Climate Downscaling

Climate downscaling aims to increase the spatial resolution of climate model outputs, providing more localized and detailed climate information crucial for impact assessments and adaptation strategies. Current research heavily utilizes deep learning, employing various architectures like convolutional neural networks (CNNs), generative adversarial networks (GANs), diffusion models, and normalizing flows, often incorporating physical constraints or observational data to improve accuracy and realism. This work is significant because high-resolution climate projections are essential for numerous applications, ranging from agriculture and hydrology to urban planning and disaster risk management, improving the reliability and precision of climate-related decision-making.