Stochastic Precipitation Downscaling with the RainFARM Method
Aggregation using box-averaging
Downscale a precipitation field
Compute spatial Fourier power spectrum
Compute logarithmic slope of a spatial power spectrum
Gaussianize field using rank ordering
Generate the spectral amplitudes for a metagaussian field
Interpolate field using nearest neighbors
Linear interpolation of longitude and latitude vectors to higher resol...
Spectral merging of a coarse field and of a fine field at a given wave...
Generate a metagaussian field
Perform RainFARM downscaling
Conservative remapping
Derive weights from a fine-scale precipitation climatology
Smoothening using convolution with a circular kernel
An implementation of the RainFARM (Rainfall Filtered Autoregressive Model) stochastic precipitation downscaling method (Rebora et al. (2006) <doi:10.1175/JHM517.1>). Adapted for climate downscaling according to D'Onofrio et al. (2018) <doi:10.1175/JHM-D-13-096.1> and for complex topography as in Terzago et al. (2018) <doi:10.5194/nhess-18-2825-2018>. The RainFARM method is based on the extrapolation to small scales of the Fourier spectrum of a large-scale precipitation field, using a fixed logarithmic slope and random phases at small scales, followed by a nonlinear transformation of the resulting linearly correlated stochastic field. RainFARM allows to generate ensembles of spatially downscaled precipitation fields which conserve precipitation at large scales and whose statistical properties are consistent with the small-scale statistics of observed precipitation, based only on knowledge of the large-scale precipitation field.