cc: Brian Soden , Panmao Zhai , Roxana Bojariu , David Parker , Jim Renwick , Albert Klein Tank , Fatemeh Rahimzadeh , Matilde Rusticucci , Peter Ambenje , Byron Gleason , Jay Lawrimore , David Levinson , Richard W Reynolds , Tom Smith date: Tue Dec 21 16:38:03 2004 from: Phil Jones subject: Re: Map questions to: Kevin Trenberth , David Easterling David, Kevin, Thanks for the EOS pdf Dick. The colour schemes that look good in its Fig 2 are D for temperature and C for precip (if the blue gets changed to a green). The blue would be better for colour blind people. I too don't see why the schemes should be symmetrical. The temperature ones certainly will not as we're choosing the periods to show warming. We seem to be agreeing on having two other colours (one for nearish to zero anomalies and for missing). If we go with your globally complete dataset, then this only means missing is over the oceans for precip - except for the confidence in the trends in data sparse regions. So I'd go for white (no colour) for missing as this will mean the oceans are blank for precip. Nearish to zero trends then have to be grey. Determining the contours is then the only issue - I prefer nice round numbers. Nearish to zero has then to be something like +/- 0.05 deg C/decade or +/- 0.2 deg C/century. Hopefully this are will encompass about 10-20% of the map. I would like to see all seasons and the year with the same scale for 1979-2003 for temperature and similarly for precip. This means that trends in the different seasons can be intercompared. All just thoughts after a long tiring day ! Cheers Phil With this if trends were per century for 1901-2003 then the scales might be quite similar. At 15:50 21/12/2004, Kevin Trenberth wrote: Hi David My own preference for color schemes is to have something that in Fig 3.4.2.2 where the near zero anomaly is white. This means some other color (speckled gray perhaps) has to be used for missing data. The idea behind this is that the values near zero are not significant. This color scheme was one recommneded in Eos a short time ago: it is good for people who are color blind. With color one can also let the colors saturate at some point, so all values beyond a certain value are one color. This can cut down on some noise at individual points. Often it is convenient to straddle the zero with the contours and thus not plot the zero line, which can meander all over the place. Other viewpoints welcomed. I don't see why the color schemes should be symmetrical, especially for a skewed field. It is desirable to have some color coding such as warm colors for positive and cold colors for negative. For precip it is often greens vs browns for wet vs dry. David Easterling wrote: All, We just had a meeting here at NCDC for the map figures that raised some questions/issues we must resolve. These are: 1. for the maps do we want a neutral color bounding + or - some value around zero to show no real trend as Kevin has suggested or have the color scheme start with the first positive or negative value (e.g. for temperature have anything positive start with the red scheme or anything negative start with the blue scheme)? 2. for the scales, do they need to be symmetrical on either side of zero or can the range from the lightest to darkest color for positive be different than for negative? For example the range for precipitation go from 0 to 100% for positive (green) but go from 0 to -50% for negative? 3. for the Smith-Reynolds temperature data set problem in the polar and data sparse regions, we can use the confidence value supplied with the data and set a confidence threshold for whether we use a particularly grid point trend value or make that grid point missing. Tom Smith is looking at what the value should be and it will take a few weeks to work this out properly. For color one can also use semitransparent and lightly mask less confident areas rather than wiping them out. 4. For the 1901-2003 maps should the trends be in deg or % per century or per decade? The 1979-2003 maps will be "per decade". Depends on quantity. For some fields one needs to know both and the area averages are not transferable from one to the other with a simple constant. That is the case for water vapor in Fig 3.4.2.2. We have both mm/decade and %/decade. The former emphasizes the tropics. It is most impt for the actual hydrological cycle. But the latter is more impt for RH and allows the plot to be global with a linear scale. I suspect the same is true for precip. 