Climate Forecasting Unit CFU R common diagnostics CFUload

Climate Forecasting Unit CFU R common diagnostics CFU_load CFU_season CFU_trend CFU_clim CFU_anocrossvalid CFU_smoothing CFU_plotclim CFU_animvsltime CFU_plotano

Climate Forecasting Unit CFU R common diagnostics CFU_load Minimum set of arguments : 1) var 2) exp 3) obs (can be obs=NULL) 4) sdates You can request a subset of leadtimes You can work on a subdomain by providing lat/lon borders You can request area-averages, longitudinal or latitudinal averages, 2 d fields You can define any region by sending masks

Climate Forecasting Unit CFU R common diagnostics CFU_season This function computes averages over extended season. It can be used to compute annual means for exemple.

Climate Forecasting Unit CFU R common diagnostics This function computes per-pair climatologies, one climatology per member or one for all the members together. CFU_clim If you have only one start date, your climatology should be computed as a simple annual cycle not with CFU_clim. If you don’t have observations, you don’t need the per-pair method. Your clim is clim=CFU_mean 1 dim(exp, 3)

Climate Forecasting Unit CFU R common diagnostics This function computes anomalies using the crossvalidation method, i. e. for each startdate, the climatology is computed using all the other startdates. It also uses the per-pair method. CFU_anocrossvalid

Climate Forecasting Unit CFU R common diagnostics CFU_trend This function provides not only the linear trend but also the linearly detrended data.

Climate Forecasting Unit CFU R common diagnostics CFU_load mod = array(dim=c(nexp, nmemb, nsdates, nltimes) to mod = array(dim=c(nexp, nmemb, nsdates, nltimes, nlat, nlon) CFU_clim Those functions work only with the common diagnostic structure. CFU_anocrossvalid obs = array(dim=c(nobs, nmemb, nsdates, nltimes) to obs = array(dim=c(nobs, nmemb, nsdates, nltimes, nlat, nlon) CFU_plotclim CFU_animvsltime CFU_plotano

Climate Forecasting Unit CFU R common diagnostics For those functions, the input structure is free. CFU_season CFU_trend Input matrix can have any number of dimensions and the dimension along which the trend, smoothing or season has to be computed should be specified. Default parameters : common diagnostic structure, leadtime dimensions for CFU_season/CFU_smoothing, nsdates for CFU_trend CFU_smoothing You can use them on any time series

Climate Forecasting Unit CFU R common diagnostics CFU_load CFU_season CFU_trend CFU_clim CFU_anocrossvalid CFU_smoothing CFU_plotclim CFU_animvsltime CFU_plotano

Climate Forecasting Unit CFU R common diagnostics CFU_anocrossvalid CFU_spread CFU_trend CFU_corr CFU_consist_trend CFU_RMS CFU_ratio. RMS CFU_RMSSS CFU_plotvsltime CFU_ratio. SDRMS CFU_animvsltime CFU_plotequimap

Climate Forecasting Unit CFU R common diagnostics For those functions, the input structure is free. Default : common diagnostic structure CFU_spread CFU_trend CFU_corr CFU_RMSSS CFU_ratio. RMS CFU_ratio. SDRMS Scores are computed for each experimental dataset versus each observational dataset in your input matrix.

Climate Forecasting Unit CFU R common diagnostics CFU_anocrossvalid Those functions expect the common diagnostic structure CFU_plotvsltime CFU_animvsltime CFU_consist_trend

Climate Forecasting Unit CFU R common diagnostics For this function, (lat, lon) expected and a second matrix of flags=T/F with the same dimensions is expected for significance level It has many functionalities to make nice plots for publication. Color levels (square or smoothed), contours, dots …, continents can be filled in grey or show as black lines. Colorbar can be drawn or not…. It can be used in a multipanel after splitting the space with layout CFU_plotequimap

Climate Forecasting Unit CFU R common diagnostics Confidence intervals or significance levels or both are systematically provided. CFU_spread CFU_corr CFU_RMSSS CFU_trend CFU_consist_trend CFU_ratio. RMS CFU_ratio. SDRMS

Climate Forecasting Unit CFU R common diagnostics For those functions, there are issues about the temporal dependance of the data for confidence intervals/significance levels. For non-parametric tests, a window of dependence has to be defined, for parametric ones, a number of independant data has to be defined. CFU_corr CFU_RMSSS CFU_ratio. RMS CFU_ratio. SDRMS Those functions currently use parametric tests with a number of independant data defined following the classical formula from Von Storch and Zwiers (2001). This might change depending on the literature. Call to CFU_eno

Climate Forecasting Unit CFU R common diagnostics bootstrap one sided T-test Fisher transform T- distribution CFU_spread CFU_trend chi 2 CFU_corr CFU_consist_trend CFU_RMSSS one-sided Fisher test CFU_ratio. RMS CFU_ratio. SDRMS one-sided Fisher test two-sided Fisher test

Climate Forecasting Unit CFU R common diagnostics CFU_anocrossvalid CFU_spread CFU_trend CFU_corr CFU_consist_trend CFU_RMS CFU_ratio. RMS CFU_RMSSS CFU_plotvsltime CFU_ratio. SDRMS CFU_animvsltime CFU_plotequimap

Climate Forecasting Unit CFU R common diagnostics CFU_eno CFU_meanlistdim For those functions, the input structure is free. CFU_mean 1 dim CFU_insertdim This function makes a colorbar if you send the levels and colors. Useful for multipanels after calling layout CFU_colorbar

