Everyday uses for the NZTR

1. Computer models that forecast the weather

Gareth Renowden at Hot Topic has finally lost whatever finger-nail grip he ever had on climate science.

He now claims (incredibly) that national temperature records “play no part in planning” since forecasts come from computer models, not carefully-kept historical records. It’s sad, really, that NIWA totally disagrees with him. And we shall prove it.

Our good friend Chris de Freitas wrote an excellent article for the NBR on 14 November reporting on the paper by him and his co-authors—a reanalysis of the NZ temperature record (NZTR), published last month in Environmental Modeling & Assessment. Renowden first quotes de Freitas:

National temperature trends are widely used for a large number of societal design and planning purposes and it is important that they should be as reliable as modern methods allow.

Renowden comments:

This is transparent nonsense. Historical temperature trends are interesting, but they play no useful part in future planning. To plan in the face of rapid climate change, we need good regional projections for temperature changes, sea level rise and increases in weather extremes. Those will come from climate models, not temperature records.

We do agree that New Zealand’s national and regional temperature projections will continue to come from NIWA’s uber-expensive super computer. But you have to feed it something. So the 64,000-dollar question is: Where does the mathematical model come from and what data goes into it?

In simple terms: what on earth does it start with? The answer is, of course, it starts with reality: with now.

So here’s what NIWA does. It takes twelve models from the 2007 IPCC CMIP3 suite, forced by the long-outdated A1B scenario for global emissions. Then, according to its web site:

Historical observations are used to develop regression equations that relate local climate fluctuations to changes at the larger scale. These historical observations are then replaced by the model changes in the regression equations to produce the fine-scale projections. [emphasis added]

So, historical observations of the New Zealand climate (for example, the 7SS temperatures) are the starting point for downscaling the global models. This is hardly surprising, as every projected trend is by definition an extension of past trends. You have to know accurately where you are to start with.

NIWA doesn’t say whether it uses the 7SS raw data or the homogenised series set out in Mullan10, but I’m betting it’s the latter. That means the downscaling project is fed with a temperature trend that’s 300% too high.

It has proved impossible to verify climate models. Their projections have a very poor record world-wide and virtually all the CMIP3 models have consistently over-stated global mean surface temperatures for the past 20 years. However, an effort is invariably made to improve performance by calibrating and tweaking the model until it is able to “hindcast” the observed temperature record for previous decades.

David Wratt and Brett Mullan say on the NIWA website:

If we are to have confidence in future projections of climate models, it is vital to first test them against observed climate to see whether they realistically simulate different climatic conditions which were observed during the earth’s past. [emphasis added]

So the downscaled NIWA model has been calibrated by reference to Mullan’s version of the 7SS, using exaggerated adjustments. If it is managing to hindcast the national temperature trend as warming at the rate of 0.91°C/century, its untestable future projections are completely useless.

An important example of misleading projections from the NIWA model is that of mean sea level rise (MSLR). We know from tide gauges that there has been no acceleration of MSLR around New Zealand during the past century. The NIWA model should therefore pick up the fact that the MSLR has not been compatible with the temperature data it has been fed from Mullan’s 7SS. But it doesn’t, because it uses a calculated history of global SLR drawn from satellite data, and ignores the actual painstaking readings that New Zealand scientists have recorded for decades, in some of the longest records in the Southern Hemisphere.

So, Gareth, what do you have to say now about the “transparent nonsense” you alleged of Chris de Freitas? What he said is a perfectly accurate reflection of what NIWA does. What you said is demonstrably wrong—you can check the references for yourself. I think the least you can do is to apologise.

Wratt and Mullan also say on the web site:

Confidence in the ability of climate models to estimate future climate changes comes from the fact that they are based on accepted physical laws such as conservation of mass, energy and momentum, as well as a wealth of observations for their more empirically-based components such as cloud reflectivities or infrared absorptive properties of greenhouse gases. [emphasis added]

Errors in any of these components render the computer model output useless and should be corrected as soon as possible.

Imagine that overseas scientists reported that NIWA was using an incorrectly stated law of physics: say, an incorrect constant value, or a wrongly formed equation. NIWA scientists would move so quickly to fix a mistake like that that they would be mere blurs in the corridors, their speed immense. They have put in an enormous amount of work to build up credibility with colleagues around the world and would spare no effort to correct such critical errors.

Well, we don’t need to imagine anything, because this is reality: what has actually happened is that local scientists have reported that NIWA has been using an incorrectly stated national temperature record. Why are they not running to fix it? Do they disagree with the laws of physics?

We haven’t seen a reply yet to our rebuttals of your debating points, Gareth.


Hits: 116

30 Thoughts on “Everyday uses for the NZTR

  1. Richard C (NZ) on 20/11/2014 at 2:54 pm said:

    >”NIWA doesn’t say whether it uses the 7SS raw data or the homogenised series set out in Mullan10, but I’m betting it’s the latter. That means the downscaling project is fed with a temperature trend that’s 300% too high.”

    No, not the 7SS or any homogenised series. It’s their Virtual Climate Station Network (VCSN):


    I very much doubt the trend in the VCSN is anything like the 7SS. The VCSN takes the raw CLiFlo data from all over NZ as a base, as I understand, and interpolates from that. NIWA are very quiet about this. It might be interesting to challenge them to graph the trend in the VCSN against the trend in the 7SS.

