The sinking of the Titanic is typically blamed on human, design and construction errors. But what of the iceberg that did the damage? Was it a lone berg, or one of many in the area? And should it have been that far south? Earth systems scientist Grant Bigg and systems engineer Steve Billings assess the data.
Titanic on her sea trials, 2 April 1912. Navigation Center, United States Coast Guard
At 11.40 pm, ship's time (0240 GMT), on the cold, moonless night of April 14th, 1912, the RMS Titanic was 400 miles from land, near 41°47’ N, 49°55’ W in the Atlantic Ocean, when the crow's nest lookouts on board sighted a large iceberg 500 metres ahead. Despite quick action on the bridge to slow the ship and turn to port, as well as the closing of the watertight doors, the slow response of a large vessel meant that the ship still struck the iceberg aft of the bows. Some 100 metres of her hull below the waterline buckled, allowing water to food into the ship across several compartments. In little more than 2 1/2 hours she had sunk, with the loss of 1514 lives. It is most likely that a combination of human errors associated with the captain not reducing speed – despite a number of ice reports reaching the vessel in the days before the collision – and possible variable rivet quality in the hull manufacture led to the tragedy. However, the question has often been raised: was the Titanic unlucky in sailing in a year with exceptional iceberg numbers?
In 1912, ice warnings largely relied on information exchanged between ships at sea. Today there is an extensive ice hazard warning service in the northwest Atlantic, provided by the International Ice Patrol (IIP) and a number of commercial and government ice or weather services. These rely on a combination of sightings, satellite imagery, radar and iceberg trajectory models, and have led to a dramatic drop in the number of collisions of ships and icebergs since 1913. But better monitoring is not the only factor involved in risk assessment. The number of icebergs flowing into shipping lanes is also important; and the more that icebergs are created from glaciers (in a separation process known as calving), the more likely that some will escape monitoring, or that human error will come into play during close encounters between icebergs and ships.
Iceberg numbers in 1912
During the first fortnight of April 1912, before the Titanic tragedy, a number of reports of ice were exchanged between ships in the northwest Atlantic. Much sea ice and many icebergs were present in the shipping lanes crossing the Labrador Sea, off Newfoundland and the Atlantic seaboard. But was 1912 an exceptional year for icebergs? To answer this, it is first necessary to examine the variation in iceberg numbers in the North Atlantic over the last century.
The IIP – part of the US Coast Guard – has operated since 1913, collecting data on iceberg locations and sea-ice extents off the Grand Banks of Newfoundland to provide ice navigation hazard warnings to shipping, and so prevent a repeat of the Titanic disaster. The observational methods have changed significantly over the years, from ship reports and dedicated cruises in the early years, through aircraft patrols in the mid-twentieth century, to satellite image analysis and iceberg modelling in recent times1.
There is scope for inaccuracies because of the inconsistent nature of the data collection. The few alternative data sources that there are also have issues of reliability, but the comparisons that are possible support the general magnitude and yearly variability of the IIP reports.
Since 1914 the IIP has recorded a simple measure of the volume of icebergs encountered in a given year: called I48N, it is the monthly number of icebergs passing 48°N across a line from Newfoundland to approximately 40°W (Figure 1). This includes any iceberg greater than 5 metres in above-surface length (see ‘Fact file', below). The data shows great variability from year to year (Figure 2), which reflects the changeable rate of ice calving from western Greenland (though there is indication of episodic increases in the calving flux in recent decades, probably due to increases in both sea surface temperatures in Greenland fjords and ice sheet surface meltwater; we explore this a little more in the box on page 10).
Fact file: Icebergs
Most North Atlantic icebergs originate from around 27 calving sites in Greenland. Floating ice hazards can be classified by size. In seafarer's parlance “growlers” are small fragments of ice that are roughly the size of a truck or grand piano. They extend less than 3 feet above the sea surface and occupy an area of about 215 square feet.
A “bergy bit” is larger. Its height is generally greater than 3 feet but less than 16 feet above sea level and its area is normally about 1000–3000 square feet. Anything larger than a bergy bit qualifies as an iceberg.
