Statistical models of glacier loss: Are they accurate?
The road leading to Vatnajökull, Iceland’s largest ice cap, from where the Bruajjökull glacier extends. (Matt Palmer Visual World/Flick)
The disappearance of glaciers is an urgent problem facing the world. The impact of melting ice Freshwater supplies, sea level rise and ocean circulation.Typically, global glacier models are employed to better understand the extent of this threat, e.g. recent models The model shows widespread glacier loss in mid-latitudes by 2100. However, according to this and other models, there are uncertainties in any linear relationship between temperature and glacier loss, especially in regions like Iceland that experience extreme temperatures away from global averages.
A study on Bruarjokull glacier in Iceland further assessed this uncertainty and found that this linear relationship cannot be easily explained by local observations alone. This demonstrates the need to study Icelandic glaciers as a network. The study, based on the authors’ thesis project at Leiden University College in the Netherlands, used satellite imagery and data from recent weather stations to model glacier area loss from 1984 to 2020. Such retrospective models, called hindcasts, can be used to validate models that predict future changes.
Color-saturated aerial view of the Vatnajokull glacier ice cap in Iceland. The Brua Glacier studied (upper right) is the outlet of this huge ice sheet. (Vatna National Park).
Traditional methods of studying glaciers involve time-consuming and resource-intensive physical measurements, so researchers sometimes use mathematical models to do their work. In this context, there are two relevant mathematical models: deterministic models and statistical models. A deterministic model is a mathematical model that uses physical laws to simulate the behavior of a system. Statistical models, on the other hand, are based on correlations between observations and are used to make predictions or estimates.
Complex deterministic models on a global scale fail to respond to local weather conditions, so statistical models have emerged as a potential alternative for studying glacier melt.An example of this kind of statistical model is a book published in 2001 Glaciers and climate change, climatologist Hans Oerlemans. He found that stable climate conditions still caused glaciers in the European Alps to melt.Another example comes from a recent study published in journal nature Glacial retreat of the Naladhu Glacier in the western Himalayas was assessed. The study, led by Professor Rajesh Kumar of Rajasthan Central University, concluded that reduced precipitation is a more important driver of glacier melt than rising temperatures.
Prior to the latest study in Iceland, members of our research team attempted to replicate Kumar and colleagues’ results using their data and methods. After several months of processing the data and trying to contact the authors for more information, we were unable to obtain any of the results published in their paper. This difficulty made us curious about the method itself, so we chose to replicate it with new data from the Brujajökull glacier in Iceland.
Iceland is home to the world’s largest ice caps, including Langjökull and Vatnajökull. These ice sheets and their outlet glaciers are an important part of the country’s freshwater supply, tourism and ecosystems.However, climate change is causing them to melt rapidly, resulting in some claims Iceland’s glaciers will disappear Next 150 years.
Retreat of the Brujökull glacier in Iceland from 1985 to 2020. Darker colors near the top show where the glacier once stood, and lighter colors show recent retreat. (Courtesy of Domino Jones)
Research based on the Brua Icefield finds precipitation, not temperature, is the key climate driver in the glacier region.This discovery is related to Report This shows that temperature is the main climate driver for Iceland’s glaciers. But the study further found that linear models of the Brua Icefield region as a function of precipitation cannot be used reliably for short- or long-term predictions of glacier extent.
The subtle difference lies in the model residuals—the differences between observed and predicted values. Use residuals when assessing the quality of a model as a diagnostic measure. In this case, the study shows that glacier-meteorological dynamics may only be modeled partially linearly, but that the model successfully explains underlying trends in the area-precipitation relationship.
While this may sound like a contradiction, it highlights something larger. A study In Greenland, it was similarly found that single glacier dynamics were explained nonlinearly. However, they show that normalized glacial changes are uniform over larger areas. In other words, glaciers do not need to be modeled individually to simulate ice loss as a function of climate in the region. This brings us back to global glacier models, which can reduce computational costs and reduce model complexity in areas with climate anomalies that exceed global averages. Whether regional models work for Icelandic glaciers cannot be rigorously determined from the recent findings, but it does highlight an exciting new opportunity.
Other techniques, notably nonlinear statistical models or artificial neural networks, can be used to study local glaciers. This highlights the need for more complex, comprehensive and expensive data more suitable for studying glaciers (rather than the general weather station data used in the study). Obstacles such as the often remote and hazardous environment of glaciers suggest that it may not be possible to obtain the data needed for statistical models to accurately predict glacier melt. In this case, given the limitations of the data, we may need to reconsider our approach to studying Icelandic glaciers. By doing so, we can explore connections between regional glacier networks and better inform global models.
Domino Jones is an undergraduate at Leiden University College in the Netherlands, majoring in Earth, Energy and Sustainability.
