AI for weather modeling
Course Overview
This course provides an introduction to the concept and application of deep learning for weather and climate modeling followed by series of videos on integrating weather and climate models with observational data using Deep Learning. The videos introduce the concept of merging framework for spatiotemporal mapping of atmospheric model simulation errors using CNNs, describe common challenges and technical solutions in preprocessing atmospheric model and reanalysis data, uncover the significance of CNN properties such as Translation Invariance and Receptive Field in daily precipitation data, and interpret results obtained from a case study revealing the impact of U-Net in enhancing modeled precipitation data over Europe. Enroll for an introductory understanding of concepts and practical applications in weather modeling, deep learning, and data preprocessing.
Details
Lecturer
Overview
This course provides an introduction to the concept and application of deep learning for weather and climate modeling followed by series of videos on integrating weather and climate models with observational data using Deep Learning. The videos introduce the concept of merging framework for spatiotemporal mapping of atmospheric model simulation errors using CNNs, describe common challenges and technical solutions in preprocessing atmospheric model and reanalysis data, uncover the significance of CNN properties such as Translation Invariance and Receptive Field in daily precipitation data, and interpret results obtained from a case study revealing the impact of U-Net in enhancing modeled precipitation data over Europe. Enroll for an introductory understanding of concepts and practical applications in weather modeling, deep learning, and data preprocessing.
Introduction - Deep Learning for weather and climate
Introduction on AI and deep learning for weather and climate, which is a scientific field where there have been tremendous developments in the past few years.
Merging Model-based Data with Observations
There are many available approaches for merging. In this video, we briefly talk about statistical methods, data assimilation, and data-driven methods.
Deep Learning Used in a Merging Framework
In this video, we will focus on using mismatch mapping in a merging framework.
Pre-processing Model and Reanalysis Data
In this video, we'll learn about preprocessing examples such as converting COSMO-REA6 grib files to NetCDF format and upscaling its original ~6kms spatial resolution grid to match the TSMP model grid at ~12km spatial resolution using bilinear resampling.
Understanding Translation Invariance and Receptive Field Properties in CNNs
In this video, we're going to talk about two important properties of CNNs, Translation Invariance and Receptive Field in precipitation data.
Results from a Case Study: Using U-Net to Improve Modeled Precipation
In this video we'll explore the results of a case study where U-Net was used in a regression network to improve modeled precipitation data over Europe.
