Wavelet analysis is a time-series method that decomposes a signal into time and frequency space at once, so that localised variations of power can be located.12
Torrence and Compo describe it as a common tool for analysing localised variations of power within a time series, and note that it can handle series containing nonstationary power at many different frequencies.1 NOAA’s Physical Sciences Laboratory gives a plainer account, saying it is a mathematical technique that decomposes a time series into time and frequency space simultaneously.2
SWEET places the term under spectral analysis, which fits a method that estimates how power is distributed across scales.3 The Torrence and Compo guide compares the method with the windowed Fourier transform, covers the choice of a wavelet basis function and warns of edge effects in finite-length series.1
Their worked examples come from the El Niño-Southern Oscillation, so the retained sources do not establish how the method is applied to satellite imagery or to gridded Earth observation products. They also give no single recommended wavelet for remote sensing work.12
Footnotes
-
NOAA PSL (AMS), “A Practical Guide to Wavelet Analysis”. https://psl.noaa.gov/people/gilbert.p.compo/Torrence_compo1998.pdf ↩ ↩2 ↩3 ↩4
-
NOAA Physical Sciences Laboratory, “WRIT Monthly Time-series: Wavelets”. https://www.psl.noaa.gov/data/atmoswrit/timeseries/wavelets/ ↩ ↩2 ↩3
-
ESIP SWEET, “reprDataServiceAnalysis.ttl”. https://raw.githubusercontent.com/ESIPFed/sweet/master/src/reprDataServiceAnalysis.ttl ↩