Temporal stability of long-term satellite and reanalysis products to monitor snow cover trends

Monitoring snow cover to infer climate change impacts is now feasible using Earth Observation data together with reanalysis products derived from Earth system models and data assimilation. Temporal stability becomes essential when these products are used to monitor snow cover changes over time. While the temporal stability of satellite products can be altered when multiple sensors are combined and due to the degradation and orbital drifts in each sensor, the stability of reanalysis datasets can be compromised when new observations are assimilated into the model. This study evaluates the stability of some of the longest satellite-based and reanalysis products (ERA5, 1950-2020, ERA5-Land, 1950-2020, and NOAA CDR, 1966-2020) by using 527 ground stations as reference data (1950-2020). Stability is assessed with the time series of the annual bias in snow depth and snow cover duration of the products at the different stations. Reanalysis datasets face a trade-off between accuracy and stability when assimilating new data to improve their estimations. The assimilation of new observations in ERA5 improved significantly its accuracy during the recent years (2005-2020) but introduced three negative step discontinuities in 1977-80, 1991-92, 2003-04. By contrast, ERA5-Land is more stable because it does not assimilate directly snow observations, but this leads to a worse accuracy despite having a finer spatial resolution. NOAA CDR shows a positive artificial trend since 1990-1995 in fall and winter that could be related to changes to the availability of satellite data. The magnitude of most of these artificial trends/discontinuities is larger than actual snow cover trends and Global Climate Observing System (GCOS) stability requirements. Using these products in seasons and regions where artificial trends/discontinuities appear should be avoided. The study also updates snow trends (1955-2015) over local sites in the North Hemisphere (NH) corroborating the retreat of snow cover, driven mainly by an earlier melt and recently by a later snow onset. In warmer regions such as Europe, snow cover decrease is coincident by a decreasing snow depth due to less snowfall, while in drier regions such as Russia earlier snowmelt occurs despite increased maximum seasonal snow depth.

URRACA VALLE Ruben;  GOBRON Nadine; 

2023-03-06

COPERNICUS GESELLSCHAFT MBH

JRC126106

1994-0416 (online),   

https://tc.copernicus.org/articles/17/1023/2023/,    https://publications.jrc.ec.europa.eu/repository/handle/JRC126106,   

10.5194/tc-17-1023-2023 (online),