Australia: The Land Where Time Began

A biography of the Australian continent 

Greenland – Extreme Temperature Events in Observations and the MAR Climate Model

Between 1993 and 2010 meltwater from the Greenland Ice Sheet contributed 1.7-6.12 mm to global sea level and is expected to contribute 20-110 mm to sea level rise in the future up to 2100. Regional climate models (RCMs), which are known to be robust at the ice sheet scale, though they occasionally miss regional- and local-scale variability, produced these estimates (e.g. Leeson et al. 2017; Medley et al., 2013. The fidelity of these models to date, in the context of short-period variability in time, i.e. intraseasonal, has not been assessed fully, such as their ability to simulate extreme temperature events. Leeson et al. used an event identification algorithm that is commonly used in the analysis of extreme values, together with observations from the Greenland Climate Network (GC-Net), in order to assess the ability of the MAR (Modèle Atmosphérique Régional) RCM to reproduces extreme positive temperature events that had been observed at 14 sites around Greenland. Their study found that MAR is able to simulate accurately the frequency and duration of these events though underestimates their magnitude by more than half a degree Celsius/Kelvin, though this bias is much less than the bias exhibited by coarse-scale Era-Interim reanalysis data. Resulting from this, melt energy in output from MAR is underestimated by between 16 and 41% depending on the regional forcing that is applied. According to Leeson et al. further work is required to determine precisely the drivers of extreme temperature events, and the reason the model underperforms in this area, though it is suggested by their findings that biases are passed into MAR from boundary forcing data. As these forcings are common between RCMs and their range of predictions of ice sheet melting in the past and future, this is important. It was proposed by Leeson et al. that examination of these extreme events should become a routine part of global and regional climate model evaluation and that shortcomings in this area should be addressed and should become a priority for development of models.

The Greenland Ice Sheet has shifted from a state of near mass balance to one of significant loss of mass (Shepherd et al., 2012; Hanna et al., 2013a; van den Broeke et al., 2106), and this loss has contributed approximately 10 % to the measured global sea level rise over the last 2 decades (church, 2013). The rate of loss of mass from Greenland has increased since 2010 and the ice sheet has experienced episodes of rare and extreme surface melt (Nghiem et al., 2012; Hanna et al., 2014; Tedesco et al., 2013). E.g. the extent of the summer melt in 2012 reached 98.6% of the entire ice sheet: believed to be the greatest extent of melt in more than 100 years (Nghiem et al., 2012). The dielectric properties of the surface of the ice sheet is also altered by these processes, which makes it more difficult to measure changes in surface height using satellite-borne radar instruments (McMillan et al., 2016). It is therefore necessary to understand the location, frequency, duration and magnitude of melting to:

1)    Understand the response of the ice sheet to climate change,

2)    Interpret contemporary measurements of the volume of the ice sheet change and

3)    Constrain the prediction of the state of the ice sheet in the future.

Mass that has been lost by runoff of meltwater and gained through snowfall together comprises the Surface Mass Balance (SMB) of the ice sheet surface which is typically assessed at the ice sheet-wide scale by the use of regional climate models (RCMs). Regional Climate Models act as physically based interpolators of climate reanalysis data that are relatively coarse resolution and produce estimates of high resolution in areas where spatial variability is exhibited by the local climate, i.e., margins of ice sheets (Noel et al., 2016). A similar purpose is filled by alternative statistical downscaling and they give results that are broadly comparable (Wilton et al., 2017; Vernon et al., 2013). High-resolution predictions of climate in the future can also be made by Regional Climate Models, when boundary forcing is applied by the output from Global Climate Model (GCM) instead of reanalysis data. It was reported by the MAR, RACMO2 and MM5 RCMs from the last IPCC report that, though the Surface Mass Balance remains positive (net mass increase due to surface processes), increases in melting were responsible for a contribution to sea level of 0.23-0.64 mm/yr over the interval 2005-2010 (Church et al., 2013). It is known that Regional Climate Models perform well when compared to integrated quantities, e.g., mean annual melt that is measured at weather stations or the total mass of loss from the ice sheet measured by Grace (van den Broeke et al., 2016). Fidelity at the regional or seasonal scales, however, does not necessarily translate to the local scale (e.g. Medley et al., 2013). There is a tendency for extreme melt events to be localised in time, typically lasting for only a day or so. While the predictions by Regional Climate Model of the extent of melt during the extreme events have been found to be reliable (Tedesco, 2011), an assessment of their ability to simulate the frequency, duration and magnitude of these events, and how their projections of ice sheet change might be affected by this, has not yet been performed.

