De-tuning a coupled Climate Ice Sheet Model to simulate the North American Ice Sheet at the Last Glacial Maximum 

Abstract

<p>Coupled climate-ice sheet models are crucial to evaluating climate-ice feedbacks' role in future ice sheet evolution. Such models are calibrated to reproduce modern-day ice sheets, but current observations alone are insufficient to constrain the strength of climate-ice feedbacks. The extent of the Northern Hemisphere ice sheets during the last glacial maximum, ~20,000 years ago, is well known and could provide a benchmark for calibrating coupled climate-ice sheet models. We test this with the FAMOUS-ice coupled Climate-Ice Sheet model (Smith et al., 2020), a fast GCM coupled to the Glimmer ice sheet model. We ran Last Glacial Maximum simulations using FAMOUS-ice with interactive North American Ice Sheet, following the PMIP4 protocol (Kageyama et al., 2018). We find that the standard model setup, calibrated to produce a good present-day Greenland (Smith et al., 2020), produced a collapsed North American ice sheet at the Last Glacial Maximum. We ran ensembles of hundreds of simulations to explore the influence of uncertain ice sheet, albedo, atmospheric, and oceanic parameters on the ice sheet extent. The North American continent deglaciated rapidly in most of our simulations, leaving only a handful of useful simulations out of 280. We thus developed a method to efficiently identify regions of the parameter space that can produce a reasonable ice-sheet extent. This involved emulating the equilibrium ice volume and area as a function of the surface mass balance at the start of our simulations. We then ran three waves of short simulations for 20-50 years to identify parameter values and surface mass balance conditions potentially suitable to grow a realistic ice sheet. This enabled us to find ~160 simulations with good ice extent.</p><p>Through analysis of these simulations, we find that albedo parameters determine the majority of uncertainty when simulating the Last Glacial Maximum North American Ice Sheets. The differences in cloud cover over the ablation zones of the North American and Greenland ice sheet explains why the ice sheets have different sensitivities to surface mass balance parameters. Based on our work, we propose that the Last Glacial Maximum can provide an “out-of-sample” target to avoid over calibrating coupled climate-ice sheet models to the present day.</p><p><strong>References:</strong></p><p>Kageyama, M. <em>et al.</em> The PMIP4 contribution to CMIP6 – Part 4: Scientific objectives and experimental design of the PMIP4-CMIP6 Last Glacial Maximum experiments and PMIP4 sensitivity experiments. <em>Geosci. Model Dev.</em> <strong>10</strong>, 4035–4055 (2017).</p><p>Smith, R. S., George, S., and Gregory, J. M.: FAMOUS version xotzt (FAMOUS-ice): a general circulation model (GCM) capable of energy- and water-conserving coupling to an ice sheet model, Geosci. Model Dev., 14, 5769–5787, https://doi.org/10.5194/gmd-14-5769-2021, 2021.</p><p> </p>

De‐Tuning Albedo Parameters in a Coupled Climate Ice Sheet Model to Simulate the North American Ice Sheet at the Last Glacial Maximum

“…The range of values for the climate and ice sheet parameters were derived from previous uncertainty quantification work with FAMOUS (Gregoire et al, 2010) and Glimmer (Gregoire et al, 2016), with the addition of the Entrainment rate coefficient varied in Murphy et al (2004). We purposefully chose wide but plausible ranges (Table 1) with the aim to identify a region of the parameter space that would produce a reasonable North American ice extent at the LGM.…”

“…We use FAMOUS‐ice (R. S. Smith et al., 2020), a coarse resolution, fast running AOGCM, which has been used in long‐palaeo simulations (Dentith et al., 2019; Gregoire et al., 2012; Gregory et al., 2012; Roberts et al., 2014; R. S. Smith, 2012) and uncertainty quantification (Gregoire et al., 2010). Rather than previous work using a PDD SMB scheme (Gregory et al., 2012), FAMOUS‐ice is coupled to the Glimmer Ice Sheet model by downscaling SMB calculations (R. S. Smith et al., 2020).…”

De-tuning a coupled Climate Ice Sheet Model to simulate the North American Ice Sheet at the Last Glacial Maximum

“…Gregory et al, 2012;Roberts et al, 2014;Dentith et al, 2019;L. J. Gregoire et al, 2012) and uncertainty quantification (L. Gregoire et al, 2010), coupled to the Glimmer Ice Sheet model by downscaling SMB calculations (R. S. Smith et al, 2020), rather than previous work using a PDD SMB scheme (J. Gregory et al, 2012).…”

“…Dentith et al, 2019;L. J. Gregoire et al, 2012) and large ensembles for calibration or uncertainty quantification purposes (L. Gregoire et al, 2010). The Glimmer ice sheet model is a fast simplified 3D dynamical ice sheet model based on the shallow ice approximation that is used to simulate continental ice sheets over glacial interglacial cycles (L. J.…”

“…We started by running a large 280-member ensemble of simulations varying parameter values of Table 1. These values were sampled using k-extended Latin Hypercube sampling (Williamson, 2015) within ranges derived from previous uncertainty quantification work with FAMOUS (L. Gregoire et al, 2010) and Glimmer (L. J. Gregoire et al, 2016), with the addition of the Entrainment rate coefficient.…”