Statistical Data Analyst – (Geo)statistics & Knowledge Transfer – Flemish Public Institution

A department of a Flemish public institution holds several datasets that have been compiled over the past few years but have not yet been systematically analysed. As a statistical expert, you analyse these untapped datasets, from soil carbon and soil sealing to land use and permits. At the same time, you strengthen the department's in-house expertise so that, in time, staff can carry out similar analyses themselves.

You carry out analyses independently and coach a team of in-house staff. Your working time is divided between your own analyses, presentations and working sessions.

• You perform analyses on the department's datasets and present and explain them.
• You go through the analyses again together with a small group of around ten to twenty staff members, so that, in time, they can carry out similar analyses independently.
• You help the department make statistical analyses publishable, with greater attention to charts and publications.
• You help set up and carry out quality control of the published data and documents.
• You analyse datasets as part of the organisation's periodic reporting.
• In 2025, you explore in a pilot project the possibilities of Natural Language Processing (NLP) for the automated processing of permit decisions from the online permit portal, in particular unstructured PDF documents. You develop and apply techniques to extract, categorise and interpret conditions and obligations from these documents. The pilot project serves as a stepping stone towards wider use within the organisation, with knowledge transfer as a key pillar.

The department's datasets include the following, each with possible analyses:

• Carbon monitoring network: to find out how much carbon is stored in Flemish soils, the organisation launched a soil carbon monitoring network in July 2021. Over a period of 10 years, soils under all land uses are being sampled, including soils in wetlands and (semi-)natural grasslands, as well as soils managed by public authorities, local governments and citizens. The network has now completed two years of operation, so the first data can gradually be presented and analysed in greater depth. In time, the results can be used to adjust reporting and projections on land use, land-use change and forestry (LULUCF), in other words to substantiate whether, for example, agricultural soils or residential areas are storing or losing carbon.
• Soil sealing: in 2022, a new method was developed that automatically generates an annual soil sealing map. This map combines the administratively known soil sealing from the Large-scale Reference Database (GRB) with modelled soil sealing based on medium-scale winter aerial imagery. As a result, more sealed surfaces are mapped more accurately and more frequently than was previously the case. The method has since proven its worth but still needs to be optimised: for example, the model will probably have to be retrained at least once a year.
• Land use: land use in Flanders per hectare cell in 2013, 2016, 2019 and 2022. This land use can also be analysed spatially. Trends are being observed, but are these trends statistically reliable? The department has built up considerable experience in this area in recent years, but detecting ‘change’ remains difficult.
• Permits: over the past 6 years, the organisation has received around 80,000 permits a year via the online permit portal. Basic reporting on permit numbers is in place. There are more and more requests for reporting on this, and the questions are also becoming increasingly interesting in terms of content. The organisation therefore wants to carry out more, and more in-depth, analyses of the projects in the future. A project can consist of several components, some of which belong together geographically, for example dewatering (component 1) for the construction of an apartment block (component 2), or the demolition of a building and the construction of a new building on the same site. The organisation would like to know which components (largely) belong together geographically. A second, still embryonic idea is to detect striking trends in the mass of information in the portal, trends that deviate from the usual course of events: for example, a notably higher or lower number of applications of a certain type in a certain period or in a certain region (municipality).
• Peat map: last year, through an external study assignment, a very first peat map covering the whole of Flanders was produced with the help of AI. This map does not yet have the level of accuracy it should have. Statistical interpolation techniques could deliver a major improvement in quality here.

The aim is to further develop in-house expertise in the (statistical) analysis of datasets over the coming years. Knowledge transfer is therefore a key focus. A previous external consultant successfully improved the department's statistical literacy, and you will build on that: the organisation is no longer looking for basic statistics training, but for someone who can help the department further in making statistical analyses publishable.

The department staff involved in the programme have diverse backgrounds. They are all university-educated and interested in statistical analysis.