Land cover classification image for the Cypress Upland ecoregion of Saskatchewan with a spatial resolution of 10m. The goal of this land cover classification was to distinguish native from tame grasslands. The classification was based on Sentinel-1 and Sentinel-2 imagery using machine learning analysis in the Google Earth Engine platform. The classification was conducted on imagery acquired in 2023, and the classification model was built with field data collected in 2023. There are seven classes in total: native grassland, tame grassland, cropland, shrubs, trees, water, and urban area. Download: here The Prairie Landscape Inventory (PLI) aims to develop improved methods of assessing land cover and land use for conservation. Native grassland has historically been one of the hardest to map at-risk ecosystems because of the challenges in distinguishing native grassland from tame grassland land cover using remotely sensed imagery. This classification distinguishes native grassland from tame grassland and will provide valuable information for habitat suitability for native grassland species, identifying high biodiversity potential and invasion risk potential.   The classification map has seven (7) classes. The mixed grassland class included in the PLI land cover classification for other prairie ecoregions was not modelled in the Cypress Upland. 1. Cropland This class represents all cultivated areas with crop commodities, including corn, pulse, soybeans, canola, grains, and summer-fallow. 2. Native grassland This class represents the native grassland areas that are composed of at least 75% native grass, sedge and forb species, such as the needle grasses and wheatgrasses along with June grass and blue grama grass. Unbroken grassland that is invaded by species like Kentucky bluegrass, crested wheatgrass or smooth brome, such that native cover is less than 75%, is not considered native for the purpose of this project.  4. Tame grassland This class represents the tame grassland areas that are composed of at least 75% seeded or planted species with introduced grasses and forb species such as crested wheatgrass, smooth brome, Kentucky bluegrass, alfalfa, and sweet clover. 5. Water This class represents permanent water locations such as lakes and rivers. 6. Shrubs This class represents the sites dominated by woody vegetation of relatively low height (generally +/-2 meters) with shrub canopy typically >20% of total vegetation cover. 7. Trees This class represents the coniferous/deciduous trees, mixed-wood area, and other trees >2 meters height with tree canopy typically >20% of total vegetation cover. 9. Urban area This class represents both urban municipalities and buffered roads. Urban municipalities was used to mask the urban/developed area class of the Annual Crop Inventory 2021 (Agriculture Agri-Food Canada). Colour Classes: Value Label Red Green Blue 1 Cropland 255 255 190 2 Native grassland 168 168 0 4 Tame grassland 245 202 122 5 Water 190 232 255 6 Shrubs 205 102 153 7 Trees 66 128 53 9 Urban area 128 128 128 Accuracy metrics This model has an overall accuracy of 92 per cent. The table below summarizes the user’s accuracy, producer’s accuracy, and F1-score of the model on the validation dataset. Class User’s accuracy (%) Producer’s accuracy (%) F1-score Cropland 96 96 0.96 Native grassland 90 93 0.92 Tame grassland 93 71 0.82 Water 100 100 1.00 Shrubs 77 88 0.83 Trees 96 996 0.96

Temporal analysis of changes in Winnipeg, Manitoba, based on GeoAI features automatically extracted from satellite images acquired in 2013 and 2023. Simple geospatial analysis enables the detection of features present in 2023 that were not already there in 2013. The addition of new buildings is a good indicator of urban development and/or sprawl. Complementarily, an analysis of changes in the forest coverage from the GeoAI datasets is done. This analysis reflects the gains and losses between both dates. ‌ GeoAI enables temporal coverage of various areas in Canada, thus providing a useful tool for change detection and trend analysis at high resolution. While the series is still fairly new, and such examples are limited for the time being, NRCan strives to gradually increase its GeoAI data offering for both spatial and temporal coverage. For more information about the GeoAI - GeoBase Series, please visit the following link: https://open.canada.ca/data/en/dataset/74738ff5-5367-5958-9aee-98fffdcd1876

The 1 cm resolution vegetation digital height model was extracted using a bare earth model and digital surface model (DSM) derived from unmanned aerial vehicle (UAV) imagery acquired from a single day survey on July 28th 2016, in Cambridge Bay, Nunavut. The mapping product covers 525m2 and was produced by Canada Centre for Remote Sensing /Canada Centre for Mapping and Earth Observation. The UAV survey was completed in collaboration with the Canadian High Arctic Research Station (CHARS) for northern vegetation monitoring research. For more information, refer to our current Arctic vegetation research: Fraser et al; "UAV photogrammetry for mapping vegetation in the low-Arctic" Arctic Science, 2016, 2(3): 79-102. http://www.nrcresearchpress.com/doi/abs/10.1139/AS-2016-0008

