What’s New

What’s New#

v3.1.0 (Unreleased)#

New Features#

  • A migration guide for updating projects from pywatershed 2.x to 3.x, version_migration_guides/v2_to_v3.md: the verified, complete list of breaking changes with scans, fixes, and symptom lookup, written to be followed by a person or handed whole to an AI assistant. Retroactive additions to the v3.0.0 Breaking Changes section below record five changes it documents that were found undocumented. (PR416) By James McCreight.

  • HRU cascading flow (Hortonian surface runoff and soilzone interflow / Dunnian flow) following PRMS cascade_flag=1 with cascadegw_flag=0: new process classes PRMSRunoffCascadesNoDprst and PRMSSoilzoneCascadesNoDprst, cascade parameter preprocessing from PRMS parameter files (preprocess_cascade_params()), and support for inactive HRUs via base.ActiveHruMixin (which HRUs are active is always derived from the hru_type parameter; results at inactive HRUs are masked to nan and excluded from mass-balance checks by the new active_mask capability of base.Budget). Verified against PRMS 5.2.1 on the sagehen_5yr (sagehen_no_gw_cascades) and new gridded sagehen_gridded_5yr (5609 cells with inactive cells) test domains, both tested in CI on all platforms. Demonstrated in the new notebook examples/11_cascading_flow.ipynb. (PR407) By James McCreight.

  • analysis.ProcessPlot reads geodatabase layers (hru_layer/seg_layer name the layer within hru_shp_file_name /seg_shp_file_name), sizes each map to the domain’s extent and sets the initial view to it, and ProcessPlot.plot_hru_var() maps a (time, nhru) array as one frame per time with a time widget (clim holds the color scale across frames; time may be omitted for a DataArray with a time coordinate). A layer whose CRS has no EPSG code raises; a layer declaring no CRS is taken as EPSG:5070. Shown in examples/11_cascading_flow.ipynb. (PR407, PR423) By James McCreight.

  • The gridded sagehen_gridded_5yr test domain generates its own CBH forcing files with PRMS from its two-station data file (the 5609-cell text files are ~600 MB, too large to distribute), making it fully reproducible from a clean clone. For this, the CBH writer in prms_src (write_climate_hru.f90) is patched to write full precision (G0 format in place of E12.4), so the forcing pywatershed reads carries the values PRMS computed rather than four-digit roundings. (PR407) By James McCreight.

  • Reading a PRMS parameter file expands the compact forms PRMS allows to the full shape pywatershed expects: a parameter given on nmonth alone is broadcast to (nmonth, nhru), and tmax_adj and tmin_adj given on nhru alone are broadcast to (nmonth, nhru) (the sagehen_gridded_5yr parameter file uses both). A monthly parameter is recognized by its declared nmonth dimension, not by having 12 values. A shape not handled passes through with the dims the file declares. (PR407, PR423) By James McCreight.

  • separate_domain_params_dis_to_ncdf() takes an optional control; when its cascade_flag is set the cascade parameters are derived before separation so the cascade process classes get complete parameter files. A write_dis switch skips the discretization files when only process files are wanted. The script test_data/generate/separate_params_to_nc.py replaces the prms_parameter_discretization_separation notebook as the single way the test domains’ parameters_*.nc files are written. (PR407) By James McCreight.

  • Restart capability (restart_read, restart_write, restart_write_freq) for PRMSSoilzoneNoDprst, which previously raised NotImplementedError, and for the new cascade classes PRMSRunoffCascadesNoDprst and PRMSSoilzoneCascadesNoDprst (which also gain input_aliases). The soilzone restart state, also for PRMSSoilzone, now includes pref_flow_stor, which is live wherever pref_flow_den > 0 (previously omitted, so such restarts were not exact; the NHM test domains have pref_flow_den = 0). The process-level perfect-restart test, now autotest/test_restart_processes.py (was test_nhm_restart.py), picks the process classes from the control file and its dates from the simulation period, so it runs on the sagehen domains and the no_dprst controls too. That exposed a second omission: PRMSSnow now saves PRMS 5.2.1’s full snowcomp restart state (adds int_alb, salb, lst, iso, mso, lso, albedo, pk_temp, snsv), whose omission made restarts of a depleting snowpack inexact. (PR407) By James McCreight.

