paper/paper.md

In VEZY/sticRs: A Package to Set, Manage and Analyze STICS Simulations https://vezy.github.io/sticRs/

title: 'sticRs: the R package for STICS users and developers' authors: - affiliation: '1,2,3' name: R. Vezy orcid: 0000-0002-0808-1461 date: "10 September 2019" output: html_document: df_print: paged pdf_document: default word_document: default bibliography: paper.bib tags: - R - STICS - agronomy - sole crop - intercrop affiliations: - index: 1 name: CIRAD, UMR AMAP, F-34398 Montpellier, France - index: 2 name: INRA, UMR SYSTEM, F-34398 Montpellier, France - index: 3 name: AMAP, Univ Montpellier, CIRAD, INRA, IRD, Montpellier SupAgro, Montpellier, France

Summary

Crop models are often written in programming languages with relatively low-level of abstraction such as FORTRAN or C because they are well-suited for scientific calculation. However, higher-level tasks such as programmatic model parameterization, data manipulation (e.g. import/export of measurements and model outputs, statistics, plots), distributed computation, model evaluation or sensitivity analyses are often not integrated into the model, but rather made as an independent, external process. The ease of use and reproducibility of these steps depends strongly on the ability to program them [@Piccolo2016]. Hence, developers tend to design modelling environments for that purpose using different tools such as java applications, Python modules or R packages.

The STICS model [@Brisson:2008] is a dynamic, generic and robust model to simulate the soil-crop-atmosphere system. It has been developed for more than twenty years [@brisson1998stics] by more than 50 authors, and is now used by hundreds of users (1100 downloads as of 2012) across academic field (e.g. INRA, CIRAD, IRSTEA, more than 230 international papers), government agencies and technical institutions (Arvalis, French chamber of agriculture), international projects (H2020 ReMIX, agMIP), and schools (SupAgro). Hence, developing a complex model with as many users implies a thorough and careful development, while remaining easy to parameterize and use.

Excluding meteorological data, the STICS model needs more than 300 parameters for sole crops. These parameters are not always easily measurable on the field, measured with high variability, or simply not readily available to the user [@tremblay2004comparison; @launay2005assimilating]. Hence, a user should always evaluate the model sensitivity to the parameters before setting an approximated value [@VARELLA2010310]. The model has also different options for some formalisms (e.g. light interception), and the user should evaluate which is better suited for his particular use case by comparing the model outputs.

A developer that modifies a new formalism or designing a new one should also evaluate its impacts on the model outputs, and make a systematic comparison between different model versions or parameterization along field observations.

Therefore, STICS users and developers often experience the same different steps in their journey:

• Read, write or add parameters and their values
• Test the model programmatically for different parameter values, formalisms or model versions
• Evaluate the model against field observations
• Make plots
• Make a sensitivity analysis
• Optimize the parameter values using observations
• Distribute the runs across multiple cores / machines

sticRs is an R package that was designed to make these steps as easy and fast as possible, and more importantly, reproducible. It has been adapted for the regular sole crop version of STICS, but also to its intercrop version [@CORREHELLOU200972]. The following diagram presents a typical workflow using the main functions from sticRs:

Evaluating the model is so fast and straightforward using sticRs that this step is often bypassed by the user to directly make a model evaluation using the stics_eval function, which is a wrapper of the previous functions. The package also automatically distributes the simulations across computer cores when the user call stics_eval and sensitive_stics.

sticRs documentation is accessible on its dedicated website (https://vezy.github.io/sticRs) along with three tutorials:

1. Get started;
2. Sensitivity analyses;
3. Parameter value optimization.

The source code of the package has been archived to Zenodo with a concept (i.e. permanent) DOI [@zenodo].

Acknowledgements

The authors acknowledge the financial support from the European H2020 funded ReMIX project (https://www.remix-intercrops.eu).

References

VEZY/sticRs documentation built on Oct. 26, 2023, 7:37 a.m.