Developing A Novel Long-Term Ml-Based Precipitation Product From Avhrr In High Latitudes Using Various Observational Datasets And Auxiliary Variables From Reanalysis
Presenter: Omid Zandi P291
Co-Author(s): Kwabena Kingsley Kumah, Ali Behrangi
Advisor(s): Ali Behrangi
1Hydrology & Atmospheric Sciences
Accurate long-term precipitation estimation is critical to study Earth's hydrologic cycle and energy budget, particularly in high latitudes, where the effect of global warming is significant and vital freshwater reserves are stored in icecaps and glaciers. Unfortunately, in situ observation of precipitation occurrence and intensity is very sparse and uncertain and, due to the lack of high-quality geostationary observations, satellite data records are relatively short in these regions. Here, we present a novel Machine Learning (ML) based precipitation retrieval method that uses brightness temperature and cloud properties from the Advanced Very High-Resolution Radiometer (AVHRR) observations via the PATMOS-X product, environmental information from the MERRA-2, and surface type data from AutoSnow as major input features. Precipitation estimates from coincident CloudSat, GMI MW-based precipitation retrievals are used for training and verification. Where needed, ERA5 precipitation estimates were also introduced. The product is developed at 0.1-degree latitude-longitude poleward of latitude 45 degrees N/S to complement the long-term precipitation record offered by geostationary satellite data. Leveraging the continuous and consistent AVHRR observations across multiple platforms over more than four decades, this product provides a comprehensive long-term precipitation data record in high latitudes that will be helpful to the Global Precipitation Climatology Project (GPCP) product as well as other precipitation products that tend to extend their record back in time, beyond that offered by passive microwave products. This presentation introduces the product and provides early results including evaluation and comparison analyses.