High Resolution Satellite Forest Information for Canada

High Resolution Satellite Forest Information for Canada (NFIS_High_Resolution_Forest_Data)

NFIS Project Office. This Web services are for forest change products that represents the first wall-to-wall characterization of wildfire and harvest in Canada at a spatial resolution commensurate with human impacts. The information outcomes represents 25 years of stand replacing change in Canada's forests derived from a single consistent spatially-explicit data source and derived in a fully automated manner.

ca_change_nochange_r1984 (changenochange)

Forest Change and Nochange bit map for Canada. The forest change data included in this product is national in scope (entire forested ecosystem) and represents the first wall-to-wall characterization of wildfire and harvest in Canada at a spatial resolution commensurate with human impacts. The information outcomes represent 25 years of stand replacing change in Canada.s forests, derived from a single, consistent spatially-explicit data source, derived in a fully automated manner. This demonstrated capacity to characterize forests at a resolution that captures human impacts is key to establishing a baseline for detailed monitoring of forested ecosystems from management and science perspectives. Time series of Landsat data were used to characterize national trends in stand replacing forest disturbances caused by wildfire and harvest for the period 1985-2011 for Canada's 650 million hectare forested ecosystems (White et al. 2017). Landsat data has a 30m spatial resolution, so the change information is highly detailed and is commensurate with that of human impacts. These data represent annual stand replacing forest changes. The stand replacing disturbances types labeled are wildfire and harvest, with lower confidence wildfire and harvest, also shared. The distinction and sharing of lower class membership likelihoods is to indicate to users that some change events were more difficult to allocate to a change type, but are generally found to be in the correct category. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016)). The data available is, 1. a binary change/no-change; 2. Change year; and, 3. Change type. When using this data, please cite as: White, J.C., M.A. Wulder, T. Hermosilla, N.C. Coops, and G. Hobart. (2017). A nationwide annual characterization of 25 years of forest disturbance and recovery for Canada using Landsat time series. Remote Sensing of Environment. 192: 303-321. DOI: 10.1016/j.rse.2017.03.035

ca_change_year_r1984 (changeyear)

Forest Change Year 1985-2011. The forest change data included in this product is national in scope (entire forested ecosystem) and represents the first wall-to-wall characterization of wildfire and harvest in Canada at a spatial resolution commensurate with human impacts. The information outcomes represent 25 years of stand replacing change in Canada.s forests, derived from a single, consistent spatially-explicit data source, derived in a fully automated manner. This demonstrated capacity to characterize forests at a resolution that captures human impacts is key to establishing a baseline for detailed monitoring of forested ecosystems from management and science perspectives. Time series of Landsat data were used to characterize national trends in stand replacing forest disturbances caused by wildfire and harvest for the period 1985-2011 for Canada's 650 million hectare forested ecosystems (White et al. 2017). Landsat data has a 30m spatial resolution, so the change information is highly detailed and is commensurate with that of human impacts. These data represent annual stand replacing forest changes. The stand replacing disturbances types labeled are wildfire and harvest, with lower confidence wildfire and harvest, also shared. The distinction and sharing of lower class membership likelihoods is to indicate to users that some change events were more difficult to allocate to a change type, but are generally found to be in the correct category. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016)). The data available is, 1. a binary change/no-change; 2. Change year; and, 3. Change type. When using this data, please cite as: White, J.C., M.A. Wulder, T. Hermosilla, N.C. Coops, and G. Hobart. (2017). A nationwide annual characterization of 25 years of forest disturbance and recovery for Canada using Landsat time series. Remote Sensing of Environment. 192: 303-321. DOI: 10.1016/j.rse.2017.03.035

ca_change_nochange_r2012 (changenochange2)

Forest Change and Nochange bit map for Canada. The Forest Change/No-change data described here is an update to previously posted open data. The date range for this data is 2012 to 2015. The forest change data included in this product is national in scope (entire forested ecosystem) and represents the first wall-to-wall characterization of wildfire and harvest in Canada at a spatial resolution commensurate with human impacts. The information outcomes represent 4 years of stand replacing change in Canada's forests, derived from a single, consistent spatially-explicit data source, derived in a fully automated manner. This demonstrated capacity to characterize forests at a resolution that captures human impacts is key to establishing a baseline for detailed monitoring of forested ecosystems from management and science perspectives. Time series of Landsat data were used to characterize national trends in stand replacing forest disturbances caused by wildfire and harvest for the period 21012-2015 for Canada's 650 million hectare forested ecosystems (https://authors.elsevier.com/sd/article/S0034425717301360 ). Landsat data has a 30m spatial resolution, so the change information is highly detailed and is commensurate with that of human impacts. These data represent annual stand replacing forest changes. The stand replacing disturbances types labeled are wildfire and harvest, with lower confidence wildfire and harvest, also shared. The distinction and sharing of lower class membership likelihoods is to indicate to users that some change events were more difficult to allocate to a change type, but are generally found to be in the correct category. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673).

