1.5.0 beta 3

FIx delta_t consistency check. How it just checks that the offsets are spaced consistently. Should be more reliably and more performant

1.5.0 beta 2

Fix netcdf loading bug with hourly

1.5.0 beta 1

• fix missing file

1.5.0 beta 0

This version includes a major rewrite of the netCDF loader:

• requires CF-compliant standard_name attributes to dynamically determine variables
• support multipart netcdf files (e.g., one nc file per day)
• the grid of each multipart does not need to remain constant, allowing for trivially changing the resolution or product used at different timesteps, for example to switch from a short term 2.5km forcing to a coarser long-term 10km forcing
• output of a per-rank grid centres used for forcing as shapefiles

This also fixes a bug associated with loading too many forcing points for a mesh.

1.4.5

• Fix the LUT code path when [email protected]
• Relax the LAI condition to <1 and set the canopy height to be grass (0.25m)
  Full Changelog: 1.4.4...1.4.5

1.4.3

• The LUT regression from 1.4.2 has been resolved and renabled.
• The MPI unique tag generation was improved to respect MPI TAG upper bound to ensure portability. This directly supports Intel MPI usage.
• A high-rank race condition was resolved with the auto checkpoint code to ensure that if any rank thinks it should auto-checkpoint, all the ranks auto-checkpoint
• Avalanche snowdepth is constrained to be 20m maximum
• FSM2 increased for 3 to 6 maximum layers
• Harder phase partitioning: Bound the ratio of 3%-97% as per pers comms with Harder
• Low LAI canopies with tall vegetation are bounded to be 0 m height as the parameterization produces unrealistic values otherwise
• New frequency options for output include specific date and daily time. Please see the documentation for more etails
• The default behaviour of mesh output has been changed and the new default is not to output any vtu files. The user must specify the output frequency or timing.

1.4.2

The thin plate spline with tension interpolant uses a look up table (LUT) optimiziation. There is a regression with how windspeeds are interpolated that may, under rare circumstances, lead to very high windspeeds. The LUT usage has been disabled until this is resolved.

1.4.1

This release adds the ability to automatically checkpoint when X minutes of wallclock time is left. In addition, CHM can now automatically resume from the most recent checkpoint file.

Combined with the the new clean_exit file that is written on final simulation end, the intent is to allow automatic suspend/resume in HPC context that have limited wall clock time.

Please see
https://chm.readthedocs.io/en/develop/configuration.html#checkpoint
for more details.

In addition, two key bug fixes:

• protect against a known compiler bug in gcc@12 when used with boost 1.85.0
• ensure variablestorage<T>::has protects against a not-found variable lookup

1.4.0

• Increase the number of diagnostic variables output from FSM
• OpenMP is now defaulted to off due to the difficulty in propogating exceptions out of parallel regions. Suggestion is to exclusively use MPI for parallelsim.
• The spack.yaml in the root directory can be used to produce a working development environment

Full Changelog: 1.3.1...1.4.0

1.3.1

Fix a missing include that impacted older boost versions