Developing a Glacial Surface Model for Greenland to Improve the
Projections of Surface Runoff
Abstract
Over the past several decades, the Greenland Ice Sheet has been losing
mass through a combination of increased surface runoff and accelerating
ice flux to the ocean. Our understanding of the surface component is
drawn heavily from satellite observations and climate models. The MAR
(Modèle Atmosphère Régional) model is a 3D regional climate model used
extensively over Greenland. Our study focuses on the surface snow and
the ice down to 15-meter in depth. A light-weighted surface model for us
to integrate the local observation data and force many simulations is
needed. Our goal is to implement a surface-only model, derived from MAR,
as a tool for understanding the glacial surface components,
correlations, and MAR biases to improve projections of surface runoff.
This model includes the ability to integrate observations from surface
weather stations, translate the data into a model forcing format, force
different simulations with various configurations or datasets, visualize
model outputs, find key correlations between atmospheric drivers and
modeled firn densification. In the model development, we extract the
surface code from the original MAR for the simulations initialized and
forced with the following snow and atmospheric fields: snow depth,
temperature, density, water volume, and grain size. We then verify that
the surface model generates the same outputs as the full MAR does if
fetched with the identical data. The bias is checked with snowpack
time-depth plots for multiple sites around Greenland, including Summit
and Swiss Camp. We have found a very small bias when compared to the
fully-coupled MAR. We perform quality control for the data inputs, such
as replacing missing data from the station measurements, defining the
max and min for each dataset, filtering out the data outliers by
statistics standard deviations. As the result, our model software can
provide multiple simulations in sequential and concurrent mode with
user-friendly interfaces, and run robustly. The model’s first release is
currently being deployed over different sites across Greenland to
understand the importance of atmospheric forcing versus snow model
biases in projections of future mass loss due to surface melt.