An increasing number of studies have demonstrated significant climatic and ecological changes occurring in the northern latitudes over the past decades. As coupled Earth-system models attempt to describe and simulate the dynamics and complex feedbacks of the Arctic environment, it is important to reduce their uncertainties in short-term predictions by improving the description of both system processes and its initial state. This study focuses on snow-related variables and makes extensive use of a historical data set (1966–1996) of field snow measurements acquired across the extent of the former Soviet Union to evaluate a range of simulated snow metrics produced by several land surface models, most of them embedded in IPCC-standard climate models. We reveal model-specific failings in simulating snowpack properties such as magnitude, inter-annual variability, timings of snow water equivalent and evolution of snow density. We develop novel and model-independent methodologies that use the field snow measurements to extract the values of fresh snow density and snowpack sublimation, and exploit them to assess model outputs. By directly forcing the surface heat exchange formulation of a land surface model with field data on snow depth and snow density, we evaluate how inaccuracies in simulating snow metrics affect soil temperature, thaw depth and soil carbon decomposition. We also show how field data can be assimilated into models using optimization techniques in order to identify model defects and improve model performance.
CORE (COnnecting REpositories)
CORE (COnnecting REpositories)