Testing snow water equivalent retrieval algorithms for passive microwave remote sensing in an alpine watershed of western Canada

Brightness temperatures (TBs) from the special sensor microwave imager (SSM/I) and advanced microwave scanning radiometer (AMSR-E) from 2003 to 2007 are utilized to retrieve and evaluate the snow water equivalent (SWE) over the complex terrain of the Quesnel River Basin (QRB), British Columbia, Canada. Various algorithms including the Environment Canada (EC) algorithms, the spectral polarization difference (SPD) algorithm, and an artificial neural network (ANN) for both SSM/I and AMSR-E are evaluated against in situ SWE observations using several statistical metrics. The results show that the EC algorithms developed specifically for the southern prairies and boreal forest perform poorly across the complex topography and generally deep snow of the QRB. For other frequency combinations of SSM/I and AMSR-E measurements, significant relationships between TB difference and in situ SWE exist only when the snow accumulation is less than a threshold of 250 or 400 mm, which varies at the different in situ stations. Overall, AMSR-E provides better estimates of retrieved SWE than SSM/I. Compared to the algorithms based on TB difference, the ANNs for SSM/I and AMSR-E perform much better. The ANNs trained with all channels of AMSR-E have the best performance in fitting SWE and are able to resolve the temporal variations of SWE at all in situ stations. However, due to the complexity of the topography and vegetation in this mountainous watershed, the ANNs based only on limited in situ stations are not able to retrieve the spatial variations of SWE in this area.