date: Thu Feb 15 11:37:12 2001 from: Tim Osborn subject: RE: FW: heavy ppn days to: "Jenkins, Geoff" At 10:24 15/02/01 +0000, you wrote: >What is the definition of heavy rainfall days? It is a data adaptive threshold, being different (in absolute, i.e. mm, terms) for each station and for each month of the year. What we do is take a daily time series of precip from 1 station, extract all the January days from 1961-1995, discard all dry days, sort remaining days into ascending order, the group them into 10 categories so that each category contributes 10% of the total January rain for the 1961-95 period that fell at that station. The 10th category consists of however many of the heaviest raindays are necessary to contribute 10% of the total rain for that month (typically the wettest 0.5-1.5% of the wetdays are sufficient to contribute 10%). We call these "heavy" rainfall days. Although all categories contribute 10% over the 1961-95 period, they are free to show year-to-year variations and trends etc. - which is what they do. We then combine time series into seasonal averages and also a UK avergae. We do this method so that we can then identify precipitation that is unusual for that location or that time of year, which is something that a fixed (say 10mm) threshold might fail to do so well. >I would appreciate a copy of the fraction one as well, as this will separate >out the simple chage in rainfall amount. I've attached a file that shows the smoothed fraction version (with the longer 1931-1997 smoothed series also shown). >Mike may have told you that Corbyn >spotted an 11 yr cycle in the fig you sent me. Could the smoothing produce >any artifacts like this? The smoothing is unlikely to produce this - you can see the 11yr cycle in the unsmoothed data. The longer time series (1931-1997) doesn't show this apparent link with sunspot activity in the pre-1961 period, though far fewer people have seen the longer record as it appeared in IJC not in the media! >Whats the line at 24mm - is this the 61-90 avg? The 1961-95 avg, which is the period we used to define the categories. The attached fraction diagram shows a line near 0.1 (i.e. 10%), which is of course the 1961-95 avg contribution because that's how we defined the thresholds! >I gave a copy of the "quantiles"trend diagram to John Murlis (properly >attributed!) recently, and mentioned the diagram you attached. He would be v >interested to have this tto, as quantiles is harder to explain. Can you send >him a .ppt diagram? (or I can if you perefre - see attached) Could you send it to him please. >Of course 61-present is the period over which NAO has steadily increased, so >I was interested to hear from mike that the trend persists even when you go >back to the 30s. In fact, the longer analysis shows little change pre-1961 and then reproduces the upward trend over 1961-1997. So it helps indicate that the recent values are unusual in a 67-year context, rather than the more limited 35-year context of the figure you've already got. It doesn't rule out the NAO as the driving factor. But the pattern of change is fairly uniform over the UK in winter, which is *not* what you would expect from a response to the NAO (which is strongly related only to the upland regions that generate orographic rainfall in response to the NAO-related increase in frequency and strength of westerly winds). In fact, I had a an extended abstract published in AMS proceedings where we did a principal components analysis of the UK regional intensity category time series (i.e., the 10 categories as explained above) and found two separate modes - one was spatially uniform with little trend pre-1961 and trend towards enhanced intensity post-1961 and was uncorrelated with the NAO index, while the second showed the characteristic non-uniform NAO pattern over the UK and had a time series with a moderate correlation (about 0.5) with the NAO. So we concluded that the recent trend might have had a regional contribution from the NAO, but couldn't be entirely explained by the NAO. It's turned out to be a long reply! Tim