Notes on replication — station data
Richard | November 29, 2009Are we feeling warmer yet?
NZ Climate Science Coalition & Climate Conversation Group
29 Nov 2009
We’ve heard from a number of people wanting to replicate the graphs. However, we never expected such a high level of interest in our study so we were somewhat unprepared. We are now putting together a posting that will specify stations and describe our methods which we hope to post in the next few hours. In the meantime, this note outlines the difficulties. It doesn’t answer your needs, and for that we apologise, but we’re working on something more substantial right now.
We faced the same questions everyone is facing—what stations were used and how were they combined? To answer them, we looked at how the US National Oceanic and Atmospheric Administration (NOAA) combined the data into their own unadjusted Global Historical Climatology Network (GHCN) datasets for each station and we followed that. In the absence of any other guidance from NIWA that’s all we could do.
Other people will find equally valid ways to combine the data, and they will obviously show different results. The whole thing is a mess. You can produce whatever you want from the raw data, simply because the stations are all over the place. Station selection and combination affects everything.
In most cases we believe we have the right stations and the lock-step appearance of most of the graphs confirms this, but we do not claim our results are definitive. Because of the deficiencies in the records, we’re surprised that NIWA even attempted their graph—the data is all over the place, stations have been closed and others opened without overlap, etc.
The real issue is that adjustments were made by NIWA and not acknowledged publicly on their website (as several commentators have rightly pointed out) and also that they only used seven climate stations to get an average for the whole country, and as we’re seeing now (confirmed by NIWA in the case of Wellington) the alignment of station readings is done by what appear to be some fairly rough-and-ready methods. We expected something more complex—actually, we’re a bit surprised.
