BACK TO SCHEDULE REGULAR

Session 30

Innovative Practices in Agricultural Statistics

4 June 2026
16:30 – 18:00
ŠIBENIK VI

Presentation title
Farm Management Information Systems for Agricultural Statistics
New data requirements in EU regulations (2023/1538) demand more, more detailed, and more timely data from farmers.

Read more Read less

If at all possible, the use of traditional survey methods to acquire these new data from farmers will lead to an increasing response burden, which in turn may lead to a higher non-response. Connecting to Farm Management Information Systems (FMIS) could provide a good alternative to traditional survey methods for National Statistical Institutes (NSI). Many farmers use FMIS to manage their farms, which they also use to comply to government regulations, and reports to other stakeholders. FMIS are therefore a high-quality data source, that can be accessed (semi-)automatically, largely reducing the response burden on the farmer. Market research of FMIS vendors has shown that most FMIS contain the required data for Statistics on agricultural inputs and output (SAIO) requirements such as the Crop Yield Statistics and the Plant Protection Statistics. The advantages of using FMIS as a data source to meet SAIO requirements has led to the collaboration of three NSIs, CBS (The Netherlands), SSP (France) and Destatis (Germany). CBS, SSP and Destatis are at a different maturity level with experimenting, integrating and analysing FMIS data, and we are learning a lot from each other. In our presentation we hope to give you an idea of what is the best feasible way to integrate the NSIs and FMIS vendors on a local and European scale. In addition, we face similar issues, like the lack of trust of the farmer and farmer’s associations and unions in governmental organisations. We hope that our experiences will help other institutions to better involve the farmers on how to improve data collection and to make it easier and more efficient for the farmer. Lastly, we will present the possible FMIS data sharing solutions (such as an API or reporting system), the advantages and disadvantages of each data sharing solution and the results of the solutions that have already been piloted. So far, the results are promising.

Main author / Presenter
Remco Paulussen
Statistics Netherlands (CBS)

Read more Read less

Remco Paulussen is project manager of several Eurostat funded projects, which are participated by various statistical organisations. His current projects are on themes like Modernisation of Agricultural Statistics (MAS), Earth Observation and Artificial Intelligence (GEOS and AIML4OS), and Smart Surveys (SSI).

CO-AUTHORS:

Presentation title
Strengthening Data Quality in Agricultural Statistics under the SAIO Framework: cooperation between Eurostat and NSI
The new Regulation on Statistics on Agricultural Input and Output (SAIO) introduces a modern and more coordinated way of producing agricultural statistics in the European Union.

Read more Read less

Because agriculture is a sector that changes quickly and faces many environmental and economic challenges, it is essential that policymakers can rely on high-quality and up-to-date official data. SAIO helps achieve this by improving how data are collected, harmonized and reported across countries.

A central part of the SAIO quality approach is to create a common methodological basis for four important areas: milk, eggs, meat and livestock statistics. To support countries in applying the Regulation in a consistent way, Eurostat produced a set of detailed methodological handbooks. These handbooks explain standard concepts, suitable data sources, recommended validation checks and quality reporting requirements. They serve as practical tools to help national authorities follow shared principles, while still allowing them to work with their own national data systems.

The handbooks were developed through a highly collaborative process involving experts from many Member States. First, Eurostat set up Discussion Groups for each of the four statistical domains. Experts participated voluntarily and exchanged information on national practices, identified unclear points in the legislation and discussed differences in methods. The information collected during these meetings formed the initial content of the handbooks.

In the next phase, the drafts were shared with the Working Group on Agricultural Statistics. This allowed a much wider group of country representatives to provide comments and suggestions. The repeated feedback helped to improve the clarity, consistency and usability of the final documents. This process also ensured that the handbooks reflect the experience and knowledge of the whole European statistical community. Through this collaborative work, important quality principles, such as accuracy, comparability, transparency and coherence, were integrated into the operational guidance for countries.

In addition to methodological alignment, SAIO strengthens data quality through clearer metadata requirements, standard formats for transmitting data and specific obligations for quality reporting. All these elements follow the European Statistics Code of Practice and help create a more transparent and reliable quality framework.

Taken together, the SAIO Regulation, the methodological handbooks and the strong cooperation between EU countries represent an important step forward in improving the quality of agricultural statistics. This paper will describe how SAIO supports better data quality and how the collaborative development of guidance contributed to harmonised and robust statistical practices across the EU.

Main author / Presenter
Colomba Sermoneta
Istat

Read more Read less

Colomba Sermoneta is a senior statistician and researcher specializing in agricultural, animal production, and fishery statistics, with more than twenty years of experience in official statistics. She has been a Senior Researcher at ISTAT, the Italian National Institute of Statistics, since 2011, leading methodological innovation, integrating administrative and survey data, and supporting policymaking through high-quality statistical analysis. She holds a PhD in Agriculture, Food and Environment and degrees in social and demographic statistics. She also was a Seconded National Expert at Eurostat, where she contributes to the implementation of the SAIO Regulation, the development of Animal Production handbooks, and the coordination of Veterinary Medicinal Products statistics at EU level. Her work focuses on data quality, digitalisation of livestock systems, agro-environmental indicators, and international statistical harmonisation. She has authored numerous scientific publications and actively collaborates with Eurostat, FAO, and national statistical authorities across Europe and beyond.

