CHANGE RECORD

Main changes from version 2.1 to version 2.2:
-  NSCAT-4 replaces NSCAT-2 as default GMF, resulting in lower wind speed estimates
   at high winds, in agreement with buoy data.
-  GMF ID now correctly written in BUFR output: so far, SDP copied the GMF ID
   from the BUFR input, which basically boils down to the ID of the QSCAT GMF
-  NWP Ocean calibration (NOC) included.
-  Logarithmic term added in inversion cost in order to make SeaWinds processing
   consistent with OSCAT processing. It improves the wind retrieval at low
   wind speeds.
-  New Quality Control mechanism which is based on the MLE of the selected wind
   solution. Quality Control is now completely independent of the JPL wind
   retrievals and QC in the input data.
-  Length of 2DVAR Ambiguity Removal batches is increased, essentially an orbit is
   split up into six batches now: SH ascending, tropics ascending, NH ascending,
   NH descending, tropics descending and SH descending.

Main changes from version 2.0 to version 2.1:
-  Bayesian ice model introduced.

Main changes from version 1.99 to version 2.0:
-  Slight change in definition of KNMI MLE flag (now also includes ice)
-  Documentation updated according to beta tester's remarks.
-  Bayesian ice model (modules SdpIcemodel in SDP and iceModelMod in genscat)
     added, though these modules are not yet operational.

Main changes from version 1.5 to version 1.99:
-  Processing of the outer swath is included.
-  The quality control after inversion has been redefined due to the changes in
     the NOAA input files.
-  New recommended parameter settings of the 2DVAR error model.
-  More flexible definition of the 2DVAR batch grid size and dimension by applying
     a mixed-radix FFT method.
-  Memory leak repaired.
-  New command line options for resolution and MSS.
-  Automatic detection of most recent BUFR library in genscat/support/bufr.

Main changes from version 1.4 to version 1.5:
-  The initial step size for the minimalisation by routine LBFGS has been
     drastically increased, resulting in much better skill to find the minimum.
     With the old settings, LBFGS may get stuck in the first iteration.
-  Only one call to minimise is applied. The previous extra call to minimise
     without Gross Error Probabilities is no longer needed.
-  If MSS is applied, the NOAA BUFR output now only contains the selected solution
     and its normalized MLE. The other solutions can only be obtained via the
     generic wind section BUFR output (-genericws command line option).
-  A new command line option for research purposes, -ocf, has been included.
-  Improved installation script InstallSDP: now it checks if the BUFR library
     file is present before compiling anything.

Main changes from version 1.3 to version 1.4:
-  2DVAR (modules TwoDvar and underlying modules) corrected
-  Changes in minimalization strategy (single search is sufficient now).
-  Inactive command line options removed.

Main changes from version 1.2 to version 1.3:
-  Some routines moved from SWS-layer to genscat
-  Code cleanup, resulting in less warning messages during compilation under GNU g95
-  Inversion module improved for very low wind speeds
-  Flag management streamlined
-  Installation scripts extended

Main changes from version 1.1 to version 1.2:
-  Own FFT routine instead of third party software
-  Improved efficiency in writing BUFR output

Main changes from version 1.0 to version 1.1:
-  New minimization routine (LBFGS by J. Nocedal; free of rights)
-  Improved minimization strategy
-  Removed FFT routine fourn (because of copyrights)
-  Installation scripts in make system translated to Bourne shell