Tracking improvements in remotely sensed snow water equivalent from GlobSnow to the ESA Snow CCI program

Snow water equivalent (SWE) estimates based on microwave remote sensing require information on snow characteristics, including snow grain size, and so often rely heavily on ancillary information. Two decades ago, Pulliainen (2006) described how information from passive microwave (PMW) brightness temperatures and in situ snow depth observations could be combined to better estimate SWE, compared to estimates solely based on PMW data. This innovation led to the GlobSnow (GS) line of SWE products. Research and development of the GS SWE algorithm has continued under the auspices of the European Space Agency Climate Change Initiative – Snow (Snow CCI) program, but version-to-version changes and differences compared to the original GS algorithm are difficult for users to trace. We will present a synthesis of key algorithm changes within the Snow CCI SWE product line and outline the key improvements compared to the GS products.GS v3 served as the baseline algorithm for Snow CCI v1 with Snow CCI v2 marking the first point of divergence between the GS and CCI product streams. Compared to its GS/CCIv1 predecessors, the most recent SnowCCI product (CRDPv4) uses better calibrated PMW brightness temperatures, accounts for seasonally evolving snow density within the SWE retrieval and applies improved masking of snow free areas. Together, these and other changes resulted in improvements of 15% or more in each of the baseline validation statistics (bias, unbiased root mean squared error, correlation) for shallow to moderate snowpacks (SWE ≤200 mm).We will demonstrate how advancements in input data and the characterization of snow properties have led to a more physically sound SWE product with improved accuracy. Notably, incorporating spatially and seasonally evolving snow densities improved the SWE climatology and the timing of peak SWE, while the newer PMW data improved the temporal stability of the SWE time series by removing a known breakpoint in 2009 associated with the move from SSM/I to SSMIS. Improvements to snow masks mean the newest SWE product retains more snow during the onset and melt seasons, when previous GS and CCI products were known to underestimate SWE and snow extent. These changes, implemented under the Snow CCI program, have resulted in a multi-decadal satellite-derived dataset with comparable performance and accuracy to the current generation of land reanalysis systems.Pulliainen, J., 2006. Remote Sens. Environ. 101, 257–269, doi: 10.1016/j.rse.2006.01.002.