From SHAUN.HARRIGAN at nuim.ie Thu Nov 11 03:43:54 2010 From: SHAUN.HARRIGAN at nuim.ie (SHAUN.HARRIGAN at nuim.ie) Date: Wed, 10 Nov 2010 16:43:54 +0000 (GMT) Subject: [Trend] Alternating the start year within TREND Message-ID: Dear sir/madam,? I am a researcher in the Irish Climate Analysis and Research Units (ICARUS) at the National University of Ireland Maynooth. My current?research?is?focused?on the detection of climate driven changes within Irish streamflow records. I have found the?TREND?software to be an excellent statistical analysis tool for such applications. I would like to open for discussion and feedback the possibility of incorporating, into?TREND, the ability to alternate the start year of a flow input file either at the file input or options screens.? Why is this useful? One method I have used so far in my study investigates the control that the section of data employed has on a trend itself, following a similar approach by Wilby (2006) in the UK. Taking each catchment in turn, the Mann-Kendall statistic was derived first for the full record 1950-2007, then 1951-2007, and so on until 1980-2007 to give sample sizes ranging between 58 and 28 years. Zs values were plotted over the 30 year period. This reveals the extent to which trends persist throughout the record (see Wilby (2006) for a more detailed explanation of the method,?Wilby, R.L. (2006) When and where might climate change be detectable in UK river flows?, Geophysical Research Letters, 33, L19407, doi:10.1029/2006GL027552). Early results have shown that both the direction and strength of trends is dependent on the length of data employed. As far as I am aware (please correct me if I have missed something!), in order to alternate the start year of the flow files, you must for each catchment manually create 30 individual flow files (the first full period 1950-2007 and so on to 1980-2007). The 30 files must then be input into?TREND?one-by-one and the statistics recorded for each file. Is it already possible to do this without creating individual files? if not, do you think it is worthwhile/possible to incorporate this into the software?(as can be very time consuming when many catchments are used)? I hope I have not confused anyone when trying to get my point across. I do see significant advantages investigating how trends change when the period of data gets both shorter and longer rather than stating results when a set period is employed. I welcome any feedback and your opinions. Thank you for reading. Kind regards,? Shaun Harrigan, ? ICARUS, NUI Maynooth, Ireland.? -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.toolkit.net.au/pipermail/trend/attachments/20101111/27547ba4/attachment.htm