Methodology:

These simulations explore the effects of spatial model resolution on sea level projections. We take the bedmachine version 3 data set at 500 metre resolution and resample in order to get the required resolution. For each resolution lower than 500 metres, there are thus numerous different bed realisations. We select a number of these at each resolution, as described in Williams et al. (2023). The bed topography is included in each netCDF file, along with a bed_ref scalar coordinate variable to identify it (numerical value used to distinguish between different bed realisations at each resolution). Details of all other datasets can be found in Williams et al. (2023).

The initial states for the forward simulations were performed using codes written in the Matlab version of WAVI, using methods described previously in Arthern et al. 2015 and Arthern and Williams 2017. This method uses inverse methods to find the basal drag and ice viscosity parameters that allow the model to agree as well as possible with satellite observations (included in each netCDF file as two-dimensional WeertmanC and damage fields). Once these flow parameters have been found, these initial states can be used to start temporally evolving forward simulations under different parameter forcings (all data needed to run forward from an initial state are included in the dataset). The forward simulation data were produced using the WAVI ice sheet model (Bradley et al., 2024), WAVI.jl, available at https://doi.org/10.5281/zenodo.14393522. The basal melt was prescribed using a depth-dependence melt rate parameterisation including in the WAVI.jl package, where the initial average melt rate at time t=0 years was calibrated for each initial state using a parameter denoted gammaT, which is included as a scalar coordinate variable in each netCDF file. Outputs include two-dimensional fields of ice surface, ice thickness, dh/dt, basal melt rate and average speeds every year for 150 years.

Data collection:

Initialisation data processing and modelling was done using MATLAB (Version 2018b). Forward simulations were performed using Julia version 1.6.2. The version of the WAVI model used is identified in https://doi.org/10.5281/zenodo.14393522.

Data quality:

The initialisation algorithm used was evaluated on idealised test cases as described by Arthern and Gudmundsson (2010). Arthern et al (2015) shows maps of discrepancy between model velocity and satellite derived velocity, and histograms of the differences between model elevation changes and elevation changes derived using satellite altimetry. The WAVI.jl package has standard tests to check the performance of the code. These are documented in WAVI.jl/test. A value of -9999 indicates there are no data there.