A new 100 m statistically downscaled Greenland SMB product using the MAR regional climate model and parallelization tools

Abstract:

Regional climate models offer the opportunity to capture the physical processes and feedbacks associated to drivers modulating and controlling the surface mass balance of the Greenland ice sheet. However, the horizontal spatial resolution (of the order of tens of kilometers) is still limiting a factor. Statistical downscaling (rather than dynamical downscaling) can be used to improve the original coarse model predictions. In this work, we follow the approach originally reported by Noel et al., 2016 to statistically downscale runoff, surface mass balance, surface temperature and liquid water production generated by the regional climate model MAR 3.11 (6.5 km) over the whole ice sheet before the period 1949 – 2020 at a spatial resolution of 100 m using the digital elevation model of Greenland provided by the Polar Geospatial Group. For each coarse resolution pixel from the model, a linear regression is computed between the parameter to be downscaled and the elevation. This dependency is, then, applied to the high-resolution (100 m) grid for the statistical downscaling. Our approach was optimized to be run in a multi-core environment using OpenMP and taking advantages of the gdal spatial libraries. The parallelization was performed across grids, so that it is possible to perform the analysis of several grids simultaneously. We assessed the statistically downscaled outputs through in-situ observations of surface mass balance, as well as surface temperature and spatial distribution of surface meltwater obtained from LANDSAT over two selected areas (Jakobshavn and Helheim glaciers). Our results indicate that, in general, the statistically downscaled outputs perform better than the original, coarse resolution outputs. The new high-resolution product will be extremely helpful for the Greenland community. Examples of applications include the combination of these outputs with the recently launched ICESAT2 (e.g., removing compaction signal from elevation changes), the assessment of the MAR outputs through the comparison with remote sensing datasets (e.g., Sentinel) or the possibility to better spatially constrain the transport of meltwater to the ocean at specific locations, to name a few examples.