CropForge

CropForge lets you define a crop simulation entirely in Python. You write the model equations; CropForge handles time-stepping, spatial state management, terrain physics, sediment dynamics, Parquet logging, and a WebGL 3D dashboard.

pip install cropforge

What's New in v0.9.5 - Visual Architecture Complete

v0.9.5 completes CropForge's visual architecture arc with First-Party Asset Bundles, Machinery Animation, Stress/Disease Visualizations, enhanced-mode rain particles, PBR rendering, morph targets, and terrain-aware GLB export.

Open-source, code-first virtual farm runtime for agricultural researchers.

What's New in v0.9.0 — Photorealistic PBR Rendering & GLTF Export

v0.9.0 closes the visual arc: the simulation now renders with photorealistic PBR materials, exports full 3D scenes as industry-standard .glb files, and resolves plants to registered GLTF stage models.

Photorealistic PBR Rendering

farm.run(days=90)
farm.visualize(quality="enhanced")   # MeshStandardMaterial, shadows, sun-angle light
farm.visualize()                     # quality="standard" — identical to v0.8.0

Plant Stage Architecture & Model Registry

from cropforge.models import ModelRegistry

ModelRegistry.register(
    species="Triticum aestivum",
    stage=4,
    gltf_path="models/wheat_heading.glb",
)
# Cylinder fallback automatic if no model registered

GLTF Scene Export — Blender / Unreal Engine Ready

farm.run(days=90)
farm.export_scene(day=45, filepath="output/farm_day45.glb")
# pip install cropforge[export]  required

The dashboard also provides a one-click ⬇ Export 3D Scene (.glb) button.

High-Resolution Plotly Terrain

The 3D terrain panel now applies 4× bicubic upsampling (scipy.ndimage.zoom) before display. Physics data is never modified — this is a pure UI enhancement.

What's New in v0.8.0 — Terrain Arc Completion

v0.8.0 closes the terrain arc begun in v0.6.0: the simulation engine now operates at arbitrary sub-metre spatial resolution, sediment moves physically across the landscape, and the 3D renderer stays performant at field scales.

Sub-Metre Resolution

terrain = Terrain.procedural(rows=200, cols=200, resolution_m=0.25)
# All engines (runoff, LS factor, D8 routing, nutrient flux) scale correctly
# resolution_m=1.0 default → zero change to all prior scripts

Sediment Dynamics & Mass Conservation

farm.use_physics(erosion=True, sediment_transport=True)
# Every gram eroded is either deposited downslope or exits the field boundary.
# SoilState gains: sediment_flux_kg_m2, sediment_deposited_kg_m2,
#                  cumulative_sediment_loss_kg_m2, cumulative_deposition_kg_m2

Geomorphological Feedback

The terrain elevation grid updates daily from net sediment flux. Slope and aspect recompute automatically for the next day's D8 routing. Layer 0 topsoil expands and contracts in response.

Advanced Conservation Land Management

from cropforge import TiedRidges, VegetativeFilterStrip

field.set_land_prep(TiedRidges(
    ridge_height_m=0.20, ridge_spacing_m=2.0,
    tie_spacing_m=6.0,   tie_height_m=0.12,
))
# Periodic tie-dams block D8 flow, forming micro-catchments.
# Proved: cumulative erosion < plain RidgeFurrow.

field.set_land_prep(VegetativeFilterStrip(
    strip_width_m=3.0, n_strip_rows=3, position="downslope"
))
# Per-cell roughness=0.95 → 95% reduction in soil detachment at strip rows.

LOD 3D Rendering — 500×500 Fields at 60 fps

The Three.js renderer now chunks terrain into 64×64-cell tiles. Distant chunks automatically downgrade to 1/16th vertex count. For a 500×500 sub-metre field:

Quick Start

from cropforge import Farm, Field, Crop, Soil, Weather, Terrain
from cropforge.plugins import StandardWheat

farm  = Farm(name="MyFarm", location=(28.6, 77.2))
field = Field(name="Plot A", rows=20, cols=30, area_ha=2.4)
field.set_crop(Crop(species="wheat"))
field.set_weather(Weather.from_csv("data/weather.csv"))
field.set_soil(Soil.from_csv("data/soil.csv", apply="uniform"))
field.set_terrain(Terrain.procedural(rows=20, cols=30, resolution_m=0.5))
farm.add_field(field)

field.use_plugin(StandardWheat)
farm.use_physics(
    et0=True, radiation=True, erosion=True, sediment_transport=True,
    slope_radiation_correction=True,
)

farm.run(days=90)
farm.visualize()

See examples/digital_twin_full_lifecycle.py for the complete v0.9.5 capstone example. See examples/photorealistic_twin_trial.py for the v0.9.0 visual export example. See examples/conservation_ag_trial.py and examples/submetre_performance_trial.py for v0.8.0 examples.

What's New in v0.7.0

• Solar Incidence Engine: slope-aspect radiation correction.
• Wind Shadow Engine: terrain-modulated wind fields.
• Clod Dynamics: exponential roughness decay under rainfall.
• Topographical Erosion: RUSLE model on the 3D terrain.

What's New in v0.6.0

• Terrain Engine: procedural, CSV, and GeoTIFF topographies.
• Land Preparation Modifiers: RidgeFurrow, ContourBund, Terrace, DeepTillage, ConservationTillage.
• D8 Hydrology Coupling: lateral water and nutrient routing over terrain.
• 3D Dashboard Modal: WebGL terrain viewer with agronomic variable overlays.

What's New in v0.5.0 – v0.4.0

• Plugin Ecosystem: StandardWheat, StandardMaize.
• Multi-Season Rotations: soil state carries over between runs.
• Compare Dashboard: side-by-side farm comparisons, CSV export.
• Opt-In Physics: FAO-56 ET0, Beer-Lambert radiation, SIR disease spread.

Documentation

Full documentation at cropforge.readthedocs.io.

Licence

MIT — Saswat Sundar Rath, ICAR-IARI Jharkhand, 2026