say there is a ‘rule’ that GHCN/GISS is working to a specific adjustment against the EC data. Also, with the older data also suffering a ‘shift’, and many stations having a similar magnitude of ‘shift’ throughout, we can’t say that there is an adjustment for a change of location or equipment. Curiouser and curiouser.

I do notice though that many of the ‘shifts’ are fairly consistent.  Since there is the usual hue and cry of “absolute temperatures do not matter” (“the anomalies will save us”) when any skeptic posts a graph with actual temperatures, worrying over the differences between the two data sources may be irrelevant.  So bear with me while I go away and calculate some anomalies and see what they say. (Answer found see Part 4)