date: Fri, 20 Feb 2009 09:51:40 +0000 from: Adrian Simmons subject: Re: Temperature trends and SST analyses to: Phil Jones Thanks, Phil. I think it is worth saying something about absolute values. I'm just looking at them for precip, for which the observation-based data do not come as anomalies. Adrian Phil Jones wrote: > > Adrian, > The absolute values for ERA-Interim are interesting. Over the > 1989-2007 period > the range is 0.55 from the coldest in 1992 (13.74) to 2005 (14.29). > These compare > with HadCRUT3v from 1992 (+0.07) to 1998 (+0.53), so a difference of > 0.46. The value > is slightly smaller if 2005 is used instead of 1998. > Using the 14K value for HadCRUT3v for 1961-90 indicates that the > absolute > values differ by about 0.3K (with HadCRUT3v warmer), a value that is > consistent with > your cold bias relative to SYNOPS of about 0.5K. I agree that the SST > differences are > an issue, but I think we are quite close in absolute terms. > This might be worth reporting in a small section. A map of the > absolute differences > may show some larger cancelling errors across the world. > > The attached figure and bit of text comes from this paper - which is > too large to email > in full. > > Jones, P.D., New, M., Parker, D.E., Martin, S. and Rigor, I.G., 1999: > Surface air temperature and its variations over the last 150 years. > Reviews of Geophysics 37, 173-199. > > The range of absolute global averages is from 12 to 16K, almost the > same as the second > of the two plots you sent yesterday. > > [Aside] The absolute temperature over the ocean in this paper is one > for air temperature. > Even though we use SST anomalies as a surrogate for air T over the > oceans, the > absolute was derived through the addition of an air minus SST climatology. > > We could possibly use ERA-Interim's average absolute fields to reduce > my presumed > uncertainty in the 14K number, by perhaps doing better over the data > sparse regions > such as the sea-ice areas and parts of the Antarctic. > > It isn't just the media that are interested in the absolute figure, a > lot of users would > like better absolute figures for driving impacts models where absolute > thresholds can > be important. > > Cheers > Phil > > > > At 17:12 19/02/2009, Adrian Simmons wrote: >> Phil >> >> All I did for the ten warmest years was to make sure that the set from >> the Hadley Centre/BBC website had the same average as the ERA-Interim >> set. ERA-Interim then shows a larger range. I did think this was >> predominantly because ERA-Interim shows stronger warming in the average >> over land (due to high-latitude coverage problems in CRUTEM3v), but with >> the SST differences discovered this morning perhaps things are more >> complicated. Maybe the SST difference is enough to allow 1995 into the >> list for ERA-Interim. Not one for the journalists, I guess, though I do >> wonder how/whether one should convey uncertainty in this sort of thing. >> >> Absolute values for ERA-Interim, for the record, are: >> >> 1989 13.78483200 >> 1990 13.99974918 >> 1991 13.94549942 >> 1992 13.73666668 >> 1993 13.76133442 >> 1994 13.83608341 >> 1995 14.01591682 >> 1996 13.87950039 >> 1997 14.00908375 >> 1998 14.20900059 >> 1999 13.91508484 >> 2000 13.94616604 >> 2001 14.12133312 >> 2002 14.16433334 >> 2003 14.16483307 >> 2004 14.08841705 >> 2005 14.28650093 >> 2006 14.21850014 >> 2007 14.16433430 >> >> See also the first panel on the attached plot. But one should treat >> these numbers with caution. Not only because of all the values after the >> decimal point, (needed to distinguish 2002 from 2007), but because over >> land we have a bias of about 0.5K cold compared with the SYNOPS. >> >> Adrian >> >> >> Phil Jones wrote: >> > >> > Adrian, >> > I think you're definitely imagining the ship tracks! These are >> > decadal averages. They potentially do show the drifter/ship offsets >> > I was talking about, though - about 0.1 to 0.2 deg C. John >> > Kennedy may have a difference for ships minus drifters. >> > >> > In the warmest year pdf, I assume you're adding 14 to the >> ERA-Interim >> > anomalies. If not these are amazingly close! >> > >> > It would be useful to see what ERA-Interim does get for the global >> > average >> > surface absolute T for your 1989-99 base period. The work that >> > estimated the 14 >> > number (it was 14.02 for the globe) made lots of assumptions over the >> > sea ice >> > areas and the