---
id: geocif
version: "0.4.908"
license: MIT
license_treatment: permissive
maintenance: active
---

# geocif — Models to visualize and forecast crop conditions and yields

License: permissive · Maintenance: active · Downloads: 103.9K/mo

## What it is and what it does

Geocif is a machine-learning toolkit for agricultural forecasting that ingests Earth observation (EO) data—satellite imagery, climate records, soil data—and derives climatic impact drivers (CIDs) to predict crop yields and monitor field conditions. It wraps a large ecosystem of geospatial (gdal, rasterio, fiona, shapely, pyproj, rtree) and ML libraries (catboost, xgboost, tabpfn, tabicl, ngboost) to automate the pipeline from raw EO extraction through model training to visualization and reporting.

The package is organized around configuration files (geobase.txt, countries.txt, crops.txt, geocif.txt) that define regions, crop calendars, extraction methods, and model settings. It provides runners for data preparation (via geoprepare), index calculation, ML training, agricultural meteorology monitoring, and yield outlook generation. It also includes cropmask optimizers (threshold-based and per-cell) to refine which agricultural areas are used in modeling, and supports multiple model architectures including gradient boosting and prior-fitted networks.

Use it for:

- Build regional crop yield forecasts by training ML models on historical yields and EO-derived climate indices.
- Monitor agricultural conditions and stress across growing seasons using agmet plots and climate metrics.
- Optimize cropland masks for a region by tuning threshold or per-cell inclusion rules to improve model accuracy.
- Generate annual yield outlooks and production reports with maps and accuracy scorecards for policy or market analysis.
- Experiment with different ML architectures (CatBoost, TabPFN, etc.) and feature sets to find the best forecasting model for a crop and region.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Generates climatic impact drivers from Earth observation data and builds machine-learning models to forecast crop yields and monitor agricultural conditions.

Yes, with conditions. Geocif is actively maintained and permissively licensed, making it suitable for agricultural forecasting research and operational monitoring. However, the 57 runtime dependencies—especially heavy geospatial and ML libraries—create substantial install complexity and disk footprint. Install only if you have a specific crop-yield or agricultural-monitoring use case and can manage the geospatial dependency chain (particularly on Linux/macOS). Not suitable for lightweight or embedded deployments.

## Install

pip install geocif
uv add geocif
poetry add geocif

## Installing geocif

Before you install: Low install friction with a wheel distribution, but carries 57 runtime dependencies including heavy geospatial libraries (gdal, rasterio, fiona, rtree) and machine-learning frameworks (catboost, xgboost, tabpfn, tabicl). Active maintenance (released 2 days ago) and requires Python 3.11+.

License in practice: MIT license is permissive—you can use, modify, and distribute geocif freely in commercial and private projects with minimal restriction.

Quickstart:

# Install via uv (recommended)
uv sync

# Configure paths and settings
config_dir = "/path/to/config"
cfg_geocif = [f"{config_dir}/geobase.txt", f"{config_dir}/countries.txt", f"{config_dir}/crops.txt", f"{config_dir}/geocif.txt"]

# Run yield forecasting
from geocif import geocif_runner
geocomcif_runner.run(cfg_geocif)

Requires Python 3.11+. On Windows, uv automatically pulls pre-built geospatial wheels; on Linux/macOS, geospatial packages install from PyPI. Geospatial dependencies (gdal, rasterio, fiona, rtree) may require system libraries.

Verify before relying:

- Whether the 57 runtime dependencies are all strictly required or if subsets can be used for specific workflows.
- Performance characteristics and typical runtime for large-scale regional yield forecasting.
- Data format requirements and expected input sizes for Earth observation datasets.
- Availability and format of pre-trained models or whether all models must be trained from scratch.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: active
- Downloads: 103.9K/month (top 15,000 on PyPI)
- Known vulnerabilities: none known

## Tags

crop yield forecasting, earth observation agriculture, climate impact drivers, agricultural monitoring ML, geospatial crop analysis, yield prediction models, agmet condition monitoring, geospatial-ml, agriculture-forecasting, earth-observation

[View on SkillFed](https://skillfed.io/packages/geocif) · [View on PyPI](https://pypi.org/project/geocif/)