-The enhancements in IT solutions and the open science movement are injecting changes in the practices dealing with data collection, collation, processing and analytics, and publishing in all the domains, including agri-food. However, in implementing these changes one of the major issues faced by the agri-food researchers is the fragmentation of the “assets” to be exploited when performing research tasks, e.g. data of interest are heterogeneous and scattered across several repositories, the tools modellers rely on are diverse and often make use of limited computing capacity, the publishing practices are various and rarely aim at making available the “whole story” with datasets, processes, workflows. This paper presents the AGINFRA PLUS endeavour to overcome these limitations by providing researchers in three designated communities with Virtual Research Environments facilitating the use of the “assets” of interest and promote collaboration.
The developments in information and communication
technologies, including big data availability and management,
web and cloud technologies, as well as open science related
practices are not yet fully embraced by Agriculture and Food
Science research domain [
The AGINFRA PLUS project has been set up to develop
an innovative approach in Agri-food digital science
practices aiming at overcoming the limitations stemming from
the above settings by leveraging on existing e-Infrastructures
and services. In particular, AGINFRA PLUS promotes the
exploitation of Virtual Research Environments (VREs) [
AGINFRA PLUS is exploiting the VREs approach for three prominent agri-food research communities, namely: (i) agro-climatic and economic modelling, focusing on use cases related to crop modelling and crop phenology estimation, (ii) food safety risk assessment, focusing on use cases to support scientists in the multidisciplinary field of risk assessment and emerging risk identification, and (iii) food security, focusing on use cases related to high-throughput phenotyping to support phenomics researchers to select the most suitable plant species and varieties for specific environments.
The remainder of the paper is organised as follows. Sec. II presents the major constituents of the AGINFRA PLUS platform. Sec. III discusses the exploitation scenarios developed by each community and the benefits resulting from the use of the platform. Finally, Sec. IV concludes the paper by reporting some future works.
In order to support the AGINFRA PLUS communities, a comprehensive and feature rich platform has been developed and operated. An overall picture of such a platform aiming at offering its facilities by the as-a-Service delivery model is given in Fig. 1.
Such a platform follows the system of systems approach
[
These basic facilities are complemented by services for the semantic-oriented management of data, data analytics, data visualization, and publishing.
The AGINFRA PLUS data & semantics facilities offer
an array of services for managing semantic resources (e.g.
ontologies, thesauri, vocabularies) and for benefitting from
such resources in tasks related with data management. The
supported facilities include: (a) an ontology engineering
service for creating, editing and managing semantic resources
and, at the same time, catering for their collaborative design,
editing and management. It is based on VocBench [
The AGINFRA PLUS analytics facilities offer a rich array
of services for the challenging task of big data analytics [
All these platforms and environments are nicely integrated each other as well as are integrated with the rest of services offered by AGINFRA PLUS. For instance, every method integrated in the data analytics platform can be easily executed by a Jupyter notebook or by a Galaxy workflow. All these tools are equipped with solutions facilitating the access to the workspace content thus to make use of it during the processing steps, e.g. to use files as inputs or to store results. It is straightforward to publish every analytics process implemented by these tools into the catalogue to share it with coworkers.
The AGINFRA PLUS data visualization & publishing
facilities provide users with feature-rich and flexible solutions
for developing representations (e.g. graphs) out of datasets and
publishing “research objects” documenting a research activity
and its results. The final goal is to provide the reader with an
effective representation of a research activity and its results
thus to enable its repeatability. The supported facilities include:
(a) a graphs management workbench for creating several
typologies of interactive graphs ranging from generic ones
(e.g. Spline, Scatter, Bar, Line, Step, Pie, Doughnut, Polar) to
very specific ones (e.g. graphs reporting the height of plants
across time with values and images). The platform provide
users with facilities to import a dataset of interest, to define
how its content has to be used to produce the graph of interest,
and to share the produced graphs; (b) a mind map workbench
for managing this typology of diagrams; (c) a network
visualisation for creating visualisations aiming at highlighting the
connections among the entities of a connected graph; (d) a
catalogue-based publishing platform to disseminate artefacts
according to the FAIR principles [
The objective is to set-up and evaluate an AGINFRA PLUS
VRE for use by agro-climatic researchers to perform crop
modelling related work. To guide the selection and
development of tools to be included in the VRE, two typical research
activities were selected: (i) performing crop model simulations
at scale; and (ii) explorative modelling focussing on crop
phenology studies. Based on data availability both have a strong
focus on The Netherlands as study area (using AgroDataCube
[
Initially the well-know WOFOST model [
A second activity that is examined is the use of explorative modelling for the estimation of crop phenology characteristics, using available agronomic data, combined with crop development indicators (e.g. the NDVI vegetation index), derived from remote sensing data. This activity uses the AGINFRA PLUS analytics facilities such as Jupyter Notebooks and RStudio, to experiment with agronomic data analytics. The aim is to test such analytics, providing insight in critical crop development indicators, and to convert these into algorithms deployed as DataMiner processes on the VRE to run them at scale. Results can then be used to more accurately estimate regional crop yields, using long-term agronomic statistics and yield prediction systems.
Early stage evaluation results from piloting both agroclimatic modelling activities indicated that the VRE already is regarded as being well equipped for collaborative research. At that point (about one year ago) there were however reservations concerning the ability of the VRE to support full agro-climatic modelling workflows, due to some limitations regarding the integration of the different processing, analytics and visualisation components available in the VRE.
Therefore recent efforts in the use case have been focussing on improving these integration capabilities and on providing better-connected prototypes for both activities, supporting the full research working process. For example by creating a dashboard that visualises crop parcels, the input data for crop model simulations (crop, soil and weather information), and on-the-fly calculated simulation results such as leaf area index and total biomass produced.