Arctic icebergs melt within 5 years of being calved from their glacier; most melt over the first year. During their lifetime most, but not all, overturn several times.
Icebergs are frozen freshwater; their density compared with that of seawater means that they float with 13% of their mass above water and 87% submerged. The pointed shape of many bergs means that does not translate exactly into heights above and below water. For many Arctic bergs the draught is more like 5 times the height above sea level.
Icebergs erode preferentially along the waterline: there is mechanical erosion by wave action, and the water itself is a source of heat transfer for melting (sea temperature will be above freezing point; air temperature need not be). This can lead to submerged platforms extending out just below the waterline, and to the characteristic shapes of bergs, especially after they have overturned.
This unequal melting above and below the waterline can lead to instability, and to the iceberg overturning. The Weeks-Mellor criterion relates the length of an iceberg along its longest axis, La, to its overall height (from lowest submerged point to peak), T. For an iceberg to be stable,
But what of the iceberg risk in 1912? In that year, 1038 icebergs were observed to cross 48°N. However, this number does not even reach the 90th percentile of the annual distribution – in the 112 years shown in Figure 2, 14 recorded an I48N exceeding this number. There are several years in surrounding decades with similar numbers, and many years since the mid-1970s have exceeded this number. Over 1200 icebergs crossing 48°N were recorded in 2009, and 2014 already looks to be a significant ice year, for example. The iceberg risk in 1912, in terms of numbers entering the northwest Atlantic shipping lanes, was therefore large, but not unprecedented.
The origin of the Titanic iceberg
Few icebergs have been tracked from their source to Newfoundland waters; but experts believe that the vast majority of icebergs in the main western Atlantic stream in the Labrador Sea originate from southern, western or northwestern Greenland. This is consistent with the (limited) distributional data that we have, with ocean circulation, and with previously modelled iceberg trajectories.
The iceberg thought to have been hit by Titanic, photographed by the chief steward of the liner Prinz Adalbert on the morning of 15 April 1912. The iceberg was reported to have a streak of red paint from a ship's hull along its waterline on one side. Navigation Center, United States Coast Guard
The iceberg that sank the Titanic at 42°N was relatively large at the time of impact. Reports from survivors estimated it to be 15–31 metres high and 122 metres long. The rescue vessel, RMS Carpathia, reported sailing through ice up to 61 metres high on the way to the rescue, and on the following day. While the density of ice relative to water suggests that only 13% of an iceberg's mass should be above water, the eroded shape of most bergs means that the depth-to-height ratio is more like 5:1, so the Titanic iceberg is likely to have been 90–185 metres deep, while being approximately 125 metres long. The Weeks–Mellor stability criterion2 enables us to tie down the iceberg's size more tightly. As an iceberg is eroded or melted preferentially from the side, its centre of gravity eventually becomes too high for the iceberg to remain upright and it rolls over (see ‘Fact file', page 7). If the reported length of 125 metres is assumed to be roughly correct, then this stability constraint suggests that the vertical thickness of the iceberg could not have been greater than 100 metres, putting the likely above-water height to be around 15–17 metres, with a mass of 2 million tonnes. This is consistent with the dimensions of an iceberg with a red paint streak photographed by Captain de Carteret of the CS Minia the day after the sinking.
The iceberg risk in 1912, in terms of numbers entering the northwest Atlantic shipping lanes, was large but not unprecedented
Arctic icebergs of course get smaller as they drift south and start to melt. For an iceberg to still be more than 100 m long as far south as 42°N suggests that it began life as a large iceberg when it calved into a Greenland fjord.
Our ocean iceberg modelling (see ‘Technical notes', page 9) produced a range of possible sources and trajectories for icebergs reaching the general area of the Titanic's sinking within 3 months, either side, of the collision. The modelled iceberg passing closest to the sinking site around the correct date was calved from southwest Greenland in early autumn 1911 and moved across the Labrador Sea, rather than being swept into the circulation system further north, in Baffin Bay. It began life roughly 500 metres in length by 300 metres in depth and 75 million tonnes in weight, but had melted to 2.1 million tonnes by mid-April 1912 – which is remarkably close to the estimated size from observations from Titanic itself and from other ships linked with the disaster.