ca_change_type_r2012 (changetype2)

Forest Change Type (Wildfire, Harvest, Low Confidence Wildfire, Low Confidence Harvest).The Forest Change Type data described here is an update to previously posted open data. The date range for this data is 2012 to 2015. The forest change data included in this product is national in scope (entire forested ecosystem) and represents the first wall-to-wall characterization of wildfire and harvest in Canada at a spatial resolution commensurate with human impacts. The information outcomes represent 25 years of stand replacing change in Canada's forests, derived from a single, consistent spatially-explicit data source, derived in a fully automated manner. This demonstrated capacity to characterize forests at a resolution that captures human impacts is key to establishing a baseline for detailed monitoring of forested ecosystems from management and science perspectives. Time series of Landsat data were used to characterize national trends in stand replacing forest disturbances caused by wildfire and harvest for the period 1985-2010 for Canada's 650 million hectare forested ecosystems (https://authors.elsevier.com/sd/article/S0034425717301360 ). Landsat data has a 30m spatial resolution, so the change information is highly detailed and is commensurate with that of human impacts. These data represent annual stand replacing forest changes. The stand replacing disturbances types labeled are wildfire and harvest, with lower confidence wildfire and harvest, also shared. The distinction and sharing of lower class membership likelihoods is to indicate to users that some change events were more difficult to allocate to a change type, but are generally found to be in the correct category. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673).

ca_vlce_1984_wkg_r (land_cover_1984)

High-resolution forest land cover for Canada (1984). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1984. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1986_wkg_r (land_cover_1986)

High-resolution forest land cover for Canada (1986). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1986. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1988_wkg_r (land_cover_1988)

High-resolution forest land cover for Canada (1988). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1988. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1990_wkg_r (land_cover_1990)

High-resolution forest land cover for Canada (1990). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1990. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1992_wkg_r (land_cover_1992)

High-resolution forest land cover for Canada (1992). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1992. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1994_wkg_r (land_cover_1994)

High-resolution forest land cover for Canada (1994). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1994. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1996_wkg_r (land_cover_1996)

High-resolution forest land cover for Canada (1996). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1996. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_1998_wkg_r (land_cover_1998)

High-resolution forest land cover for Canada (1998). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 1998. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2000_wkg_r (land_cover_2000)

High-resolution forest land cover for Canada (2000). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2000. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2002_wkg_r (land_cover_2002)

High-resolution forest land cover for Canada (2002). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2002. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2004_wkg_r (land_cover_2004)

High-resolution forest land cover for Canada (2004). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2004. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2006_wkg_r (land_cover_2006)

High-resolution forest land cover for Canada (2006). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2006. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2008_wkg_r (land_cover_2008)

High-resolution forest land cover for Canada (2008). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2008. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2010_wkg_r (land_cover_2010)

High-resolution forest land cover for Canada (2010). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2010. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2012_wkg_r (land_cover_2012)

High-resolution forest land cover for Canada (2012). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2012. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2014_wkg_r (land_cover_2014)

High-resolution forest land cover for Canada (2014). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2014. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2016_wkg_r (land_cover_2016)

High-resolution forest land cover for Canada (2016). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2016. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2018_wkg_r (land_cover_2018)

High-resolution forest land cover for Canada (2018). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2018. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

ca_vlce_2015_v1_wkg_r (land_cover_2015_v1)

High-resolution forest land cover for Canada (2015). The forest land cover data included in this product is national in scope (entire forested ecosystem) and represents the a wall to wall land cover characterization for 2015. This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018; https://www.tandfonline.com/doi/full/10.1080/07038992.2018.1437719).

wetlands_year_sum (wetlands_year_sum)

