Researchers from Karlsruhe Institute of Technology (KIT) Advance Precipitation Mapping with Deep Learning for Improved Spatial and Temporal Resolution

Due to climate change, extreme weather, particularly heavy precipitation events, is expected to become more frequent. Many natural disasters, such as floods or landslides, are directly caused by extreme precipitation. Models based on climate prediction are frequently used. The existing climate models must improve their ability to accurately represent highly variable atmospheric phenomena. Researchers expect that increasing average temperatures will cause extreme precipitation events to further.

The researchers of Karlsruhe Institute of Technology (KIT) have harnessed the power of artificial intelligence (AI) to enhance the precision of coarse precipitation maps generated by global climate models.

Researchers emphasized that this model shortened the temporal resolution of the precipitation fields from one hour to ten minutes, and the spatial resolution was increased from 32 to two kilometers. They said a higher resolution is necessary to predict the future occurrence of heavy local precipitation events and the ensuing natural disasters.

This method involves the application of a generative neural network, specifically a Generative Adversarial Network (GAN), a form of AI. This GAN is trained using high-resolution radar precipitation data, allowing it to learn and mimic realistic precipitation fields with significantly higher spatial and temporal resolutions.

The existing global climate models use a grid that lacks the necessary fine detail to capture precipitation variability precisely. Also, producing highly resolved precipitation maps traditionally requires computationally expensive models, leading to spatial or temporal limitations.

According to researchers, this is the reason for developing GAN, an AI-based generative neural network trained using high-resolution radar precipitation fields. In this manner, from coarsely resolved data, the GAN learns how to produce realistic precipitation fields and determine their temporal sequence.

Compared to trilinear interpolation and a classical convolutional neural network, the generative model reconstructs the resolution-dependent extreme value distribution with high skill. It showed a high fractions skill score of 0.6 on rainfall intensities over 15 mm h−1 and a low relative bias of 3.35%.

According to the researchers, their approach produces an ensemble of various possible precipitation fields. This is significant because numerous physically possible, highly resolved solutions exist for every coarsely resolved precipitation field.

They explained that the higher resolution of precipitation events simulated with this method will allow for a better estimation of the impacts the weather conditions that caused the flooding of the river Ahr in 2021 would have had in a world warmer by 2 degrees.

In conclusion, this model offers a solution to enhance the precision of global climate models in predicting precipitation. This advancement contributes to more accurate climate forecasts. It holds the potential to understand better and prepare for the consequences of extreme weather events in the face of a changing climate.

Check out the Paper. All credit for this research goes to the researchers of this project. Also, don’t forget to join our 34k+ ML SubReddit, 41k+ Facebook Community, Discord Channel, and Email Newsletter, where we share the latest AI research news, cool AI projects, and more.

If you like our work, you will love our newsletter..