Example 04: NEW ZEALAND Tmax

UPDATE December 2021: This example temperature reconstruction has been updated to latest available data (November 2021). The start date has been amended to 1863, due to the inconsistency of the data from the two stations with data before that date. Some stations have been added, and some removed. Recent data is now  only from stations “near” Wellington, roughly in the centre of the country, stations further afield are only used to extend the reconstruction as far back as possible.

This post documents a reconstruction of changes in monthly average surface air daily maximum temperatures (Tmax) for New Zealand back to 1863, see below for why this start date was adopted. Data is averaged democratically, excluding periods deemed to be distorted by non-climatic influences such as station moves and equipment changes. Data is not averaged from the far South West of the South Island, or from the far North and North East of the North Island, in order to limit the geographical spread of stations in the average.

Temperature Data Sources: All temperature data (monthly average Tmax) was downloaded from NIWA CliFlo (see the Data page for link) in late 2021. It is not necessary to deal with version numbers, as it is all raw data, and NIWA does not alter its raw data. A list of stations used is given at the end of this post.

Metadata Sources: Online “Sensor and Site History” pages, accessed via the “agent number” of the stations, and “NIWA7S”, the NIWA “Report on the Review of the NIWA 7-stations”, December 2010 (see the Data page for links).

Methodology: The standard methodology of this website was used, see the pages above, starting with OUTLINE.

RESULTS

The following figure shows the 12-month moving average of regional-average Tmax variations, together with its 11-year moving average:

The seasonal dependence of the Tmax variations is shown in the following figure:

START DATE

There is data from the 1850s from 2 sites, Auckland and Dunedin. If those 2 sets of data were consistent (maybe with a step change inhomogeneity) then they could be included in the reconstruction. The following figure shows the 12-month moving-average differences between recent data from those cities, indicating what consistency looks like:

The following figure shows the differences between the early station data:

There is insufficient consistency between the two sets of early data, and insufficient metadata to allow one of the sets to be trusted and the other to be discarded.

VALIDATION

The following set of figures show 12-month and 11-year moving averages of “regional weather-corrected” Tmax data for all of the stations examined, in order of increasing start date. The stations can be identified using the following  lists of stations, from top to bottom on each figure, together with the alphabetical list of stations to be found at the end of the post.

Fig. 1: station ids 10, 2, 9, 15, 17, 33
Fig. 2: station ids  7, 19, 11, 10, 31, 36
Fig. 3: station ids 24, 13, 8, 25, 26, 35
Fig. 4: station ids 34, 14, 32, 21, 1, 5
Fig. 5: station ids 28, 27, 4, 16, 23, 30
Fig. 6: station ids 12, 20, 18, 22, 6, 29

The bold lines are the station regional-weather-corrected moving averages used to obtain the regional moving average, shown as the dashed black lines. Validation follows from the identification and clean exclusion of periods of anomalous temperature changes, which are the periods without bold lines.

SAMPLE STATION ANALYSIS

Each station record underwent a visual inspection to detect/confirm/mark inhomogeneities. The following two figures shows the main analysis plots for an example station: NELSON (agent 4244).

The following figure shows 12-month and 11-year moving averages, for the regional-weather-corrected station data and for the regional average:

Inhomogeneities are clearly visible at around 1867 and 1920, the latter expected from the metadata, which indicates a move in December 1920. However, examination of the monthly data version of the figure above reveals a possible error in the metadata:

The figure above gives strong evidence that the move occurred in January (rather than December) 1920, with data before (such as October to December 1919) being “high” and data after being “low”. Maybe the metadata date was when the move was recorded, rather than when it happened.

STATION DATA

The following figure shows the “nominal” number of stations being averaged in each year of the reconstruction, the actual number of stations in the average is somewhat lower because some periods of data are omitted:

The list of stations used, together with the periods excluded, is given in the following list, copied and pasted from the configuration file for the region.