Technical notes: Modelling iceberg distribution
We have studied the distribution of icebergs in the Atlantic during the twentieth century by using a coupled ocean iceberg model3. This is basically a daily-forced ocean circulation model with an in-built dynamical and thermodynamical iceberg trajectory model, in which the icebergs, seeded into the ocean from glacier outlets, are regarded as points carried by the currents of the ocean model, and supplying the ocean model with freshwater as they melt. The iceberg model has been well tested in both the Arctic and Antarctic. The annual calving rate in the ocean iceberg model from the 27 major calving sites around Greenland was set proportional to the magnitude of the I48N series shown in Figure 2, as this produces an excellent correlation of the model iceberg flux at 48°N with this series (Figure 3). We are thus able to model the likely iceberg trajectories of 1912, within the limitations of the forcing and the model. According to our model 75% of icebergs reaching 48°N during the first third of the twentieth century came from southern Greenland, making this route the norm at the time rather than an exception.
The iceberg hazard
The year 1912 was one of raised but not – in the long term – exceptional hazard, in terms of the number of icebergs counted. As we have seen, 1038 icebergs crossed 48°N that year. In the surrounding decades (1901–1920) there were five years with at least 700 icebergs crossing 48°N, and 1909 recorded slightly higher numbers than 1912. More recently, the risk has been much greater – between 1991 and 2000, eight of the 10 years recorded more than 700 icebergs and five of those years exceeded the 1912 total. Several other periods during the twentieth century experienced iceberg risk at a level similar to, or greater than, 1912. While the uncertainty in the early numbers will be higher, the continuous need to accurately monitor this hazard for shipping suggests that the I48N series for icebergs is generally reliable.
But compounding the iceberg numbers in 1912, the weather, ice conditions and time of year combined to increase the hazard on that fateful day. High pressure had dominated the mid-latitude, central Atlantic atmosphere for several days and, by the time of the collision, a ridge linking two high-pressure centres – over Nova Scotia and the south of Ireland respectively – extended across the entire Atlantic. This resulted in north-to-northwesterly winds carrying near freezing air from northeastern Canada over the western Atlantic, south of Newfoundland. These winds and temperatures, assisted by the prevailing southward flow of the ocean's Labrador Current on the Grand Banks, led to icebergs and sea ice being transported further south than normal for the time of year – but not beyond the known limits for icebergs during the twentieth century (Figure 1). It is important to note, however, that even in years of extreme sea ice early in the twentieth century, the sea-ice edge rarely extended south of 46°N. A number of reports of extensive sea-ice fields and icebergs ahead had reached the Titanic earlier on the day of the collision. April and May are the peak of the iceberg hazard season in the western North Atlantic (Figure 2), partly because of the release of icebergs previously held fast within the pack ice. In 1912, the peak number of icebergs for the year was recorded in April, while normally this occurs in May, and there were nearly two-and-a-half times as many icebergs as in an average year.
Compounding the iceberg numbers in 1912, the weather, ice conditions and time of year combined to increase the hazard on that fateful day
Thus, two unfavourable factors had combined: there were a greater (though not exceptionally greater) number of icebergs than normal that year; and weather conditions had driven them further south, and earlier in the year, than was usual. We may add a third: the stresses on the crew of the Titanic's maiden voyage.
More than a century on from that tragic day in 1912, icebergs still remain a navigation hazard. The IIP has largely removed the risk of an unexpected iceberg encounter in the northwest Atlantic, but the cruise ship MV Explorer was holed by an iceberg in the Weddell Sea of Antarctica in 2007 and the MS Fram collided with a glacier in 2008, although it was not sunk. A Russian fishing boat, however, was sunk of Antarctica in 2011. As use of the Arctic, in particular, increases in the future, with declining summer sea ice the ice hazard will increase in waters not previously used for shipping. As polar ice sheets are increasingly losing mass as well, iceberg discharge is increasing – not every year, but there are more heavy iceberg years since the 1980s than before – and increasing global warming will likely cause this trend to continue.
