RainFARM Julia module Documentation

rainfarm(r,slope,nf,weight=1.;fglob=false, fsmooth=false, verbose=false)

Perform general RainFARM downscaling.

#Arguments

• r : large-scale array to downscale
• slope : spatial spectral slope
• nf : refinement factor for spatial downscaling
• weight : weights for orographic downscaling
• fglob : conserve global average over domain
• fsmooth: use smoothing instead of gp conservation
• verbose: provide some progress report

ww = rfweights(orofile, reffile, nf; weightsfn="", varname="", fsmooth=false)

Compute orographic weights from a fine-scale precipitation climatology file.

#Arguments

• orofile : filename of input climatology
• reffile : filename of reference file (for metadata, e.g. the file to downscale)
• nf : refinement factor for spatial downscaling
• weightsfn: write weights to file weightsfn
• varname : variable name in climatology
• fsmooth : use smoothing instead of gp conservation

#Returns

• ww : a weight matrix also saved to weightsfn

#Depends

This function uses external system calls using the "cdo" command (https://code.mpimet.mpg.de/projects/cdo/wiki/Cdo) which needs to be available on your system.

(fx,ft)=fft3d(z)

Compute spatial (fx) and temporal (ft) Fourier spectra of field z.

sx = fitslopex(fx; kmin=1)

Return spectral slope (minus 1) of spatial spectrum fx.

(lon_f, lat_f) = lon_lat_fine(lon, lat,nf)

Interpolate longitude and latitude arrays (2d or 1d) lon and lat to higher resolution by a factor nf.

(data, lon, lat, time) = read_netcdf2d(file, varname="")

Read variable varname (optional) from netcdf file. Returns data and coordinate and time axes.

write_netcdf2d(fname,var,lon,lat,varname,filenc)

Write field varto netcdf file fname, with coordinates lon and lat and variable name varname. File filenc is used to copy metadata from.

za = agg(z,nas,nat)

Aggregate field z to an array za of size (nas,nas,nat)

zi = interpola(z,ns,nt)

Interpolate field z to size (ns,ns,nt) using nearest neighbors.

f = initmetagauss(sx,st,nso,nto)

Generate the spectral amplitudes f for a metagaussian field of size nso * nso * nto with slopes sx and st.

fr = metagauss(f)

Generate a metagaussian field multiplying the spectralfield f with random phases and performing an inverse FFT transform to real space.

g = gaussianize(z)

Gaussianize field z using a sorted random sequence.

zs = smoothconv(z,nas)

Smoothen field z(ns,ns) with a circular kernel of diameter ns/nas using convolution. Takes into account missing values.

fm = mergespec_spaceonly(ra,f,kmax)

Spectral merging of coarse field ra and fine field f at wavenumber kmax.

rf = downscale_spaceonly(r,f,weight=1.;fglob=false, fsmooth=false )

Downscale precipitation field r using metagaussian spectral field f. An optional weights array can be specified. Precipitation can be conserved globally (fglob) or using convolution (fsmooth).