NOTES

• TR = transition, the name given to all periods excluded from regional averaging
• W = 1 means this station record was included in regional averaging of weather fluctuations (slow transitions, such as gradual urban heating, or shading from growing vegetation, might be included in weather averaging, but excluded from averaging of station moving averages)
• M = 1 means this station record was included in the regional moving averages
• The first flag  in transition definitions allows temperature fluctuations to be included in regional weather averaging, 1 means exclude, 0 means include.
• The other transition parameters are the start year/month then the end year/month
• QC NaNs are data that were deemed to be invalid, set to NaN (Not a Number) by the software, triggering their automatic infilling using valid data either side and the regional average weather fluctuations
• The information below is ALL the configuration for this region, allowing reproduction of the results shown, given the software (available on request)

*********************************************************

% WELLINGTON_Tmax.m, N_half = 5; % Moving average window size = 2*N_half + 1 = 11 years

stations = […
% Agent W M From Norm To (Agent = NIWA Agent number)
01 4239 1 1 1932 1995 1996;… % APPLEBY (no TR)
02 1427 1 1 1853 1980 1989;… % AUCKLAND ALBERT
03 4638 1 0 1867 1879 1880;… % BEALEY (no TR) [SOUTH ISLAND]
04 4322 1 1 1941 1980 1987;… % _Blenheim_Aero_Tmax
05 4331 1 1 1932 1984 1985;… % BLENHEIM [SOUTH ISLAND]
06 4395 1 0 1997 2015 2021;… % _Brothers_Island_Tmax
07 4858 1 1 1863 2016 2017;… % CHRISTCHURCH GARDENS
08 5380 1 1 1886 1912 1913;… % DUNEDIN LEITH
09 5378 1 1 1862 1885 1886;… % DUNEDIN ROSLYN
10 22645 0 0 1852 1863 1864;… % DUNEDIN PRINCES
11 3907 0 0 1866 1930 1946;… % HOKITIKA
12 1340 1 1 1966 1990 2021;… % _Leigh_EWS_Tmax
13 4881 1 1 1881 1986 1987;… % LINCOLN (nr Christchurch)
14 2446 1 1 1923 1952 1953;… % MASTERTON BAGSHOT
15 2982 1 1 1862 1864 1866;… % Napier_Tmax
16 4241 1 1 1943 2015 2021;… % NELSON AERO
17 4244 1 1 1862 1950 1951;… % NELSON
18 2283 1 1 1991 2016 2021;… % NEW PLYMOUTH AWS (no TR)
19 2276 1 1 1864 1875 1973;… % NEW PLYMOUTH
20 3232 1 1 1970 1985 1990;… % _Palmerston_North_Ki
21 3238 1 1 1928 2000 2001;… % PALMERSTON N (no TR)
22 8567 1 1 1993 2015 2021;… % _Paraparaumu_Aero_Tmax
23 3145 1 1 1953 2015 2021;… % _Paraparauma_Tmax
24 5446 0 0 1871 2016 2021;… % QUEENSTOWN
25 1565 1 1 1888 1894 1895;… % TE AROHA (NI) (ONLY 1888–>1895 USED)
26 2473 1 1 1906 1990 1991;… % WAINGAWA (NI, near Masterton)
27 3477 1 1 1939 2000 2006;… % _Wallaceville_Tmax
28 3715 1 1 1937 2016 2021;… % WANGANUI EWS (NORTH ISLAND)
29 25354 1 1 2004 2015 2021;… % _WELLINGTON_KELBURN_AWS_Tmax
30 3445 1 1 1962 2016 2021;… % WELLINGTON AERO (no TR)
31 3389 1 1 1868 1868 1869;… % WELLINGTON Bowen_Tmax
32 3385 1 1 1927 2004 2005;… % WELLINGTON KELBURN (no TR)
33 3383 1 1 1862 1865 1868;… % WELLINGTON KNOWLES
34 3391 1 1 1912 1926 1927;… % WELLINGTON THORNDON (no TR)
35 3431 1 1 1906 1911 1912;… % WELLINGTON BUCKLE (no TR)
36 3390 1 1 1869 1905 1906]; % WELLINGTON BOLTON

END OF POST