MaDiPHS WP5 Models

This repository, developed by icipe (International Centre of Insect Physiology and Ecology), hosts the official pest models developed under Work Package 5 (WP5) of the MaDiPHS (Malawi Digital Plant Health Services) project. The MaDiPHS project is officially supported by NORAD (NORwegian Agency for Development), and implemented in partnership with key and strategic partners.

This repository provides access to pest models covering phenology, habitat suitability, and risk, along with the associated model development code & data to ensure reproducibility. It also delivers streamlined APIs (Library & Web) for model inference, validation, and deployment in both local and production environments. Both the Library API (Python package) and the Web API (REST service) are maintained here to guarantee consistency across use cases and environments.

Features

🛠 Model Development Repo

This repository is not just a package or API, but also the development home for WP5 pest models. It includes:

• Source code for phenology, habitat suitability, and risk models.
• Example training and validation datasets (raster, vector, and tabular).
• Reproducible workflows for model building, testing, evaluation, and benchmarking.
• Documentation of methods and assumptions to support transparency and reuse.
• A foundation for partners or anyone to extend or adapt models for new crops, regions, or pests.

📦 Library API (Python package)

Import and use pest models directly in Python.

• Local and offline access to inference-ready MaDiPHS WP5 pest models across multiple crops.
• Public functions, methods, and objects for model inference, validation, and geospatial utilities.
• Supports raster and vector formats as direct inputs and outputs.
• Extensible to new pests and workflows.

🌐 Web API (REST service)

Access the same pest models via HTTP requests (no Python required).

• REST endpoints for phenology, habitat suitability, and risk models.
• Geospatial input/output support via file upload or JSON payloads.
• Built for smooth deployment in cloud or local environments.
• Ensures consistency with the Library API.

🔬 Shared Capabilities

• 🧪 Validation – ensures consistency and reliability through model performance evaluation, benchmarking and workflow testing.
• 🚀 Deployment – encapsulates models and workflows for consistent, reusable, and smooth integration into production environments.
• 🧩 Extensible – can accommodate new models and functionalities, ensuring models and workflows remain adaptable as needs evolve.
• ♻ Reproducibility – provides transparent access to code, data, and workflows, enabling results to be reliably replicated and independently verified.

Pests Covered

The package currently supports access to the following pest models, organized by their respective main host crop:

🌽 Maize

1. Fall armyworm (Spodoptera frugiperda)
2. Stem borers (Chilo partellus & Busseola fusca)
3. African armyworm (Spodoptera exempta)
4. Witchweed (Striga)

🍅 Tomato

5. Tomato leafminer (Tuta absoluta / Phthorimaea absoluta)
6. Tomato red spider mite (Tetranychus evansi)

🥔 Cassava

7. Cassava Whitefly (Bemisia tabaci)

🍌 Banana

8. Banana weevil (Cosmopolites sordidus)

🥜 Groundnut

9. Groundnut Aphid (Aphis craccivora)

Overall Project Context

This project was developed under Work Package 5 (WP5) of the MaDiPHS project. WP5 delivers pest models for early warning systems and production environments, enhancing plant health monitoring, pest risk assessment, and decision-making to strengthen digital plant health solutions in Malawi.