    The paper you need to refer to is:

    Regional climate modelling in New Zealand: Comparison to gridded and satellite observations D. Ackerley, S. Dean, A. Sood and A.B. Mullan (2012) http://www.metsoc.org.nz/publications/journals

    Page 4 pdf,

    2.3 Observational data
    2.3.1 Virtual Climate Station Network (VCSN) data.

    In this study, we use the gridded fields of the
    Virtual Climate Station Network (VCSN) at
    NIWA, which was produced from observed
    station data using the methods described in
    Tait et al. (2006) and Tait (2008). We
    compare the daily maximum temperature
    (Tmax), daily minimum temperature (Tmin) and
    the daily accumulation of precipitation from
    the VCSN data with the RCM output over the
    1980–1999 period. This is diffferent to the
    method employed in Drost et al. (2007), who
    calculated the mean temperature from the
    VCSN Tmax and Tmin values and compared
    that to their RCM mean temperatures. Biases
    in the modelled values of Tmax and Tmin may
    cancel out when the mean is calculated, which
    can mask errors in the model data. Treating
    Tmax and Tmin separately removes this issue.
    The major differences noted by Drost et al.
    (2007) between the gridded observational data
    and their RCM simulations were too much
    rainfall over the western South Island and too
    little on the eastern South Island, along with
    cold biases in western areas and warm biases
    to the east. Drost et al. (2007) attributed these
    to differences in the grid-point heights used
    by the RCM and the gridded data set (VCSN).
    To avoid this possibility in this study, the
    spline method used by Tait et al. (2006) and
    Tait (2008) was applied to the climatological
    station data using the grid-spacing
    specifications of the model (30km resolution)
    and the model’s topography. This ensured
    temperatures were being compared at the
    same altitude in both VCSN and RCM data
    sets. Despite reducing the errors due to
    previously using different topographical
    datasets, other errors do still occur in the
    VCSN data. One example occurs in upland
    areas where winter temperature inversions can
    alter the relationship between Tmin and
    elevation due to deviations from the
    atmospheric lapse rate (Tait, 2008). However,
    by using the model topography, the spline
    method (Tait et al 2006; Tait, 2008)
    reproduces the observed data on the model
    grid and allows a fairer comparison with the
    model data.


    Tait, A., Henderson, R. and Turner, R. and
    Zheng, X.G., 2006. Thin plate smoothing
    spline interpolation of daily rainfall for New
    Zealand using a climatological rainfall
    scheme. International Journal of
    Climatology, 26(14), 2097-2115:

    Tait, A.B., 2008. Future projections of
    growing degree days and frost in New
    Zealand and some implications for grape
    growing. Weather and Climate, 28, 17-36.

    From Tait et al (2006):

    2.4. Spatial interpolation methods
    The underlying usefulness of spatial interpolation methods is directly related to the quality and quantity
    of the surface observations. In New Zealand, as has been mentioned in Section 1, there is an excellent and
    extensive network of currently open climate stations that record rainfall totals every day. Historically, the
    network has been more than double the current number of stations (the first meteorological observations for
    New Zealand were made in 1852 and a maximum of 1652 open rainfall stations was reached in October,
    1970); however, this was in an era before government funding levels required a rationalisation of the network.
    Nevertheless, unlike the radar- and satellite-based methods described above, there are sufficient observational
    data for generating historic (1960–2004) daily rainfall estimates for the entire country. In addition, access to
    the data is very efficient. Most of New Zealand’s rainfall data are stored in the National Climate Database
    (CLIDB), an Oracle relational database. This is maintained by the National Institute of Water and Atmospheric
    Research (NIWA) in Wellington, New Zealand. Data on daily rainfalls are quality checked and stored as 24-h
    totals from 9 to 9 A.M., New Zealand Local Time (NZLT).

    I don’t know the year VCSN temperature commences but presumably it would be prior to the 7SS.

    # # #

    RT, you need to call NIWA out on this.

    • Richard C (NZ) on 20/11/2014 at 3:05 pm said:

      Ackerley et al (2012), page 14 pdf:

      Figure 5: The annual, New Zealand mean (a) Tmax (oC), (b) Tmin (oC) and (c) daily mean
      precipitation (mm day-1) from VCSN (black lines), RCM1 (red lines) and RCM2 (yellow lines)
      data between 1980 – 1999. The correlation values (given in the key for each graph) are between
      the VCSN time series and each of the model time series separately. Statistically significant
      correlations (p≤0.05 using a student t-test) are given an asterisk.

      “RCM” is Regional Climate Model.

      1980 – 1999 trend is irrelevant to the 7SS contention so it’s the VCSN trend prior to 1980 that needs to be published, and preferably graphed against the 7SS.

      Shouldn’t be difficult.

    • Richard C (NZ) on 20/11/2014 at 3:11 pm said:

      >”The VCSN takes the raw CLiFlo data from all over NZ as a base, as I understand, and interpolates from that.”

      This product predates BEST but BEST made a song-and-dance about their revolutionary approach. It would also be interesting to see the VCSN trend against BEST’s NZ trend:


      BEST NZ trend 1910 – 2013: 0.87 ± 0.26

    • Richard C (NZ) on 20/11/2014 at 4:00 pm said:

      >”I very much doubt the trend in the VCSN is anything like the 7SS”

      I suspect it is more like GHCN from 1880, 0.2 °C / Century:


      From the source of the graph, Not PC:

      “And just for your interest, here’s a graph [above] from the Global Historical Climatology Network dataset for NZ’s latitude and longitude (unlike NIWA, GCHN has no political axe to grind) so you can see for yourself the “general warming trend in NZ” that Frogblog is talking about. Does 0.02 degrees/decade sound catastrophic to you?”


      GHCN NZ trend from 1880: 0.2 °C / Century
      BEST NZ trend from 1860: 0.71 °C / Century

      Given BEST is just below NIWA’s 7SS from 1910, 0.87 °C / Century vs 0.91 °C / Century, then I think the BEST vs GHCN difference is a good indication of what the 7SS vs VCSN difference would be from 1909.

  2. Richard C (NZ) on 20/11/2014 at 4:37 pm said:

    >”NIWA doesn’t say whether it uses the 7SS raw data or the homogenised series set out in Mullan10, but I’m betting it’s the latter. That means the downscaling project is fed with a temperature trend that’s 300% too high.”

    The fact that NIWA calibrates their RCM against VCSN (as upthread) not 7SS was pointed out in ‘Paper adds interesting perspective on NZ temperature trend’ here:


    Your response RT, was:

    [RC – as I mentioned, this info is technical and I find it hard to absorb in one bite. That means others will struggle, too. Now that’s fine, it’s a technical topic, but you make the bite larger by increasing the number of lines. Would you kindly paste text that will flow in the normal way, please? I’ve put this slab through NoteTab to join the lines and show what I mean. You’re helping to create a quite unique research tool here, but I’d ask you to respect the prime purpose of this blog, which is to make the information accessible. Thanks for your help. – RT]

    Apparently your reformatting did not aid your understanding of the underlying information. Please take the time to get to grips with it this time around.