Uneven flow: Causes of variability in iceberg calving
It was noted earlier that iceberg numbers in the northwest Atlantic shipping lanes have varied substantially since 1900, from year to year, but also over longer cycles (Figure 2). By allowing the iceberg numbers seeding our ocean iceberg model around Greenland to vary exactly as the I48N time series over the twentieth-century simulation, we found that this variability was highly correlated with the modelled iceberg numbers at 48°N (Figure 3). The inter-annual change in iceberg numbers reaching 48°N is therefore almost completely due to variation in iceberg calving around Greenland, and not to change from year to year in the ocean and atmospheric conditions guiding the icebergs across the sea.
What, then, causes the iceberg calving from Greenland to vary so dramatically over time, as shown in Figure 3? Many factors might contribute. Possibilities include the balance between snow falling onto and ice melting from the ice sheet that covers Greenland – the surface mass balance. The large-scale state of the atmosphere may be another feature. A third might be the surface temperature on the Labrador Sea, which will be related to water temperatures in fjords where icebergs are calving.
We examined this question. We used non-linear systems identification3 to investigate iceberg calving as a dynamic non-linear function of these three factors. Technically, the system identification uses a forward regression orthogonal least-squares algorithm to build models term by term from recorded data sets. It does so by finding the contribution that each selected model term makes to the variance of the dependent variable (I48N here) expressed as a percentage, taking account of the noise in the data. The method searches through an initial library of model terms, which typically includes linear and non-linear time-lagged variables, and selects the most significant terms to include in the final model. These methods have been widely applied to many problems across a wide range of scientific disciplines, from engineering to space weather, electroencephalography, visual systems in insects and many others. We used a procedure with a 30-year sliding window, to allow for a temporally evolving solution3.
For the early part of the twentieth century, at the time of the Titanic disaster, the overwhelmingly dominant term in the evolving non-linear systems model relating icebergs crossing I48N, and hence west Greenland calving, to this combined glaciological, atmospheric and oceanic forcing is a linear expression of the Greenland ice sheet surface mass balance, with a lag of 4 years. This term alone has a correlation of about 0.6 over 1900–1930. In other words, heavy snowfall onto Greenland translates four years later into greater numbers of icebergs. Other non-linear terms, including the Labrador Sea surface temperature at similar time-lags, help to explain the variance during this period more completely, but the majority of the explained variance at this time is due to the four-year-lagged surface mass balance. The physics underlying these links is not yet well understood. However, the component of the iceberg risk to the Titanic due to enhanced iceberg numbers is likely to have developed around 1908, when a moderately warm and wet year over Greenland produced enhanced snow accumulation. We believe that this gradually soaked through cracks in the ice sheet, accumulating around its margins, and probably leading to enhanced short-term sliding of glaciers at their outlets, with resulting greater calving into the sea.
Considering the non-linear systems model results over the twentieth century – an evolving model that shows a correlation coefficient of 0.84 with the I48N time series (Figure 3) – our modelling also suggests that what causes mean calving numbers to fluctuate from decade to decade is climate variability. This can be in terms of the amount of melting on the Greenland ice cap, where greater melting destabilises glaciers, or the ocean temperature, where higher temperatures act in a similar fashion but underneath the calving edge of the glaciers3. These effects vary in time, but what has changed significantly in recent decades is the increase in importance of both factors as global warming has developed.
This article is strongly based on “The iceberg risk in the Titanic year of 1912: was it exceptional?” by G. R. Bigg and D. J. Wilton and published in Weather, volume 69, pp. 100–104. Our research was carried out with the help of a range of colleagues at Sheffield – Dr David Wilton, Dr Hualiang Wei, Dr Yifan Zhou, Prof. Visakan Kadirkamanathan and Prof. Edward Hanna. It was funded by the Natural Environmental Research Council (NERC) grant NE/H023402/1, EPSRC Platform Grant EP/H00453X/1 and European Research Council Advanced Investigator Grant NSYS 226037.
There are more heavy iceberg years since the 1980s than before – and global warming will likely cause this trend to continue