Geophysical data tend to be heterogeneous in measurement characteristics and irregularly
sampled over space and time. For example, in satellite remote sensing of the sea surface
temperature, the microwave (MW) sensors have typically coarser 25-km resolution than the
infra-red (IR) sensors which can resolve down to a 1-km scale. However, the MW sensors
are less prone to cloud contamination which affects the IR data and leads to data
voids. Moreover, geostationary satellites have fine temporal resolutions but cover
limited geographical regions, while orbiting satellites can have global coverages by
compromising temporal sampling. To merge such data sets with drastically different spatial
resolution and coverage into a coherent reconstruction of the geophysical field, the
multiresolution variational analysis (MRVA) method has been used. MRVA is a
hybrid of the variational interpolation technique used commonly in meteorology and
oceanography and multiresolution analysis (MRA) technique known for wavelet-based
orthonormal signal transformation. Apparent advantages of MRVA include: (1) the
interpolation scale (correlation scale) can be optimized for each input data set; (2)
inter-sensor bias corrections are facilitated; (3) the spectral characteristics (e.g.,
power law) of the output can be controlled to match a desired specification. These
are demonstrated with remote sensing data for the sea surface temperature and
wind. |