Advanced statistical techniques for the identification of extreme events were used in this paper in order to compile a statistical climatology of extreme temperature events on Greenland since the 1990s using data acquired from 14 automatic weather stations (AWSs) that form part of the Greenland Climate Network (GC-Net: Steffen et al., 1996). It should be noted that these are distinct from years of extreme melt events as it is possible to have multipole extreme temperature events in a year. These data were then used together with estimates of the temperature from the MAR regional climate model (Fettweis et al., 2017), to evaluate the ability of the model to capture the frequency, duration and magnitude of these events when forced by climate reanalysis and by GCM data. Finally, Leeson et al. estimated the melt energy available at the GC-Net stations during this time by using a positive degree-day (PDD) sum and assess the degree to which the discrepancies between characteristics observed and modelled of extreme events affects estimates that are MAR-based of melt energy.

Extreme temperature events in observations of the GC Net

Given the size of the ice sheet, clear relationship between the characteristics of extreme events and elevation, latitude and melt regime can be seen by Leeson et al. in spite of the relatively small size of the sample. According to Leeson et al. it is not surprising that extreme temperature events exhibit a stronger magnitude at low-lying locations, given the temperature lapse rate of the atmosphere, though it is interesting that this relationship is not as strong for the 5 percolation zone stations in the zones of ablation and dry snow. Leeson et al. speculated that this resulted from the exchange of heat at the surface of the snow moderating near-surface temperatures in this region; sublimation is a known sink for heat in the percolation zone in summer (Ettema et al., 2010). There is a tendency for extreme events at lower elevations in southern Greenland to be more frequent and of shorter duration than those higher up in the ice sheet. Cloudiness (reflecting upwelling longwave radiation back down to the surface) can be associated with temperature anomalies and it is more likely that stations that are lower lying will experience short-term periods of orographic cloud cover. West Greenland is particularly likely to be affected, as it lies in the path of the prevailing summer pattern of circulation and as a consequence receives onshore flow that is moisture-laden during summer (Ohmura & Reeh, 1991). However, in the north of Greenland, extreme events become both longer lasting and more frequent with increase of elevation. It is likely longer extreme temperature events will be associated with high pressure conditions which are relatively persistent. Years of extreme melt on Greenland have been attributed to an increase in the frequency and duration of high pressure conditions promoted by wider-scale atmospheric pressure gradients such as the North Atlantic Oscillation and Greenland Blocking Index (e.g. Nghiem et al., 2012; Hanna et al., 2013b; Lim et al., 2016; Hanna et al., 2016). In the percolation and dry snow zones on the ice sheet extreme temperature events are responsible for the vast majority of melt energy produced, though they contribute a much smaller proportion to overall melt energy in the ablation done. Further work is required to assess whether this is a general property of the ablation zone or restricted to this location, as data is available only for 2 stations in the ablation zone which are located in close proximity; in general temperatures are much warmer here, and extreme events are not required to generate melting.

