Understanding the Phases of Precipitation

Different phases of precipitation would bring different hydrological impacts. For example, rainfall would become streamflow and increase surface runoffs, while snow would accumulate and increase surface albedo, acting as a crucial driver in the climate system. Therefore, understanding whether precipitation would fall in solid or liquid phase is fundamental to hydrological modeling and forecasting. The precipitation phase is also a critical factor for the surface precipitation retrieval by satellite-based radars, since solid and liquid particles feature different scattering properties and correspond to different relationships between the radar reflectivity and the precipitation rate. With surface temperature warming over the past century, the precipitation phase is expected to shift from solid to liquid, but a global view of the trends has not been established. Moreover, in order to obtain an improved phase classification scheme for satellite snowfall retrieval, it is crucial to investigate the factors that impact the phase transition process and understand the underlying physics. In this thesis, we first investigated the global means and trends of annual precipitation frequency and the ratio of snow events to precipitation events (SE/PE ratio) using land station and shipboard synoptic present weather reports from 1973 to 2019. Results show that when averaged over all qualified land stations, the annual rain frequency and annual snow frequency both have an increasing trend. When averaged over the shipboard reports, the annual rain frequency increases while the annual snow frequency decreases over the 47 years. Over both land and ocean, the averaged SE/PE ratio has a significant decreasing trend. Moreover, the trend of SE/PE ratio shows a strong latitudinal dependence. At the mid and low latitudes in the Northern Hemisphere, the snow frequency increases slower than the rain frequency, and the SE/PE ratio has a decreasing trend. In contrast, at high latitudes, the snow frequency increases faster than the rain frequency, and the SE/PE ratio has an increasing trend. Then, we focused on the physics of the phase transition process. Since the Global Precipitation Measurement (GPM) uses ECMWF Reanalysis Version 5 (ERA5) for their snowfall retrieval, we combined ERA5 with surface observations to investigate the performance of reanalysis dataset in determining precipitation phase. Results based on the station and ship observations were also presented for understanding the physics. The impact of temperature and wet-bulb temperature on the precipitation phase was first examined. On average, the temperature threshold, defined as the conditional probability of snow at 50%, is 1.36°C for land and 1.61°C for ocean based on the ERA5 reanalysis, both around 0.3°C colder than the thresholds derived from observations. The wet-bulb temperature thresholds for land and ocean are 0.72°C and 0.14°C based on the reanalysis, and 0.73°C and 0.83°C for the observations. Using wet-bulb temperature, which incorporates moisture in the calculation, reduces uncertainties in the phase classification. The spatial pattern of the wet-bulb temperature thresholds is also established. Over most of the regions, the wet-bulb temperature threshold falls between 0°C and 1°C. However, this threshold has considerable variations over the globe; derivations from global mean value are particularly large over mountainous areas, coastal regions, and warm ocean currents. Over land, colder wet-bulb temperature thresholds are observed in most high-elevated regions except for the Midwestern U.S., where the thresholds are generally warmer than 2°C. Over ocean, the wet-bulb temperature threshold is colder where there is a warm current, and is warmer at very high latitudes. The influence of different geophysical variables on the precipitation phase was examined. Lapse rate is found to impact the snow conditional probability significantly. At a given wet-bulb temperature, a larger lapse rate leads to a larger conditional probability of snow. In the Midwestern U.S., a smaller pressure would lead to larger snow conditional probability since hydrometeors fall faster in thin air, but this pressure dependence is not evident outside the U.S. For the skin temperature and the difference between the near-surface temperature and skin temperature, there are discrepancies between results based on reanalysis and that based on observations. Further investigation into this issue is expected in the future, in order to develop an improved phase classification scheme based on the reanalysis data.