Our Research

The work in this project is grouped into four work packages as listed below:

1. Data handling and tools    

“create an unprecedented information hub on air quality related data, which can then be explored in future projects to improve the understanding of atmospheric composition processes”

GeoDataServices

Figure 1: Europe’s nighttime light brightness as exemplary illustration of GeoDataServices’ underlying high resolution data (own graph). Image and Dat

2. Interpolation 

“address interpolation of air quality data through the unique combination of relevant datasets and through careful choice of the machine learning techniques that will be applied including the optimisation of neural network architectures based on inspection of the learned system state”

3. Forecasting

 “develop improved methods for air quality predictions with (deep) neural networks, and possibly demonstrate a novel concept based on movie frame prediction methods”

4. Quality assurance

“explore novel concepts for robust, automated outlier detection and data screening of air quality measurements with deep neural networks”

Statistics

To evaluate the model performance, we base our analysis on the full joint distribution of forecasts and observations (Murphy and Winkler, 1987). We ac