Skip to content

Latest commit

 

History

History
75 lines (48 loc) · 7.31 KB

CHANGELOG.md

File metadata and controls

75 lines (48 loc) · 7.31 KB
Raw

Changelog

All notable changes to the model will be documented in this file. Breaking changes (i.e. those that cause changes to the model's interface) are denoted by ⚠️. Until v1.0.0, breaking changes might occur between minor versions and patches. Thereafter, semantic versioning will be followed.

Unreleased

Changed

  • ANSI escape character now uses char(27) instead of backslash escape sequences, so -fbackslash (GFortran) and -assume bscc (Intel) compiler options no longer needed.
  • ⚠️ The example config files now save output data to output, instead of data/output.

0.0.4 - 2021-12-08

Added

  • Config option to include bank erosion or not: &water > include_bank_erosion. Defaults to true.

Changed

  • ⚠️ Model version number is now obtained from the src/VersionModule.f90 file. This is dynamically updated at build time by the example Makefile or CMakeLists.txt file, and if you are not using one of these build processes, should ensure that yours does the same. In short, the output of git describe --tags should be used for the modelVersion variable in src/VersionModule.f90. For example, this can be achieved in a Makefile by the use of sed: sed -i "s/\".*\"/\"${git describe --tags}\"/g" src/VersionModule.f90. Note that the version number in this file and the CITATION.cff file still needs bumping on each release. Build process will need updating.
  • ️️️️Lots of file, module and type names have been updated to follow a coherent convention. See the updated Makefile.example.

0.0.3 - 2021-08-05

Added

  • Warm up period: You can now run the model before inputting NM using the &run > warm_up_period config option, which runs the model for the number of time steps specified. The data that is used is sliced from the start of the dataset (e.g. warm_up_period = 10 would use the first 10 time steps' data), but no NM is input to the environment. The warm up period must be fewer time steps than the model run (or the first chunk of the run, if running in batch mode).
  • Checkpointing now has the ability to save a checkpoint after the warm up period, specified by the &checkpoint > save_checkpoint_after_warm_up option. If both save_checkpoint and save_checkpoint_after_warm_up are specified, the checkpoint at the end of the model run will overwrite the checkpoint file saved after the warm up.
  • Documentation for running the model for different time periods.
  • Added output_hash config option. The value specified here will be appended to output file names (e.g. "output_water.csv" will become "output_water{output_hash}.csv"). Particularly useful when performing parallel model runs to avoid multiple model runs writing to the same file.
  • make fast target added to example Makefile, which turns on -Ofast -march=native -mtune=native flags. Use with caution.
  • Config option to not include estuary: &waterbody > include_estuary. Defaults to true. If include_estuary = .false., the is_estuary NetCDF variable is ignored.

Changed

  • Changed "untransformed" to "pristine" in output data nomenclature, to describe NM that hasn't undergone any chemical transformation. This brings the model nomenclature in line with literature conventions.
  • ⚠️ NM transfer between bed sediment layers uses a NxN mass transfer matrix, where N is the number of sediment layers. This tends to end up quite a sparse matrix, and is the biggest bottleneck in the model. To speed things up, CSR storage is now being used for this matrix, requiring the compilation of two new files: vendor/spoof/sparskit.f90 and vendor/spoof/Spoof.f90. Your Makefile might need updating. SPOOF is a new library: SParse matrices in Object-Oriented Fortran, which acts as a modern Fortran interface over the SPARSKIT library (so far only CSR and diagonal matrices, and matrix by vector multiplication, are implemented).

Fixed

  • Fixed bug where output_soil__land_use and output_soil__bulk_density attribute of NetCDFOutput class weren't being deallocated before being allocated between chunks in a batch run.

Removed

  • ️️⚠️ Removed depracated calibration method, including config file calibrate group. You will need to remove this from any config files to avoid errors. The src/classSampleSite.f90 has been removed, and may need removing from your Makefile.

0.0.2 - 2021-05-21

Added

  • ⚠️ Added option to write to NetCDF file. Specify &output > write_netcdf = .true. to use. This requires the compilation of a new Fortran source file, src/Data/NetCDFOutputModule.f90, and your compilation process might need ammending. The example Makefile has been updated accordingly. NetCDF output variables are stored in memory for the entire model run (as opposed to CSV files, which are written iteratively on each time step), which means choosing to write to NetCDF may use significantly more memory.
  • Along with this, added the option &output > netcdf_write_mode, to choose when NetCDF files are written to. If itr, the NetCDF file is written to on every timestep. If end, the NetCDF file is written to at the end of the each chunk (which is the end of the model run, if not running in batch mode). itr is slower, but output variables aren't stored in memory for the duration of the run, and so it uses less memory. end is faster, but for particularly long chunks or geographically extensive runs, memory use might be prohibitively high.
  • Added option to aggregate CSV output for bed sediments at grid cell level (extending functionality added in 0.0.1 to do so for waterbodies). This option can be used by specifying &output > include_waterbody_breakdown = .false. in the model config file
  • Added option to compile with cmake using the CMakeLists.txt file. In theory this should work with GFortran and Intel Fortran, though testing on the latter has been limited.

Changed

  • ⚠️ Renamed files - your compilation process may need ammending. The example Makefile and CMakeLists.txt have been updated.
    • src/CheckpointModule1.f90 → src/CheckpointModule.f90
    • src/DataOutputModule1.f90 → src/DataOutputModule.f90
  • Submodule vendor/mo_netcdf updated to latest version to allow the creation of scalar variables. You might need to git submodule update.
  • Submodule vendor/json2netcdf removed. You might need to remove this folder manually after pulling changes.

0.0.1 - 2021-03-20

Added

  • Added option to aggregate CSV output for waterbodies at grid cell level, rather than breaking it down to waterbody level. Internal functions for aggregating to grid cell added (e.g. weighted means, fetching outflow reaches). This option can be used by specifying &output > include_waterbody_breakdown = .false. in the model config file. Default is .true..
  • This changelog.