Agricultural Model Adapter A MAD Toolbelt
A premium, zero-dependency agricultural crop modeling companion. Evaluate dataset completeness, map vocabulary schemas, and test model suitability in real-time.
Core Capability Suite
Dataset Ranking (V1 & V2)
Classifies dataset quality across Management, Phenology, Site, Initial Values, Soil, and Weather. Translates observation density and layer structure into Platinum, Gold, Silver, and Copper medals.
AgMIP Vocabulary Mapping
Couples local parameters to international ICASA and AgMIP vocabulary dictionaries (e.g. PDATE, SABD, GWAM), preparing metadata for standard translation pipelines.
Model Suitability matching
Instantly evaluates dataset compatibility against 7 crop models: HERMES, DSSAT, APSIM, WOFOST, MONICA, AquaCrop, and STICS.
Dirichlet Uncertainty Engine
Runs client-side Monte Carlo simulations using a Dirichlet prior distribution to extract uncertainty bounds and confidence percentiles (p05, p50, p95). Click to learn how it works →
Origins & Modernization
The Agricultural Model Adapter is the web-native continuation of the desktop QT/C++ tool belt originally built by Dr. Jason S. Jorgenson for the European MACSUR research network.
It has been completely rewritten for the web as an individual, passionate effort to preserve the integrity of the original project while scaling it with modern Open Science frameworks.
What was the FACCE MACSUR Network?
FACCE MACSUR (Modelling European Agriculture with Climate Change for Food Security) was a massive joint programming network spanning 17 countries, 71 organizations, and over 230 crop, livestock, and economic modelers.
The network faced a massive hurdle: 75+ heterogeneous models written in various programming languages, run on multiple operating systems, and executed against data subsets that were fragmented, inconsistently formatted, or restricted under complex intellectual property rights (IPR).
MAD was developed under MACSUR Work Package 1 (WP1) to establish standardized data adapters, check dataset completeness, and catalog available simulation models.
The Legacy of Dr. Jason S. Jorgenson
Dr. Jason S. Jorgenson (University of Reading) conceived the Toolbelt as three core components working alongside Aarhus University's data platform:
- Model Adaptor: An execution wrapper (incorporating the Python-based "Mishelle" scripting shell) to integrate and run models dynamically.
- Dataset Ranking: A mathematical point system scoring data quality based on temporal frequency and resolution.
- Model Repository: A catalog mapping required inputs and outputs for various models.
While original Qt desktop files relied on the defunct Qt Enginio cloud synchronizer, this browser port maintains the dataset and model repository components, substituting cloud dependencies with full local JSON import/export compatibility.
What is the Müncheberg Dataset?
The default crop-rotation preset pre-loaded in the Agricultural Model Adapter (A MAD Toolbelt) is derived from historical field data from **Müncheberg, Germany**, provided to the MACSUR WP1 inter-comparison pilot by Chris Kollas and Kurt Christian Kersebaum (ZALF).
This dataset serves as the standard mathematical baseline. It tracks a 13-season crop rotation scheme and evaluates site parameters (128m elevation, sandy loam texture), crop phenology observations, and soil moisture boundaries.
Modernizing to the V2 Framework
The MAD Web Toolbelt is now maintained as an individual effort to keep agricultural data science accessible. The tool has been upgraded from the legacy 2015-Kersebaum metrics to the modern DataRanking V2 architecture:
FAIR Multipliers: Evaluates dataset compliance against Findability, Accessibility, Interoperability, and Reusability metrics, scoring sections with custom multipliers.
Advanced Soil PTFs & Sensors: Maps high-resolution remote sensing registries, in-situ telemetry feeds, and genetic cultivar coefficients.