Antarctic. I always thought this was accurate to about >> > +/- 0.5 deg C. >> > It was impossible to explain this accuracy of the absolute vs the much >> > smaller error bars >> > on the anomalies to any journalists. >> > >> > Cheers >> > Phil >> > >> > >> > Cheers >> > Phil >> > >> > >> > At 16:09 19/02/2009, Adrian Simmons wrote: >> >> Hi everyone >> >> >> >> Here are the differences in map form. Shown in the first attachment >> are >> >> ten-year means relative to 1989-1999 (except the last month of 2008 is >> >> missing, as ERA-Interim is still working on it). Please discount what >> >> happens right near the coast and sea-ice regions, as I may not have >> >> things quite right there. >> >> >> >> What is clear is that relative to 1989-1999, HadCRUT3v (or HadSST2) is >> >> almost universally a little warmer than the SSTs we use in >> ERA-Interim. >> >> I think I can see ship tracks also, but maybe I'm over-interpreting. >> >> >> >> See also the second attachment. We should know by Saturday whether >> 2008 >> >> makes it to the ERA-Interim list - I doubt it. ERA-Interim is >> currently >> >> analysing 18 December. >> >> >> >> Adrian >> >> >> >> >> >> Phil Jones wrote: >> >> > >> >> > Adrian, >> >> > I'm going down to the Hadley Centre next week on an unrelated >> >> matter, >> >> > and have arranged to talk to Kate and Peter about the HadCRUH >> >> analyses, >> >> > so we can add this one in as well. It would be worth showing these >> >> > plots to >> >> > John Kennedy and Nick Rayner to get their thoughts. >> >> > >> >> > HadCRUT3v uses HadSST2 anomalies over the ocean. HadSST2 >> should be >> >> > similar to HadISST1 where there are data, but you require a >> complete >> >> > field so >> >> > the latter is infilled - and there is also the sea ice component. >> >> > >> >> > The problem with the various SST forcing fields used by ERA and >> >> NCEP >> >> > is that >> >> > it is made up of different components and these may have slightly >> >> > different >> >> > absolute fields. There is probably a document explaining how the >> SST >> >> > fields >> >> > for Reanalyses was produced - maybe Nick will know about this. >> >> Overall the >> >> > differences will be small for most of the world's oceans, but are >> >> > likely to be >> >> > larger in data sparse regions and near the sea ice edges. So, if >> >> you redid >> >> > your plot for NH oceans, I'd expect it to be closer than a >> similar one >> >> > for the SH. >> >> > >> >> > As for the recent 8 years or so, there is an issue of the >> dramatic >> >> > increase in >> >> > drifter based SST values, which may be nearer the true value, but >> >> which >> >> > seem >> >> > about 0.1 to 0.2 deg C cooler than the ships. The HC are working >> on a >> >> > revised >> >> > SST dataset (HadSST3 presumably) which I'd like to think would >> >> solve these >> >> > problems. The fact that precip looks better in more recent years >> >> > suggests that >> >> > the drifter-based SST may be nearer the truth. >> >> > >> >> > The important message here is that ERA-Interim must be >> getting very >> >> > good to >> >> > be able to respond to barely detectable differences in SST >> across the >> >> > world's >> >> > oceans. >> >> > >> >> > Cheers >> >> > Phil >> >> > >> >> > >> >> > At 09:35 19/02/2009, Adrian Simmons wrote: >> >> >> Dear co-authors (also Saki, Per and Tim Stockdale) >> >> >> >> >> >> More interesting results. This time Kate can relax, as I'm back >> on to >> >> >> temperature trends. >> >> >> >> >> >> I wanted to produce a plot contrasting the recent temperature >> changes >> >> >> over land and sea. So I looked at the full HadCRUT3v dataset, >> which is >> >> >> based on Hadley Centre SST over the oceans, and compared the >> >> ocean-wide >> >> >> averages with the corresponding averages of ERA background 2m >> >> >> temperatures. I chose the background fields not the analyses, as >> the >> >> >> later took in ship air temperatures and cannot be relied upon, as >> >> >> discussed in Simmons et al.