Object-based land use classification of West Ottawa neighbourhoods to inform up-to-date tick habitat suitability models
Zvonimir Burcul, Anders Knudby
Department of Earth and Environmental Sciences, University of Ottawa
Zvonimir Burcul, Anders Knudby
Department of Earth and Environmental Sciences, University of Ottawa
Background
The growing West Ottawa neighbourhoods of Kanata, Stittsville, and Carp offer suitable tick habitat. An existing ecological niche model (ENM) uses modified SOLRIS 2.0 land use data to inform tick habitat suitability across Ottawa. The most recent SOLRIS 3.0 dates to 2015. In this study, we produce 2020 land use data for the West Ottawa neighbourhoods of interest to allow for an update of the ENM for 2020 and provide a methodology for future updates that does not rely on the availability of SOLRIS data.
Methods
High resolution Sentinel-2 MSI L1C imagery of the study area for 2015 and 2020 is obtained from Google Earth Engine and preprocessed to allow for classification of 9 land use types. Using eCognition software, object-based classification is used to classify land use in 2015 and 2020. The 2015 classification is validated against a modified SORLIS 3.0 classification using the same land use types.
Results
Overall accuracy of the 2015 classification is 58%. Built-up impervious, tilled agriculture and barren were classified with producer’s accuracy ranging from 70-78%, and user’s accuracy of 68-76%. Treed areas were classified with a producer’s accuracy of 77% and user’s accuracy of 45%. The remaining land use types were classified with accuracy <54%. The 2020 land use classification was produced using the same methodology and is assumed to have the same accuracy.
Conclusions
The classification methodology used allows for classification of built-up impervious, tilled agriculture, and barren areas, but other land use types lack accuracy. The 2020 land use classification accurately shows areas of urban sprawl and agricultural loss in neighbourhoods in West Ottawa. The methodology must be reworked to allow for proper classification of all land use types in the study area.
The growing West Ottawa neighbourhoods of Kanata, Stittsville, and Carp offer suitable tick habitat. An existing ecological niche model (ENM) uses modified SOLRIS 2.0 land use data to inform tick habitat suitability across Ottawa. The most recent SOLRIS 3.0 dates to 2015. In this study, we produce 2020 land use data for the West Ottawa neighbourhoods of interest to allow for an update of the ENM for 2020 and provide a methodology for future updates that does not rely on the availability of SOLRIS data.
Methods
High resolution Sentinel-2 MSI L1C imagery of the study area for 2015 and 2020 is obtained from Google Earth Engine and preprocessed to allow for classification of 9 land use types. Using eCognition software, object-based classification is used to classify land use in 2015 and 2020. The 2015 classification is validated against a modified SORLIS 3.0 classification using the same land use types.
Results
Overall accuracy of the 2015 classification is 58%. Built-up impervious, tilled agriculture and barren were classified with producer’s accuracy ranging from 70-78%, and user’s accuracy of 68-76%. Treed areas were classified with a producer’s accuracy of 77% and user’s accuracy of 45%. The remaining land use types were classified with accuracy <54%. The 2020 land use classification was produced using the same methodology and is assumed to have the same accuracy.
Conclusions
The classification methodology used allows for classification of built-up impervious, tilled agriculture, and barren areas, but other land use types lack accuracy. The 2020 land use classification accurately shows areas of urban sprawl and agricultural loss in neighbourhoods in West Ottawa. The methodology must be reworked to allow for proper classification of all land use types in the study area.