Integration of Active and Passive Multifrequency Data from AMSR-2 and Cosmo SkyMed for Snow Depth Monitoring at High Resolution in Alpine Environments

• Published in: IGARSS 2024 - 2024 IEEE International Geoscience and Remote Sensing Symposium pp. 631 - 634
• Main Authors: Santi, Emanuele, Pettinato, Simone, Paloscia, Simonetta, Pilia, Simone, Baroni, Fabrizio, Ramat, Giuliano
• Format: Conference Proceeding
• Language: English
• Published: IEEE 07-07-2024
• Subjects: Accuracy; AMSR-2; Artificial Neural Network; Artificial neural networks; Cosmo SkyMed; Radio frequency; Radiometers; Random Forest; Satellite broadcasting; Sensors; Snow; Snow Depth
• ISSN: 2153-7003
• DOI: 10.1109/IGARSS53475.2024.10642383

This study aims at improving the spatial resolution of snow depth (SD) products derived from microwave satellite radiometers by proposing a disaggregation method based on X-band SAR data. The method has been developed and tested in the Western part of Italian Alps, by involving Cosmo SkyMed (CSK) and AMSR-2 data. Machine learning methods play a twofold role in the proposed active/passive (A/P) implementation: the AMSR-2 data disaggregation process is indeed based on Artificial Neural Networks (ANN), while the SD retrieval using the disaggregated data is based on ANN and Random Forest (RF) algorithms. To assess the effectiveness of the proposed A/P technique, the SD retrievals have been compared with those obtained by estimating SD directly from CSK data. Taking advantage of the multifrequency information, the retrievals based on A/P method clearly outperformed those based on CSK data only: correlation increased from R=0.77 to R= 0.85 for the ANN based retrievals and from 0.76 to 0.86 for the RF based retrievals. The corresponding RMSE decreases from 34 cm to 28 cm and from 34 cm to 27 cm for ANN and RF, respectively, in a SD range between 0 and ≃ 220 cm.