GeoAI are buildings, hydrography, forests, and roads automatically extracted using Deep Learning models applied to a source dataset, typically aerial or satellite images. The primary aim of GeoAI is to increase Canada's availability of high-resolution foundational geospatial data for both spatial and temporal coverage. ‌ The infrastructure and expertise put in place by NRCan enables a rapid, efficient, and scalable data creation process through the use of leading-edge technology and Artificial Intelligence models. Published datasets for a given source can be revisited at a later date as more accurate models are developed and put into production. For now, only static files are available, but as the series develops, new products and services will be added. Several use cases of GeoAI data for change detection have been created. To view them, please follow the links below: - Iqaluit: https://open.canada.ca/data/en/dataset/ef29513d-a007-408c-beba-65f86e7b5b01 - Calgary: https://open.canada.ca/data/en/dataset/bd07b597-5166-4b62-ae9c-245b3c62ea03 - Québec City: https://open.canada.ca/data/en/dataset/4026720e-f7fb-478e-a58f-c62ebe20005d - Winnipeg: https://open.canada.ca/data/en/dataset/a6874aac-cfcd-4b92-bb3f-6031eedb5b57 - Victoria: https://open.canada.ca/data/en/dataset/f011c7b9-f79e-42f8-9481-9bf9788bb189 - Trois-Rivières: https://open.canada.ca/data/en/dataset/2b11a323-52c2-43d5-9c09-bc4b15b1bcc8

This layer represents Land use polygons as determined by a combination of analytic techniques, mostly using Landsat 5 image mosaics . BTM 1 was done on a federal satellite image base that was only accurate to about 250m. The images were geo-corrected, not ortho-corrected, so there is distortion in areas of high relief. This is not a multipart feature

Temporal analysis of changes for Victoria, British Columbia, based on GeoAI features automatically extracted from satellite images acquired in the 2000s, 2010s and 2020s. Simple geospatial analysis intersecting Statistics Canada's Open Database of Buildings, version 3 (ODB v3) with GeoAI multidate building features enables the detection of buildings observed in the 2000s, 2010s and 2020s. The addition of new buildings is a good indicator of urban development and/or sprawl. Using the same approach, GeoAI multidate roads enable the detection of Statistics Canada's National Roads Network (NRN) segments present in the 2000s, 2010s and 2020s. The development of new roads is also indicator of urban development and/or sprawl. Complementarily, an analysis of changes in the forest coverage from the GeoAI datasets is done between the 2000s and the 2020s. This analysis reflects the gains and losses between both dates. ‌ GeoAI enables temporal coverage of various areas in Canada, thus providing a useful tool for change detection and trend analysis at high resolution. While the series is still fairly new, and such examples are limited for the time being, NRCan strives to gradually increase its GeoAI data offering for both spatial and temporal coverage. For more information about the GeoAI - GeoBase Series, please visit the following link: https://open.canada.ca/data/en/dataset/74738ff5-5367-5958-9aee-98fffdcd1876

Each pixel value corresponds to the mean historical “Best-quality” Max-NDVI value for a given week, as calculated from the previous 20 years in the MODIS historical record (i.e. does not include data from the current year). These data are also often referred to as “weekly baselines” or “weekly normals”.

The MODIS surface albedo dataset was produced by the Canada Center for Remote Sensing (CCRS), Natural Resources Canada. The dataset represents the solar shortwave broadband surface albedo and it is at a 10-day interval covering the entire Canadian landmass as well as northern USA, Alaska, and the Greenland. The dataset was derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the TERRA satellite which provides a global coverage every 1-2 days in 36 spectral bands ranging from visible to infrared and to thermal wavelengths between 405 and 14,385 nm, and was available since 2000. For the estimation of surface albedo, the first seven spectral bands of B1 to B7 ranging from 459 nm to 2155 nm were used. B1 and B2 have a 250 meter resolution and B3 to B7 have a 500 meter resolution. A downscaling method using a regression and normalization scheme was employed to downscale the bands B3 to B7 to 250 meter resolution while preserving radiometric properties of the original data. To obtain clear-sky observations from MODIS, composite images for a 10 day period were generated by using a series of advanced algorithms (Luo et al., 2008). The 10-day composites of B1-B7 reflectance were then used to retrieve spatially continuous spectral albedo by using a combined land/snow BRDF (Bi-directional Reflectance Distribution Function) model. In that method, the modified RossThick-LiSparse BRDF model (Maignan et al., 2004) for land and Kokhanovsky and Zege’s model (2004) for snow are linearly combined for mixed surface conditions. They are weighted by snow fraction (0.0 ~ 1.0). The seven spectral albedo were then converted into the shortwave broadband surface albedo using the empirical MODIS polynomial conversion equation of Liang et al. (1999). The data product is in LCC (Lambert Conformal Conic) projection with a 250m pixel resolution. There are 36 albedo images per year. A dataset representing the pixel state (e.g. cloud/shadow, snow/ice, water, land, et al.) was also generated for each 10-day corresponding to the surface albedo product. References: Kokhanovsky, A. A. and Zege, E. P., 2004, Scattering Optics of Snow, Applied Optics, 43, 1589-1602, doi:10.1364/AO.43.001589, 20. Liang, S., Strahler, A.H., Walthall, C., 1999. Retrieval of land surface albedo from satellite observations: a simulation study. J. Appl. Meteorol. 38, 712–725. Luo, Y., Trishchenko, A.P., Khlopenkov, K.V., 2008. Developing clear-sky, cloud and cloud shadow mask for producing clear-sky composites at 250-meter spatial resolution for the seven MODIS land bands over Canada and North America. Remote Sens. Environ. 112, 4167–4185. Maignan, F., F.M. Bréon and R. Lacaze, 2004, Bidirectional reflectance of Earth targets : evaluation of analytical models using a large set of spaceborne measurements with emphasis with the hot spot, Remote Sens. Environ., 90, 210-220.