Glacier loss is a pressing problem worldwide as melting ice affects freshwater supplies, sea level rise, and ocean circulation. Global glacier models are frequently used to better understand the scale of this threat; for example, a new model showing widespread ice decline in mid-latitudes by 2100. However, these and other models will not be able to determine a linear relationship between temperature and glacier loss, especially in regions such as Iceland that experience extreme temperatures that differ from the global average.

A study on the Bruarjokutl glacier in Iceland further evaluated this uncertainty and found that the linear relationship cannot be explained by local observations alone. This suggests the need to study Icelandic glaciers as a network.

The study, based on this author’s thesis project at Leiden University College in the Netherlands, used satellite imagery and data from a nearby weather station to create a model of glacier area loss from 1984 to 2020. Such hindcast models, called hindcast models, can be used to test models for predictions of future changes.

Traditional methods of studying glaciers involve physical measurements that are time-consuming and resource-intensive, so researchers sometimes use mathematical models for their studies. In this context, there are two types of mathematical models: deterministic and statistical models. Deterministic models are a type of mathematical model that uses physical laws to model the behavior of a system. Statistical models, on the other hand, are based on correlations between observed data and are applied for predictive or predictive purposes.

Because complex deterministic models on a global scale do not respond to local weather conditions, statistical models have emerged as a potential alternative to studying melting glaciers. An example of such a statistical model is the 2001 book Glaciers and Climate Change by climatologist Hans Erlemans. He found that stable climatic conditions were still causing the glaciers in the European Alps to melt.

Another example is a study recently published in a journal. Scientific ReportsAssessing the retreat of the Naradu Glacier in the Western Himalayas. This study, led by professor Rajesh Kumar of the Central University of Rajasthan, concluded that decreasing rainfall is a more important factor in melting glaciers than increasing temperatures.

Until the recent study in Iceland, members of our research team attempted to reproduce Kumar and colleagues’ results using their data and methods. After several months of manipulating 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 challenge led us to become interested in the method itself, and we decided to reproduce it with new data for the Bruarjokutl glacier in Iceland.

Iceland is home to some of the largest glaciers in the world, including Langjökull and Vatnajökull. These glaciers and outlet glaciers are critical components of the country’s freshwater supply, tourism industry, and ecosystem. But climate change is causing glaciers to melt rapidly, leading some to claim that Iceland’s glaciers will disappear within the next 150 years.

A study conducted in Bruarjökull revealed that the main driver of the glacial climate is rainfall, not temperature. This finding contradicts reports showing that temperature is an important climate factor for Icelandic glaciers. But the study also found that linear modeling of the Bruarjocutl area as a function of precipitation cannot be used reliably to predict glacier size in the short or long term.

The nuance here comes when looking at model residuals (the difference between observed and predicted data values). Residuals are used as a diagnostic measure in model quality assessment. In this case, the study shows that glacier dynamics and meteorological dynamics can only be partially modeled linearly, but that this model successfully explains the trends underlying the relationship between area and precipitation.

While this may seem like a contradiction, it illuminates something larger. Research in Greenland also found that the dynamics of individual glaciers are described in a non-linear way. But they show that normalized glacier change is the same across large regions. In other words, there is no need to model individual glaciers to model ice loss as a function of the climate in that region. This brings us back to global glacier models, which can work with less computational cost and model complexity for areas with climate anomalies beyond the global average. It is unclear from the latest findings whether regional modeling can be applied to Icelandic glaciers, but it highlights an exciting new possibility.

Other methods, including nonlinear statistical models or artificial neural networks, can be used to study local glaciers. This underscores the need for more complex, comprehensive and expensive data that is better suited to studying glaciers (as opposed to the public weather station data used in the study). Obstacles, including the often remote and dangerous environment of glaciers, mean it is impossible to obtain the data needed for statistical models to accurately predict glacier melt.

In this case, given the limited data, we may have to reconsider the methods used to study Icelandic glaciers. By doing this, we can explore connections between regional glacier networks and better inform global models. Source