Deep neural networks enhance urban temperature forecasting with high-resolution satellite data

Researchers have developed deep neural network models to improve the resolution and forecasting of urban land surface temperatures. By combining data from geostationary and polar-orbiting satellites, they created models capable of providing temperature data at 1 km resolution every 15 minutes. A U-Net model was trained to map lower-resolution satellite data to higher-resolution data, achieving an RMSE of 1.92°C. A subsequent ConvLSTM model was used for nowcasting, outperforming baseline models with RMSEs between 0.57°C and 1.15°C for lead times of 15 to 75 minutes. AI

IMPACT This research demonstrates how AI can improve the granularity and predictive accuracy of environmental monitoring, potentially aiding urban planning and climate studies.

RANK_REASON The cluster contains an academic paper detailing a new methodology using deep neural networks for spatiotemporal downscaling and nowcasting of urban land surface temperatures. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• SEVIRI/MSG
• Solomiia Kurchaba
• Terra/Aqua MODIS
• U-Net

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →