Warning
Page under construction.
ecTrans transforms a batch of meteorological fields from a grid point space representation , where is the longitude, is the latitude, and is the index which ranges over the batch of fields, to a spectral space representation , where is the zonal wavenumber, and is the total wavenumber. This constitutes a direct spectral transform. ecTrans can also carry out the inverse spectral transform.
Beginning with the direct spectral transform (grid point space to spectral space), this is accomplished in two computational steps. Firstly, a Fourier transform is performed in the longitudinal direction at each latitude and for each field, independently,
Here, is the number of longitudinal points at latitude and is the imaginary identity. The input meteorological fields, , are always entirely real. Thus, although a full Fourier transform would be evaluated for , terms are simply the conjugate of terms so we needn't calculate them. This is referred to as an "RFFT" algorithm.
The second step is the Legendre transform which is performed in the latitudinal direction. This is performed by Gaussian quadrature with Gaussian weights ,
where is the associated Legendre polynomial of zonal wavenumber and total wavenumber which is evaluated at the point . The Legendre transform is implemented as a matrix-matrix multiplication, collapsing over the latitudinal dimension with the fields and total wavenumber dimensions "free". This operation is performed independently at each zonal wavenumber . Notably, the fact that s is symmetric when is even and is antisymmetric when is odd is exploited to halve the required number of floating- point operations. The Legendre transform is split into two transforms over the Northern hemisphere: one for each of the even and odd polynomials.