Enhancing Snowfall Rate Estimates from Passive Microwave Measurements through Improving Snow Representation in Algorithm

Snowfall plays a crucial role in the global hydrological cycle in mid- and high latitudes. Understanding and quantifying snowfall are essential for gaining insights into Earth’s current weather patterns and climate systems. Utilizing Space-borne microwave measurements, the NOAA NESDIS Snowfall Rate (SFR) product provides snowfall retrievals in two steps: a machine-learning-based algorithm for snowfall detection and a 1DVAR-based algorithm for snowfall rate estimation.

Evaluation has shown good performance of this product, which uses spherical ice habit to represent snow optical properties. In this presentation, we will emphasize the importance of proper ice habit selection in the SFR algorithm for the accuracy of the product. The results indicate that sphere ice habit is the best candidate for most cases. However, it produces significant positive bias under certain conditions, necessitating the use of other ice habits to reduce the bias. We will demonstrate that the most fitting ice habit depends on snow macrophysics, meteorological conditions and the radiances observed by passive microwave sensors. We used advanced machine learning methods to choose the best ice habits for different conditions, which greatly help reduce the biases of snowfall rate estimations from passive microwave observations.