Central tropical Pacic convection drives extreme high temperatures and surface melt on the Larsen ice shelf

The northern parts of the Larsen ice shelf, eastern Antarctic Peninsula (AP) have experienced 23 dramatic break-up since the early 1990s as a result of strong summertime surface melt, which has 24 been linked to stronger circumpolar westerly winds. Here we show extreme summertime surface 25 melt and high temperatures over the eastern AP and Larsen C ice shelf occur because of 26 enhanced deep convection in the central tropical Pacific, which produces cyclonic conditions 27 across the middle and high-latitude South Pacific, and a strong high pressure anomaly over 28 Drake Passage. Together these transport extreme heat and moisture from low latitudes to the AP, 29 at times in the form of "atmospheric rivers", producing strong surface warming and melt on the 30 Larsen ice shelf by the Foehn effect. Therefore, variability in central tropical Pacific convection 31 is crucial for interpreting past and projecting future AP surface mass balance and extreme 32 temperature events.


Introduction 35
The Larsen ice shelf, located on the eastern Antarctic Peninsula (AP) (Fig. 1a) has lost around 36 18,000 km 2 (20%) of its surface area since 1995 1-3 . The northernmost section, the Larsen A ice 37 shelf, disintegrated in the summer of 1995 4 , which was shortly followed by the collapse of the 38 larger Larsen B ice shelf to its south in the summer of 2002 5,6 . Farther south, the Larsen C ice 39 shelf is the largest remaining section, which has been thinning 7 and experiencing dramatic 40 rifting and calving in recent years, including a 5,800 km 2 section that broke away in 2017 41 forming what was the planet's largest iceberg (A-68) 3 . The loss of these ice shelves has caused 42 thinning and acceleration of the AP's interior glaciers 6,8,9 resulting in an increasing rate of 43 contribution to global sea level rise; approximately 20% (2.5 mm) of Antarctica's total sea level 44 rise contribution since 1979 has come from the AP 10 , and mass loss from the AP has increased 45 by around 15 Gt yr -1 since 2000 11 . 46

47
The dramatic loss of ice shelves and subsequent ice sheet retreat on the eastern AP is linked to 48 extreme high summer surface air temperatures and surface melt events 12-14 , which can lead to 49 ponding and hydrofracturing 15 due to the relatively thin and dry snowpack on the Larsen ice 50 shelf 16 . Indeed, the loss of the Larsen A and B sections were preceded by anomalous high 51 summer surface air temperatures and surface melt 13,14 , which led to the collapse of the Larsen B 52 ice shelf over a period of just a few weeks 17 . 53 54 These extreme high surface air temperature events on the eastern AP are commonly associated 55 with Foehn winds 18-20 , during which warm, moist westerly flow ascending the steep 56 mountainous slope of the western AP cools slowly at the saturated adiabatic lapse rate (when 57 saturated) and acquires a high potential temperature, and then warms rapidly at the dry adiabatic 58 lapse rate while descending the eastern slope of the AP. This can lead to localized regions of 59 very high surface air temperature and surface melt on the Larsen ice shelf 19,21,22 . Therefore, the 60 strong surface melt events that triggered the collapse of Larsen A and B have been qualitatively 61 shelf. As the circumpolar westerly winds are projected to continue strengthening over the 76 remainder of this century due to increasing greenhouse gases 30 , it is crucial to reconcile the 77 dominant processes leading to summer surface melt on the remaining Larsen C ice shelf in order 78 to interpret past, and make reliable future projections of, AP surface mass balance and associated 79 sea level rise.   (Table 1). 92

93
The AP climate is known to be influenced by teleconnections originating from the tropics 31,32 . 94 However, these relationships are strongest on the western AP and generally associated with 95 anomalous thermal advection tied to variations in the depth of the Amundsen Sea Low 29 , and 96 they are generally weak during the austral summer season due to weak Rossby wave sources in 97 the sub-tropics and unfavorable jet stream configurations which inhibit Rossby wave propagation 98 from the tropics into the southern high-latitudes 33 . Instead, studies have found the AP climate 99 during summer to be more strongly tied to variability in the SAM and associated fluctuations in 100 the strength and position of the mid-latitude jet 24,29 . The caveat is that this relationship between 101 AP climate and the tropics/SAM is derived from a limited number of weather stations that are 102 confined to the northern tip of the AP 34 (Fig. 1a), over 100km north of the Larsen C ice shelf. 103

