Spatiotemporal downscaling and nowcasting of urban land surface temperatures with deep neural networks

Datasets for the article: "Spatiotemporal downscaling and nowcasting of urban land surface temperatures with deep neural networks".

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 = $1.92$~°C and near-zero bias MBE = $0.01$~°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 $1.15$~°C for the considered lead times and biases ranging from $-0.1$ to $0.14$~°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.

Dataset description:
- downscaling_dataset_Europe.zip: Dataset used to train a U-Net model 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) in our first step.
- downscaled_for_nowcasting_{city}.zip/downscaled_for_nowcasting_test_{city}.zip: Training and test data for 3 cities (Bucharest, Antwerp, and Berlin) used to train and apply in our second step an LST nowcasting model based on ConvLSTM architecture, trained across downscaled LST fields with forecast lead times of 15 to 75 minutes.
- downscaling_model.zip: contains the trained downscaling model with its weights and the means and standard deviations applied during data normalization