KNMI DataLab

The KNMI DataLab leads KNMI into the Big Data & Data Science era.

Climate change research, weather forecasting and seismology rely more and more on (Big) data science. Besides increasing data from traditional sources such as observation networks, radars and satellites, the use of open data, crowd sourced data and the Internet of Things (IoT) is emerging. To deploy these sources of data optimally in our services and products, data-driven innovations are performed in collaboration with public and private sector partners.  KNMI DataLab facilitates and coordinates these innovations since the January 2016. Big data management, data analytics including machine learning and deep learning are playing an important role in the DataLab.

From idea to implementation
Ideas for data-driven innovations need to be pitched successfully to obtain support from the KNMI Datalab. Support can consist of data science effort and advice by the Datalab as well as data science training of the job. Data science activities include development of Proof-of-Concepts (PoC) and Minimum Viable Products (MVP) that are ready for implementation.

Data science facilities, data sources and expertise
The DataLab offers a working environment for data science with facilities for data integration (data wrangling), data analysis (e.g., machine learning and deep learning modeling) and data visualization. The KNMI Datalab offers also access to (big) data sources, training material and a network of experts from other datalabs, universities and  knowledge institutions. Outreach and sharing of results is performed by facts sheets, presentations, demonstrations and tools that are publicly made available.


Initial focus has been given to the themes “high resolution weather and atmospheric conditions”  and “weather and traffic (i.e. smart mobility)”.  New themes such as “impact based forecasting”  become more important for the development of an Early Warning Centre (EWC).

Examples of projects that are performed with support of the KNMI Datalab.

  • Fog detection from camera images
    Fog has impact on the safety and capacity of transportation and logistics activities on land, water and in the air. Many of cameras have been placed along these highways and waterways for traffic and water management purposes. In this project camera images are used to train a Deep Learning models to recognize reduced visibility conditions. This can be applied to decide on reduced visibility warnings for highways and waterways. More details: CIMO-TECO2018-Paper,  EGU2018-2988, Flyer
  • Sky View Factor from Elevation Model
    The sky view factor is essential to describe the urban climatology and its spatial variations. It is a proxy of radiation fluxes, depending on the height of the obstacles in its surroundings. The goal is of this project is to determine the sensitivity of the sky view factor computations for its different parameters: variations in grid-resolution, search radius and number of directions. The resulting dataset for the whole Netherlands can be found here.  More details: Flyer
  • Ice formation on overhead lines
    To explore the impact of severe weather events, there is a need for high resolution weather data. A use case that concerns ice formation on overhead lines of the railway network during the night is investigated by this project. Models are trained to predict the probability of ice formation on the overhead lines since this is a source of disruption during the winter season. An application is envisaged that offers the ablity to forecast the probability of ice formation hence assisting ice-mitigation operations. More details: Ice formation on overhead lines  
  • Car sensors as mobile meteorological network
    The availability of car sensor data is emerging. Over time, quality and applicability of car sensor data for meteorological and air quality purposes will increase as well as the temporal and spatial resolution (any time, any place). Car sensor data, for example, can be used to detect (dangerous) weather conditions on the road using car sensor data. More details: Flyer

More information about the KNMI datalab and related projects is also available on KNMI datalab.

De Bilt, 18 December 2018.