Breaking Changes#

  • PRMSCanopy takes the input pkwater_ante in place of the inputs pk_ice_prev and freeh2o_prev, and PRMSSnow declares the variable pkwater_ante to supply it. Code which supplies PRMSCanopy its inputs individually must be updated; models assembled from process lists or model dictionaries are unaffected. (PR414) By James McCreight.

  • hru_type is now a declared parameter of PRMSAtmosphere, PRMSAtmosphereTranspFrost, PRMSAtmosphereTranspFrostDynamic, PRMSSolarGeometry, PRMSCanopy, PRMSGroundwater and PRMSGroundwaterNoDprst, which use it (through base.ActiveHruMixin) to identify inactive HRUs. A Parameters object built by hand for one of these processes must now include hru_type; PRMS parameter files and the parameters_dis_hru.nc discretization file written by separate_domain_params_dis_to_ncdf() already carry it. As a consequence, PRMSCanopy now applies PRMS’s lake treatment to HRUs with hru_type lake (no interception), where it previously treated every HRU as land; results change on domains with lake HRUs. Note that lake HRUs are currently untested in pywatershed: no test domain contains one. (PR407) By James McCreight.

  • Keyword arguments were inserted mid-signature: stream_seg_in=None now precedes dprst_flag in PRMSSoilzone, and active_mask=None precedes unit_desc in base.Budget. Code passing those or any later arguments positionally must switch to keywords. (PR407, PR423) By James McCreight.

  • Every process is named for its class: Process.__init__ sets self.name = type(self).__name__ and no subclass assigns name itself. The subclasses’ own assignments all matched their class names already, except that PRMSAtmosphereTranspFrost and PRMSAtmosphereTranspFrostDynamic inherited their parent’s "PRMSAtmosphere"; their single-file netCDF output (when netcdf_output_separate_files is off) and budget descriptions now carry their own class names. (PR423) By James McCreight.

  • The variable metadata entry hru_hortn_cascflow (PRMS’s name, declared by no process) is replaced by hru_horton_cascflow, the variable PRMSRunoffCascadesNoDprst declares; the unused strm_seg_in entry is removed in favor of stream_seg_in, declared by both cascade classes. PRMS output files are renamed on conversion to netCDF. (PR407, PR423) By James McCreight.

Bug fixes#

  • NetCdfWrite silently dropped a string coordinate whose maximum length matched an earlier string coordinate’s (the two share a char<N> dimension, and the guard against creating that dimension twice also skipped creating the variable). In FlowGraph output node_maker_id went missing whenever its ids were as long as the node_maker_name values. Reported in GH421. (PR422) By James McCreight.

  • Loading a parameter netCDF file with netCDF4 (the default for Parameters.from_netcdf()) dropped a coordinate that no data variable uses (recorded in the file’s global coordinates attribute), so a process declaring a dimension none of its parameters use, e.g. nsegment for PRMSRunoffCascadesNoDprst, could not write netCDF output on that dimension. (PR407) By James McCreight.

  • PRMSCanopy gates rain interception by grasses on the antecedent snowpack, pkwater_ante, as PRMS does (intcp.f90:416, which tests Pkwater_equiv before snowcomp updates it for the timestep; snowcomp.f90:946 publishes that value as Pkwater_ante). pywatershed instead reconstructed the pack as pk_ice_prev + freeh2o_prev. PRMS carries Pkwater_equiv as its own accumulated double rather than recomputing it as that sum, and in the vanishing tail of a melting pack the two differ by orders of magnitude and fall on opposite sides of the dnearzero (2.23e-16) threshold, flipping whether a day’s rain is intercepted. PRMSSnow now declares pkwater_ante, as snowcomp does (snowcomp.f90:346-349), so coupled models supply it. (PR414) By James McCreight.