High-resolution wetland year count for Canada (1984-2016). Count of number of years a pixel is classified as wetland. The wetland year count data included in this product is national in scope (entire forested ecosystem) and represents a wall to wall wetland characterization for 1984-2016 (Wulder et al. 2018). This product was generated using both annual gap free composite reflectance images and annual forest change maps following the Virtual Land Cover Engine (VLCE) process (see Hermosilla et al. 2018), over the 650 million ha forested ecosystems of Canada. Elements of the VLCE classification approach are inclusion of disturbance information in the processes as well as ensuring class transitions over time are logical. Further, a Hidden Markov Model is implemented to assess individual year class likelihoods to reduce variability and possible noise in year-on-year class assignments (for instances when class likelihoods are similar). The values can range from 0 to 33 denoting the number of years between 1984 and 2016 that a pixel was classified as wetland or wetland-treed in the VLCE data cube. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; http://www.tandfonline.com/doi/full/10.1080/17538947.2016.1187673). A detailed description of the VLCE process and the subsequently generated land cover product, including an accuracy assessment, please see Hermosilla et al. (2018). The focused wetland analyses can be found described in Wulder et al (2018).

elev_cv_r (elev_cv)

Coefficient of variation of first returns height (%). Represents the variability in canopy heights relative to the mean canopy height. Products relating the structure of Canada's forested ecosystems have been generated and made openly accessible. The shared products are based upon peer-reviewed science and relate aspects of forest structure including: (i) metrics calculated directly from the lidar point cloud with heights normalized to heights above the ground surface (e.g., canopy cover, height), and (ii) modelled inventory attributes, derived using an area-based approach generated by using co-located ground plot and ALS data (e.g., volume, biomass). Forest structure estimates were generated by combining information from 'lidar plots' (Wulder et al. 2012) with Landsat pixel-based composites (White et al. 2014; Hermosilla et al. 2016) using a nearest neighbour imputation approach with a Random Forests-based distance metric. These products were generated for strategic-level forest monitoring information needs and are not intended to support operational-level forest management. All products have a spatial resolution of 30 m. For a detailed description of the data, methods applied, and accuracy assessment results see Matasci et al. (2018). When using this data, please cite as follows: Matasci, G., Hermosilla, T., Wulder, M.A., White, J.C., Coops, N.C., Hobart, G.W., Bolton, D.K., Tompalski, P., Bater, C.W., 2018b. Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots. Remote Sensing of Environment 216, 697-714. https://doi.org/10.1016/j.rse.2018.07.024

stem_vol_r (stem_vol)

Gross stem volume. Individual tree gross volumes are calculated using species-specific allometric equations. In the measured ground plots, gross total volume per hectare is calculated by summing the gross total volume of all trees and dividing by the area of the plot (units = m3ha-1). Products relating the structure of Canada's forested ecosystems have been generated and made openly accessible. The shared products are based upon peer-reviewed science and relate aspects of forest structure including: (i) metrics calculated directly from the lidar point cloud with heights normalized to heights above the ground surface (e.g., canopy cover, height), and (ii) modelled inventory attributes, derived using an area-based approach generated by using co-located ground plot and ALS data (e.g., volume, biomass). Forest structure estimates were generated by combining information from 'lidar plots' (Wulder et al. 2012) with Landsat pixel-based composites (White et al. 2014; Hermosilla et al. 2016) using a nearest neighbour imputation approach with a Random Forests-based distance metric. These products were generated for strategic-level forest monitoring information needs and are not intended to support operational-level forest management. All products have a spatial resolution of 30 m. For a detailed description of the data, methods applied, and accuracy assessment results see Matasci et al. (2018). When using this data, please cite as follows: Matasci, G., Hermosilla, T., Wulder, M.A., White, J.C., Coops, N.C., Hobart, G.W., Bolton, D.K., Tompalski, P., Bater, C.W., 2018b. Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots. Remote Sensing of Environment 216, 697-714. https://doi.org/10.1016/j.rse.2018.07.024

pct_2m_r (pct_2m)

