Comparing spatial interpolation approaches for mapping peatlands
PhD candidate Victoria Steblaj compares machine learning and traditional methods to help aid wetland conservation.
Peatlands are important wetland ecosystems for carbon sequestration and species at risk. Quantifying carbon storage, mapping carbon decomposition risk, and determining habitat suitability require accurate peatland thickness and surface elevation information. This is accomplished by using field measurements to interpolate a continuous surface model. There are many different spatial interpolation approaches, and it can be difficult to determine which method should be used.
Victoria Steblaj has always been interested in wetlands and the species within them. As her knowledge of wetlands increased, she recognized the importance of understanding how these ecosystems function to support their long-term conservation. She enrolled in the Master of Science program in the Department of Geography and Environmental Management to study peatlands in the intersection of ecology, hydrology, spatial analysis and environment science. Supervised by Dr. Chantel Markle, Steblaj compared spatial interpolation approaches for mapping habitat and peatland ecohydrological metrics.
The analysis compared two common interpolation methods, as well as a new approach that uses machine learning. “As machine learning continues to advance, it is important to evaluate its potential applications in environmental mapping alongside more established spatial interpolation methods. Including a machine learning approach allowed us to assess its ability to capture spatial patterns and improve mapping accuracy in a small, basin-confined peatland system.”
Steblaj compared continuous surface models, created from the three interpolation approaches, from peatlands in the Georgian Bay Mnidoo Gamii Biosphere. As part of her ongoing research, she has had the opportunity to visit the site to collect further data. “Field work is the most rewarding part of the research experience. From wading through wetlands to spending long days outdoors in all kinds of weather, it offers a unique opportunity to experience these ecosystems first-hand.”
The results show that the three interpolation approaches produced broadly similar spatial predictions, although differences among these methods were more pronounced for peat thickness than for surface elevation. Choosing which interpolation method to choose should be guided by the specific goals of the analysis, the quality and sampling distribution of the available data, and the degree to which the assumptions underlying the chosen method are met.
This research will have a direct impact on conservation and quantifying carbon storage in peatlands. Robust spatial modelling can lead to better conservation decisions, and utilizing the best interpolation method will ensure these decisions are accurate and impactful.
The scope of Steblaj’s research increased, and she transitioned into the PhD program. Her doctoral research will build upon her initial work and investigates whether constructed wetlands within an industrial site can provide suitable habitat for reptile species at risk, particularly turtles and snakes.
“I’m especially interested in whether these wetlands provide the temperature and moisture conditions reptiles need to survive the winter, as well as whether they can support reptiles during the active season. I love that the research connects so many related concepts, allowing me to look at constructed wetlands from multiple perspectives and ultimately better understand how we can design and manage them to support biodiversity.”
A comparison of spatial interpolation approaches for mapping habitat and peatland ecohydrological metrics, authored by Steblaj and others, is published in Wetlands.