Extreme temperature events in MAR simulations

The frequency of extreme events is underestimated by all 4 MAR model variants though they simulate their duration well. It is suggested by this that MAR is capable of reproduce persistence of conditions that drive extreme temperature events when they arise in the model. The magnitude of extreme temperature events at most stations is underestimated by all variants of MAR, in most cases by >0.5^oC. In part this can be explained by a general low bias in summer temperatures that are modelled, though the magnitude of this bias is not sufficient to account for the magnitude of the data-model mismatch during extreme periods. E.g., a bias of -0.35^oC during summer and -0.76^oC in extreme temperature events, is exhibited by MAR-Era. A low bias during extreme temperature events at most of the CC-Net stations is also exhibited by raw Era-Interim output, with a notable exception being northeast Greenland and stations that are the most marginal in which temperatures are overestimated during extremes. It is suggested by this that the low bias seen in the MAR model during extreme periods could be an artefact of the forcing data. This is important because Era-Interim and the GCMs that were examined in this study are commonly used to force other regional scale and local scale models (e.g. RACMO2); their use is not restricted to MAR. It is known that the version of MAR which is analysed in this study (v3.5) underestimated the atmospheric liquid water content and therefore cloudiness (Fettweis et. al., 2017) which Leeson et al. also suggest could contribute to the cold bias in temperature extremes. Leeson et al. repeated the analysis with the most recent version of MAR (v3.7) in which a correction had been incorporated for this but there was no noticeable difference in the result. All of the variants of the MAR model and Era-Interim overestimate event magnitude at stations in the ablation zone, JAR and Swiss Camp. This was attributed to albedo differences between the bare ice in the ablation zone and surfaces that were snow covered at high elevations. In areas of bare ice energy exchange is generally more sensitive to sunny conditions; this, according to Leeson et al., is the likely explanation of why the biases are opposite in this area compared to the percolation and dry snow zones where the albedo is high enough to prevent this sensitivity.

MAR simulated energy is underestimated by 19, when forcing is provided by Era-Interim, MICROC5 25, NorESM1 41 and CanESM2 16%. During extreme events, however, model biases in terms of melt energy are double those that are calculated at times of non-extreme, positive, temperature, conditions. As approximately half of all melt energy is generated in extreme events, this is important. In general, MAR simulations that are forced by GCM perform more poorly than the simulation that is forced by Era-Interim, apart from MAR-Can (bias = 16% vs 19% for MAR-Era). According to Leeson et al. they expect the reanalysis-forced simulation to perform the best, given its assimilations of observations; they note, however, that the difference is not large.

Leeson et al. observe that the melt energy that is generated at the 2 inland stations that are the highest/furthest, Summit and NGRIP, but none of the variants of MAR simulate any melting at either of these stations during the period of this study. This results from extreme temperatures being underestimated by ⁓1.0^oC by MAR at these stations (e.g. MAR-Era exhibits a bias f -0.91 at Summit and -0.76 at NGRIP). It is important to note that these are very small quantities and ice-sheet-wide estimates of melting would not be impacted by them; melting is also important, however, because of its role in the albedo of the ice sheet: wet snow is less reflective than dry snow. According to Leeson et al. a significant melt event can be defined as achieving >1 mm w.e./day (Franco et al., 2013), and with the exception of Summit in 2012 this was not achieved at either of these stations during the period of the study. Nonetheless, as the climate warms, it is likely that there will be more abundant melting at these stations, and it will become even more important to properly capture the variability of the temperature.

Conclusions

It was shown by analysis of GC-Net temperature that the frequency, magnitude and duration of extreme temperature events on Greenland are controlled strongly by geography, e.g., elevation, latitude, though further work is required to determine the relative contributions of potential physical drivers of extreme events at different locations and over different periods of time. The duration of extreme temperature events on Greenland was predicted accurately by the MAR regional model, though it underestimates the frequency by about 1 day/yr and underestimates the magnitude by >0.5^oC. Though this is an improvement over the coarse scale reanalysis data, it leads, nonetheless, to an underestimation of melt energy, which was calculated to be 16-41% over the study period, depending on the model forcing that was chosen. Prediction of future melting that was based on MAR is calculated by use of an energy balance method which has been shown to perform well against observations in the past (Fettweis t al., 2917). Since a significant role is played by temperature in the energy balance equations, though melt does not increase linearly with temperature, however, these predictions are likely to be affected by the exaggerated model bias found in this study during extreme events. The determination of the reason the model underperforms in this area and whether other similar models have the same limitation requires further work. This was identified by this study to be a model development priority to ensure that estimates based on MAR of ice sheet change are both comprehensive and robust.

Sources & Further reading

1. Leeson, A. A., et al. (2018). "Extreme temperature events on Greenland in observations and the MAR regional climate model." The Cryosphere 12(3): 1091-1102.