Re: SVD analysis Display problem

From: Dennis Shea <shea_at_nyahnyahspammersnyahnyah>
Date: Tue Feb 08 2011 - 07:29:30 MST

This would require time and the data to look at this
and don't have either.

[1]
The method used by Bretherton, Smith and Wallace
and implemented in NCL as:
http://www.ncl.ucar.edu/Document/Functions/Built-in/svdcov.shtml

has some potential serious drawbacks. As noted in the documentation,
at least 2 journal articles [Newman and Sardeshmukh (1995) and
Cherry (1996)] urge caution when interpreting results. I have
not seen this methods used much as a result.

[2]
; attach geographic coordinaes
   copy_VarCoords(ppr(0,:,:),xL(0,:,:))
   printVarSummary(xL)

   copy_VarCoords(stt(0,:,:),xR(0,:,:))
   printVarSummary(xR

Look at (say) Example 3
http://www.ncl.ucar.edu/Applications/cylineq.shtml

Good luck

On 2/7/11 2:18 PM, Siraj ul Islam wrote:
> Hi All,
>
> I am very new to NCL and working hard to learn all about NCL. I started
> some basic analysis by using codes available on Internet. I made some
> changes in the code for SVD analysis and I think have made every thing
> fine. I just need some help to display right and left singular vectors
> in NCL using this code. I do not now how to plot output (spatial
> patterns and time series). Can someone further edit this script to plot
> patterns. I did some thing for plotting but I received an error.
>
> here is the script
> -----------------------------------
>
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_code.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/gsn_csm.ncl"
> load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl"
> begin
> nsvd = 3
>
> f1 = addfile ("cmap_pre_obs_jjas.nc <http://cmap_pre_obs_jjas.nc>", "r")
> ppr = f1->prec(:,{-5:45},{55:110}) ; (time,lat,lon)
> pr = dim_rmvmean_n(ppr,0)
>
> f2 = addfile ("sst_mean_obs_jjas.nc <http://sst_mean_obs_jjas.nc>", "r")
> stt = f2->ssto(:,{-60:60},{-180:180}) ; (time,lat,lon)
> st = dim_rmvmean_n(stt,0)
>
> dimtsm = dimsizes(pr)
> ntim = dimtsm(0)
> nlat = dimtsm(1)
> mlon = dimtsm(2)
>
> dimu10 = dimsizes(st)
> ntim1 = dimu10(0)
> nlat1 = dimu10(1)
> mlon1 = dimu10(2)
>
> printVarSummary(pr)
> printVarSummary(st)
>
> if (ntim.ne.ntim1) then
> print("time dimensions must be the same: ntim="+ntim+" ,
> ntim1="+ntim1)
> exit
> end if
>
> P = onedtond( ndtooned(pr), (/ntim ,nlat *mlon /))
> S = onedtond( ndtooned(st) , (/ntim1,nlat1*mlon1/))
> P!0 = "time"
> P!1 = "space"
> S!0 = "time"
> S!1 = "space"
>
> PP = P(space|:,time|:)
> SS = S(space|:,time|:)
>
> ; covar = svdcov(PP,SS,nsvd,homlft,hetlft,homrgt,hetrgt)
> ; print(covar)
>
> svLeft = new((/nsvd,nlat *mlon /),float)
> svRight = new((/nsvd,nlat1*mlon1/),float)
>
> svdec = svdstd_sv(PP,SS,nsvd,svLeft,svRight)
> print (svdec)
>
> Le = svLeft
> Le!0 = "time"
> Le!1 = "space"
> Left = Le(space|:,time|:)
>
> Ri = svRight
> Ri!0 = "time"
> Ri!1 = "space"
> Right = Ri(space|:,time|:)
>
> transsvLeft = transpose(Left)
> transsst = transpose(P)
> xL = onedtond(ndtooned(transsvLeft),(/2,nlat,mlon/))
>
> transsvRight = transpose(Right)
> transV = transpose(S)
> xR = onedtond(ndtooned(transsvRight),(/2,nlat1,mlon1/))
>
> expcoeficientsL = transsvLeft#transsst
> expcoeficientsR = transsvRight#transV
>
> printVarSummary(transsvLeft)
> printVarSummary(xL)
> printVarSummary(transsvRight)
> printVarSummary(xR)
> printVarSummary(expcoeficientsR)
> printVarSummary(expcoeficientsL)
>
> end
>
> ------------------------------
> here is the output
> ------------------------------
> SVD: Warning all Y values in column are missing or are constant
>
>
> Variable: svdec
> Type: float
> Total Size: 8 bytes
> 2 values
> Number of Dimensions: 1
> Dimensions and sizes: [2]
> Coordinates:
> Number Of Attributes: 1
> sv : <ARRAY of 380 elements>
> (0) 29.16271
> (1) 14.85151
>
>
> Variable: transsvLeft
> Type: float
> Total Size: 3040 bytes
> 760 values
> Number of Dimensions: 2
> Dimensions and sizes: [time | 2] x [space | 380]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
> Variable: xL
> Type: float
> Total Size: 3040 bytes
> 760 values
> Number of Dimensions: 3
> Dimensions and sizes: [2] x [19] x [20]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
> Variable: transsvRight
> Type: float
> Total Size: 87840 bytes
> 21960 values
> Number of Dimensions: 2
> Dimensions and sizes: [time | 2] x [space | 10980]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
> Variable: xR
> Type: float
> Total Size: 87840 bytes
> 21960 values
> Number of Dimensions: 3
> Dimensions and sizes: [2] x [61] x [180]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
> Variable: expcoeficientsR
> Type: float
> Total Size: 160 bytes
> 40 values
> Number of Dimensions: 2
> Dimensions and sizes: [2] x [20]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
> Variable: expcoeficientsL
> Type: float
> Total Size: 160 bytes
> 40 values
> Number of Dimensions: 2
> Dimensions and sizes: [2] x [20]
> Coordinates:
> Number Of Attributes: 1
> _FillValue : -999
>
>
>
>
> Cheers
> Siraj
>
> --
> Siraj Ul Islam
>
> PhD student / Research Assistant
> Environmental Science and Engineering,
> University of Northern British Columbia,
> Prince George, BC, Canada
>
> -------------------------------------------------------------------------------------------
> Please consider the environment before printing this e-mail
> -------------------------------------------------------------------------------------------
>
>
>
> _______________________________________________
> ncl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/ncl-talk
_______________________________________________
ncl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/ncl-talk
Received on Tue Feb 8 07:29:36 2011

This archive was generated by hypermail 2.1.8 : Fri Feb 11 2011 - 16:11:42 MST