Deep Learning Based Quantitative Precipitation Estimation over Complex Terrain in the San Francisco Bay Area

The complex terrain of the San Francisco Bay Area presents significant challenges in precipitation estimation from weather radars. The Bay Area has insufficient coverage from the operational NEXRAD radars because of gaps due to terrain blockage and radar beam overshooting due to which the NEXRADs are unable to detect low-level precipitation. The Advanced Quantitative Precipitation Information (AQPI) [1] project aims to enhance precipitation monitoring and estimation in this region by deploying gap-filling scanning X-band radar systems. These X-band systems have dual-polarization capability, can be deployed in regions to fill gaps, and are able to scan the lower atmosphere more rapidly, thereby providing observations at higher spatial and temporal resolution compared to the NEXRADs. Present radar-based rainfall estimation algorithms at the X-band use Specific Differential Phase (K_dp) since, being a phase measurement, K_dp is immune from effects of attenuation and partial beam blockage and is more sensitive to the differential propagation phase compared to NEXRAD S-band systems. The estimation is performed based on rainfall rate versus K_dp (R-K_dp) relationships, which may not yield correct results in areas with bright-band contamination like the San Francisco Bay area. This work addresses these challenges and aims to improve radar-based rainfall estimation using deep learning techniques. Machine learning has been found to improve QPE in recent studies [2]. The study region is focused on Santa Clara valley in California which has coverage from an AQPI X-Band radar named XSCV. Ground-based precipitation gauges are used as ground reference measurements to train the machine learning model. The final QPE product is found to improve on traditional radar-rainfall products. This presentation will discuss the formulation and application of the machine learning model and results are evaluated against independent rainfall products.

References

[1] Cifelli, R., and Coauthors, 2024: Advanced Quantitative Precipitation Information: Improving Monitoring and Forecasts of Precipitation, Streamflow, and Coastal Flooding in the San Francisco Bay Area. Bull. Amer. Meteor. Soc., 105, E313–E331, https://doi.org/10.1175/BAMS-D-21-0121.1.

[2] EunYeol Kim, V.Chandrasekar, Chandrasekar Radhakrishnan, Luca Delle Monache, and Duncan Axisa, “Radar Based Quantitative Precipitation Estimation Using Deep Learning over the UAE region”, 104th American Meteorological Society Annual Meeting, 2024