  3. Richard C (NZ) on 22/11/2014 at 1:52 pm said:

    Change to VCSN Data Access: From 13-Oct-2014 public access to VCSN (Virtual Climate Station Network) data has been discontinued. Please contact [email protected] if you wish to discuss data access.

    de Freitas et al (2014) publication:
    Received: 29 December 2013 /Accepted: 7 October 2014
    Published online: 26 October 2014

    Pre-emptive strike by NIWA? I didn’t know there was any public access to VCSN, the blurb indicated that access was fee-based. Just wish I’d extracted T Max/Min/Mean when there was public access, if it were possible.

    Still, anyone can “discuss” data access apparently, but Climate-Smart Farmers aren’t happy:

    Virtual Climate Station Network

    Well from 13th October 2014 we have a problem. We no longer have access to the Virtual Climate Station network that drives the weekly climate updates. NIWA, the Crown-owned agency that holds this data has decided to terminate access and deny this information to landusers, unless you pay a fee to use their NIWA FarmMet subscription service. We think we had developed, in collaboration with NIWA, a pretty useful service that helped land users make better decisions using information that was local. And we had plans for working with NIWA to further develop and expand the applications and value of this data for business management. The irony is that a whole lot of the real climate stations that provide the data that NIWA uses for these models is collected across the rural community by farmers.

    So we apologise for this cessation of service around the VCS network – but keep using the other parts of the site. We do intend to try and provide real information that helps make a difference, one way or another.

    So in the interim, none of the instructions below will function!

    Across New Zealand there are 11,491 virtual climate stations. [continues……]


    • Richard C (NZ) on 22/11/2014 at 3:26 pm said:

      From Climate-Smart Farmers:

      >”1. Click on this Map browser and zoom into your location.”

      Works, gets you to the ESRI ArcGIS:


      Go to Masterton.

      The 7SS reference station for Masterton is “Site 7 East Taratahi AWS (2612) 8 km southeast of Masterton” (see map link below) but is actually south and a little west of Hood Aerodrome according to the map. Oddly, the 7SS reference station East Taratahi AWS does not appear in VCSN.

      Masterton Aero, not 7SS but mentioned in pdf below, is VCSN Hood Aerodrome:


      The VCSN grid containing East Taratahi AWS appears to be the VCSN node (south of Hood Aero):

      AGENT_NO 29831
      NETWORK P193132
      LONGT 175.63
      LAT -41.02
      NAME Carterton District

      The P193132 Mean series should conform to the 7SS Masterton series in toto because all of the Masterton 7SS stations are homogenized to East Taratahi AWS and East Taratahi AWS is within P193132.

      But “No information available”. Climate-Smart Farmers old workaround no longer works. That was to take the above information to the default plot and enter it here:

      DEFAULT PLOT LINKS (will open in new windows)
      Air Temperature and Frost: http://wrenz.niwa.co.nz/vcsoper/P203148_air.png

      [NiWA] – “Important Notice: This (VCSN plots) service is provided by NIWA as a prototype and temporary service for test purposes only and no guarantee is given on this service. NIWA reserves the right to discontinue this service at any time. NIWA is currently working on a robustly supported, web-based subscription service to replace this service in the future. Comments / Suggestions please to [email protected]

      You don’t have permission to access /vcsoper/P203148_air.png on this server.”

      OK, so now try CliFlo:

      STEPS FOR EXTRACTING Virtual Climate Station DATA

      1. Go to http://cliflo.niwa.co.nz
      2. If you do not have a registered cliflo username and password, click on ‘subscribe on-line’ (there is no cost)
      3. Login
      4. On the database query form, click on ‘select datatype’ then ‘special datasets’ then ‘VCSN’
      5. On the database query form, type in the agent number of this VCS point
      6. On the database query form, choose a date range (VCS rainfall data begin on 1/1/1960; VCS wind data begin on 1/1/1997; other VCS data variables begin on 1/1/1972. The data are updated every day at approximately 1pm local time for the 24-hour period up to 9am local time on the same day)
      7. On the database query form, choose an output data format (e.g. ExCel file)
      8. On the database query form, click ‘send query’.

      Got to 4. but:

      “Virtual Climate Network (VCSN) Access is restricted – see Changes in VCSN Data Access.”

      Next step is “discuss” with NIWA I suppose to get VCSN NETWORK P193132/AGENT_NO 29831 data.

  4. Richard C (NZ) on 22/11/2014 at 3:46 pm said:

    Comparing VCSN to 7SS directly is just a matter of obtaining VCSN Max/Min data for the 7 VCSN nodes corresponding to each of the 7SS reference stations.

    That negates the mission of obtaining the VCSN average Mean series for all of NZ (a mammoth task – probably expensive). But only useful if VCSN has data prior to 1980 back to 1909.

  5. Richard C (NZ) on 22/11/2014 at 7:45 pm said:

    VCSN nodes corresponding to each of the 7SS reference stations.

    Nodes are not all available i.e. VCSN has missing grids for 3 reference stations. The nearest and most appropriate nodes were chosen as representative of the 7SS reference station.

