Coupling Remote Sensing and Modelling to Monitor the Spatial Distribution and Trends of Surface Temperature and Ice Thickness on Sub-Arctic Lakes

Lake surface temperature (LST), lake ice thickness (LIT), and lake ice phenology (LIP) play significant roles in the diverse regional processes of freshwater in cold regions. They offer direct indications of regional weather and climate conditions, and their interactions with the atmosphere impact climate processes. Furthermore, lake ice is valuable to northern communities, such as those in the Northwest Territories (NWT). Ice roads, including the longest ice road in the NWT, spanning over 80 lakes, are constructed during winter to haul goods to and from industrial establishments (e.g., mines) and for travel within and between communities. A significant challenge to lakes and the ongoing use of ice roads are the changes in LST, LIT and LIP due to climate warming. Knowledge of LST, LIT, and LIP is crucial to understanding how lakes respond to climate change and determining how much weight an ice cover can safely sustain for winter travel on frozen lakes. This knowledge, however, is minimal due to the logistical difficulties in traditionally collecting measurements directly. In recent years, satellite-based observations have gained significant traction for studying lakes. However, multispectral sensors are not equipped to measure ice thickness directly, as it is a subsurface feature, which poses a limitation. Furthermore, other methods, such as one-dimensional thermodynamic lake ice models, which rely on weather station input data, are limited by the sparse availability of weather station and in-situ data, especially at high latitudes. This research adopts a multimodal monitoring approach to address these limitations by combining remote sensing data with spatially distributed modelling to study and monitor the trends and spatial distribution of LST, LIT and LIP.\nIn this study, a retrieval algorithm was applied to the thermal bands of Landsat archives to generate a lake-specific surface temperature dataset (North Slave LST dataset) for 535 lakes in the North Slave Region (NSR), NWT, Canada, from 1984 to 2021. Cloud masks were applied to Landsat images to eliminate cloud cover. In addition, a 100 m inward buffer was used on lakes to prevent pixel mixing with shorelines. A good agreement was observed between in-situ observations and North Slave LST, with a mean bias of 0.12 °C and a root mean squared deviation (RMSD) of 1.7 °C. The North Slave LST dataset contains more available data for warmer months (May to September; 57.3 %) than colder months (October to April). The North Slave LST dataset is available at https://doi.org/10.5683/SP3/J4GMC2 (Attiah et al., 2022).\nBased on the North Slave LST data, LST trends and spatial distribution across the 535 predominantly small to medium lakes across NSR were studied. LST was analyzed in four distinct periods: open water season (OW), ice cover season (IC), and the transitional months of May (TM) and October (TO). The trend and relationships of LST were analyzed using the Mann-Kendall test and a multilinear regression model. The analysis revealed an overall increase in LST, with average rates (max) of 0.03 °C/year (0.05 °C/year), 0.03 °C/year (0.06 °C/year), and 0.13 °C/year (0.27 °C/year) for OW, TM, and TO, respectively across study lakes. A faster rate of change was observed in October compared to other periods.\nUsing the North Slave LST data generated as input, a comprehensive approach was adopted to simulate the spatial variability of ice thickness on lakes at a high resolution by spatially distributing a one-dimensional thermodynamic lake ice model. The spatial distribution of LIT was modelled for study lakes from 1984 to 2022. The generated LST data, in combination with the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) data, were used as inputs for the model. The model simulates the spatial distribution of daily lake ice thickness on a 50-meter spatial grid as well as the annual freeze-up and break-up dates. Results showed a root mean square deviation of LIT from 2.7 cm to 7 cm compared to in-situ data. Further analysis of ice cover on study lakes from 1984 to 2022 revealed decreasing trends in LIT (-0.26 cm/year to -0.10 cm/year) and ice cover duration (ICD) (-0.40 day/year to -0.15 day/year). Simulated LIT and freeze-up proved sensitive to morphometry (depth), while location properties (latitude/longitude) primarily drove the break-up process.\nThis dissertation provides comprehensive approaches to deriving LST, LIT, and LIP information from small and medium lakes in data-sparse regions. A multimodal approach combining remote sensing and spatially distributed modelling is adopted to address the insufficiency of in-situ data and the sparse distribution of weather station data. The methods utilized can be replicated in other regions, providing a broader understanding of the trends and spatial distribution of LST, LIT, and LIP on sub-arctic lakes with varying physical, geographical, and morphometrical properties.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.