Homogeneity testing of climate data and homogenization for
Contents
1. 23401061 unsw UNSW Climate Change Research Centre 5 Change point detection The lt filename gt U pdf file shows the tested data series with identified breakpoints overlaid Base anomaly series and regression fit 1970 1980 1990 2000 2010 The lt filename gt _1Cs txt file lists all changepoints that are significant at the 5 level even without metadata support Type 1 changepoints P value confidence Test statistic Confidence Date range PFmax interval of PFmax deme lest nadi_tmax tt 31 5628 2 29 6976 32 5887 38 6015 35 9421 37 7651 38 0014 confidence Si gnificance level 6 Edit change point file for known changes add using Notepad any known changepoints to the lt filename gt _mCs txt file by considering station history metadata information type 0 changepoints changepoints that can only be significant if supported by metadata In this example metadata records for Nadi are relatively complete and it is known that the station underwent site moves in June 1965 May 1971 December 1985 and April 1998 gt edited lt filename gt _mCs txt file 7 changepoints in Series srv ccrc data05 z3401061 Pacific emacs RHTest nadi tmax txt O Yes 19650601 1 0000 1 0000 0 950 32 2282 28 0852 29 6976 Yes 19710501 1 0000 1 0000 0 950 54 9394 36 2768 38 6015 Yes 19851201 1 0000 1 0000 0 950 63 8630 33 8336 35 9421 Yes 19980401 Yes 20090106 1 0000 1 0000 02. make a homogeneity test detect an inhomogeneity that does not really exist UNSW Climate Whanga ___ Research Centre Quickguide source RHtestsV3 r StartGUI TransformData FindU FindUD Stepsize repeat until all changepoints significant i AHtestsV3 RHtests V3 Transfonn Data Change Pars Quit PMF and F tests FindU FindUD Step Size PMT and t tests FindU wRef FindUD wRef Step Size wRef To adjust daily Gaussian data QMadj_GaussDLY Current Missing Value Code 99 9 Current nominal level of confidence p lev 0 95 segment to which to adjust the senes ladj 10000 current Mo OT points Tor evaluating PDF 10 Current Hy4a max of years of data for estimating PDF 0 Current input Base seres filename HA Current input Reference series filename HA Current data directory HA Current output directory Now hands on 1 Run Quality Control in RClimDex 2 Test homogeneity of data using RHTest I Keep a copy of your data in a safe place Recommendation create a work directory copy your data files there and only work on this copy UNSW Climate Change ___ Research Centre
3. the data have been measured UNSW F Climate Change Research Centre Step by step instructions to use RHtest 0 Set up Ron your computer using the R set up file available at http www r project org and create a new working folder Copy RHtestV3 r and station data to be tested for homogeneity to your new working folder 1 Start R by clicking the icon on your desktop and change the working directory to your new folder Click File Change dir Then start RHtest by s RHtestV3 r Click File Source R code putestsV3 EE RHtests V3 Transform Data Change Pars Quit 2 To start the user interface PMF and F tests mu mao Stpsie PMT and t tests FindU w Ref FindUD wRef Step Size w Ref enter StartGUI at the R prompt To adjust daily Gaussian data oN cassi Sogment to which to adjust the sere a 1000 Current Mg of points for evaluating PDF 10 Current ci rs A oak s gt i for estimating PDF a Sree ape decry Ha 4 convert RClimdex data format to RHtest data format Any data that is in the RClimdex 6 column years months days Tmax Tmin Precip format will need to be converted to RHtest format before it can be tested for homogeneity To do this a From the RHTest GUI click Transform Data b Select the station that you want to transform click Open c This will produce 9 files tmaxMLy txt tminMLY txt prcoMLy txt tminDLy txt _tmaxDLy txt p
4. 38 3996 Q YifD 19851201 1 0000 1 0000 0 950 36 0256 38 4693 40 9792 No 19980401 0 8809 0 9066 0 950 3 9025 36 9877 39 3639 1 Yes 20090106 1 0000 1 0000 0 950 94 3808 32 8130 34 8125 1 Yes 20100112 1 0000 1 0000 0 950 257 7519 29 1818 30 8537 1 Yes 20101010 1 0000 1 0000 0 950 105 4842 28 5294 30 1567 Remove any changepoints found to be not statistically significant from lt filename gt _mCs txt Deleting any changepoints will affect the significance of the remaining changepoints So click the StepSize button again to reassess their significance and magnitude E UNSW Climate Change Research Centre 6 changepoints in Series srv ccrc data05 z3401061 Pacific emacs RHTest nadi_tmax txt Yes 19650601 1 0000 1 0000 0 950 45 1182 32 3767 34 3359 Yes 19710501 1 0000 1 0000 0 950 63 9958 36 1375 38 4354 Q YifD 19851201 1 0000 1 0000 0 950 22 0735 42 4176 45 2812 1 Yes 20090106 1 0000 1 0000 0 950 75 0505 37 3992 39 8107 1 Yes 20100112 1 0000 1 0000 0 950 260 0614 29 2067 30 8804 1 Yes 20101010 1 0000 1 0000 0 950 104 5662 28 5538 30 1826 10 Review significant change points repeat StepSize until all remaining change points are statistically significant Once you are satisfied with the final changepoints the following four final output files will be in the output directory i lt filename gt _fCs txt lists the changepoints identified and