    North of Mangere Airport. not same node as Mangere EWS
    AGENT_NO 25397
    NETWORK P176213
    LONGT 174.78
    LAT -36.98
    NAME Manukau City

    South of Hood Aerodrome, same node as East Taratahi AWS
    AGENT_NO 29831
    NETWORK P193132
    LONGT 175.63
    LAT -41.02
    NAME Carterton District

    Wellington CBD altitude, not same as Kelburn altitude. Same node as Kelburn AWS
    AGENT_NO 28602
    NETWORK P176127
    LONGT 174.78
    LAT -41.27
    NAME Wellington City
    Karori altitude, same as Kelburn altitude. Not same node as Kelburn AWS
    AGENT_NO 24930
    NETWORK P175127
    LONGT 174.72
    LAT -41.27
    NAME Wellington City

    Stoke, south of Nelson Aero. Not same node as Nelson Aero.
    AGENT_NO 20596
    NETWORK P145126
    LONGT 173.22
    LAT -41.33
    NAME Nelson City

    Hokitika Town, same node as Hokitika Aero
    AGENT_NO 19484
    NETWORK P100098
    LONGT 170.97
    LAT -42.73
    NAME Westland District

    North of Lincoln, same node and location as Lincoln Broadfield EWS
    AGENT_NO 20993
    NETWORK P130080
    LONGT 172.47
    LAT -43.63
    NAME Selwyn District

    Belleknowes, not the same node as Musselburgh EWS
    NETWORK P090035
    AGENT_NO 19446
    LONGT 170.47
    LAT -45.88
    NAME Dunedin City

    Nodes can be viewed in GIS here:

    ‘Report on the Review of NIWA’s ‘Seven-Station’ Temperature Series’ shows correlation between VCSN and each 7SS location:

    Page 25 pdf:
    Figure 5: Map of correlation between interannual temperature changes at the grid-point
    nearest to Auckland Aero (Site 5, location marked by asterisk) near Auckland, 1972-73 to
    2007-08, and all other grid-points in the NIWA 0.05° gridded “Virtual Climate Station” data set.

    Page 52 pdf:
    Figure 5: Map of correlation between interannual temperature changes at the East Taratahi
    grid-point (Site 7, agent 2612, location marked by asterisk) near Masterton, 1972-73 to
    2007-08, and all other grid-points in the NIWA 0.05° gridded “Virtual Climate Station” data set.

    Page 51 pdf:
    “Over the past few years, NIWA research scientists have developed gridded data sets of daily climate
    parameters, on a 0.05° latitude by 0.05° longitude grid covering the whole country (a total of
    approximately 11,500 grid-points). The “Virtual Climate Station” (VCS) data set for daily maximum
    and minimum temperatures begins on 1 January 1972, and interpolates data from between 150 and 200
    climate stations using a sophisticated interpolation technique developed at the Australian National
    University in Canberra (Tait 2008).”

    # # #

    VCSN data only begins 1 January 1972. At 1972 there is very little difference between NIWA and NZCSC/de Freitas et al (2014), so not able corroborate 7SS in the early years prior to 1972. The VCSN linear trend, if projected back past 1972, should conform to NZCSC/de Freitas et al (2014) +0.34 C/century and raw unadjusted +0.3 C/century back to 1900:


    But given we only have VCSN 1980 onwards to look at so far (Ackerley et al, 2011) it is worth comparing the VCSN profile to the 7SS above. See Figure 5 page 10 Ackerley et al:

    Figure 5: The annual, New Zealand mean (a) Tmax (oC), (b) Tmin (oC)……. from VCSN (black lines), ……..data between 1980 – 1999

    The 7SS profile is very clear:
    1980 up 0.5 to 1987
    1997 down 0.7 to 1995
    1995 up 0.7 to 1999

    The VCSN Max/Min profiles (similar) look nothing like that and I suspect the respective trends, VCSN Mean vs 7SS would be very different even over those 2 decades. A series comparison from 1972 to present (or 4 decades to 2011) would be very enlightening.

    • Richard C (NZ) on 22/11/2014 at 9:55 pm said:

      BEST looks nothing like VCSN either over 1990 – 1999. Resembles 7SS up-down-up but somehow they manage to introduce a ridiculous trend:


      Data table:

      Year, Month, Annual Anomaly
      1980 6 -0.040
      1985 11 0.557 (up 0.6)
      1992 4 -0.848 (down 1.4)
      1999 5 1.063 (up 1.9)

      1980 12,0
      1987 12.5 (up 0.5)
      1995 11.8 (down 0.7)
      1999 12.4 (up 0.6)

      Except by their methodology, BEST should resemble VCSN.

      It will be possible, but painstaking, to re-compile the VCSN Mean from Ackerley et al Figure 5 to plot against 7SS and BEST 1980 – 1999. Should be interesting.

    • Richard C (NZ) on 23/11/2014 at 10:50 am said:

      >”BEST should resemble VCSN”

      It does. Both VCSN and BEST are at altitude therefore cooler than 7SS. The 7SS is near sea level mostly.

      1980 10.10 12.25 10.6
      1981 10.75 12.86 11.05
      1982 9.80 12.12 10.37
      1983 9.85 12.03 10.30
      1984 10.55 12.68 10.97
      1985 10.60 12.85 11.03
      1986 10.50 12.65 10.94
      1987 10.55 12.77 11.03
      1988 10.80 12.93 11.09
      1989 10.80 12.97 11.28
      1990 10.85 12.99 11.27
      1991 10.00 12.16 10.53
      1992 9.45 11.49 9.94
      1993 9.80 11.84 10.24
      1994 10.15 12.33 10.54
      1995 10.45 12.59 10.82
      1996 10.20 12.45 10.74
      1997 10.10 12.27 10.54
      1998 11.25 13.41 11.63
      1999 11.15 13.35 11.613

      Mean trends
      NiWA-VCS: 0.147 C/decade
      NIWA-7SS: 0.135 C/decade
      BEST-NZT: 0.147 C/decade

      Y intercepts
      NiWA-VCS: 10.23
      NIWA-7SS: 12.408
      BEST-NZT: 10.673

      VCS trends
      VCS-Max: 0.25 C/decade
      VCS-Min: 0.045 C/decade

      In all 3 series it is only the 1998 and 1999 data that skews the trend:

      Mean trends 1980 – 1997
      NiWA-VCS: -0.133 C/decade
      NIWA-7SS: -0.157 C/decade
      BEST-NZT: -0.127 C/decade

      VCS trends 1980 – 1997
      VCS-Max: -0.042 C/decade
      VCS-Min: -0.224 C/decade

    • Richard C (NZ) on 23/11/2014 at 11:09 am said:


      “Warming, over New Zealand through the past century, is unequivocal”

      NIWA, corroborated by BEST:

      “Cooling, over New Zealand through 1980 to 1997, is unequivocal”

    • Richard C (NZ) on 23/11/2014 at 5:05 pm said:


      “Cooling, over New Zealand through [1972] to 1997, is unequivocal” (2.6 decades)

      1972 is when NIWA’s VCSN begins but I’ve no access to data 1972 – 1980 and 1999 – Present. But BEST serves as a VCS proxy from 1972 because the methodologies are similar i.e. we can consider a “BEST-VCS” series instead of VCSN. And VCS, 7SS, and BEST profiles are identical (near enough) from 1972, just the altitudes (therefore absolute temperatures) and sensitivities differ (7SS most sensitive).

