Solvers#

Note

Use this page when the question is: “Which process do I need, which backend implements it, and which numerical option matters before I trust the result?”

HydroModPy is a process-first modelling stack. Pick the physical or operational process to execute, pick one or more solvers for that process, and let the planner expand the request into concrete process/solver runs. The same backend can play different roles depending on context: modflow6 is a flow solver when used as flow/modflow6, modflow6gwt is a transport solver when used as transport/modflow6gwt.

Pick a flow solver#

HydroModPy currently exposes three flow solvers. They are not interchangeable: each one constrains the mesh family, the available packages, and the calibration budget.

Solver

Best fit

Mesh support

Notes

modflow_nwt

Legacy MODFLOW-family flow

Structured sgrid only

Continuity with historical studies; gateway to the MODPATH and MT3DMS ecosystem.

modflow6

Modern MODFLOW-family flow

Structured sgrid and runtime DISV-style irregular meshes

Preferred when irregular meshes, modern package semantics, or MODFLOW 6 GWT transport coupling matter.

boussinesq

In-house shallow-groundwater flow

Triangular runtime meshes from the Gmsh pipeline

Useful for solver-to-solver comparisons and explicit Boussinesq formulations; still under active validation.

The active backend is declared in [solver]. Backend-specific options live under [modflow6] or [modflownwt]. Process binding remains in [[simulation.process]].

[solver]
backend = { backend = "modflow6" }

[modflow6.runtime]
xt3d = true

[[simulation.process]]
id = "flow_main"
type = "flow"
solvers = ["modflow6"]

Process families#

The planner reads each [[simulation.process]] block, binds it to one or more solvers, and orders dependent processes (transport after flow). Four process families exist today.

Process type

Role

Current solver names

Status

flow

Groundwater-flow simulation. Produces heads, storage changes, boundary exchanges, and flow-budget outputs.

modflow_nwt, modflow6, boussinesq

Active numerical process.

transport

Particle tracking or concentration transport driven by a previous flow run.

modpath, mt3dms, modflow6gwt

Active numerical process, with explicit upstream flow requirements.

postprocess

Derive secondary products after a simulation run.

timeseries, netcdf

Registry stubs today; intended extension point.

display

Generate presentation outputs from stored results.

flow, transport

Registry stubs today; intended extension point.

Minimal flow plan:

[simulation.time]
start_datetime = "2000-01-01"
end_datetime = "2000-12-31"
step_value = "1 month"

[[simulation.process]]
id = "flow_main"
type = "flow"
solvers = ["modflow6"]

Transport solvers#

Use a transport process when the question depends on a previously computed flow field: travel time, particle paths, concentration advection, dispersion, or decay.

Solver

Transport type

Requires

Notes

modpath

Particle tracking.

Earlier flow/modflow_nwt run.

Uses the MODFLOW-NWT flow model as its velocity source.

mt3dms

Concentration transport.

Earlier flow/modflow_nwt run.

Uses MT3DMS-style species, dispersivity, diffusion, and decay parameters.

modflow6gwt

Concentration transport.

Earlier flow/modflow6 run.

MODFLOW 6 GWT route aligned with a MODFLOW 6 GWF flow model.

modflow6prt

Particle tracking.

Earlier flow/modflow6 run.

MODFLOW 6 PRT route aligned with a MODFLOW 6 GWF flow model. It writes PRT track output, which HydroModPy ingests as pathlines/ arrays.

The planner does not reorder processes. Declare the upstream flow process before the downstream transport process so dependency resolution can bind the transport adapter to the correct flow model.

[[simulation.process]]
id = "flow_main"
type = "flow"
solvers = ["modflow_nwt"]

[[simulation.process]]
id = "transport_main"
type = "transport"
solvers = ["modpath", "mt3dms"]

[[simulation.process]]
id = "flow_main"
type = "flow"
solvers = ["modflow6"]

[[simulation.process]]
id = "transport_main"
type = "transport"
solvers = ["modflow6gwt"]

[[simulation.process]]
id = "flow_main"
type = "flow"
solvers = ["modflow6"]

[[simulation.process]]
id = "prt_main"
type = "transport"
solvers = ["modflow6_prt"]

[transport.modflow6prt.parameters]
release_zone = "upstream_nonriver"
track_dir = "forward"
track_time_step_days = 10.0
stop_time_days = 3650.0

Solver-specific transport parameters#

Transport parameters live under [transport.<solver>.parameters]. This keeps process-wide orchestration separate from solver-specific numerical options.

[transport.modpath.parameters]
zone_partic = "domain"
track_dir = "forward"
cell_div = 2

[transport.mt3dms.parameters]
spc_name = "NO3"
sconc_init = 0.0
sconc_input = 30.0
disp_long = 10.0
rate_decay = 0.0

[transport.modflow6gwt.parameters]
spc_name = "NO3"
sconc_init = 0.0
sconc_input = 30.0
disp_long = 10.0

[transport.modflow6prt.parameters]
release_zone = "upstream_nonriver"
porosity = 0.01
release_times_days = [0.0]
track_time_step_days = 10.0
stop_time_days = 3650.0

Generalized categories#

The same structure can cover future processes without changing the mental model.

Category

Example process types

Example solvers or adapters

Expected dependency style

Physical PDE processes

flow, transport, future heat or reactive_transport

Numerical engines.

Often depend on mesh, domain, forcing, and sometimes earlier process outputs.

Hydrological or forcing processes

Future recharge or surface_runoff

Forcing builders or hydrological models.

Usually feed flow rather than consume flow outputs.

Analysis processes

postprocess, future metrics or uncertainty

Catalog readers, metric builders, exporters.

Usually depend on completed simulation outputs.

Presentation processes

display, future report

Figure and report generators.

Usually depend on catalog data and derived analysis outputs.

Practical selection rules#

  • Start from the question, not from the backend name.

  • Use flow when the output of interest is the water table or flow budget.

  • Add transport only when the transport result needs a previously solved flow model.

  • Keep each process block focused: process type declares the modelling task; solver name declares the implementation.

  • Prefer the scientific capability matrix when comparing assumptions, mesh support, and validation status.

Decision matrix#

Question

Best entry point

What can each solver family represent?

Solver Capability Matrix

How do MODFLOW 6 and MODFLOW-NWT differ scientifically?

MODFLOW 6 Versus MODFLOW-NWT: Scientific Comparison

Why is XT3D important on irregular meshes?

XT3D On Irregular DISV Meshes

Where are the analytical and semi-analytical validation cases?

Validation Benchmarks

Where are stable solver-to-solver comparison cases?

Simulation Comparison

Where is the solver package architecture documented?

Solver Architecture

Read more#

Theory

Solver families, governing equations, package semantics, and cross-cutting numerical notes.

Solvers
Validation

Analytical and semi-analytical comparisons rendered as reproducible teaching figures.

Validation Benchmarks
Comparison

Shared-case studies that quantify solver, mesh, or option differences.

Comparison Workflow
Architecture

Package boundaries, adapter contracts, and runtime handoff.

Solver Architecture

See also#