The Virtual Research Environment developed for supporting this scenario is available at https://aginfra.d4science.org/web/ agroclimaticmodelling.
In the domain of food safety modelling two exploitation
scenarios were identified where scientific data analysis
workflows and software based resources for knowledge sharing and
integration are of extraordinary importance. Both scenarios
nicely complement the activities the community is promoting
to harmonise the knowledge produced [
The DEMETER scenario is aiming at developing a working environment supporting the early identification of issues in the food (and feed) chain. This scenario largely build upon the workspace, the data analytics and the catalogue to demonstrate how KNIME-based data mining workflows can be efficiently shared and applied from within the VRE.
The RAKIP scenario aims at providing risk assessors and
risk modellers with an environment supporting their efforts to
share their knowledge (data, mathematical model, simulation
results) in a harmonized way. A distinguishing feature of
this environment is a community-driven food safety model
repository, that contains mathematical models from the area
of predictive microbial modelling and quantitative microbial
risk assessment (QMRA). This repository builds upon the
FSK-Lab [
The AGINFRA PLUS platform support these scenarios by providing: (a) facilities for developing the ontology underlying the FSK-Lab solution for models representation (VocBench, see Sec. II-A); (b) two specific processes integrated into the data analytics platform to respectively support the publishing of a model into the catalogue and the execution of any model; (c) a catalogue where the models are published according to the community ontology and endowed each with three actionable resources enabling users to respectively download the model, perform a model simulation by using the default parameters, perform a model simulation by tuning the parameters; (d) a mind map development and dissemination solution facilitating the communication among the members; (e) a journal-based approach for publishing the models. The Food Modeling Journal1 has been designed and launched to support the needs emerging in this community. It promotes 1https://fmj.pensoft.net/ the publishing of Models, Data analytics, Applied study, Data paper, and Software description. Thanks to the integration of the publishing platform into the VREs (see Sec. II-C) it is straightforward to produce papers linking the available artifacts, e.g. the models in their actionable form.
The Virtual Research Environments developed for supporting these scenarios are available at https://aginfra.d4science. org/web/demeter and https://aginfra.d4science.org/web/rakip portal.
The Food Security Community is focusing on a highthroughput plant phenotyping scenario. This scenario can help to select crop varieties that better adapt to global changes in order to respond to the food security challenges. Highthroughput phenotyping produces a large amount of data which need to be integrated and analysed right away. For example, in a greenhouse platform, a lot of images of plants are taken: 13 images per plants per day are taken in the Montpellier platform which works on 1600 plants (more than 20,000 images per day). Field platforms produce and need a lot of images including UAV or satellite. High-throughput phenotyping platforms produce complex data (sensors data, human reading) at different scales (e.g. population, individuals, molecular).
The phenomics community needs tools to easily access
to large datasets and to be able to visualize and analyse
them. Moreover, sharing data, analytics process and results is
essential. The objective of this use case is to develop a VRE
for phenomics researchers where these users: have access to
relevant ontologies; collaborate on building and share semantic
resources; have access to phenomics platforms data from the
information system OpenSILEX-PHIS [
The first evaluation results of the Food Security VRE indicated that the VRE is useful for collaborative work. The diversity of tools that are available has also shown interest from the users. However, there were some reservations on the integration of these which made difficult the execution of certain data analysis workflows. Another concern on big data manipulation and data access has also been noted. Considering this, recent work has been made to improve these integration capabilities in order to provide better connected tools. Web Services based on the Breeding API standards2 had also been implemented into the OpenSILEX-PHIS system in order to easily access phenotyping data in the VRE.
The Virtual Research Environment developed for supporting this scenario is available at https://aginfra.d4science.org/web/ foodsecurity.
This paper presented the AGINFRA PLUS platform, a science gateway providing the Agri-food community with a rich array of services oriented to promote the implementation of open science practices. Such a platform is currently supporting three designated communities dealing with crops simulation, food safety risk assessment, and high-throughput plant phenotyping scenarios.
The platform is bringing into these communities and their working practices a number of benefits including (a) the simplicity for coworkers to perform collaborative work, e.g. the workspace is a working area users can count on to collaborate, the social networking is a means to have informed dialogues; (b) the easiness to share results of any form within and across the boundaries of their communities and the platform itself, e.g. the catalogue is a valuable service for disseminating research artefacts and enable users to access them, the integration with the OpenAIRE dashboard and the scholarly communication platform reduces the gaps with the scholarly communication domain; (c) the attention dedicated to ease the flowing of existing artefacts into the platform thus to reduce fragmentation and facilitate their reuse, e.g. the plethora of programming languages and approaches supported by the analytics facilities make it possible to easily integrate almost any existing analytics method, the array of solutions for ontology management facilitate their reuse.
Overall, the AGINFRA PLUS platform is currently serving hundreds of users (more than 340 in Feb. 2019) by 13 active VREs. In the coming months these figures are going to improve because the project will enter into the community validation and uptake phase. In the period Mar. 2018 - Feb. 2019 the users served by this platform and its VREs performed: a total of 24,439 working sessions, with an average of circa 2,036 sessions per month; a total of 1,959 social interactions, with an average of circa 163 interactions per month; a total of 1,842 analytics tasks, with an average of circa 153 tasks per month; a total of 387 items have been published into the catalogue including models, research objects, methods, services, terms, and datasets.
Future developments includes the development of a catalogue supporting semantic queries, the development of tools easing the discovery and access to geospatial datasets, the development of recommender systems, the development of tools supporting the identification of suitable licences for the produced artifacts.
This work has received funding from the European Union’s Horizon 2020 research and innovation programme under AGINFRA PLUS project (grant agreement No. 731001).