Logger

The Parquet state logger (cropforge/logger.py) serialises the full simulation state at every timestep.

---

Output Location

Parquet data is written to:

cropforge_output/<session_name>/

Where session_name is <farm_name>_<UTC_timestamp>, e.g. WheatBasic_20260626T183000.

---

Table Schemas

The logger writes three partitioned Parquet tables:

plants/

One row per plant per day per field.

Column	Type	Description
day	int32	Simulation day.
field_name	string	Field identifier.
plant_id	string	e.g. "r00c03".
row, col	int16	Grid position.
age_days	int16	Days since sowing.
lai	float32	Leaf Area Index.
biomass_g	float32	Dry biomass (g).
height_cm	float32	Canopy height (cm).
root_depth_cm	float32	Root depth (cm).
stress_index	float32	Cumulative stress (0–1).
alive	bool	Plant viability.
phenological_stage	string	Growth stage label.
custom_json	string	JSON-serialised PlantState.custom dict.

soil/

One row per soil voxel (row × col × layer) per day per field.

Column	Type	Description
day	int32	Simulation day.
field_name	string	Field identifier.
row, col	int16	Grid position.
layer	int8	Layer index.
depth_top_cm, depth_bottom_cm	float32	Layer boundaries (cm).
moisture_pct	float32	Volumetric water content (%).
nitrogen_kg_ha	float32	Available N (kg ha⁻¹).
bulk_density	float32	Bulk density (g cm⁻³).
penetration_resistance	float32	Soil strength (MPa).
custom_json	string	JSON-serialised custom dict.

environment/

One row per field per day.

Column	Type	Description
day	int32	Simulation day.
field_name	string	Field identifier.
doy	int16	Day of year.
temp_max_c, temp_min_c, temp_mean_c	float32	Temperature (°C).
radiation_mj_m2	float32	Solar radiation (MJ m⁻²).
rainfall_mm	float32	Rainfall (mm).
et0_mm	float32	Reference ET (mm).
wind_speed_ms	float32	Wind speed (m s⁻¹).
humidity_pct	float32	Relative humidity (%).
co2_ppm	float32	CO₂ concentration (ppm).
events_fired	string	JSON array of event names that fired this day.
custom_json	string	JSON-serialised custom dict.

---

Reading Logs Externally

import pandas as pd

plants = pd.read_parquet("cropforge_output/MyFarm_20260626T120000/plants/")
soil   = pd.read_parquet("cropforge_output/MyFarm_20260626T120000/soil/")
env    = pd.read_parquet("cropforge_output/MyFarm_20260626T120000/environment/")

# Filter to one field and one day
day45_plotA = plants[
    (plants["field_name"] == "Plot_A") & (plants["day"] == 45)
]

Partitioning by field_name and day means queries on a single day/field are very fast — PyArrow only reads the relevant partition files.