(2004), based on a comment from Simon >> Tett. >> >> >> >> >> >> The upper panel in the attachment shows the result - >> temperatures are >> >> >> adjusted to have the same average for the period 1989-1998 as >> before. >> >> >> The most obvious feature is a shift between HadCRUT3v and ERA >> >> around the >> >> >> turn of the century. The lower panel shows the corresponding >> >> comparison >> >> >> between HadCRUT3v and the SST used in ERA (the same for >> ERA-Interim as >> >> >> ERA-40). Clearly there is a shift in recent years between the >> SST used >> >> >> in ERA and that used in HadCRUT3v. >> >> >> >> >> >> Now, ERA used the following SSTs: >> >> >> >> >> >> (i) HadISST1 monthly up until November 1981 >> >> >> >> >> >> (ii) Then Reynolds 2D-Var weekly up until June 2001 >> >> >> >> >> >> (iii) Then NCEP operations until September 2008 >> >> >> >> >> >> (iv) Then OSTIA >> >> >> >> >> >> The lower panel clearly indicates a shift of between 0.1 and 0.2C >> >> in SST >> >> >> in 2001. So apparently there is a shift in going from the >> 2D-Var to >> >> >> NCEP operations, assuming that HadCRUT3v uses a homogeneous SST >> >> product. >> >> >> >> >> >> Then there is a question of whether it is the 2D-Var or the NCEP >> >> >> operations that is closer to HadCRUT3v. Remember the anomaly curves >> >> are >> >> >> normalised to 1989-1999 (just because that's what's convenient for >> >> other >> >> >> plots in the paper), but it could be that the absolute values are >> >> closer >> >> >> for the end of the period. >> >> >> >> >> >> Finally, I wonder whether this is related to the shift in >> >> precipitation >> >> >> over land in ERA around the turn of the century relative to >> GPCC. It's >> >> >> in the right sense (ERA shifting to relatively cooler SSTs, less >> >> >> evaporation, and thus less precipitation over land). As Dick and I >> >> >> discussed yesterday after looking at one of Per's plots for the >> >> >> comparison of ERA with GPCP and CMAP, when looking at the absolute >> >> >> values the precip is in better agreement in recent years than in >> the >> >> >> 1980s and 1990s, so perhaps the more recent SSTs are better also. >> >> >> >> >> >> I'd appreciate some input on this from the SST experts, at ECMWF >> >> and the >> >> >> Hadley Centre. Clearly we need to resolve this before the next >> >> >> reanalysis, and I hope ERACLIM will get funded to enable us to >> do this >> >> >> thoroughly. >> >> >> >> >> >> Adrian >> >> >> >> >> >> PS Yesterday's interesting result was that the temperature >> trends over >> >> >> land from the ERA-Interim background fields are significantly >> >> closer to >> >> >> CRUTEM3v than are the trends from the ERA-40 background. The >> analyses >> >> >> themselves don't differ much, as there is plenty of 2m T data >> >> >> assimilated by the OI analysis in grid boxes for which there are >> >> >> CRUTEM3v values, and the OI analysis for 2m T is the same in ERA-40 >> >> and >> >> >> ERA-Interim. But the result does suggest that more confidence >> can be >> >> >> placed in the ERA-Interim analysis than the ERA-40 analysis in data >> >> >> sparse regions, where the background comes more into play. >> >> >> >> >> >> -- >> >> >> -------------------------------------------------- >> >> >> Adrian Simmons >> >> >> European Centre for Medium-Range Weather Forecasts >> >> >> Shinfield Park, Reading, RG2 9AX, UK >> >> >> Phone: +44 118 949 9700 >> >> >> Fax: +44 118 986 9450 >> >> >> -------------------------------------------------- >> >> >> >> >> > >> >> > 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 >> >> > >> >> >> ---------------------------------------------------------------------------- >> >> >> >> >> > >> >> >> >> -- >> >> -------------------------------------------------- >> >> Adrian Simmons >> >> European Centre for Medium-Range Weather Forecasts >> >> Shinfield Park, Reading, RG2 9AX, UK >> >> Phone: +44 118 949 9700 >> >> Fax: +44 118 986 9450 >> >> -------------------------------------------------- >> >> >> >> >> > >> > 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 >> > >> ---------------------------------------------------------------------------- >> >> > >> >> -- >> -------------------------------------------------- >> Adrian Simmons >> European Centre for Medium-Range Weather Forecasts >> Shinfield Park, Reading, RG2 9AX, UK >> Phone: +44 118 949 9700 >> Fax: +44 118 986 9450 >> -------------------------------------------------- >> > > 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 > ---------------------------------------------------------------------------- > -- -------------------------------------------------- Adrian Simmons European Centre for Medium-Range Weather Forecasts Shinfield Park, Reading, RG2 9AX, UK Phone: +44 118 949 9700 Fax: +44 118 986 9450 --------------------------------------------------