Universitas Helsingiensis

Frost turns into statistical models

Frost turns into statistical modelsImagine six hundred square kilometres of fells and uplands in front of you in the wilds of Lapland. You have two summers to map all the frost formed landforms in that area. The terrain is rough: steep valleys, bogs, waist-high shrubbery. When not scorched by the ruthless sun, you are drenched to the bone by a cloudburst or tormented by a swarm of black flies. When you finally return to your desk in autumn, it is time to start processing the material. Would you do it?

Jan Hjort, PhD, did, and his trips into the wilderness and hard work at the office resulted in a doctoral dissertation in physical geography entitled Environmental factors affecting the occurrence of periglacial landforms in Finnish Lapland: a numerical approach. It was well worth the effort: Hjort was awarded one of the three 2006 Dissertation Awards presented at the University of Helsinki. “It was quite a surprise,” he says.

“After my first week in the wild I had many doubts. But once I had the material and started to get interesting results, I also started to believe in the dissertation.”

Periglacial landforms and their modelling

Hjort modelled the occurrence of periglacial landforms in the Paistunturi fells in Utsjoki with Geographical Information System (GIS), Remote Sense Data, and statistical methods. While the main objective of his research was to identify environmental factors affecting the occurrence of periglacial landforms, he also wanted to assess the suitability of statistical methods for such research.

“This approach has been rare in geomorphology, but in ecology it is quite popular,” Hjort says. “After a break of decades, quantitative methods are on the rise again, probably because of new digital data and the increased computational power in today’s computers.”

Periglacial areas such as the Paistunturi fells in Utsjoki are climatically cold regions where the most important earth surface processes are the freezing and thawing of the ground. Factors contributing to periglacial phenomena include climate, topography, rock material, soil moisture, vegetation, time, and the human factor. Periglacial processes result in various landforms, such as palsa mines, mounds of perennially frozen peat and mineral soil.

Hjort focused on the application of statistical methods in geomorphological modelling, but he was also motivated by global climate change. Periglacial areas are vulnerable to the effects of climate change. Mapping and researching vast areas, however, is expensive and time-consuming. Statistical methods create both savings and prognoses for future development in periglacial areas.

A 1,500-kilometre walk

Even though his dissertation is the result of much more than just walking, Hjort did walk a lot. He wanted to map all the periglacial landform types and subtypes in the area, of which there were 40 different kinds. He would set camp at certain points in the area, and then make day trips that he planned with aerial photos that had been taken earlier. Once he had found the landforms, he would locate them with GPS and mark them on the map. Through the two summers, he walked a total of 1,500 kilometres.

“Sometimes it was mentally demanding to spend days by myself in the wild,” Hjort says. “I am not a particularly social person, but there I would talk to any trekker who had wandered off the trekking route or gotten lost.”

Digitisation and statistical analysis

GIS is a heading for various software, databases, and users who use, produce, or process geographical information. This information usually includes a certain feature, such as a periglacial landform, and its geographical coordinates.

The geographical database necessary for the research was drafted in phases. First he digitised the periglacial landforms on the basis of terrain observations and aerial photos; then he computed the environmental variables based on topographical, soil property and vegetation material. The final modelling containing the data on the periglacial landforms and environmental factors was drafted at a 25 ha resolution.

Hjort used statistical analysis to find the environmental factors associated with certain landforms, twelve of which were included in the final part of his study. The statistical methods used were generalized linear modelling (GLM) and hierarchical partitioning (HP).

GLM is like a generalised, ordinary regression method on response material containing just zeros and ones. HP was used to find the significance of a single environmental factor to the occurrence of a certain landform, since these factors are often interdependent. HP also served as a control method to the functionality of GLM.

Good results

In his research, Hjort demonstrated that GIS and statistical methods can be efficiently used for the modelling of periglacial phenomena. Certain factors predicted certain landforms well when his predictions were compared to the landforms Hjort had mapped on the terrain. Based on his analyses, he could also make prediction maps of the occurrence probability of periglacial landforms.

“In the future, I’m planning to obtain more accurate data on environmental variables with satellite images, test other statistical methods besides HP and GLM, and use these methods in climate change studies,” Hjort says. “My main objective is to promote research in geomorphological processes on various scales by combining topographical measurements, GIS data and statistical modelling methods.”

As the climate changes, will the permafrost remain?

Professor of Physical Geography Matti Seppälä has studied, measured and tested frost in several ways for four decades. His research topics include the permafrost on palsa mires in Lapland. Palsas are frozen peat mounds rising up from open peatlands.

“The palsa mires are sensitive to any environmental changes, which is why we keep an eye on them. Currently there are fewer palsas than a hundred years ago, but new ones are also formed all the time,” Seppälä says.

According to present knowledge, climate change is raising the temperature of the surface water in the North Atlantic Ocean. As a result, there will be more snowfall in Lapland, and the palsas will be covered in snow; currently, the mounds usually remain almost snowless even in winter.

“We used a snow fence to ensure that a palsa is covered by snow in winter to see how it reacts. It took several years before the palsa reacted in any way,” he says.

In another experiment, Seppälä and his colleagues removed snow from the mire surface to create an artificial palsa. In natural conditions, the wind thins the snow cover of the mire, and a palsa begins to form.

Kerttu Aitola