In collaboration with researchers from WUR, VU Amsterdam, TU Delft and KNMI. ©Vincent de Feiter.
The Dutch atmosphere is among the most monitored in the world, thanks in large part to the infrastructure of the Ruisdael Observatory. Leveraging this unique network, a team of researchers launched a pilot study to evaluate how well different atmospheric models represent surface weather and carbon exchange processes across key Dutch ecosystems. This intercomparison effort, led by Mary-Rose Mangan and Wouter Mol since September 2024, remains a work in progress but marks an important step toward building a national framework for climate and weather model validation.
A Virtual Testbed for the Atmosphere
The pilot centers on observations collected during a field campaign on 17–18 May 2022, spanning three sites in the Netherlands:
- Cabauw (grassland)
- Loobos (forest)
- Veenkampen (grassland)
These locations form a triangle that captures the diversity of land cover and surface-atmosphere interactions in the Dutch landscape. Using a rich dataset — from surface flux towers and ground-based sensors to radiosondes and aircraft measurements — the project compares real-world observations against simulations from six different atmospheric models.
These models span from global numerical weather prediction systems like the IFS (ECMWF) to high-resolution large-eddy simulations (LES) such as DALES, MicroHH, and ASPIRE. Importantly, some models include advanced land surface and carbon cycle processes, allowing researchers to evaluate both meteorological and biogeochemical dynamics.
First Insights
The study focuses on how models simulate:
- Radiation and surface energy balances
- Turbulent fluxes (heat, moisture, and CO₂)
- Atmospheric boundary layer dynamics
- Cloud and wind profiles
At Cabauw, for instance, most models overestimated daytime sensible heat (H) and underestimated latent heat (LE), resulting in a higher Bowen ratio than observed. This has implications for how well models simulate evapotranspiration and surface moisture dynamics. In terms of CO₂ flux, models that include vegetation and respiration processes (like MicroHH-Ags) better matched observations, though some timing mismatches were apparent.
At Loobos, the forest canopy introduced further challenges. The IFS and MicroHH-Ags models captured radiation patterns relatively well but diverged in simulating carbon fluxes. MicroHH-Ags, which includes a more detailed representation of photosynthesis and respiration, more closely matched observed CO₂ dynamics during the day.
Veenkampen offered a second grassland comparison. While radiation fluxes were generally well captured, differences emerged in ground heat flux and the partitioning of turbulent energy. These results underline the importance of land surface representation and soil parameterisation across ecosystems.
Toward a National Benchmark
This intercomparison is more than an academic exercise—it is a first test of what a Ruisdael Model Evaluation Testbed could look like. By combining long-term observational infrastructure with coordinated model experiments, the initiative aims to build a robust and transparent system for evaluating atmospheric models across scales.
As the analysis progresses, the team plans to:
- Incorporate more mesoscale models (e.g., WRF, HARMONIE)
- Improve observational corrections (e.g., soil heat storage)
- Expand to other land covers (e.g., urban, coastal)
- Develop standard metrics and interactive tools for comparison
Learn More
To dive deeper into the data, methods, and preliminary results, you can access the full report here: Ruisdael_Intercomparison.