5. The color schemes will be modified to make the darker colors more bold (e.g. brighter red). Check out the article in Eos a few weeks ago. Byron Gleason is making up some examples of maps addressing these questions for review. In particular the symmetrical vs non-symmetrical scale, and the neutral color to indicate no trend. Jay Lawrimore did the maps for the TAR and may have more information or questions. Dave Phil Jones wrote: Kevin, These use the spatially infilled dataset of Smith and Reynolds (2005). This has problems in earlier years. I've yet to see the paper, but it is due out in J. Climate quite soon. The problem in the earlier years is commented upon in section 3.2. We can work on these plots a little more. The break could be 120E - this keeps the Pacific and Atlantic together. My hope was for the 1901-2003 plots to be in change per century with those for 1979-2003 in change per decade (change being deg C or % for precip). In change per decade for 1901-2003 the overall change is 10.3 times the change shown. I did comment about the precip trends looking odd. They do look a lot better than the first draft ! There is another plot from Dave in the other file - page before Fig 3.4.1.1. This shows how hairy these trends are. The temperature ones use the same colour scheme for all seasons for the same period. I quite the temperature ones , but the precip scale needs to be done like the temperature - clear breaks between each colour, rather than the gradual change at present. Dave - is there a chance for another iteration before Jan 14 - or earlier ? I know you don't have so much time off as we do here ! I'll be off from Dec24-Jan3 inclusive. I will be checking email from home. We haven't just had Thanksgiving though ! Cheers Phil At 15:45 16/12/2004, Kevin Trenberth wrote: Phil, Dave This is the first time I have seen these figures. Firstly let me say I like the projection, and it works great with land data. I personally prefer to have the dateline in the middle as this highlights the ENSO stuff. However, I am not enarmoured with the color scheme and some units must be wrong??? Are there no missing data in the first figure on T trends from 1901-2003? There must be, over Antarctica, southern oceans etc???? Then for the precip trends the units are blocks of 20%/decade??? So for 1901 to 2003 it means over 10 decades and the values for the lowest contour is 200%??? Even for the last 30 years the lowest contour range is 0 to 60%!!! I do note that mostly only one color makes it on either side of zero, but these units make no sense whatsoever. My own preference on all these kinds of plots is to have some nodescript color straddling zero where trends are not significant. But these need to be distinguished from no data. Why is it labeled "IPCC trend"? This is my quick feedback Kevin Phil Jones wrote: Same email with NCDC figures Dear All, Later today I'll send the current version of the text and the references. I'm sending the Figures now. There are 2 sets of figures. The first is all those currently numbered in the draft text. Most have captions (which repeat in the text in Kevin's sections). I will add those captions from 3.2/3.3/3.8 in this fashion in early January. I'll send this again in zipped format. The second file is 20 global plots of temp/precip trends for 1901-2003 and 1979-2003 (4 seasons + year =5 *2 periods*2 variables =20). Not all these will be in the ZOD, but a selection will. This file doesn't compress at all well, so if you don't get it email Dave Easterling to see if he can put it on a server at NCDC. All were produced by Dave and Byron Gleason. Once NCDC (Russ Vose) have finished their next version of the max/min/DTR dataset, plots like this will be produced for that as well. The temperature plots are all spatially infilled (see Smith and Reynolds, 2005). This has its problems for the 1901-2003 period, but for 1979-2003 is OK (and for this it should use blended SST estimates for satellites). So 3 emails (including this one coming now). Sending now to allow a few hours for bounces to occur. The text and figs are much, much smaller. This email - numbered figs in doc format Next email - same file zipped Third email - NCDC plots in doc format. The email with the text will have more details about the timetable. Cheers Phil 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 ---------------------------------------------------------------------------- -- **************** Kevin E. Trenberth e-mail: trenbert@ucar.edu Climate Analysis Section, NCAR [1]www.cgd.ucar.edu/cas/ P. O. Box 3000, (303) 497 1318 Boulder, CO 80307 (303) 497 1333 (fax) Street address: 1850 Table Mesa Drive, Boulder, CO 80303 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 ---------------------------------------------------------------------------- -- **************** Kevin E. Trenberth e-mail: trenbert@ucar.edu Climate Analysis Section, NCAR [2]www.cgd.ucar.edu/cas/ P. O. Box 3000, (303) 497 1318 Boulder, CO 80307 (303) 497 1333 (fax) Street address: 1850 Table Mesa Drive, Boulder, CO 80303 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 ----------------------------------------------------------------------------