      There was cooling from 1972 to 1997 then an abrupt climate shift at the time of the 1998/99 El Nino. In other words, the only warming that has occurred in New Zealand over the last 4 decades has been natural variation. Numbers and trends another time.

    • Richard C (NZ) on 23/11/2014 at 8:41 pm said:

      BEST-NZT 1972-1987 (2.6 decades) trend: -0.002 C/decade
      NIWA-7SS 1972-1987 (2.6 decades) trend: -0.024 C/decade

      El Nino
      BEST-NZT 1988/89 shift: +0.40
      NIWA-7SS 1988/89 shift: +0.39

      BEST-NZT 1987-2013.25 (1.625 decades) trend: -0.156 C/decade
      NIWA-7SS 1987-2013 (1.7 decades) trend: -0.017 C/decade

      BEST is moving annual mean on monthly data. NiWA data is annual mean.

    • Richard C (NZ) on 24/11/2014 at 10:18 am said:

      Corrections (hopefully got this right now):

      BEST-NZT 1972 y intercept: 10.776
      NIWA-7SS 1972 y intercept: 12.469

      BEST-NZT 1972-1997 (2.6 decades) trend: -0.024 C/decade
      NIWA-7SS 1972-1997 (2.6 decades) trend: -0.024 C/decade

      BEST-NZT 1997 y value: 10.7136
      NIWA-7SS 1997 y value: 12.4066

      BEST-NZT 1998/99 El Nino shift from 1997: +0.46
      NIWA-7SS 1998/99 El Nino shift from 1997: +0.39

      BEST-NZT 1997 y intercept: 11.174
      NIWA-7SS 1997 y intercept: 12.798

      BEST-NZT 1997-2013.25 (1.625 decades) trend: -0.156 C/decade
      NIWA-7SS 1997-2013 (1.7 decades) trend: -0.017 C/decade

      BEST is moving annual mean on monthly data and approximates NIWA-VCS but 0.44 C warmer, both at altitude. NIWA-7SS is annual mean at near-sea level.

  6. Richard C (NZ) on 23/11/2014 at 9:08 pm said:

    The Regional Climate Model (RCM1), forced with ERA-40 re-analysis, conforms to 7SS, VCS, and BEST-NZ over 1980-1999 (including the 1998/99 El Nino) because ERA-40 is essentially the same data (see Ackerley et al, 2011). So RCM1 is reasonably reliable as long as it is constrained by the observations.

    But if Gareth Renowden thinks RCM1 projections past the observations are anything to go by for policy, he is deluded.

  7. Richard C (NZ) on 24/11/2014 at 11:33 am said:

    ‘Climate change scenarios for New Zealand’ [NIWA]

    The New Zealand downscaled projections follow the IPCC Fourth Assessment approach. That is, changes are relative to 1980-1999, which we abbreviate as “1990” for convenience.

    Mean Temperature:

    Downscaled projections of mean temperature changes over New Zealand are shown in Figure 3 (annual-average changes).

    Averaging over all models and all 6 illustrative emissions scenarios gives a New Zealand-average warming of 0.2–2.0°C by 2040…..[….]. For just the A1B scenario alone, the projected warming is 0.3–1.4°C by 2040 ……[…], with a 12-model average (or “best estimate”) of 0.9°C …[…] for 2040


    # # #

    So how are they going?

    1990 – 2040 is 5 decades. Their mid-range “best estimate” projection is 0.9°C or 0.164 C/decade. BEST-NZT is similar to NIWA-VCS methodology and tracks NIWA-VCS but 0.44 C warmer:

    NIWA-7SS 1980-1999 mean: 12.550
    BEST-NZT 1980-1999 mean: 10.827

    NIWA-7SS 1989/90-2008/9 is 1.9 decades
    BEST-NZT 1989/90-2008.25 is 1.825 decades

    NIWA-7SS 2004-2013 mean: 12.729 actual (12.862 projection)
    BEST-NZT 2003.33-2013.25 mean: 10.960 actual (11.126 projection)

    Already by 2013, NIWA’s mid-range temperature projections are warmer than observations:

    NIWA-projection vs NIWA-7SS to end of 2013: +0.133
    NIWA-projection vs BEST-NZT to 2013.25: +0.166

    • Richard C (NZ) on 24/11/2014 at 3:40 pm said:

      Curiously, although BEST-NZT and NIWA-7SS exhibit almost identical profiles and trends 1980-1997, that is not the case 2004-2013. BEST is very much flatter and the trend is half that of the 7SS (0.3 C/decade vs 0.6 C/decade)). There is no significance in the trends because the start year was relatively cool and the end year relatively warm in each case. It’s just the difference that is significant.