Climate Forecasting Unit CFU R common diagnostics [vguemas@bor ~]$ R Ø source(‘/cfu/pub/scripts/R/common_diagnostics. txt’) [1] List of functions : [1] CFU_load [1] CFU_season [1] CFU_clim [1] CFU_ano_crossvalid [1] CFU_smoothing [1] CFU_plotano [1] CFU_plotclim [1] CFU_spread [1] CFU_plotvsltime [1] CFU_corr [1] CFU_RMSSS

Climate Forecasting Unit CFU R common diagnostics [1] CFU_ratio. RMS [1] CFU_ratio. SDRMS [1] CFU_trend [1] CFU_consist_trend [1] CFU_plotequimap [1] CFU_colorbar [1] CFU_animvsltime [1] CFU_eno [1] CFU_enlarge [1] CFU_insertdim [1] CFU_mean 1 dim [1] CFU_meanlistdim [1] CFU_inilistdims [1] For more information about any function, type info_cd('function name') Ø info_cd(‘CFU_load’)

Climate Forecasting Unit CFU R common diagnostics [1] Description [1] ~~~~~~~ [1] Load experimental data and corresponding observed ones in 2 matrix with similar structures [1] If loading EC-Earth experiments, PUT FIRST THE EXPERIMENT ID WITH THE LARGEST NUMBER [1] OF MEMBERS & if possible, THE LARGEST NUMBER OF LEADTIMES. If not possible, fill up the nleatime argument. [1] Inputs [1] ~~~~ [1] - var= 'tas', 'prlr', 'tos', 'g 500', 'g 200', 'ta 50', 'psl', 'hflsd', 'hfssd', 'rls', 'rsds', 'uas', 'vas'

Climate Forecasting Unit CFU R common diagnostics [1] exp=c('ecmwf', 'ukmo', 'cerfacs', 'ifm', 'De. Pre. Sys. Asim. Dec', 'De. Pre. Sys. No. Asim. Dec', 'De. Pr e. Sys. Asim. Seas', 'ECMWF_S 3 Seas', 'ECMWF_S 4 Seas. QWe. CI', 'hadcm 3 dec', 'miroc 4 dec', 'miroc 5 dec', 'mri-cgcm 3 dec', 'cancm 4 dec 1', 'cancm 4 dec 2', 'cnrmcm 5 dec', 'knmidec', 'mpimdec', 'gfdldec', 'cmcccmdec', 'hadcm 3 his', 'miroc 4 his', 'miroc 5 his', 'mri-cgcm 3 his', 'cancm 4 his', 'cnrmcm 5 his', 'knmihis', 'i 00 k', 'b 013', 'b 014', 'yve 2'. . . ) [1] obs=c('ERA 40', 'NCEP', 'ERAint', 'GHCN', 'ERSST', 'HADISST', 'GPCP', 'GPCC', 'CRU', 'DS 9 4', 'OAFlux', 'DFS 4. 3', 'NCDCglo', 'NCDCland', 'NCDCoc', 'GISSglo', 'GISSland', 'GISSoc', 'H ad. CRUT 3 glo', 'Had. SST 2 oc', 'CRUTEM 3 land') [1] - sdates=c('YYYYMMDD', 'YYYYMMDD') [1] - lonmin, lonmax, latmin, latmax : domain border 0 <= lonmin, lonmax <= 360 [1] default : world [1] - nleadtime : optional argument needed only if the first exp does not have the largest number of leadtimes. [1] default : number of leadtimes of the first experiment.

Climate Forecasting Unit CFU R common diagnostics [1] - leadtimemin : output only the leadtimes from leadtimemin. default = 1 [1] - leadtimemax : output only the leadtimes before leadtimemax. default = nleadtime [1] - output = 'areave' / 'lon' / 'lat' / 'lonlat' [1] 1) Time series of area-averaged variables over the specified domain [1] 2) Time series of meridional averages as a function of longitudes [1] 3) Time series of zonal averages as a function of latitudes [1] 4) Time series of 2 d fields [1] default : 'areave' [1] - method = 'bilinear' / 'bicubic' / 'conservative' / 'distance-weighted' [1] Method of interpolation for 'lon' / 'lat' / 'lonlat' output options [1] default : 'conservative' [1] - grid = to choose the output grid [1] possible options : r. NXx. NY or t. TRgrid, ex: r 96 x 72, t 106 grid [1] default : model grid, argument need to be filled if various exp on various grids

Climate Forecasting Unit CFU R common diagnostics [1] - maskmod=list(mask[lon, lat]) = 1/0 : kept/removed grid cell over the entire model domains [1] Warning : list() compulsory even if 1 model !!! [1] default : 1 everywhere [1] - maskobs=list(mask[lon, lat]) = 1/0 : kept/removed grid cell over the entire [1] observed domains, only necessary for 'areave' output option [1] Warning : list() compulsory even if 1 dataset !!! [1] default : 1 everywhere [1]

Climate Forecasting Unit CFU R common diagnostics [1] Outputs [1] ~~~~~ [1] $mod = model outputs [1] $obs = observations [1] $lat = latitudes of the model grid [1] $lon = longitudes of the model grid [1] 2 matrix with dimensions [1] c(nmod/nexp or nobs, nmemb/nparam, nsdates, nltime) if output = 'areave' [1] c(nmod/nexp or nobs, nmemb/nparam, nsdates, nltime, nlat ) if = 'lat' [1] c(nmod/nexp or nobs, nmemb/nparam, nsdates, nltime, nlon ) if = 'lon' [1] c(nmod/nexp or nobs, nmemb/nparam, nsdates, nltime, nlat, nlon) = 'lonlat' [1] Author [1] ~~~~ [1] CFUers <vguemas@ic 3. cat> March 2011