The increasing impact of climate change and variability on food security and agriculture, and the need for agroecological transition require improving the performance of crop simulation models and to model new biophysical processes to tackle new challenges. Recently, the Agricultural Models Exchange Initiative (AMEI) proposed Crop2ML, an open-source modeling framework for exchanging and reusing crop model components between modeling platforms. This framework provides a unified description of model components at a high level of abstraction based on shared concepts, lifting constraints of modeling platforms through an automatic system that transforms high-level models into platform-compliant components. This paper presents an approach based on reverse engineering to automatically extract and transform meta-information and algorithms of existing crop model processes into a platform-independent model component. A search algorithm using Crop2ML concepts, and a many-to-one transformation system was adopted as the main reverse engineering techniques to produce high-level models. The system consists in parsing the codebase of model components from different high-level languages (Python, R, Java, C++, C#, and Fortran) using the ANother Tool for Language Recognition (ANTLR) parser generator and processing the generated syntax trees to produce the various model implementations. It is implemented and evaluated on three crop model components: an energy balance model provided by BioMA platform and two soil temperature models provided by SIMPLACE and DSSAT platforms. We demonstrated the extensibility of our approach with the STICS, OpenAlea, and SiriusQuality modeling platforms. CyMLTx is a significant contribution towards the interoperability of crop modeling platforms and the reuse of crop model components beyond programming languages. It will thus contribute to foster model intercomparison and improvement activities.

If you use this software, please cite it using the metadata from this file.