      So why is BEST-NZ considerably different to the 7SS now after being identical 16 – 33 years ago? I think because there’s a totally different set of stations contributing. BEST states on the NZ page that there are currently 33 “active” stations but I think there are 37 as follows::

      Station Name Months Distance (km) Earliest Most Recent
      AUCKLAND, ALBERT PARK 1904 Inside Jun 1853 Oct 2013
      CHRISTCHURCH AP/HAREWOOD N Z 1789 Inside Jan 1864 Oct 2013
      INVERCARGILL AERODROME 1423 Inside Jan 1865 Oct 2013
      NAPIER AERODROM AWS 1417 Inside Jan 1870 Oct 2013
      GISBORNE AERODROME AWS 1049 Inside Feb 1905 Oct 2013
      KAITAIA OBSERVATORY 849 Inside Oct 1942 Oct 2013
      TARA HILLS AWS 713 Inside Nov 1949 Sep 2013
      DUNEDIN AERODROME A 463 Inside Dec 1962 Oct 2013
      WAIOURU AIRSTRIP 461 Inside Jan 1973 Oct 2013
      HAAST AWS 453 Inside Jan 1973 Oct 2013
      CAPE REINGA AWS 443 Inside Mar 1954 Oct 2013
      WELLINGTON AIRPORT 406 Inside Jan 1960 Oct 2013
      TAURANGA AERO AWS 274 Inside Aug 1990 Oct 2013
      KAIKOURA AWS 269 Inside May 1991 Oct 2013
      HICKS BAY AWS 258 Inside Oct 1991 Oct 2013
      NELSON AERO AWS 257 Inside Apr 1992 Oct 2013
      PALMERSTON NORTH AW 253 Inside Sep 1992 Oct 2013
      AUCKLAND AERO AWS 229 Inside Aug 1994 Oct 2013
      CHRISTCHURCH AERO AWS 229 Inside Aug 1994 Oct 2013
      SECRETARY ISLAND AWS 223 Inside Sep 1994 Oct 2013
      WELLINGTON AERO AWS 219 Inside Jul 1995 Oct 2013
      QUEENSTOWN AERODROM 218 Inside Jul 1995 Oct 2013
      PURERUA AWS 209 Inside Jul 1995 Sep 2013
      TIMARU AERODROME AW 208 Inside Jun 1996 Oct 2013
      SOUTH WEST CAPE AWS 207 Inside Jun 1996 Oct 2013
      TAUPO AWS 206 Inside Sep 1996 Oct 2013
      CASTLEPOINT AWS 193 Inside Sep 1995 Oct 2013
      INVERCARGILL AERO AWS 72 Inside Jul 2005 Sep 2013
      FAREWELL SPIT AWS 451 0.14 Mar 1954 Oct 2013 [Inside NZ]
      PUYSEGUR POINT AWS 262 0.39 Aug 1991 Oct 2013 [Inside NZ]
      MANIHIKI AWS 31 0.40 Oct 1997 Sep 2013 [Inside NZ]
      MOKOHINAU AWS 220 0.58 Sep 1994 Oct 2013 [Inside NZ]
      NEW PLYMOUTH AERODROME 1495 0.76 Jan 1864 Oct 2013 [Inside NZ]
      WESTPORT AERODROME 259 0.81 Jan 1992 Oct 2013 [Inside NZ]
      PARAPARAUMU AERODROME 502 0.89 Jan 1972 Oct 2013 [Inside NZ]
      CAPE CAMPBELL AWS 436 0.97 Jan 1973 Oct 2013 [Inside NZ]
      HOKITIKA AERODROME AWS 1615 29.59 Feb 1866 Oct 2013 [Inside NZ]


      I count 37 “inside” NZ and BEST’s “Google Map of Stations” states: “37 station(s) are recently active” i.e. a 37SS. Apparently a number of NZ mainland locations are not “inside” NZ which is rather odd.

      It looks like the diversity in these 37 stations are not only “damping” the fluctuations in temperature inherent in the 7SS but the diversity produces a trend half that of the 7SS:

      NIWA-7SS 2004-2013 trend: +0.6 C/decade
      BEST-37SS 2004-2013 trend: +0.3 C/decade

      NIWA’s VCS should be similar to BEST’s 37SS. Obviously this is not good for NIWA’s projections because their proj-obs divergence is greater from BEST than from 7SS:

      NIWA-projection vs NIWA-7SS to end of 2013: +0.133
      NIWA-projection vs BEST-37SS to Sept 2013: +0.166

    • Richard C (NZ) on 24/11/2014 at 10:58 pm said:

      >”BEST is very much flatter and the trend is half that of the 7SS” [2004 to 2013]

      Presumably VCS is similar. Except NIWA publishes annual means each year that correspond to 7SS temperatures (appears to be preliminary) not VCS, although they don’t state this explicitly. Preliminary it seems because they don’t update the 7SS immediately, for example the last available datapoint tp download is 2011 and the value differs from the published figure given for that year.

      NIWA has published a 2013 figure of 13.4 which is a marked increase from 12.5 in 2012 (+0.9 C) i.e. an apparently exceptional year-to-year change. This jump makes a big difference to the trend obviously.

      BEST does not yet have Oct, Nov, and Dec for 2013 but given the “damped” nature of their 37SS, those 3 months will not make much difference from that of the 12 months to Sept. BEST annual data is a centred average i.e. last datapoint 2013.25 is the centred mean of the 12 months Oct 2012 to Sept 2013.

      So looking at what BEST does have, there is quite a difference to the 7SS 2010 to 2013 (nominal means):

      2010: 13.07
      2011: 12.8
      2012: 12.5
      2013: 13.4 (+0.33 from 2010)

      2010: 11.24
      2011: 11.1
      2012: 10.78
      2013: 11.27 (+0.03 from 2010)

      Except these nominal annual averages are somewhat misleading. 2010/11 was El Nino and warmer than 2013. This can be seen in BEST’s moving centred annual mean data rather than aggregating to nominal years. The exceptional 2013 disappears in BEST’s 37SS month-by-month moving annual means:

      BEST 37SS moving annual means
      2010 12 11.480 June 2010 to May 2011
      2011 1 11.493 July 2010 to June 2011
      2013 3 11.266 Oct 2012 to Sept 2013 (–0.227 from July 2010 to June 2011)

      Data source

      This is exactly opposite to the impression given by the 7SS. We can see the the highest ranking temperatures in the BEST-37SS month-by-month moving annual mean series compared to 7SS:

      NIWA-7SS nominal annual means
      1998 13.41 (#1)
      2005 13.11 (#3)
      2010 13.07 (#4)
      2013 13.40 (#2)

