The article “Explainable machine learning reveals capabilities, reundancy, and limitations of a geospatial air quality benchmark dataset” is now available in the journal Machine Learning and Knowledge Extraction.
Here you go: https://www.mdpi.com/2504-4990/4/1/8
This work is based upon AQ-Bench and was inspired by the high-resolution mapping. Thanks to KI:STE the authors got together and were able to gain new insights into how machine learning models represent the air quality data and how their predictions function.