  • Model reports both candidate filenames when an input file is missing. Input discovery looks for <variable>.nc in the input directory and, failing that, retries the same name as <variable>.param; only the second name reached the error, so a missing netCDF input was reported as a not-found dynamic parameter file. Both lookups are now checked up front and a missing input raises naming both candidates. (PR407) By James McCreight.

  • Integer variables written by dd_to_nc4_ds() (and so by DatasetDict and separate_nhm_params()) no longer carry a default _FillValue of -9999, which made xarray promote them to float on read and turn legitimate -9999 values into NaN. The in-memory fill used to mask inactive HRUs is now the separate mask_fill_values_dict. (PR407) By James McCreight.

  • ProcessPlot.plot_hru_var() no longer errors on a var_name that contains a known variable or parameter name as a substring (the metadata lookup passed the string where a list of names is expected, so "sroff (cascade - no cascade)" matched sroff and then failed to find itself). (PR407) By James McCreight.

  • NetCDF output no longer floods the console with NumPy’s Setting the shape on a NumPy array has been deprecated warning (one per variable per timestep with netCDF4 <= 1.7.4 and NumPy >= 2.5). The suppression is applied around each write in code, not only as a pytest filter, because flopy sets warnings.simplefilter("always", DeprecationWarning) at import and overrides any earlier filter; see MAINTENANCE.md for when it can be removed. (PR419) By James McCreight.

Internal changes#

  • The public API surface (exports, __init__ signatures, declared input/variable/parameter name sets, control options, and variable/parameter metadata) is snapshotted in autotest/api_surface.txt, generated by autotest/api_surface.py and checked by autotest/test_api_surface.py (domainless). A surface change fails CI until the baseline is regenerated with the script’s --write option — and, if any existing line changed or was removed, documented under Breaking Changes. (PR416) By James McCreight.

  • Retire the last intel-built artifact, the GSFLOW binary for MacOS, which was x86_64 and required Rosetta 2 on Apple Silicon. All three checked-in GSFLOW 2.4.0 binaries (macOS arm64, Linux, Windows) are now gfortran double-precision builds from a single GSFLOW source commit and CI run, with provenance recorded in bin/README.md for the first time. (PR415) By James McCreight.

  • Quiet the test suite’s NumPy >= 2.5 deprecation warnings, which numbered tens of thousands per run. constants.nat is now np.datetime64("NaT", "ns"), the generic form being deprecated and ns being what its only use already cast to; environment.yml requires holoviews >=1.23.0, the release which dropped the nat_as_integer the warning came from; and autotest/pytest.ini ignores the ndarray.shape assignment that netCDF4 <= 1.7.4 makes on every variable write, which is fixed upstream by netcdf4-python PR #1469 but unreleased and so is tracked in MAINTENANCE.md. (PR414) By James McCreight.

  • Retire ifort and Intel MacOS. The PRMS binaries are no longer checked in to bin/; they are compiled from prms_src/ with gfortran (supplied by environment.yml) the first time they are needed, by pywatershed.utils.compile_prms(), which ci.yaml, autotest/ci_local.sh, and the test-data fixtures all call. The prms_src makelists gained -std=gnu17 for gcc, without which PRMS 5.2.1’s C sources fail to build under the C23 default of gcc 15. Apple Silicon binaries are tagged mac_arm rather than m1, and Intel MacOS is no longer detected. GSFLOW binaries remain checked in because their source is not part of this repository. (PR414) By James McCreight.

  • Lint Jupyter notebooks. [tool.ruff] include covered only *.py, so ruff check . and CI never saw notebooks, while the pre-commit hook passed *.ipynb paths explicitly (which overrides include) and did. Notebook problems were therefore invisible until someone edited a notebook and was met with errors they had not caused. Notebooks are now in include, with E501 and I001 exempted: they are narrative, and several import a module purely for its side effect (hvplot.xarray registers a .hvplot accessor) and must do so after the module they extend, which sorting undoes. Notebooks are linted but not auto-formatted, for the same line-length reason. This surfaced four broken cells that no test covered: two stray ), a string opened with " and closed with ', and a use of pl.Path with no import pathlib. Three notebooks needed further cleanup once the syntax errors stopped masking it: unused imports, unused assignments, a semicolon-joined statement, and bare except clauses narrowed to except AssertionError. Also repairs four malformed # noqa directives (including a # noaq typo) that ruff silently ignored, which meant the unused-import fixer would have deleted the hvplot imports they were meant to protect. (PR412) By James McCreight.

