Assimilation of Satellite-Based Rainfall Estimations in the Canadian Precipitation Analysis

The Canadian Precipitation Analysis (CaPA) produces a gridded product by assimilating data from surface stations and radar, using a background field provided by the Global Environmental Multiscale (GEM) model. This study assesses the performance of two satellite-based rainfall estimates for Canada and the results of their assimilation within CaPA. Evaluations are completed on 10 years of rainfall data for June, July, and August. The satellite-based rainfall estimates considered are those from the Climate Prediction Center morphing (CMORPH) method and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN). Relative to the second generation of adjusted daily precipitation for Canada (APC2), it is found that both satellite products possess unique characteristics within central Canada in the form of greater skill on par with the greatest observed rainfall across all of Canada, and an unrivalled ability to detect large rainfall events (between 2 and 50 mm/day) within the scope of this study. This exceptional performance is correlated in space and time with the occurrence of convective rainfall events. Comparing the two satellite products' categorical scores, it was found that CMORPH is better at detecting rainfall events and estimating the events' magnitude than PERSIANN. Two CaPA configurations are tested: one where CMORPH is used as a background field in place of the GEM model and one where it is used as an additional data source. Combining CMORPH with stations results in an overall increase in skill, except for the detection of light rainfall events less than 1 mm/6 h. When CMORPH is used alongside stations to adjust the GEM model, the resulting analysis is a combination of the strengths of both gridded datasets used, in that light rainfall events are influenced by the GEM model and larger events by CMORPH. The ability of CMORPH to detect and estimate convective rainfall events within central Canada, specifically events greater than 2 mm/6 h, has proven particularly useful when assimilated with the GEM model. (C) 2017 American Society of Civil Engineers.