v4.7.2

• Allow for variable and file names with space.
• Bug correction for data cleaning an generation of CL field.
• Fixed indentation is some scripts.

v4.7.1

New features

• Major feature : bottom analysis is now possible. The distance is counted from the bottom ocean, derived from the interpolation of the topography topo_fine.grd. This topography can be different (and finer) than topo.grd used for the creation of contours.
• Major feature : conversion of EMODnet bathymetry to Diva-readable format with the tool emobath2ghertopo.
• Major feature : Variable correlation length depending on the gradient of the depth. Advection field adapted to this relative length field. Suited for bottom analyses. Updated programs : diva3Ddat, divarlvargraddepth, rlvargraddepth.f90, divadoall, divaUVtopo, UVtopogen.f, divadocommit.
• Acknowledgements field in 3D and 4D netCDFs.
• New compilation option DIVAHUGEMEMORY. Enabled by default in divacompile_options, it allows the use of a finer mesh, and/or a larger domain. Particularly useful with variables characterized by low correlation lengths.

Bug fixes

• divacutNCDF : correction on climatology bounds + dealing with very big obsid vector
• divacompile_options : new tests on nc-config, due to recent change in its behaviour
• dv3Dreadnc.F : warning if dimensions are incoherent between GridInfo.dat and the netcdf file
• divadoall, divadoNCDF , divadoNCYR : corrected handling of 3DNCinfo and 3DNClist files so that recreating a new 4D NC file is much easier
• divadoxml : removed because deprecated (use divadoxml-gui instead)

Other

Update of the user guide

diva-4.7.1-test

diva-4.6.12-test

release 4.6.12-test

diva-4.6.11

New features

• Creation of very large 4D netCDF files (> 2Go)
• divacutNCDF: a tool to cut the domain of your final NetCDF
• divaselectorODV4: more general detection of non-sdn profiles through the use of position
• dvvinterpodv: warning in case of varying time inside a single profile

Bug fixes

• anatransf.F: generation of a correct reference field when a transformation (log, logit,...) is asked for the reference
• dv3Dinvtrsh: fieldatdatapoint.anl files are now correctly "inverse transformed" when a transfo is asked
• dv3Drefoanl, dvrefetransf : cleaning of the code (path, log files)
• dvrefetransf: correct transfo of reference field when a transfo is asked for the analyse
• refetransf.F : handling of exclusion values in ref transfo
• diva3Dwrt.F: correction of the NetCDF time axis in some winter season (ex. 1101) crossing new years (no problem if 1202)
• dv3Drefoanl: always recompute databins files, in order to avoid old files impacting the final netcdf dimensions
• boudarypoints.f : minor correction avoiding diva crash in very rare cases (if ispec=-113)
• divaselectorODV4 : bug fix for winter season was forgotten in 4.6.10

Other

Update of the user guide

diva-4.6.9

New features

• Creation of contours if they are absent (+ warning) but needed for cleaning/optimization/analysis
• Use of whitespaces in variable names is allowed
• Safeguards added when optimizing parameters with too few data
• obsid,obslon,obslat,obsdepth and obstime now included in the global 4D netcdf file
• Optimization of parameters + vertical filtering in Example4D
• Increased speed of SN optimization in some cases, by avoiding the useless calculation of analysis errors
• Diva now also handles time series (profile and time series have to be in separated ODV files)
• gebco2diva : subsampling 10x10 by default
• better compatibility with Mac OS

Bug fixes

ncdf2gher.f90 : generalization to variables other than "bat"
dv3Drefoanl : now handles variable names with blanks
dvdatatransf,dv3Dintrsh,datatransf.F,anatransf.F : handles variables names with blanks (when using transformation flags 11,12,...)
main_tomvec.f90 : bug fixed when filtering CL vertically in some cases
diva3Dwrt.F : correction of the time axis when month period is across two years (ex : dec-feb)
dv3Dncwrt.F,diva3Dwrt.F : bug solved with climatology bounds when using 3D version

v4.6.6

New features

• Check for severe errors in DIVA 3D/4D (script godiva) + simple errors and warnings
• Possibility of binning the data before the parameters optimization (script divabin + program binning_lines.f90)
• Variable correlation length, depending on depth (script "divarlvardepth" + program "rlvardepth.f90")

New bug fixes

• Correction of the example in 4D (datasource)
• Correction of the script divaguessformODV4
• Exact match needed between variable name in "varlist" and its real name in the data fil

diva-4.6.5

New features

New bug fixes

• "end of line" problems under Windows (file "datasource")
• Portability of scripts using the "sort" command
• Vertical filtering of correlation length : case of 1 and 2 layer(s)
• Wrong min and max values in the netcdf output file (error and analyzed field) when using some values of ispec
• Error field not written in the netcdf output file under some values of ispec
  Other small fixes

v4.6.2

• Use of tee to redirect echo messages to screen and log files.
• Implementation of divaR (mimic a non-diagonal R matrix).
• Inflation approach

v4.6.1

• Two additional solvers: a parallel and iterative version
• Optimisations for large data sets
• Optimisations of file exchanges for use with ODV
• Highly optimised new version of the grid generator (typical reduction of a factor 10 in CPU time)
• Integration of faster error calculations into the 3D/4D version (using new ispec values)
• Clearer log files for debugging.

Note: with the availability of gfortran, we decided to abandon the support of pure f77 compilations and this release needs Fortran 90 or Fortran 95 and -cpp preprocessing tools).

Example of performance gain using divabigtest (test with 1.5 million data points and a very fine mesh):

CPU time in seconds in	4.5.1	4.6.1
mesh	27	2
analysis	318	159

The results presented below show how the computation time evolves:

• for the computation when the number of observations increases.
• for the mesh generation when the correlation length is modified.