  • Add MAINTENANCE.md, a ledger of maintenance todos blocked on external events (dependency releases, cross-repo work), each with a mechanically checkable unblock condition; the /maintenance Claude skill checks them live and reports what is actionable. (PR409) By James McCreight.

  • Reduce CI footprint with a skeleton/full split: pushes to any branch (in this repository or on forks) run a skeleton — installs, linting, domainless tests, the docs build, and example notebooks on ubuntu only — while the full suite (domain test jobs, all platforms) runs for pull requests (including drafts), pushes to develop/main, and workflow_dispatch. Domain jobs can be opted in on branch pushes with a ci-<token> (e.g. ci-fgr, ci-all) in the branch name or head commit message — see DEVELOPER.md. concurrency groups cancel in-flight runs superseded by a newer push on the same non-mainline ref. (PR408) By James McCreight.

  • PRMSCanopy, PRMSSnow, PRMSRunoff and PRMSSoilzone (and their no-dprst and cascade subclasses) loop over active HRUs only, the cascade classes in routing order; PRMSAtmosphere, PRMSSolarGeometry and PRMSGroundwater compute every HRU and mask the inactive ones at initialization, as does PRMSRunoffAg, which loops over all HRUs. All-active domains are unaffected. (PR407, PR423) By James McCreight.

  • compile_prms() prefers the active Python environment’s gfortran/gcc (e.g. conda-forge’s) over ones earlier on the PATH, and fails up front naming whichever of make, gcc and gfortran is missing. (PR407) By James McCreight.

  • Require pyPRMS >=0.10.0 and remove the temporary packaging <26.3 pin it supersedes (pyPRMS 0.9.10 crashed on import of metadata with packaging >=26.3). Also remove calls to pyPRMS methods deprecated in 0.10.0: Parameters.adjust_bounded_parameters() in domain subsetting and DataFile.data_by_variable() in the obsin flow node test. (PR406) By James McCreight.

v3.0.0 (13 July 2026)#

A migration guide for updating a project from v2.x, written to be followed by a person or handed to an AI assistant, is at version_migration_guides/v2_to_v3.md in the repository.

New Features#

  • The new staticmethod Model.solve_inputs() determines where each process input comes from — another process or a file — from a process list or model dictionary, without instantiating a Model or requiring any files to exist. Useful for determining the file inputs a model configuration requires, e.g. when forcing a sub-model from another model’s outputs. Model construction uses the same implementation internally. (PR396) By James McCreight.

  • The base.ConservativeProcess class now supports both mass and energy budgets. Processes can specify which quantity to budget using the quantity parameter in _set_budget(). The new mass_budget and energy_budget properties provide explicit access to each budget type. The legacy budget property is deprecated and will be removed in the next major release - use mass_budget instead. (PR343) By James McCreight.