      BEST-37SS moving annual means ranked above 11.5 (all 1998/99 El Nino)
      1998 4 11.527 (#11) vvvvv El Nino
      1998 5 11.556 (#10)
      1998 6 11.634 (#9)
      1998 7 11.720 (#3) <<<<< Highest ranking
      1998 9 11.680 (#5=)
      1998 10 11.670 (#6)
      1998 11 11.747 (#1) <<<<< Highest ranking
      1998 12 11.738 (#2) <<<<< Highest ranking
      1999 1 11.650 (#9)
      1999 2 11.666 (#7)
      1999 3 11.655 (#8)
      1999 4 11.680 (#5=)
      1999 5 11.703 (#4) ^^^^^El Nino

      1999 7 11.448
      1999 8 11.364

      2001 7 11.289

      2001 12 11.314
      2002 1 11.415
      2002 2 11.396
      2002 3 11.379

      2005 6 11.376
      2005 7 11.371
      2005 8 11.28
      2005 9 11.158
      2005 10 11.308

      2007 10 11.286

      2010 6 11.337 vvvvv El Nino
      2010 7 11.394
      2010 8 11.375
      2010 9 11.360
      2010 10 11.271
      2010 11 11.391 <<<<< Next ranking below 1998/9 El Nino, 7SS annual ranks as #4)
      2010 12 11.48
      2011 1 11.493
      2011 2 11.381
      2011 3 11.296
      2011 4 11.339 ^^^^^El Nino

      2013 3 11.266 Doesn't rate, 7SS ranks as #2

      # # #

      I think we have to dispense with the NIWA-7SS nominal annual mean – it's meaningless. And NIWA are providing a misleading impression by publishing the nominal annual 7SS means each year.

      The BEST month-by-month moving annual average is the only way to go to get the picture in my opinion. That is unless a month-by-month moving annual mean can be compiled for the 7SS.

      de Freitas et al (2014) and the NZCSC 'Statistical Audit' used monthly data too.

  8. HemiMck on 24/11/2014 at 2:28 pm said:

    For those of us not keeping up with the detail, could we have an occasional summary or editors overview of the significance of the above.

    • Excellent idea! How about it, RC?

    • Richard C (NZ) on 24/11/2014 at 4:56 pm said:

      [HemiMck] >”For those of us not keeping up with the detail, could we have an occasional summary or editors overview of the significance of the above.”
      [Richard Treadgold] >”Excellent idea! How about it, RC?”

      Yes, needed. It’s taken me a bit of work to get to this.

      In terms of RT’s original post, and Renowden’s assertions, we get to NIWA’s ‘Climate change scenarios for New Zealand’ which is my last thread header just upthread here:


      That deals with how NIWA’s down-scaled Climate Model temperature projections are tracking warmer than observations at end of 2013 and the following comment deals with how their model projection divergence is different depending on whether observations are by 7SS or by BEST 37SS. The 37SS exhibits a much flatter current decadal trend than the 7SS which makes NIWA;s projections worse by that measure. Given BEST’s 37SS is similar methodology to NIWA’s VCS then those are the major observational benchmarks that must be considered i.e. the 7SS is a minor benchmark because of the much lower diversity of observations. We don’t have access to VCS, unless we pay for it, but BEST’s 37SS is a good proxy going by 1980-1999 comparison upthread.

      Obviously if it is only observations are to be modeled then a Regional Climate Model (RCM) is redundant, NIWA’s VCS or BEST’s 37SS do that job. But it’s the RCM that is used for projections past the observations.

      The RCMs are initialized and constrained (trained) by observations, despite Gareth Renowden’s raving, e.g. RCM1 by ERA-40 re-analysis in the Ackerley et al (2011) paper. Neither the 7SS nor NIWA’s VCS are used for that purpose. But Ackerley et al do validate the RCM against the VCS (not 7SS) over the observation period 1980-1999. The correlation is very good – as it should be. The model doesn’t exhibit enough diurnal range (Max to Min not wide enough) but the mean profile is all but identical.

      I doubt the RCM is initialized with observations earlier than 1972 (when the VCS begins). I think the 1972 date is significant. There is no contention over the quality of observations after that date (NIWA 7SS and NZCSC 7SS agree after 1972 for example) and for some other reason NIWA has opted for 1972 to begin their VCS. I think this is because of the quality issue and it is reasonable to assume they wouldn’t go back prior to that date for their RCM either. But they may go back to Sept 1957 because that is when ERA-40 begins:


      Prior to 1972 is when the observation quality goes downhill and the trends become contentious. So the question arises: was there any ERA-40 initialization in the RCM prior to 1972? There may have been, I don’t know, I’m doubtful there was, but if so it would only be from Sept 1957.

      RT’s contention, I infer, is that the RCM may have been initialized by data exhibiting an excessive trend, possibly the 7SS. This is not the case because the 7SS is not used for the model initialization, ERA-40 is. The early ERA-40 trend may be doubtful if from 1957 but that isn’t too much of an issue because the model is constrained by recent observations after that anyway.

      I’ve commented upthread on the UKMO parallel of global “decadal farecasts” (actually running 5-yr) and how those forecasts are not tracking observations. I think in the NZ context the focus should be on NIWA’s 1990 to 2040 forecast and how that is tracking over time (or not). This takes us back to the top of this comment.

      I hope this helps.

      P.S. Note too that both the 7SS and 37SS have 2 periods of cooling from 1980 – 2013 broken at 1997 by an abrupt +0.45 C shift brought about by the 1998/99 El Nino.

    • Richard C (NZ) on 24/11/2014 at 11:24 pm said:

      Excel Workbook NIWA-VCS vs NIWA-7SS vs BEST-NZT.xls with graphs in Dropbox here:


      Caveat emptor.

    • Richard C (NZ) on 25/11/2014 at 3:33 pm said:

      >”Prior to 1972 is when the observation quality goes downhill and the trends become contentious. So the question arises: was there any ERA-40 initialization in the RCM prior to 1972? There may have been, I don’t know, I’m doubtful there was, but if so it would only be from Sept 1957.”

