Keywords
Spatio-Temporal; Land Cover Changes; Statistical Assumptions; Models Reliability
Start Date
5-7-2022 12:00 PM
End Date
8-7-2022 9:59 AM
Abstract
The oil sands zone (OSZ) is a vast area (142,200 km2) of northern Canada, best known for oil and gas extraction, and characterized by a delicate natural and human environment, owing to its diverse flora and fauna, as well as its cultural diversity, indigenous population, and resource economy. A spatio-temporal study is conducted to detect environmental changes in OSZ, based on DEM input and Landsat composites every fifth year between 2000 and 2020. Once changes are identified, statistical models will be implemented to probabilistically assess the association of disturbances with specific process categories, i.e., natural vs. anthropogenic. Spatial data, including high resolution imagery, are known to exhibit spatial dependence and heterogeneity, which violate standard statistical assumptions, inducing so-called spatial effects, i.e., decreased reliability of model estimates. The problems can be addressed through specialized methods (e.g., spatially autoregressive or geographically weighted); however, spatial analyses remain prone to the modifiable areal unit problem (MAUP), whereby analytical results depend on the scale and aggregation of spatial units. To date, MAUP has no known solution, but it implies that results obtained for one scale and aggregation cannot be inferred to different aggregations/scales. While we do not have a solution to these problems, we make two practical suggestions. For our analysis, we present multiscale, hierarchical analytical tools, which help uncover the variation of results embedded in changing scale/aggregation. Further, we illustrate the need for environmental software tools to incorporate spatial methods and consider locational matters when integrating LULC dynamics in models.
Space matters: spatial effects, MAUP, and spatial statistics in the analysis of environmental change in the oil sands region of Alberta (Canada)
The oil sands zone (OSZ) is a vast area (142,200 km2) of northern Canada, best known for oil and gas extraction, and characterized by a delicate natural and human environment, owing to its diverse flora and fauna, as well as its cultural diversity, indigenous population, and resource economy. A spatio-temporal study is conducted to detect environmental changes in OSZ, based on DEM input and Landsat composites every fifth year between 2000 and 2020. Once changes are identified, statistical models will be implemented to probabilistically assess the association of disturbances with specific process categories, i.e., natural vs. anthropogenic. Spatial data, including high resolution imagery, are known to exhibit spatial dependence and heterogeneity, which violate standard statistical assumptions, inducing so-called spatial effects, i.e., decreased reliability of model estimates. The problems can be addressed through specialized methods (e.g., spatially autoregressive or geographically weighted); however, spatial analyses remain prone to the modifiable areal unit problem (MAUP), whereby analytical results depend on the scale and aggregation of spatial units. To date, MAUP has no known solution, but it implies that results obtained for one scale and aggregation cannot be inferred to different aggregations/scales. While we do not have a solution to these problems, we make two practical suggestions. For our analysis, we present multiscale, hierarchical analytical tools, which help uncover the variation of results embedded in changing scale/aggregation. Further, we illustrate the need for environmental software tools to incorporate spatial methods and consider locational matters when integrating LULC dynamics in models.
Stream and Session
false