5. 950 82 2340 Yes 20100112 1 0000 1 0000 0 950 261 0596 Yes 20101010 1 0000 1 0000 0 950 106 7205 35 7255 38 0014 28 7500 30 4086 28 1094 29 7238 Pre OD oN ZN ON Climate Change Be fee Research Cente 7 test for type 0 change points can be skipped if station history is complete or no station history available click FindUD button this does not need to be done in this example as the dates of all known site moves have already been added to the lt filename gt _mCs txt file UNSW F Climate Change eee Research Centrenmia 8 re asses the significance and magnitude of all change points Changing the number of changepoints will affect the significance of all changepoints So click the StepSize button to re assess the significance and magnitude of the new changepoints Click OK to proceed Ste pSize Do not choose daily precipitation data Input Data filename EelDesktopitest Folder Wadi_MinTtrans Input changepoints Filename oo Polderfoutpuk Madi_MinTkrans mcs txt Sai ie I 4 gt UNSW tao pW ae A ee I wis gt Climate Change __ Research Centre 9 Determine which change points are significant examine results in the file lt filename gt _fCs txt 7 changepoints in Series srv ccrc dataQ 5 z3401061 Pacific emacs RHTest nadi_tmax txt Yes 19650601 1 0000 1 0000 0 950 51 5468 32 3474 34 3050 Q YifD 19710501 1 0000 1 0000 0 950 18 2208 36 1032
6. Homogeneity testing of climate data and homogenization for RClimDex Markus Donat Climate Change Research Centre University of New South Wales Sydney Australia a f as E UNSW a F ie Research Centre F Climate Change ga eN Homogeneity Testing Homogenisation Background The aim of climate data homogenisation is to adjust observations if necessary so that all variations in the data series are caused by real changes in the climate and not due to changes in the way the data have been recorded Aim To test data series for homogeneity This procedure will involve testing daily monthly temperature data series without the use of a homogeneous reference series the most common scenario for stations in the North African Arab region Instructions for homogeneity testing with a reference series can be found in the RHtest User Manual UNSW Climate Change as Research Centre Homogeneity Testing Homogenisation Software Daily monthly temperature data is homogenised with RHtestV3 This is an R application written by Xiaolan Wang and Yang Feng Environment Canada It includes functions to detect potential discontinuities in a candidate or base climate series The software is available at http cccma seos uvic ca ETCCDMI software shtml Metadata If available completed station history sheets will assist in assessing whether any of the detected breakpoints could be due to changes in the way
7. rcpDLy txt _prcoMLYimm txt ae z LogprcpMLY txt and _LogprcpMLY1mm txt Zn n t h e j n p u t fi e n a m e Ourrent input aiea series filename DLY daily data MLY monthly data MILY1 daily precipitation 1mm d The following files will be used in subsequent sections tminMLy txt and _tmaxMLy txt _tminDLy txt and _tmaxDLy txt UNSW Climate Change RH tests V3 __ Research Centre 5 Change point detection click the FindU button to start testing the data series without a reference series Bele Do not choose daily precipitation data Input Data filename EejDeskkopitest Folder Nadi_MinTtrans Change The function will then detect all changepoints that are significant at the 5 level in your data even without metadata support Type 1 changepoints A series of output files describing the test results will be produced in an output folder in your working folder Name Size File Type Modified a Permissions Owner Group nadi_tmax_Ustat txt 2 9KB Plain Text Docu 06 03 12 10 48 am rw 23401061 UNSW nadi tmax_mCs txt 744 B Plain Text Docu 06 03 12 10 48 am rw 23401061 UNSW nadi_tmax_1Cs txt 7448 Plain Text Docu 06 03 12 10 48 am rw 273401061 UNSW Anadi tmax U pdf 1 9 MB PDF Document 06 03 12 10 48 am TW 73401061 unsw 2 inadi tmax Udat 1 2 MB Plain Text Docu 06 03 12 10 48 am rw
8. their significance ii lt filename gt _Fstat txt stores the estimated shift sizes and a copy of the content from the previous file iii lt filename gt _F dat stores the original base series in its 3 column the mean adjusted series in its 5t column and Quantile Matching QM adjusted series in its 9th column and iv lt filename gt _F pdf shows plots of the original base series with breakpoints overlaid and adjusted series UNSW Climate Change ___ Research Centre lt 7 10 Review significant change points Count Station WMO Variable CPs Final CPs Reasons Base trend Adjusted Y Name No tested yyyymm for CPs trend 19641112 19650909 49650601 Site move 19651025 19721225 19710501 Site move 19851201 Site move 0 00048 C 0 00024 C iji Nadi Max Temp 19880517 decade decade 19881016 20100112 20090106 20101010 20100112 20101010 Base anomaly series and regression fit Mean adjusted base series QM adjusted base series 1970 1980 1970 1980 1990 2000 2010 Discussion Do any of the calculated trend values differ substantially between raw and adjusted series for the stations you tested RHtestV3 can also be used for detecting discontinuities in base series that is compared to reference series What could be used as a reference series in your country What factors might prevent a homogeneity test from detecting an inhomogeneity What factors might
Download Pdf Manuals
Related Search