Still under major development: At the moment many placeholders, and very basic functionality (proof of concept).

Overview/Goals of rmb

rmb (rapid model builder) is a Python package designed to streamline the creation and manipulation of groundwater model input files. It does this by leveraging a user-defined YAML input file and a set of backend templates.

• Generic to any model software (provided a template file exists)
• Rapid model framework: YAML input file
• Visualize and capture pre-processing pipelines through to model file creation in a single place
• Uses network graph models as backend to solve for dependancy (similar to snakemake)
• Hopefully future integration with data versioning CICD (ie. DVC)

Core Functionality

Model Template Selection: The input file specifies a model_type, which rmb uses to pull the appropriate backend templates.

Graph-Based Execution: Using the input and backend dependencies, rmb constructs a directed graph where nodes represent processing modules (e.g. creating arrays, applying spatial data) and edges represent dependencies between them.

Model Building: Traversing this graph, rmb generates the required model input files for engines like MODFLOW or SWAT.

Simulation Object (sim): Allows investigation of intermediate or gridded data. Supports plotting and diagnostics.

Enables modular edits: if a node is changed (e.g. a boundary condition or array), only downstream dependent nodes are recomputed.

Zoom in showing the pipeline nodes for ghb build.

Usage

input_yaml = r"examples\simple_freyburg\freyburg_1lyr_stress.yaml"

sim = create_simulation(input_yaml)

#visualize model
sim.graph.plot() # all nodes in the model (including template, pipeline, default nodes)
sim.graph.plot(subgraph=True) # this is only the nodes which are built (ie. upstream of the module nodes)

sim.build() # resolves all the data for the nodes upstream of module nodes (ie. runs the pipelines)

#view module data
dis = sim.nodes['module.dis'].data
dis.top # view top input data for flopy

#view pipeline/pipe specific data
sim.nodes['pipeline.ghb.stress_period_data'].data

sim.write() # writes the simulation files
sim.nodes['module.sim'].data.run_simulation() # you can run the model

Nodes

Types of nodes/Edges

• Input nodes:
  • User input: from the user config file (ie. yamls)
  • Template: Specific to the template model files (ie. mf6, swat). related to module nodes. This is also the 'output' of rmb, but the input to the model.
• Module nodes: Derived based on core inputs in user input file and backend model type template files (ie. generic, mf6, swat).
• Mesh nodes: Derived based on core inputs in user input file. Represents spatial discretization.
• Pipeline nodes: Represents operational processes. Will have input and output nodes. Maybe this is a type of edge?

Node Naming Convention

Node Type	Description	Suggested Naming Prefix	Example ID
input	From frontend YAML; scope is the dict path	input.<usr_key_path>.<param>.<hash>	input.mesh.resolution input.module.gwf.modelname
template	Specific to backend model template; cmd kwargs in module inputs	template.<mtype>.<module>.<param>	template.mf6.drn.stress_period_data
module	Logical building block from templates	module.<kind>.<usrname>	module.sfr.mysfr
mesh	Spatial discretization	mesh.<dimension/element>	mesh.grid
pipeline	Operation/process node	pipeline.<name>	pipeline.interpolate_rainfall