  • The new PRMSStreamTemp and PRMSStreamTempHumidityCBH classes provide stream temperature simulation using the PRMS stream temperature methodology, computing water temperatures based on energy balance in stream segments. The latter class accepts time-varying humidity inputs on the HRUs while the former accepts a mean monthly humidity for each segment. The classes support optional energy flux tracking and budgeting via the track_energy_fluxes parameter (default: True). When enabled, it computes and tracks 11 energy flux components including advective heat transport (upstream, lateral, outflow), surface energy exchange (solar radiation, longwave emission/absorption, evaporative cooling, convective exchange), and internal sources (friction heating, groundwater conduction). These fluxes are available as output variables and included in the energy budget. When disabled (track_energy_fluxes=False), energy flux variables are set to None and excluded from NetCDF output, with imbalance_behavior required to be None. The classes PRMSStreamTemp and PRMSStreamTempHumidityCBH take a stream shade class as input on initialization. Two stream shade classes have been implemented, PRMSStreamShadeConstant and PRMSStreamShadeDynamic. The former works based on 3 parameters: summer shade fraction, winter shade fraction, and segment latitude. The latter class computes shade dynamically based on topographic and vegetation parameters using solar geometry calculations. This is the default PRMS behavior when stream_temp_shade_flag = 0 and requires 13 parameters describing topography and vegetation characteristics for each stream segment. The classes PRMSStreamTemp and PRMSStreamTempHumidityCBH also require one of PRMSHydraulicGeometryFull or PRMSHydraulicGeometryWidthOnly as an upstream process to provide hydraulic geometry variables needed for energy balance calculations. PRMSHydraulicGeometryFull computes flow-dependent hydraulic geometry (width, depth, area, velocity) using power-law relationships when all parameters are provided, while PRMSHydraulicGeometryWidthOnly uses PRMS default values for depth parameters (depth_alpha=0.27, depth_m=0.39) when they are missing from the parameter file, matching PRMS 5.2.1 behavior. These capabilities are demonstrated in notebooks examples/01_multi-process_models.ipynb and examples/02_prms_legacy_models.ipynb as part of the NHM configuration in pywatershed. (PR343) By James McCreight.

  • Option for Model class to read from a single netcdf file or (not and, the existing option,) from a directory containing multiple netcdf files. (PR333) By James McCreight.

  • A new SourceSinkFlowNode class adds or removes flow above some minimum flow parameter as specified by an input data file. (PR327) By James McCreight.

  • The Control class has new method edit_init_start_times to manage changing these times. (PR335) By James McCreight.

  • The FlowGraph class has new method plot to show an abstract plot of the FlowGraph. (PR351) By James McCreight.

  • The base.Process class and subclasses have a new restart capability. See notebook examples/08_restart_streamflow.ipynb for examples. (PR349, PR362) By James McCreight.

  • The PRMSAtmosphereTranspFrost implements the transp_frost module of PRMS. (PR354) By James McCreight.

  • The PRMSAtmosphereTranspFrostDynamic extends PRMSAtmosphereTranspFrost to accept dynamic (time-varying) fall_frost and spring_frost dates from PRMS dynamic parameter files, reproducing PRMS/GSFLOW runs with dyn_fallfrost_flag and/or dyn_springfrost_flag set. (PR392) By James McCreight.

  • The load() method of parameters.PrmsParameters now supports reading multiple parameter files which are treated as addenda to the first parameter file in the list which contains the dimension information. (PR354) By James McCreight.

  • The StarfitSourceSinkFlowNode allows sources and sinks to interact with storage of a Starfit reservoir/FlowNode. (PR348) By James McCreight.

  • The new Output class provides flexible output collection and statistical analysis for models, supporting HRUs of interest (HOI), segments/nodes of interest (NOI), and monthly accumulations. Includes Zarr chunked output capability for efficient large-scale data writing (~6x faster than NetCDF). See notebook examples/09_model_output.ipynb for examples. (PR363) By James McCreight.

  • New agricultural water use classes enable simulation of irrigated agriculture based on GSFLOW. PRMSRunoffAg extends PRMSRunoff to calculate infiltration separately for pervious and agricultural areas. PRMSSoilzoneAgObsET provides dual-area soil moisture accounting with iterative adjustment of irrigation to match observed actual ET. PRMSSoilzoneAg is a simplified version without the observed ET iteration, suitable when ET observations are not available. See notebook examples/10_ag_irrigation_use.ipynb for examples. (PR362) By James McCreight.

  • Add pre-commit hook to run security review on staged files or on entire repository, checks for: 1. Absolute paths, 2. IP addresses, 3. Internal server hostnames, and 4. Usernames/passwords or credentials. See .github/scripts/check_security.py. (PR384) By James McCreight.

