Conveners
Deep/Machine learning and data science
- Peter Braesicke (KIT, IMK-ASF)
We present a new empirical model of electron density in the ionosphere, which is a crucial parameter impacting radio signal propagation and GNSS systems. Our model utilizes radio occultation profiles obtained from CHAMP, GRACE, and COSMIC missions. We assume a linear decrease in scale height with altitude and consider four parameters: F2-peak density and height (NmF2 and hmF2), as well as the...
While clouds are a key component of the climate system, cloud processes and aerosol-cloud interactions are still poorly understood. This is also the case for fog and low stratus clouds, which are particularly affected by aerosol emissions at the surface and by land-atmosphere interactions.
In this study, Extreme gradient boosting (XGBoost) and xAI methods such as SHapley Additive...
Climate models are limited in their ability to accurately represent highly variable atmospheric phenomena on small scales. Downscaling techniques are employed to resolve fine-scale physical processes and reveal local impacts.
We present spateGAN, a conditional generative adversarial network for spatio-temporal precipitation downscaling in Germany. As a so-called video-superresolution...
As a novel remote sensing approach, GNSS Reflectometry (GNSS-R) offers unique potential for characterizing the complex Earth system with its different spheres on various spatiotemporal scales with numerous geoscientific applications. With the continuous increase of space-borne GNSS-R observation data volume, Artificial Intelligence (AI) offers an alternative data-driven direction of achieving...
Machine Learning (ML) model is widely used to make predictions of air pollutant concentration in the ‘business as usual’ scenario. Compared with chemical transport model, ML model cannot be limited by its spatial resolution and potentially outdated emission inventories. To quantify the contribution of the meteorological driver of air pollutants during the COVID-19 lockdown period in China,...
Machine-learning (ML) techniques have been recently applied to the calibration of low-cost sensors (LCS). Many studies report successes in calibration with ML techniques such as random forests (RF), neural networks (NN), and support vector regression (SVR). We find that calibrating LCS for the measurement of nitrogen dioxide (NO2) and particulate matter (PM) with ML techniques is not as...