104
In contrast to temperature variability across the northern tip of the AP, summertime surface melt 105 on Larsen C is not significantly correlated with the SAM index (Table 1;  C. On the other hand, the decrease in summer surface melt is consistent with changes in the 113 tropical Pacific and regional circulation over Drake Passage: concurrent with the decrease in 114 surface melt, EPAC SST significantly cooled ( Fig. 1c; p=0.05), there was a slight reduction in 115 CPAC convection ( Fig. 1e; p>0.10), and Drake Z500 slightly decreased ( Fig. 1i; p>0.10). 116 117 The regional circulation around the AP associated with enhanced CPAC convection and Larsen 118 C surface melt (Fig. 2) is consistent with Foehn events on the eastern AP. Both Larsen C surface 119 melt and CPAC OLR are associated with a positive pressure anomaly over Drake Passage 120    anticyclone. With the lower latitude cyclonic anomaly sourcing an air mass from nearly sub-182 tropical latitudes, an atmospheric river was again triggered that made landfall on the western AP 183 on 6 February 2020, nine hours prior to the record-high temperature event (Fig. 3f). In addition 184 to setting the new record high surface temperature, the AP experienced its highest surface melt 185 on record for early February during this event affecting more than 50% of the region 35 .    Atmospheric circulation was investigated with the European Centre for Medium-range Weather 415 Forecasting (ECMWF) fifth generation atmospheric reanalysis ERA5 42 . ERA5 replaces the 416 ECMWF Interim (ERA-Interim) reanalysis, which was considered to be the best reanalysis for 417 depicting Southern Hemisphere high-latitude climate 43 , with improved model physics, core 418 dynamics, and data assimilation as well as a higher horizonal resolution. We use monthly-mean 419 ERA5 fields for the period to compute the seasonal means employed in the correlation analysis, 420 and 6-hourly fields to compute the five-day group means investigated for the two case studies.

Detection of atmospheric rivers 498
The atmospheric river (AR) detection algorithm used in this study is a modified version of the 60 499 method. Identification of ARs is based on the intensity and geometry of integrated water vapor 500 transport (IVT) sourced from the ERA5 reanalysis dataset. The algorithm uses 6-hourly IVT data 501 with a horizontal grid resolution of 0.25°x0.25°. The algorithm identifies regions of enhanced 502 IVT whose shape and IVT direction are consistent with the AR definition. Regions of enhanced 503 IVT are defined where IVT magnitude in a given month exceeds the 85th percentile IVT for the 504 3-month period centered on that month. A fixed lower limit of 150 kg m -1 s -1 is also imposed. 505 Contiguous regions of grid cells with IVT values above the percentile threshold and fixed lower 506 limit are isolated and labelled. An AR is considered to be a landfalling AR if a region of 507 enhanced IVT intersects a grid cell occupied by the Antarctic Peninsula. The object axis is then 508 calculated following 61 . Axis calculation can be described as follows: 3. If the IVT threshold is exceeded this grid cell is labelled as the new target grid cell e, and 517 we repeat step 2). This process is continued until the upstream grid cell fails to exceed the 518 threshold or a grid cell is detected twice. 519 520 The length of the object is computed as the sum of the distances between neighboring axis cells. 521 To be identified as an AR, the following geometry criteria must be satisfied: 522 523 1. Length Check: The length of the object must exceed 2,000 km. 524 2. Narrowness Check: The width of an object is defined as its surface area divided by its 525 length. An object is discarded if its length/width ratio is less than 2; 526 3. Mean Meridional IVT Criterion: An object is discarded if the mean IVT does not have a 527 poleward component greater than 50 kg m -1 s -1 . This filters objects that do not transport 528 moisture toward higher latitudes. Date Store (https://cds.climate.copernicus.eu/#!/search?text=ERA5&type=dataset). ERSSTv5 571 and OLR data are available online from the NOAA Physical Sciences Laboratory 572