  • Bug fixes for PRMS 5.2.1.1: 1) errant code skipped humidity CBH files entirely when they were selected to be used, 2) code deletion resulted in seg_humid not being zeroed each timestep and erroneously accumulating. Both fixes are extensively documented. The PRMS code was modified (compared to the released 5.2.1.1) and the pywatershed code was made to match PRMS stream temperature when using humidity inputs from 1. CBH, 2. scalar parameter, and 3. monthly spatially distributed parameters. (PR386) By James McCreight.

  • Add a weekly security scan using Safety CLI in a GitHub Actions workflow (.github/workflows/security_check.yaml): checks conda-installed packages, pip-installed packages, and pyproject.toml dependencies separately. (PR387) By James McCreight.

Breaking Changes#

  • The budget_type parameter has been renamed to imbalance_behavior in base.ConservativeProcess and all its subclasses, in base.FlowGraph, and in control options. Update all budget_type references to imbalance_behavior in your code and configuration files. This breaking change clarifies what the parameter does and is intentionally distinct from the budget quantity parameter. (PR343) By James McCreight.

  • Budget netcdf output filenames have changed to include the quantity type. Mass budgets are now named ProcessName_mass_budget.nc instead of ProcessName_budget.nc. Energy budgets use ProcessName_energy_budget.nc. (PR343) By James McCreight.

  • The variable seg_width was renamed seg_flow_width. It is the only variable removed from the metadata in this release; code selecting it for output, or reading it from output files, must use the new name. (PR343) By James McCreight.

  • PRMSAtmosphere no longer overrides get_variables() and so inherits base.Process’s implementation, which returns the keys of get_init_values(). The result is a list of 15 names where it was a tuple of 14; tmax_sum is now included. Code which iterates the result, to select output variables for example, gets a variable it did not get before and raises no error. (PR349) By James McCreight.

  • The method PRMSAtmosphere.calculate_transp_tindex was renamed calc_transp_tindex. (PR354) By James McCreight.

  • FlowNodeMaker.get_node takes self as its first argument; the base class previously declared get_node(control, index) without it. The base.FlowNodeMaker subclasses in pywatershed already supplied self, but custom subclasses which copied the base class signature must add it. (PR348) By James McCreight.

  • Control.load_prms() changed in two ways. The options parameter_file, netcdf_output_dir and streamflow_module are unwrapped from their single-element lists only when exactly one value is present; two or more values are now retained as a list. And keep_unused_options=True no longer forces warn_unused_options to True, so the two are set independently. (PR354) By James McCreight.

Bug fixes#

  • The PRMSRunoffAg mass budget did not balance on days with canopy changeover (intcp_changeover > 0, at canopy density transitions), for two reasons: the budget input term intcp_changeover_budget was assigned only in dead code and remained zero, and depression storage inflow was missing changeover water when changeover is carried separately from net_rain (intcp_changeover_in_net_rain=False). GSFLOW carries changeover inside net_rain, whereby its depression storage receives this water. With the fix the budget closes to machine precision. (PR401) By James McCreight.

  • utils.PrmsDynamicParameter daily_data_array left fill values for days between daily_start_date and the first dynamic parameter date inside that window, instead of forward-filling from the most recent date at or before daily_start_date as PRMS applies dynamic updates. Only runs starting between dynamic parameter dates were affected. (PR393) By James McCreight.

  • PRMS 5.2.1.1 had a bug in stream temperature where division by HRU area was repeated multiple times. In the old code this occurred in routing.f90 on lines 764 and 765 and then again on 789 and 790, where seginc_swrad and seginc_potet were divided despite this having already occurred on lines 744 and 745. Comments regarding the fix are found on lines 764 and 793 in the fixed code. (PR383) By James McCreight.

Internal changes#

  • The base.ConservativeProcess class now uses _mass_budget and _energy_budget attributes internally instead of budget. The budget property remains as a deprecated alias for _mass_budget for backward compatibility.

