Validation of precipitation phase estimates from CloudSat-CPR across Canada

Abstract

Snow is an important component of the global climate system with significant impacts on local weather, fresh water resources, and energy balance in high latitude cold countries. Therefore precise snowfall monitoring is essential for cold countries such as Canada. Apart from the sampling issues related to access and climate in cold regions, a further significant issue that impacts snowfall monitoring is the accurate detection of precipitation phase. The CloudSat Cloud Profiling Radar (CPR) instrument is highly useful because it provides an estimate of precipitation phase along with retrievals of solid precipitation intensity. Furthermore, the sun-synchronous orbit of CloudSat allows it to have enhanced coverage over the Canadian Arctic. In this study, we validate the precipitation phase retrievals from CloudSat using the present weather information recorded on the ground by human observers (ECCC hourly weather data) from 27 stations across Canada and Precipitation Occurrence Sensor System (POSS) radar at Eureka, both maintained by Environment and Climate Change Canada (ECCC). Probability of Detection (POD), defined as the percentage of coincident CloudSat and ground observations that agree on the precipitation phase (solid, liquid or no precipitation), is used as the metric for validation. Mean POD values of CloudSat in classifying solid, liquid and non-precipitating weather at the 27 stations are 80.8%±1.5, 83.2%±1.9 and 69.8%±0.8 respectively. Binning the collocated CloudSat-ECCC hourly weather observations across Canada by the snowfall rate information available from CloudSat, we find that the accuracy of CloudSat in classifying precipitation phase increases with snowfall rate with a maximum accuracy of 85% for snowfall rates >1 mm/hr. We find that the POD varies with precipitation type, and is inversely proportional to cloud cover, with the lowest POD obtained under the heaviest cloud cover. Also, using binomial and multinomial logistic regression analysis of different physical factors, it is seen that POD of CloudSat is influenced by near-surface reflectivity, near-surface temperature and altitude of the lowest cloud layer. The results from this study imply that CloudSat has high accuracy in classifying precipitation phase and can be used to improve snowfall monitoring in cold regions.