Percentage of first returns above 2 m (%). Represents canopy cover. Products relating the structure of Canada's forested ecosystems have been generated and made openly accessible. The shared products are based upon peer-reviewed science and relate aspects of forest structure including: (i) metrics calculated directly from the lidar point cloud with heights normalized to heights above the ground surface (e.g., canopy cover, height), and (ii) modelled inventory attributes, derived using an area-based approach generated by using co-located ground plot and ALS data (e.g., volume, biomass). Forest structure estimates were generated by combining information from 'lidar plots' (Wulder et al. 2012) with Landsat pixel-based composites (White et al. 2014; Hermosilla et al. 2016) using a nearest neighbour imputation approach with a Random Forests-based distance metric. These products were generated for strategic-level forest monitoring information needs and are not intended to support operational-level forest management. All products have a spatial resolution of 30 m. For a detailed description of the data, methods applied, and accuracy assessment results see Matasci et al. (2018). When using this data, please cite as follows: Matasci, G., Hermosilla, T., Wulder, M.A., White, J.C., Coops, N.C., Hobart, G.W., Bolton, D.K., Tompalski, P., Bater, C.W., 2018b. Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots. Remote Sensing of Environment 216, 697-714. https://doi.org/10.1016/j.rse.2018.07.024

95pct_mean_r (95pct_mean)

95th percentile of first returns height (m). Products relating the structure of Canada's forested ecosystems have been generated and made openly accessible. The shared products are based upon peer-reviewed science and relate aspects of forest structure including: (i) metrics calculated directly from the lidar point cloud with heights normalized to heights above the ground surface (e.g., canopy cover, height), and (ii) modelled inventory attributes, derived using an area-based approach generated by using co-located ground plot and ALS data (e.g., volume, biomass). Forest structure estimates were generated by combining information from 'lidar plots' (Wulder et al. 2012) with Landsat pixel-based composites (White et al. 2014; Hermosilla et al. 2016) using a nearest neighbour imputation approach with a Random Forests-based distance metric. These products were generated for strategic-level forest monitoring information needs and are not intended to support operational-level forest management. All products have a spatial resolution of 30 m. For a detailed description of the data, methods applied, and accuracy assessment results see Matasci et al. (2018). When using this data, please cite as follows: Matasci, G., Hermosilla, T., Wulder, M.A., White, J.C., Coops, N.C., Hobart, G.W., Bolton, D.K., Tompalski, P., Bater, C.W., 2018b. Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots. Remote Sensing of Environment 216, 697-714. https://doi.org/10.1016/j.rse.2018.07.024

basal_area (basal_area)

Basal area. Cross-sectional area of tree stems at breast height. The sum of the cross-sectional area (i.e. basal area) of each tree in square metres in a plot, divided by the area of the plot (ha) (units = m2ha). Products relating the structure of Canada's forested ecosystems have been generated and made openly accessible. The shared products are based upon peer-reviewed science and relate aspects of forest structure including: (i) metrics calculated directly from the lidar point cloud with heights normalized to heights above the ground surface (e.g., canopy cover, height), and (ii) modelled inventory attributes, derived using an area-based approach generated by using co-located ground plot and ALS data (e.g., volume, biomass). Forest structure estimates were generated by combining information from 'lidar plots' (Wulder et al. 2012) with Landsat pixel-based composites (White et al. 2014; Hermosilla et al. 2016) using a nearest neighbour imputation approach with a Random Forests-based distance metric. These products were generated for strategic-level forest monitoring information needs and are not intended to support operational-level forest management. All products have a spatial resolution of 30 m. For a detailed description of the data, methods applied, and accuracy assessment results see Matasci et al. (2018). When using this data, please cite as follows: Matasci, G., Hermosilla, T., Wulder, M.A., White, J.C., Coops, N.C., Hobart, G.W., Bolton, D.K., Tompalski, P., Bater, C.W., 2018b. Three decades of forest structural dynamics over Canada's forested ecosystems using Landsat time-series and lidar plots. Remote Sensing of Environment 216, 697-714. https://doi.org/10.1016/j.rse.2018.07.024

Harvest recovery rate (HarvRec)

Post-disturbance forest recovery data for Canada's forested ecosystems, representing a total area of ~650 million ha, captures the return of forests following wildfire and harvest that occurred between 1986 and 2012. These spatially-explicit outputs represent the rate of spectral recovery — the rate at which a pixel returns to 80% of its pre-disturbance value (White et al. 2017) within the observation period (1985–2017) using the Y2R or Years-to-Recovery metric derived from Landsat times series data. Baseline rates of spectral recovery (Y2R) were defined for each of Canada's 12 forested ecozones. These baselines were then used to identify spatial clusters of recovering pixels on the landscape where Y2R were either significantly faster or slower than their ecozonal baseline. Finally, areas that were disturbed by wildfire and harvest (1986-2012), but which had not recovered by the end of the observation period (2017) are also provided. Note that these areas are still recovering, but they had not yet recovered according to our metric of spectral recovery, by the end of the time series in 2017. For an overview of the methods, the validation of the Y2R metric, and interpretation of the derived trends, see White et al. (2022) and White et al. (2017).

