Use SI units for internal computations

[SouthEndMusic]

Issue addressed

Fixes #326
Fixes #769
Fixes #810

Checklist

• Updated tests or added new tests
• Branch is up to date with master
• Tests & pre-commit hooks pass
• Updated documentation if needed
• Updated changelog.qmd if needed

Comments

Please when reviewing the code keep in the back of your mind the question whether everything would still work with a non-default timestep.

Changes in the order in which I think the review makes the most sense:

• A few small sediment documentation fixes
• Have a good look at the unit implementation in units.jl
• Have a good look at io.jl and the ncread function where the unit conversions for IO are done
• For most numerical computations I added the units as I understood them. Please check whether they make sense.

Please see this comment for some the causes I identified for changes in outcome (other than the change in units)

Also note that for some outputs which are denoted in the TOML by the path in the model object the unit is not available, which will result in warnings like this:

┌ Warning: No unit was found for routing.overland_flow.sediment_flux.variables.clay, so the output will be expressed in standard SI units (K, s, m, kg, J).
└ @ Wflow D:\repositories\Deltares\Wflow.jl\Wflow\src\io.jl:703
┌ Warning: No unit was found for routing.overland_flow.transport_capacity.variables.clay, so the output will be expressed in standard SI units (K, s, m, kg, J).
└ @ Wflow D:\repositories\Deltares\Wflow.jl\Wflow\src\io.jl:703
┌ Warning: No unit was found for routing.river_flow.sediment_flux.boundary_conditions.transport_capacity, so the output will be expressed in standard SI units (K, s, m, kg, J).
└ @ Wflow D:\repositories\Deltares\Wflow.jl\Wflow\src\io.jl:703
┌ Warning: No unit was found for routing.river_flow.sediment_flux.boundary_conditions.erosion_land_clay, so the output will be expressed in standard SI units (K, s, m, kg, J).

[SouthEndMusic]

A note on the integration tests that are still failing:
I identified several places in the code which are extremely sensitive to small floating point differences in the variables:

• When some flows here turn from zero to very small positive:

Wflow.jl/Wflow/src/routing/surface_kinwave.jl

Line 538 in 402be35

if q[i] > 0.0

• When the difference here is very slightly bigger than a multiple of 0.1 where it was exactly a multiple before:

Wflow.jl/Wflow/src/routing/routing_process.jl

Line 170 in 402be35

its = Int(cld(abs(max(zi, 0.0) - zi_prev), max_delta_zi))

The first one I couldn't really do anything about, for the second one I changed the sensitivity a bit so that it behaves at before, but we might just accept the change in behavior to keep the code simple.

[SouthEndMusic]

In the latest commit I refactored the Writer struct, I thought it was a bit messy (which probably happened slowly over time). There also wasn't a uniform approach for collecting fields from the model (either during initialization or after each simulation step). It's now also easier to extend, if at some time we want to add more CSV or NetCDF files.