  • Release procedures were revamped: .github/RELEASE.md rewritten as a concrete step-by-step guide with a running example, guarded release automation jobs (checks, package build/publish, frozen conda environment exports per platform), a preflight script shared between local use and CI, and a version-consistency test. (PR395) By James McCreight.

  • CI no longer uses fortran-lang/setup-fortran; gfortran is provided by the conda environment. (PR394) By James McCreight.

  • CI and the conda environments temporarily install flopy from its develop branch (with codegen options) to accommodate a flopy API change, until the next flopy release. (PR397) By James McCreight.

  • The autotests now require an explicit --domain option; the silent drb_2yr default was removed. (PR393) By James McCreight.

  • autotest/ci_local.sh keeps previously downloaded mf6 binaries on PATH so mf6-dependent tests run even when the modflow section is skipped. (PR398) By James McCreight.

  • Refactor of test_data/generate/convert_prms_output_to_nc.py to put final variables into a separate file to run by pytests both after all other variables are generated and so the final variables are run serially. (PR331) By James McCreight.

v2.0.4 (23 February 2026)#

New Features#

Fixes to release workflow, pypi publishing.

v2.0.3 (22 February 2026)#

New Features#

Some minor fixes. This is a data release for the upcoming major release, new data will be an asset on this release.

v2.0.2 (14 March 2025)#

Bug fixes#

  • Fixed setup.py to allow editable installs, keeping up with changes in the pythonverse. Deprecated all fortran code built and interfaced using f2py as it was not popular and had only maybe very slight speed advantages compared to numba. This was not considered a breaking change because there are redundant alternatives to the fortran. (PR331) By James McCreight.

v2.0.1 (19 December 2024)#

New Features#

  • Corrected disclaimer on top-level README.md. Other minor fixes not to code base (CI, envs, etc).

v2.0.0 (16 December 2024)#

New Features#

Breaking Changes#

  • pref_flow_infil_frac now a required parameter input for PRMSSoilzone. The NHM values assumed previously are zeros on all HRUs. (PR288) By James McCreight.

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. (PR288) By James McCreight.

  • Fixed calculation of variable dprst_area_open which was not being checked but was affecting no other variables. (PR288) By James McCreight.

  • The variable pptmix was incorrectly calculated in certain situations not covered by the NHM configurations. (PR288) By James McCreight.

Internal changes#

  • Testing system refactor to handle pairs of domains and control files allowing much more flexibility in configuration/control testing. (PR278) By James McCreight.

  • 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. (PR288) By James McCreight.

  • Tests are now marked as “domain” or “domainless” to avoid redundant runs of domainless tests across test domains. (PR288) By James McCreight.

  • 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. (PR288) By James McCreight.

  • Migration to Numpy 2.0+. (PR310) By James McCreight.

v1.1.0 (25 June 2024)#

New features#

  • Minor enhancement to ensure PRMSSnow hru_deplcrv parameter is integer or coercable. (PR296) By James McCreight.

  • Release assests to include new GIS files and an additional domain to support the upcoming major release. By James McCreight.

v1.0.0 (18 December 2023)#

New features#

Breaking changes#

  • The control.options “netcdf_output_dir”, “netcdf_output_var_names”, and “netcdf_output_separate_files” match the keyword arguments “output_dir”, “output_vars”, and “separate_files” for both process.intitalize_netcdf() and model.initialize_netcdf(). None of these arguments can be supplied in both places (control and method call). It used to be that calling initialize_netcdf() would override what is supplied in control.options but this will now throw an error. The suggestion is to use control.options and not pass arguments to intialize_netcdf(). When using Control.load() (deprecated) or Control.load_prms() from a PRMS control file, note that the “control.options” of “netcdf_output_dir” and “netcdf_output_var_names” are set by values in the PRMS control file. You can edit these, but be aware that they are now set in that load. (PR257) By James McCreight.

Deprecations#

Performance#

Bug fixes#

  • Mass balance fix in PRMS snow for rain on snow followed by evaporation consuming the entire snow pack. (PR248) By James McCreight.

