AI Insight
Researchers developed a deep learning system that combines satellite data from two different sources to produce high-resolution land surface temperature maps of European cities at both fine spatial detail (1 kilometer) and frequent time intervals (every 15 minutes). The system uses a U-Net neural network to enhance lower-resolution geostationary satellite imagery to match the quality of polar-orbiting satellites, then employs a ConvLSTM model to forecast temperatures up to 75 minutes ahead with errors ranging from 0.57 to 1.15 degrees Celsius depending on forecast length.
Why it matters
This technique overcomes a longstanding limitation in satellite remote sensing where datasets typically offer either high spatial or high temporal resolution but not both. The ability to monitor and forecast urban temperatures at high spatiotemporal resolution has direct applications for urban heat management, climate studies, and operational environmental monitoring systems.
Understand the Science
Abstract: Land Surface Temperature (LST) is a key variable for various applications, such as urban climate and ecology studies. Yet, existing satellite-derived LST products provide either high spatial or high temporal resolution, resulting in a fundamental trade-off between the two. To address this trade-off, we combine observations from a geostationary and a polar orbiting satellite and provide LST fields at high spatial and high temporal resolution (1 km at 15-min intervals). We demonstrate their application for intraday forecasting of LSTs. To estimate LST fields at high spatiotemporal resolution, a U-Net model is trained to map LST fields from SEVIRI/MSG (3 km and 15 min resolution) to LST fields from Terra/Aqua MODIS (1 km, 4 overpasses per day) that are collocated in space and time. The presented model has been trained on LSTs across large European cities with a population exceeding 1 million inhabitants, and achieves an RMSE = $1.92${deg}C and near-zero bias MBE = $0.01${deg}C on the hold-out test set. As a second step, we present an LST nowcasting model based on ConvLSTM architecture, trained across downscaled LST fields with forecast lead times of 15 to 75 minutes. The nowcasting model outperforms a persistence and a Climatological Rolling Median benchmarks, with RMSEs of $0.57$ to $1.15${deg}C for the considered lead times and biases ranging from $-0.1$ to $0.14${deg}C. An additional validation conducted against independent MODIS overpasses confirms robust performance. Our LST forecast model at high spatiotemporal resolution is directly applicable to operational satellite-based LST monitoring.