Presentation title
Earth Observation Data and Machine Learning for Crop Yield Statistics
Estimated yield of important crops is highly relevant for agricultural statistics.

Read more Read less

In Germany crop yield estimation relies on knowledge-based yield estimations of experienced experts in the field. In Germany this work is mostly done by volunteers. The number of volunteers decreases continuously and data gaps are likely throughout the next years, which will in turn lead to fewer and possibly less accurate estimations. Furthermore, in terms of harmonization of statistical production workflows within the European Union, uniform data acquisition would be desirable.

Earth Observation (EO) data from satellites might be a suitable data source for future crop yield estimation. For agricultural statistics as well as for ecosystem extent and ecosystem condition accounting this data source has already been researched and in parts successfully explored. The Sentinel satellites, in case of crop yield specifically Sentinel-2, from the European Copernicus Program provide a high quality, public available data source with frequent revisit rates throughout the growth stage of important crop types. Thereby they can be used for consistent data acquisition over large areas, potentially for whole Europe.

In conjunction with the use of Sentinel-2 data, cloud coverage needs to be addressed. In areas with frequent and high cloud cover the number of usable satellite images can be limited.

The Land Statistical Office of the German Federal State Hesse has developed a method for crop yield estimation that is applied to several German federal states in a productive stage. This method employs a pre-processing pipeline for open-access Sentinel-2 Copernicus satellite data and uses a machine-learning model trained on known crop yields to retrospectively estimate crop yields per type.

To support the dissemination and application of the crop yield processing pipeline as well as to reduce parallel developments, the land statistical office Hesse, the German Federal Statistical Institute (Destatis) and the National Statistical Institute of the Netherlands (CBS) aim to (1) transfer the pre-processing pipeline to the Netherlands, (2) try to deploy the pre-processing pipeline within the Copernicus Data Space Ecosystem, and (3) to try to improve the pre-processing pipeline through the reduction of manual steps. This work is part of the project COSEO - Converging official statistics with Earth Observation, which is financed by the European Commission and is set to start in February 2026.

Main author / Presenter
Maren Koehlmann
German Federal Statistical Office

Read more Read less

.

CO-AUTHORS:

Presentation title
A SAS-Based Framework for Statistical Disclosure Control in Integrated Farm Statistics
Integrated Farm Statistics (IFS) is the EU framework for collecting harmonised microdata on the structure, production methods and environmental characteristics of agricultural holdings.

Read more Read less

As with any statistical data collection, ensuring statistical disclosure control (SDC) is essential before publishing results. SDC typically includes primary protection, which identifies cells at high risk of disclosure and suppresses or perturbs them, and secondary protection, which assesses whether these suppressed cells can be recalculated from published information, such as linear relationships between categories and their totals.

While primary protection is relatively straightforward to implement with standard programming tools, secondary protection is considerably more complex and usually requires specialised software. In developing an SDC procedure for IFS, we initially tested the τ-Argus tool. However, several limitations emerged, particularly related to the large volume of IFS microdata and the fact that IFS outputs are rounded, providing an initial layer of confidentiality that τ-Argus could not properly incorporate. These constraints indicated that relying on τ-Argus for IFS secondary protection would be impractical.

Consequently, we designed an alternative SAS-based procedure for SDC that can be applied to standalone tables as well as linked tables. Although developed for IFS, the procedure was designed to be adaptable with limited effort to other statistical domains requiring secondary confidentiality protection. It consists of three macros in a SAS program (SC_macro.sas) and uses an Excel template (SC_metafile.xlsx) which provides metadata, allowing customisation for specific data collections. The free solution is available on GitHub for use and further development by the statistical community. Detailed methodology descriptions and user instructions are provided in separate documents, SC_SAS_Methodology.docx, respectively SC_SAS_UserGuide.docx, also accessible on GitHub.

The procedure addresses key methodological and operational requirements:

• It accounts for the rounding applied in the dissemination process, which already reduces disclosure risk.

• It uses a flexible cost function typically used in SDC to prioritise suppression, enabling, for example, the preferential suppression of aggregates with lower reliability.

• Compared to other frequently used approaches, such as the hypercube method in τ-Argus, it results in a significantly smaller number of secondarily suppressed cells.

This paper presents the theoretical basis and technical design of the SAS-based procedure, outlining the assumptions regarding the intruder’s behaviour (which differ slightly from those used in τ-Argus). It describes the implementation within the SAS environment and includes examples showing rates and processing times for secondary suppression in specific IFS tables.

Main author / Presenter
Denisa Camelia Florescu
European Commission, Eurostat E.1

Read more Read less

Denisa Florescu is a methodologist at Eurostat, where she is an integral member of the farm statistics team. Her work focuses on agricultural censuses and sample-based data collections conducted in inter-census periods. She holds a master’s degree in marketing research from the Academy of Economic Studies of Romania. She began her career at the National Statistical Institute of Romania, specializing in sampling techniques within the business directorate. At Eurostat, she has nearly two decades of experience, contributing to both the statistical methodology and agricultural statistics units. Denisa will present an innovative SAS-based framework on secondary confidentiality tailored for farm statistics which can be easily adapted to other domains. This framework was developed by her team colleague Rudi Seljak, with contributions from Wladimir Raymond and herself.

CO-AUTHORS: