cc: john.kennedy@metoffice.gov.uk, Philip Brohan date: Wed Mar 2 16:42:00 2005 from: Phil Jones subject: Re: Query to: Ian Harris , Peter Thorne Dear All, I'll elaborate on Harry's email a little and then make a suggestion. The deleted stations were mainly duplicate ones, or ones with very little data (but had normals from the WMO source). The changed stations may have had from 1 to N changes. You don't yet have these files (Philip may have) and need the appropriate normals and SD files. So, I hope you've not mixed any files sent by Harry - as you'll need a full new set. I don't think this is the reason, by the way. Far more likely is that the month you're now doing is January - a new year. This always gave me problems as the Anders file has to be expanded to have one more year for each station. I also needed to expand some arrays to read in one more year. As Harry says we've not altered anything in NZ. The fact that the error manifest itself in NZ may be irrelevant and it is just masking where it really occurred. Hope you find it ! Peter will be back with you on Friday ! Cheers Phil At 13:34 02/03/2005, Ian Harris wrote: On 2 Mar 2005, at 13:15, Peter Thorne wrote: Phil, Philip, Harry, John, do you know of any changes made to a NZ station(s) in the last month? I'll try to investigate further upon my return on Friday, but it looks like this can't be dismissed as a simple bug in the system. I doubt a change at UEA would have led to this problem unless it had been deliberately propogated to the Met Office archive which is conceivable. I have no info as to the relative magnitude of the change. No changes to NZ data from my end, Peter. Here's a map showing the stations I've 'affected' ;-) [1]http://www.cru.uea.ac.uk/~harry/for_philip_b/changedstationstorev10.pdf Keep me posted! Harry Thanks Peter Note: forwarded message attached. From: "Hardwick, Jen" Date: 2 March 2005 08:49:34 GMT To: Peter Thorne Subject: HadCRUTtttttttttttttttttewwwe ew HadCRUT fails at test 3# from the test_landsstmerge program, after the negrhadcrut2 script. "Test 3 failed, values before the current month are too dissimilar. Check random years Either SST or land in error for hadcrut2 The proportion of values changed before the current month is: 0.0063" The criteria are 0.001 for a normal run through and 0.005 if Phil updates the anders file. Looking at the data there are a number of differences for the previous month and subsequent months. This is interesting because the land values have not changed and the SST values have not changes. Looking more closely at the 'dodgy' file you can see that the change in values, for almost all years back to 1856, are localised to the lat, long grid box areas (27, 66+). This is around New Zealand. I'm baffled. Not an unusual turn of events there! Don't worry too hard about it, Happy day Jen -- Jen Hardwick, Climate Information Scientist Met Office Hadley Centre for Climate Prediction and Research Tel: +44 (0)1392 884288 Fax: +44 (0)1392 885681 Email: jen.hardwick@metoffice.gov.uk [2]http://www.metoffice.gov.uk Global climate data sets are available from [3]http://www.hadobs.org Ian "Harry" Harris Climatic Research Unit University of East Anglia Norwich NR4 7TJ United Kingdom Prof. Phil Jones Climatic Research Unit Telephone +44 (0) 1603 592090 School of Environmental Sciences Fax +44 (0) 1603 507784 University of East Anglia Norwich Email p.jones@uea.ac.uk NR4 7TJ UK ----------------------------------------------------------------------------