Harvest Years To Recovery (HarvY2R)

Post-disturbance forest recovery data for Canada's forested ecosystems, representing a total area of ~650 million ha, captures the return of forests following wildfire and harvest that occurred between 1986 and 2012. These spatially-explicit outputs represent the rate of spectral recovery — the rate at which a pixel returns to 80% of its pre-disturbance value (White et al. 2017) within the observation period (1985–2017) using the Y2R or Years-to-Recovery metric derived from Landsat times series data. Baseline rates of spectral recovery (Y2R) were defined for each of Canada's 12 forested ecozones. These baselines were then used to identify spatial clusters of recovering pixels on the landscape where Y2R were either significantly faster or slower than their ecozonal baseline. Finally, areas that were disturbed by wildfire and harvest (1986-2012), but which had not recovered by the end of the observation period (2017) are also provided. Note that these areas are still recovering, but they had not yet recovered according to our metric of spectral recovery, by the end of the time series in 2017. For an overview of the methods, the validation of the Y2R metric, and interpretation of the derived trends, see White et al. (2022) and White et al. (2017).

wildfire_year_r (wildfire_year)

Spectral change magnitude for wildfires that occurred from 1985 and 2015. The wildfire change magnitude included in this product is expressed via differenced Normalized Burn Ratio (dNBR), computed as the variation between the spectral values before and after the change event. This dataset is composed of year of greatest wildfire disturbance. The information outcomes represent 30 years of wildfires in Canada's forests, derived from a single, consistent spatially-explicit data source in a fully automated manner. Time series of Landsat data with 30-m spatial resolution were used to characterize national trends in stand replacing forest disturbances caused by wildfire for the period 1985-2015 for Canada's 650 million hectare forested ecosystems (Hermosilla et al. 2017).

harvest_r (harvest)

The information outcomes represent 31 years of harvesting activity in Canada’s forests, derived from a single, consistent, spatially-explicit data source in an automated manner. Time series of Landsat data with 30-m spatial resolution were used to characterize national trends in stand replacing forest disturbances, including those attributed to harvest for the period 1985–2015 for Canada's 650 million hectare forested ecosystems (Hermosilla et al. 2016). See references below for an overview regarding the data, image processing, and time-series change detection methods applied, as well as information on independent accuracy assessment of the data.

DHI_lyr_r (DHI_lyr)

Dynamic Habitat Index. (2000-2005) Satellite derived estimates of photosynthetically active radiation can be obtained from satellites such as MODIS. Knowledge of the land cover allows for calculation the fraction of incoming solar radiation that is absorbed by vegetation. This fraction of photosynthetically active radiation (fPAR) absorbed by vegetation describes rate at which carbon dioxide and energy from sunlight are assimilated into carbohydrates during photosynthesis of plant tissues. The summation of carbon assimilated by the vegetation canopy over time yields the landscape's gross primary productivity. Daily MODIS imagery is the basis for periodic composites and monthly data products. Over the 6 year period from 2000-2005, we calculate the annual average cumulative total of 72 monthly fPAR measurements, to describe the integrated annual vegetative production of the landscape, the integrated average annual minimum monthly fPAR measurement, which describes the annual minimum green cover of the observed landscape, and the integrated average of the annual covariance of fPAR, which describes the seasonality of the observed landscape. We also share the combination of the annual integrated values for visualization and analysis as the Dynamic Habitat Index (with additional information in Coops et al. 2008).

eco40_r (eco40)

bc_tree_spec_r (bctreespec)

The data represent dominant tree species for British Columbia forests in 2015, are based upon Landsat data and modeling, with results mapped at 30 m spatial resolution. The map was generated with the Random Forests classifier that used predictor variables derived from Landsat time series including surface reflectance, land cover, forest disturbance, and forest structure, and ancillary variables describing the topography and position. Training and validation samples were derived from the Vegetation Resources Inventory (VRI), from a pool of polygons with homogeneous internal conditions and with low discrepancies with the remotely sensed predictions. Local models were applied over 100x100 km tiles that considered training samples from the 5x5 neighbouring tiles to avoid edge effects. An overall accuracy of 72% was found for the species which occupy 80% of the forested areas. Satellite data and modeling have demonstrated the capacity for up-to-date, wall-to-wall, forest attribute maps at sub-stand level for British Columbia, Canada.