      Oooo, nasty:

      ‘Can Climate Trends be Calculated from Re-Analysis Data?’ [ERA-40]

      Lennart Bengtsson, Stefan Hagemann and Kevin I. Hodges, MPI Report, January 2004


      3. Temperature Trends [page 3 pdf]

      For the 1979-2001 period the MSU and ERA40 TLT (temperature of the lower
      troposphere) are shown in Figure 1. The equivalent TLT has been calculated from the
      ERA40 data following the method used by Stendel and Bengtsson (1997). The ERA40
      data show a slightly stronger warming trend in the lower troposphere than the MSU data.
      The calculated anomaly trend for the TLT in ERA40 is +0.11 K per decade versus +0.06
      K per decade forMSU. The larger trend in ERA40 is still within the estimated error range
      as given by Christy et al. (2003) but likely overestimated. We will be discuss this further
      in section 6.

      For the longer period 1958-2001 results are shown in Figure 2 which yield that in
      ERA40 the TLT warming trend (0.14 K per decade) is higher than for the period 1979-
      2001. We have excluded data for the period 1972-1978 as we believe the influence of
      VTPR data in ERA40 in these years have been detrimental for the trend calculations of
      TLT and IWV. While the ERA40 TLT trend is just within the error bars of the observed
      MSU temperatures for the time period 1979-2001 the question is whether the even larger
      trend for the longer period 1958-2001 is reliable?


      Could the large TLT trend be an artefact of the changes in the global observing system?
      These were substantial in the late 1970s with the change from a terrestrial based system
      to a system including satellite observations with global coverage. To investigate such a
      bias recent data-assimilation experiments with ERA40 (Bengtsson et al., 2004) were
      considered. These experiments have been designed to explore the characteristic
      differences in the ERA40 data for the 1979-2001 period when selected observing systems
      are omitted.

      In the NOSAT experiment (all satellite data removed) TLT was 0.16 degree K lower
      than for the ERA40 control for the three experimental periods suggesting that the ERA40
      data using the observing system prior to 1972 may have had a systematic cold bias in the
      troposphere. Assuming that such a bias would be representative for the whole period
      1958-1972, the effect would be to reduce the warming trend to a value similar to the
      period after 1979 and also to become more in agreement with the overall SST trend. A
      more detailed estimate would require to repeat the data-assimilation for the whole period
      without the satellite and other novel observing systems. Such a calculation is presently
      not feasible to undertake, but is strongly recommended for the future. Nevertheless, there
      are indications that the ERA40 tropospheric warming trend for 1958-2001 is too large. A
      more credible warming trend is about 40% less (see Figure 2).

      Page 16 pdf:

      Figure 2. TLT calculated from ERA40 for the period 1958-2001. The dashed line shows
      the corresponding warming trend. The full line indicates a corrected warming trend
      obtained by adding a factor to the data for the period 1958-1972 obtained from the
      difference between ERA40 and the NO SAT experiment, and by excluding data for the
      years 1972-1978.

      # # #

      >”ERA40 data using the observing system prior to 1972 may have had a systematic cold bias in the

      So 1972 is a very important date. NIWA begin their VCS at 1972 but what about their simulations? OK if they do but not if those begin 1958 with ERA-40.

      >A more credible warming trend is about 40% less (see Figure 2).”

      This is VERY wrong. it is 60% less in Figure 2:

      1958 – 2001 (4.4 decades)
      ERA-40 actual trend: 0.145 C/decade or 1.45 C/century
      Corrected trend: 0.05 C/decade or 0.5 C/century

      Basically, Richard Treadgold is half right but with ERA-40, not 7SS, and only if the initialization is prior to 1972 with ERA-40 i.e. 1957/8. NIWA’s RCMs, when initialized by ERA-40 (e.g. RCM1 Ackerley et al, 2011), are being forced by observations that have a trend 150% too high by the above assessment if ERA-40 is used from 1958.

      And the corrected ERA-40 trend is half NIWA’s 7SS trend over the same period:

      1958 – 2001 (4.4 decades)
      ERA-40 actual trend: 0.145 C/decade
      NIWA-7SS trend: 0.108 C/decade
      ERA-40 corrected trend: 0.05 C/decade.

    • Richard C (NZ) on 25/11/2014 at 3:43 pm said:

      >”This is VERY wrong. it is 60% less in Figure 2″

      Actually 65.5% less i.e. the corrected trend is only 34.5% of the actual ERA-40 trend.

    • Richard C (NZ) on 25/11/2014 at 4:40 pm said:


      >”NIWA’s RCMs, when initialized by ERA-40 (e.g. RCM1 Ackerley et al, 2011), are being forced by observations that have a trend [????] % too high by the above assessment if ERA-40 is used from 1958.

      In terms of ERA-40:

      1.45/0.5 = 2.9 x 100 = 290%

      Just as Richard Treadgold asserts but not in terms of NIWA’s 7SS, which is also warming too much in terms of corrected ERA-40 so in those terms:

      1.08/0.5 = 2.16 x 100 = 216%

  9. Richard C (NZ) on 25/11/2014 at 11:57 am said:

    A fun thing to do is to plot BEST’s moving centred annual mean 2004-2013 against NIWA’s forecast by adding 0.166 to the observations.

    I’ve graphed that here:

    • Richard C (NZ) on 25/11/2014 at 12:49 pm said:

      Once the end of 2013 data and 2014 data comes in, the projection-observation discrepancy will not be as pronounced because the cool 2004 year drops out. And the recent temperatures have been relatively high.

      NIWA’s forecast may even match observations for a year or so but the fluctuation will swing back cool again eventually as it has all decade. That will leave NIWA high again. And if the regime can shift up +0.45 C abruptly as it did 1998/9 then it can shift down too. Temperatures are still trending down from the 1998/9 El Nino even with the recent warmth and 2010/11 El Nino.

  10. HemiMck on 25/11/2014 at 9:12 pm said:

    Thanks RC RT for the summary up thread. I can at least follow the tread of the conversation now.

Leave a Reply

Your email address will not be published. Required fields are marked *

Post Navigation