NCL Website header
NCL Home > Documentation > Functions > Interpolation, Regridding


Iterative improvement objective analysis and returns meta data.

Available in version 5.2.0 and later.


load "$NCARG_ROOT/lib/ncarg/nclscripts/csm/contributed.ncl"  ; not needed from 6.2.0 onward
                                                           ; contributed.ncl is automatically loaded

	function obj_anal_ic_Wrap (
		zlon   [*] : numeric,  
		zlat   [*] : numeric,  
		z          : numeric,  
		glon   [*] : numeric,  
		glat   [*] : numeric,  
		rscan  [*] : numeric,  
		option [1] : logical   



One-dimensional arrays containing the longitudes and latitudes associated with the z values. They must be the same length as the rightmost dimension of z.


An array, whose rightmost dimension is the same length as zlon and zlat, containing the values associated with the zlon and zlat coordinates. Missing values, indicated via z@_FillValue, may be present but will be ignored.


A one-dimensional array of length M containing the longitude coordinates associated with the returned two-dimensional grid.


A one-dimensional array of length N containing the latitude coordinates associated with the returned two-dimensional grid.


A one-dimensional array of length K specifying the successive radii of influence. The maximum size of rscan is 10. Typically, rscan contains one-to-four elements. Must be expressed in degrees of latitude and must be monotonically decreasing. eg: rscan = (/10, 5, 3/)


If option=False, this function will operate under default mode. If option=True, then this variable may have associated with it the attribute blend_wgt. This specifies how successive new estimates are 'blended'(local smoother) with previously derived estimates. These values must be 0.0 < blend_wgt < 1.0. A value of 1.0 means that the current interpolated value will be used. Otherwise:

      new_value(n) = blend_wgt(n)*current_estimate + (1-blend_wgt(n))*value(n-1)
The size of blend_wgt must be the same size as rscan.

Return value

If z is one dimensional,the return array be N x M; if multidimensional the return array will be K x N x M, where K represents the leftmost dimensions of z. It will be of type double if z is double, and float otherwise.


This performs an iterative improvement type objective analysis using an input triplet set ( zlon, zlat, z). For each gridpoint, multiple passes of descending radii of influence are used to obtain better estimates.

Note: It is possible that some returned grid points could be _FillValue. If this is not desired, increase the radii of incluence.

See Also

obj_anal_ic, cssgrid, cssgrid_Wrap, natgrid, natgrid_Wrap, triple2grid, triple2grid2d


Example 1

Assume zlon, zlat and zVal are one-dimensional (1D) arrays; let lon and lat be 1D arrays specifying grid locations. Then:

   rscan = (/10, 5, 2/)
   grid = obj_anal_ic(zlon,zlat,zVal, lon,lat, rscan, False)

will use default behavior and return a 2D array of size N x M M.

Example 2

Let zVal be a two-dimensional array of size (T,KPTS). Let lon and lat be 1D arrays of lengths N and M specifying grid locations. Then:

   rscan = (/20, 12.5, 7.5/)
   grid = obj_anal_ic_Wrap(xlon,ylat,zVal, lon,lat, False)        ; grid(T,N,M)

will use default behavior and return a 3D array of size T x N x M.

Example 3

   rscan         = (/20, 10, 5/)
   opt           = True          
   opt@blend_wgt = (/1.0, 0.5, 0.25/)
   grid = obj_anal_ic_Wrap(xlon,ylat,zVal, lon,lat, opt)