green_r (Greenness)

Urban Greenness Score (1984-2016) for 18 selected major Canadian urban areas The Urban Greenness Score data included in this product covers 33 years and all contiguous census dissemination areas of 18 selected major Canadian urban areas. The 18 urban areas represent over half of Canada’s population in 2016 (Czekajlo et al. 2020). The Urban Greenness Score uses greenness fractions from an annual time series (1984-2016) of spectrally unmixed Landsat satellite image composites (White et al. 2014; https://doi.org/10.1080/07038992.2014.945827; Hermosilla et al. 2016, https://doi.org/10.1080/17538947.2016.1187673) to characterize greenness and its overall change, summarized by census dissemination area. Image Code; Urban Greenness Score; Description 1; -L; Decrease in greenness resulting in a low final greenness 2; 0L; Stable low level of greenness 3; +L; Increase in greenness resulting in a low final greenness 4; -M; Decrease in greenness resulting in a moderate final greenness 5; 0M; Stable moderate level of greenness 6; +M; Increase in greenness resulting in a moderate final greenness 7; -H; Decrease in greenness resulting in a high final greenness 8; 0H; Stable high level of greenness 9; +H; Increase in greenness resulting in a high final greenness For more information about the data, image processing and spectral unmixing methods applied, development of the urban greenness score, and information on independent accuracy assessment of the data, as well as to cite this data, please use: Czekajlo, A., Coops, N.C., Wulder, M.A., Hermosilla, T., Lu, Y., White, J.C., van den Bosch, M., 2020. The urban greenness score: A satellite-based metric for multi-decadal characterization of urban land dynamics. International Journal of Applied Earth Observation and Geoinformation. 93, 102210. https://doi.org/10.1016/j.jag.2020.10221

LeadTreeSpec (LeadingTreeSpecies)

ca_change_year_r1984 (harvestTo2020)

The annual forest change data included in this product is national in scope (entire forested ecosystem) and represents the wall-to-wall characterization of harvest in Canada at a 30-m spatial resolution. The information outcomes represent 36 years of harvest change over Canada’s forests, derived from a single, consistent, spatially explicit data source, derived in a fully automated manner. This demonstrated capacity to characterize forests at a resolution that captures human impacts is key to establishing a baseline for detailed monitoring of forested ecosystems from management and science perspectives. Time series of Landsat data were used to characterize national trends in stand replacing forest disturbances caused by wildfire and harvest for the period 1985–2020 for Canada's 650 Mha forested ecosystems. Landsat data has a 30 m spatial resolution, so the change information is highly detailed and informative regarding both natural and human driven changes. These data represent annual stand replacing forest changes. The stand replacing disturbance types labeled are wildfire and harvest, with lower confidence wildfire and harvest, also shared. The distinction and sharing of lower class membership likelihoods is to indicate to users that some change events were more difficult to allocate to a change type, but are generally found to be in the correct category. For an overview on the data, image processing, and time series change detection methods applied, as well as information on independent accuracy assessment of the data, see Hermosilla et al. (2016; https://doi.org/10.1080/17538947.2016.1187673).

age_2019 (Age_2019)

Satellite-based forest age map for 2019 across Canada’s forested ecozones at a 30-m spatial resolution. Remotely sensed data from Landsat (disturbances, surface reflectance composites, forest structure) and MODIS (Gross Primary Production) are utilized to determine age. Age can be determined where disturbance can be identified directly (disturbance approach) or inferred using spectral information (recovery approach) or using inverted allometric equations to model age where there is no evidence of disturbance (allometric approach). The disturbance approach is based upon satellite data and mapped changes and is the most accurate. The recovery approach also avails upon satellite data plus logic regarding forest succession, with an accuracy that is greater than pure modeling. Given the lack of widespread recent disturbance over Canada’s forests, the allometric approach is required over the greatest area (86.6%). Using information regarding realized heights and growth and yield modeling, ages are estimated where none are otherwise possible. Trees of all ages are mapped, with trees >150 years old combined in an “old tree” category. See Maltman et al. (2023) for an overview of the methods, data, image processing, as well as information on agreement assessment using Canada’s National Inventory (NFI). Maltman, J.C., Hermosilla, T., Wulder, M.A., Coops, N.C., White, J.C., 2023. Estimating and mapping forest age across Canada’s forested ecosystems. Remote Sensing of Environment 290, 113529.