What problem does it solve? Running the RZWQM2 process-based agricultural model requires deep operational expertise scattered across Fortran binaries, fixed-format input files, and unit conventions, where silent errors (wrong units, wrong paths) produce scientifically invalid results that look successful. ## Core Features & Use Cases - End-to-end staged pipeline: Twelve stages (S0-S10) covering global data acquisition (SoilGrids, HWSD, MSWX, CMFD), site configuration, meteorological file generation, soil property setup, scenario assembly, binary execution, and result parsing. - Diagnostic recovery: 29 symptom-diagnosis-remedy triplets in diagnostics/triplets.yaml map known failure modes (unit traps, path issues, case sensitivity) to validated fixes. - Batch and calibration support: Mass scenario generation from a CSV of sites and automated calibration via LHS/Sobol sampling with NSE/RMSE/PBIAS metrics. - Use Case: An agent simulates maize yield, nitrate leaching, and tile drainage for a site in Iowa by retrieving SoilGrids soil and MSWX weather, assembling a scenario from the validated template, running the Fortran binary, and parsing the .ana output. ## Quick Start Run python preflight_check.py in this directory, then ask the agent to set up and run an RZWQM2 maize simulation for your site coordinates and date range.