Application of the wavelet transform for modelling non-stationarity and local anisotropy in precipitation fields
Robert W. Scovell · Repository for Publications and Research Data (ETH Zurich) · 2019
Methods for scale-selective blending of precipitation fields, e.g. as implemented in STEPS (Bowler et al., 2006), typically use band-pass filtering methods to achieve scale separation.For certain choices of the set of band-pass filters (the filter bank), this is equivalent to a discretized Continuous Wavelet Transform (CWT).The filters cannot be localized to an arbitrarily small support simultaneously in both frequency and position space, so it is useful to consider the optimal tradeoff.Examples of the localization properties of some well-known (continuous) wavelets, as in Torrence and Compo (1998), will be given and compared to those of the non-standard filters in STEPS.