NCL專輯 | 自帶常用統計學函數集錦

NCL作為一門氣象專業語言，自帶了很多氣象屆常用的算法和命令， 比如計算相關係數、均方根誤差等統計量，相關係數顯著性檢驗，EOF，以及各種插值。這裡將一些常用函數做了歸納：NCL中計算相關係數的函數有好幾個，常用的為escorc系列和regCoef系列。escorc(x,y)是計算兩個數組最右側一列的相關係數。如果想指定計算的列數，比如想計算x(lat,lon)和y(lat,lon,year)在lat這一列上的相關係數，就要用到escorc_n: escorc_n(x,y,0)。相關係數的檢驗方法包括t檢驗、f檢驗和顯著性檢驗，分別為ttest，ftest，rtest。regCoef系列的命令比escorc強大很多，因為它除了返回相關係數rc，還捎帶手把兩個序列的標準差（rc@rstd）和t檢驗結果（rc@tval）都給算了。函數名為dim_rmsd。它是計算兩個變量在所有其他維度上最右側維度的均方根誤差。而對於想要計算指定維度的情況，則要用dim_rmsd_n。計算EOF分解的函數為eofunc開頭的系列函數，一般用eofunc_Wrap計算空間模態，用eofunc_ts_Wrap計算時間係數。以下為針對變量x計算EOF分解並繪圖的NCL子程序：

undef("draw_eof_plot")procedure draw_eof_plot(dir_plot,file_plot,type_plot,x,year,neof)begin;--EOF analysisoptEof = Trueeof    = eofunc_Wrap( x, neof, optEof)eof_ts = eofunc_ts_Wrap( x, eof, False)lat = x&latlon = x&lon;--Begin plotting section.wks  = gsn_open_wks(type_plot,dir_plot+file_plot)      ; Opens a ps file gsn_define_colormap(wks,"rainbow")   plot = new(neof,graphic)   res              = True                       ; plot mods desired;************************************************; original data;************************************************   res@gsnDraw                    = False        ; don't draw yet   res@gsnFrame                   = False        ; don't advance frame yet   res@gsnAddCyclic             = False;--map plot resources  res@mpFillOn                   = False        ; no grey continents  res@mpCenterLonF      = 180.  res@mpDataBaseVersion     = "MediumRes"    ; or "Ncarg4_1"  res@mpDataSetName="Earth..4"  res@mpOutlineSpecifiers=(/"China:states","Taiwan"/)  res@mpOutlineBoundarySets = "AllBoundaries"  res@mpMinLatF  = min(lat)            ; range to zoom in on  res@mpMaxLatF  = max(lat)  res@mpMinLonF  = min(lon)  res@mpMaxLonF  = max(lon)
;--contour resources  res@cnFillOn             = True              ; turn on contour fill  res@cnLineLabelsOn       = False              ; turn off contour  res@cnLinesOn            = False        ; add countor or not,True is default  res@gsnLeftString = " "  res@gsnRightString = " "
;--tickmark resources  res@tmXTOn            = False  res@tmYROn            = False  res@tmYLLabelFontHeightF =0.02  res@tmXBLabelFontHeightF =0.018  res@tmXTOn            = False  res@tmYROn            = False
;--labelbar resources  res@lbLabelBarOn         = False  res@lbLabelFontHeightF   = 0.02  res@lbOrientation       = "Vertical"          ; vertical label bar
  symMinMaxPlt(eof, 16, False, res); contributed.ncl; panel plot only resources  resP                     = True         ; modify the panel plot  resP@gsnMaximize         = True         ; large format  resP@gsnPanelLabelBar    = True         ; add common colorbar  resP@txString  = "EOF"do n=0,neof-1        res@gsnLeftString  = " EOF "+(n+1)        res@gsnRightString = sprintf("%5.1f", eof@pcvar(n)) +"%"        plot(n) = gsn_csm_contour_map(wks,eof(n,:,:),res)end do   gsn_panel(wks,plot,(/neof/1,2/),resP)     ; draw all 'neof' as one plot;*******************************************; time series (principal component) plot;*******************************************  eof_ts@long_name = "Amplitude"  rts           = True  rts@gsnDraw   = False       ; don't draw yet  rts@gsnFrame  = False       ; don't advance frame yet
; decide exactly where on the page to draw it.  rts@vpHeightF = 0.40        ; Changes the aspect ratio  rts@vpWidthF  = 0.85  rts@vpXF      = 0.10        ; change start locations  rts@vpYF      = 0.75        ; the plot  rts@gsnYRefLine           = 0.              ; reference line  rts@gsnAboveYRefLineColor = "red"           ; above ref line fill red  rts@gsnBelowYRefLineColor = "blue"          ; below ref line fill blue; panel plot only resources  rtsP                     = True             ; modify the panel plot  rtsP@gsnMaximize         = True             ; large format  rtsP@txString  = "EOF"  do n=0,neof-1        rts@gsnLeftString  = " EOF "+(n+1)        rts@gsnRightString = sprintf("%5.1f", eof@pcvar(n)) +"%"        plot(n) = gsn_csm_xy (wks,year,eof_ts(n,:),rts)  end do  gsn_panel(wks,plot,(/neof/1,2/),rtsP)        ; draw all 'neof' as one plotend
數據處理·機器學習·可視化
行業資訊·學習資料
長按關注不迷路