  • Fix mass balance issue in PRMSSnow is also present in PRMS, snow evap is not taken from freeh2o when there is no pk_ice. (PR236) By James McCreight.

  • Resolve issues with different ways of specifying necdf output options. (PR230) By James McCreight.

  • Resolve issues with different ways of specifiying netcdf output options. (PR230) By James McCreight.

  • PRMSSoilzone remove soil_moist_prev because soil_moist is not prognotic and PRMSRunoff was needing it in the advance and not getting the correct value. PRMSRunoff now depends on soil_lower_prev and soil_rechr_prev instead. (PR244) By James McCreight.

Documentation#

Internal changes#

  • New system for generating test_data, by calling generate_test_data.py from autotest/. The system helps autotest know if test data were generated and if they are up to date. (PR253) By James McCreight.

  • Apply pylint and flake8 everywhere as much as possible. (PR251) By James McCreight.

  • Remove diagnostic variables pkwater_equiv_change, pkwater_ante (PR248) By James McCreight.

  • Use v1 instead of main for fortran-lang/setup-fortran. (PR242, PR243) By Wes Bonelli.

  • Refactor test data generation to solve race condition for dependent tests. (PR237) By Wes Bonelli.

  • Refactor tests against PRMS for consistency, flexibility, and thoroughness. (PR244) By James McCreight.

v0.2.1 (19 July 2023)#

Bug fixes#

v0.2.0 (18 July 2023)#

New features#

  • New example notebooks. Moved old notebooks to examples/developer. (PR204) By James McCreight.

  • New way to specify Model instantiation either in-memory or from yaml files using a model dictionary. The approach is loosely based on MODFLOW 6 input organization. See Model documentation. Introduced the concept of discretizations for PRMS, defining “dis_hru” and “dis_seg”. These are components of how model dictionaries are specified. (PR188) By James McCreight.

  • New Control.from_yaml() method. (PR188) By James McCreight.

  • What’s new workflow (behold!) per GH180 and PR181 By James McCreight.

  • Add automatic release workflow to PyPi as per GH178. Associated implementation of gitflow and semver conventions. Overhauled CONTRIBUTING.md, DEVELOPER.md, README.md, and .github/RELEASE.md to document adopted practices. Adoption of git-cliff to generate change logs by filtering comitt messages, see cliff.toml. Clean up of environment files and streamlining against pyproject.toml. Symlink gfortran dylibs to /usr/local/lib on macOS CI so PRMS binaries included in this repo can find them. (PR179) By Wes Bonelli.

Breaking changes#

  • Move Control attribute “config” to “options” for handling global options. (PR188) By James McCreight.

  • Remove arguments from Model initialization. Options pass via control, new set_options() method on Process and ConservativeProcess (PR188) By James McCreight.

  • Control no longer takes a Parameter object as an initialization argument. Process subclasses now require arguments discretization and parameters. The firstargument of Model not a indefinite number of processes, it is now either a list of Process subclasses or a model dictionary (see Model documentation. (PR188) By James McCreight.

Deprecations#

Performance#

  • Introduce ASV performance benchmarks for import and various NHM configurations in pywatershed. (GH170 and PR184) By James McCreight.

Bug fixes#

Documentation#

Internal changes#

  • Introduce precommit hooks: nbstripout, blackdoc, and doctoc. (PR197) By James McCreight.

  • Rename StorageUnit to ConservativeProcess that subclasses from a new Process class that contains most of the StorageUnit functionality. (PR188) By James McCreight.

  • New set_options() method on Process and ConservativeProcess to set their initialization options as ‘_` atrributes. (PR188) By James McCreight.

  • Clean up of how the calc_method option assigns function names to reduce the total amount of code and do it upon initialization. (PR188) By James McCreight.

  • Rename many modules to use lower-snake-case names including those in base/, atmoshpere/, and hydrology/ (PR188) By James McCreight.

  • NHM “self-driving” tests (PR160) By James McCreight.

  • Refactor dependencies for standard pypi installation. (PR164, GH178) By Joseph Hughes.

v0.1.1 (27 April 2023)#

Initial release.