pywatershed 2.0.0
Overview
This section looks at some highlights of this version. The following section gives a (somewhat redundant though) more detailed change log.
FlowGraph
The primary new capability introduced is the FlowGraph
and its related classes. The FlowGraph
allows users to combine different kinds flow solutions on FlowNodes
in arbitrary order. In the figure below, a node of class B is inserted into the original graph. Class B may have a different flow solution than the class A nodes in the original graph, but FlowGraph
will handle new FlowNodes
wherever you want to put them. This allows users to mix and match various flow solutions, or introduce new flow solutions at new locations. The figure below links to the FlowGraph
documentation for further details.
While FlowGraph
s may be constructed from scratch, a common use case (demonstrated in a new example notebook,
06_flow_graph_starfit.ipynb) is to add a reservoir representation into an existing FlowGraph
which is equivalent to PRMSChannel
, using PRMSChannelFlowNode
. The STARFIT reservoir representation is implemented as StarfitFlowNode
which can be added into the FlowGraph
. Other FlowNodes
implemented are PassThroughFlowNode
and ObsInFlowNode
. The second reproduces the "obsin" capabilities present in PRMS. Please see the FlowGraph
documentation (linked throughout above) for additional details.
Additional PRMS Functionality
- Subclasses of PRMSRunoff, PRMSSoilzone, and PRMSGroundwater are available
which do not include depression storage: PRMSRunoffNoDprst, PRMSSoilzoneNoDprst, and PRMSGroundwaterNoDprst. - Dunnian flow is implemented in PRMSSoilzone and is 2-way coupled to PRMSRunoff.
- Preferential flow is implemented in PRMSSoilzone.
MmrToMf6Dfw
The MmrToMf6Dfw
class builds a MF6 1-D channel (CHF) simulation using diffusive wave (DFW) routing from PRMS NHM input files and a few simple assumptions. The lateral (to-channel) fluxes from a PRMS run are used as time varying boundary conditions. A new example notebook (07_mmr_to_mf6_chf_dfw.ipynb) runs the Delaware River Basin using MF6 CFH-DFW based on PRMS input files.
Breaking Changes
The parameter pref_flow_infil_frac
is now a required parameter input for PRMSSoilzone. The NHM values assumed previously are zeros on all HRUs.
Detailed Change Log
For additional details, see the What's New Documentation:
New Features
- The
FlowGraph
capabilities are introduced. These allow users to combine different kinds flow solutions in arbitrary order on a "flow graph". The accompanying base classesFlowNode
andFlowNodeMaker
are introduced along with their subclasses for modelingPassThroughFlowNode
s,ObsInFlowNode
s (flow replacement by observations with sink and source tracking in mass balance),PRMSChannelFlowNode
s, andStarfitFlowNode
s. A new example notebook, examples/06_flow_graph_starfit.ipynb demonstrates adding STARFIT reservoir nodes into a FlowGraph otherwise simulatingPRMSChannel
and highlights helper functions for this use case. - The
MmrToMf6Dfw
class builds a MF6 simulation with Diffusive Wave (DFW) routing from PRMS NHM input files and a few simple assumptions. The lateral (to-channel) fluxes from a PRMS are used as time varying boundary conditions. A new notebook runs the Delaware River Basin using MF6 DFW: examples/07_mmr_to_mf6_chf_dfw.ipynb. - No depression storage subclasses are available for PRMSRunoff, PRMSSoilzone, and PRMSGroundwater by adding "NoDprst" to the end of the names. Depression storage is switched off in sagehen_5yr and in new nhm_no_dprst configurations.
- Dunnian flow is implemented (in PRMSSoilzone) and tested for sagehen_5yr.
- Preferential flow is implemented (in PRMSSoilzone) and tested for sagehen_5yr.
- Control instances have a diff method to compare with other instances.
- Feature to standardize subsetting input data (parameters and forcings) in space and time either from file (
utils.netcdf_utils.subset_netcdf_file
) or in memory (utils.netcdf_utils.subset_xr
).
Breaking Changes
- pref_flow_infil_frac now a required parameter input for PRMSSoilzone. The NHM values assumed previously are zeros on all HRUs.
Bug fixes
- Fixed calculation of the variable transp_on was incorrectly calculated in certain situations not covered by NHM configuratons but covered by sagehen_5yr.
- Fixed calculation of variable dprst_area_open which was not being checked but was affecting no other variables.
- The variable pptmix was incorrectly calculated in certain situations not covered by the NHM configurations.
Internal changes
- Testing system refactor to handle pairs of domains and control files allowing much more flexibility in configuration/control testing.
- New testing domain "sagehen_5yr" is added to test_data directory with configuration sagehen_no_cascades. This domain introduces multiple PRMS capabilities (noted indvidually in this PR) not used in the NHM configuration and provides a test for these.
- Tests are now marked as "domain" or "domainless" to avoid redundant runs of domainless tests across test domains.
- New tests test_prms_above_snow and test_prms_below_snow replace test_model and are extremely close to PRMS (PRMSSolarGeometry: 1.0e-8, PRMSAtmosphere: 1.0e-5, PRMSCanopy: 1.0e-6, PRMSRunoff: 1.0e-8, PRMSRunoffNoDprst: 1.0e-8, PRMSSoilzone: 1.0e-8, PRMSSoilzoneNoDprst: 1.0e-8, PRMSGroundwater: 1.0e-8, PRMSGroundwaterNoDprst: 1.0e-8, PRMSChannel: 5.0e-7) for all test domains.
- Migration to Numpy 2.0+.