date: Wed, 6 Aug 2008 16:58:32 +0100 (BST) from: David Lister subject: Draft report (fwd) to: p.jones@uea.ac.uk ---------- Forwarded message ---------- Date: Fri, 25 Jul 2008 10:23:08 +0200 From: "Buishand, Adri (KNMI)" To: "David Lister, env (School of Environmental Sciences) " Subject: Draft report Dear David, I read the draft report on your work for WP5.4 last weekend. I also consulted two of my colleagues about a number of points this week. The meaning of the values for the decadal trends in Figs. 3 - 7 is not entirely clear to me. Do these values refer to the slope of a linear regression fit? Does the brown colour in e.g. the upper right panel of Fig. 5 indicate that there is an upward trend of 1 to 1.5 degrees per decade (or 4.5 to 6.7 degrees over the 45 year period) in parts of northern and eastern Europe during spring? I think it may be good to say something about the statistical significance of the trends. Further, some references could be given to other results on trends in extremes. I wonder what is already known and what is new. A striking point in Fig. 5 is the negative trend over Turkey. One of our Greek partners in ENSEMBLES mailed me once a Climate Research publication (Vol. 30, 161-174, 2006) in which negative trends in Tn_10 were found over parts of Greece. My colleague Albert Klein Tank compared your results with his trends in the number of days > Tx_90 and number of days < Tn_10 in the ECA data set. Looks good. The only difference is that the large positive trends in Tn_05 over Finland, the Baltic states, Russia and the Ukraine in spring (rose , red and brown colours in the upper right panel of Fig. 5) are not found in the number of days < Tn_10 (most station trends are not significant in thist area). These relatively large values do also not show up in your Fig. 6 for the trends in Tn in the CRUTS dataset. Figures 8 - 11 show considerable differences between the trends in the WP5.1 gridded observed dataset and those in the ERA40 driven output. I wonder how far these difference could be explained. I showed Fig. 8 to Erik van Meijgaard who is involved in the RACMO2 simulations. The trend differences are sometimes more than 1 degree per decade (in parts of northern Europe in spring, and in southern Norway and small parts of central Europe in autumn). Does this imply that there is no trend, or even a decreasing trend, in the RACMO2 simulation? Is there a bias in Tn_-05 that changes over time? This may occur in spring in northern Scandinavia if the model is unable to reproduce snow melt. Another point is that RACMO2 has difficulties with low temperatures in stable boundary layers. This likely leads to a positive bias in Tn_05 (RACMO2 warmer than observed), especially during winter. But it is unclear how this should lead to a trend in Tn-_05. Furthermore, unlike the other regional climate models, RACMO2 has the same physics as ERA40 reanalysis and biases in the RACMO2 simulation may therefore also show up in the ERA40 reanalysis. Because the ERA40 reanalysis may suffer from the same shortcomings as the RACMO2 simulation, it might be interesting to see whether the trends in Tn-05 in the ERA40 reanalysis differ from those in the WP5.1 gridded observations. But please keep in mind the difference between the trends in Tn-05 and those in other Tn indices noted above for Finland, the Baltic states, Russia and the Ukraine in spring. Concluding, it might be useful to think a bit about the causes of trend differences. Maybe we can have some discussion on it at an ENSEMBLES meeting. Are you going to attend the Santander General Assembly? Kind regards, Adri