How to do Lanczos Filtering of 3-D geospatial data?

I have 3-D geospatial data of Geopotential anomalies that has 8658 time points, and 31x31 spatial points ( 8658time x 31 Lat x 31 Lon). I need to filter the data to retain only high frequency anomalies, using LANCZOS FILTER. The errors on execution point out that Lanczos filtering is done on 2-D time series (any way past this issue?). Now I have hustled through a variety of permutations; closest one being converting that data into (31x31 = 961) 2-D arrays of 8658x1 elements (i.e. separate time-series for each spatial point), applied the filter and then concatenated the resulting arrays into the original 8658x31x31. I am totally unsure about the authenticity of this. In a nutshell I need to apply high pass filter on 3-D data of dimensions [time x latitude x longitude]. Any thoughts?

Answer 1

[Xin Xiao]

Based on ndnanfilter.m - File Exchange - MATLAB Central (mathworks.cn), a Guassian filter was bulit by filt2 2D geospatial data filter - File Exchange - MATLAB Central (mathworks.cn). You can change line <f = fspecial('gaussian',2*ceil(2.6*sigma)+1,sigma);> in filt2.m to build a Lanzos filter.

Hope this help for you.