Portions of Universal Transverse Mercator Zones 7 - 12 which cover British Columbia, Northern Hemisphere only, formed into polygons, in BC Albers projection

AAFC’s Canadian Ag-Land Monitoring System (CALMS), operational since 2009, was developed by AAFC’s Earth Observation Service (EOS) to deliver weekly NDVI-based maps of crop condition in near-real-time. The CALMS uses data collected by the Moderate Resolution Imaging Spectro-radiometer (MODIS), a sensor mounted onboard NASA’s Terra satellite that has been acquiring data since February 2000. The state-of-the-art radiometric, spectral and spatial resolutions of MODIS Terra make it particularly well-suited for large-scale vegetation mapping and assessment. Crop condition (NDVI) maps are generated weekly by AAFC throughout Canada’s growing season, the period defined as the six-month period stretching from the start of Julian week 12 (end of March) to the end of Julian week 44 (late October). Weeks of the year are defined according to the ISO 8601 week-numbering standard, where weeks start on a Monday and end the following Sunday. CALMS products are generated in the MODIS native Integrated Sinusoidal (ISIN) projection for the region covering the twelve MODIS tiles h09v03 to h14v03 and h09v04 to h14v04.

Each pixel value corresponds to the actual number (count) of valid Best-quality Max-NDVI values used to calculate the mean weekly values for that pixel. Since 2020, the maximum number of possible observations used to create the Mean Best-Quality Max-NDVI for the 2000-2014 period is n=20. However, because data quality varies both temporally and geographically (e.g. cloud cover and snow cover in spring; cloud near large water bodies all year), the actual number (count) of observations used to create baselines can vary significantly for any given week and year.

CHS offers 500-metre bathymetric gridded data for users interested in the topography of the seafloor. This data provides seafloor depth in metres and is accessible for download as predefined areas.

Each pixel value corresponds to the best quality maximum NDVI recorded within that pixel over the week specified. Poor quality pixel observations are removed from this product. Observations whose quality is degraded by snow cover, shadow, cloud, aerosols, and/or low sensor zenith angles are removed (and are assigned a value of “missing data”). In addition, negative Max-NDVI values, occurring where R reflectance > NIR reflectance, are considered non-vegetated and assigned a value of 0. This results in a Max-NDVI product that should (mostly) contain vegetation-covered pixels. Max-NDVI values are considered high quality and span a biomass gradient ranging from 0 (no/low biomass) to 1 (high biomass).

Data represents surface water occurrence frequency (percentage), which describes the frequency for each grid appeared as water in the 30 years time period of 1991 to 2020. The data covers Canada’s entire landmass including all transboundary watersheds, and is at 30-meter spatial resolution. The surface water occurrence frequency is derived using the surface water model of Wang et al. (2023) from all-available monthly water data observed by the Landsat satellites (Pekel et al., 2016). Here, permanent waters are represented by 100%, and permanent land surfaces by 0%, of water occurrence for a 30-meter by 30-meter grid. References: Pekel, J.-F., A. Cottam, N. Gorelick, A.S. Belward, 2016, High-resolution mapping of global surface water and its long-term changes. Nature, 540, 418-422. Wang, S., J. Li, and H. A. J. Russell, 2023, Methods for Estimating Surface Water Storage Changes and Their Evaluations. Journal of Hydrometeorology, DOI: https://doi.org/10.1175/JHM-D-22-0098.1.

The ‘Circa 1995 Landcover of the Prairies’dataset is a geospatial raster data layer portraying the rudimentaryland cover types of all grain-growing areas of Manitoba, Saskatchewan, Alberta and northeastern British Columbia at a 30-metre resolution for the 1995 timeframe. It is the collection of all the classified imagery (1993 to 1995) of the Western Grain Transition Payment Program (WGTPP) assembled into a single seamless raster data layer.

The 2000 AAFC Land Use is a culmination and curated metaanalysis of several high-quality spatial datasets produced between 1990 and 2021 using a variety of methods by teams of researchers as techniques and capabilities have evolved. The information from the input datasets was consolidated and embedded within each 30m x 30m pixel to create consolidated pixel histories, resulting in thousands of unique combinations of evidence ready for careful consideration. Informed by many sources of high-quality evidence and visual observation of imagery in Google Earth, we apply an incremental strategy to develop a coherent best current understanding of what has happened in each pixel through the time series.

This data publication contains a set of files in which areas affected by fire or by harvest from 1984 to 2015 are identified at the level of individual 30m pixels on the Landsat grid. Details of the product development can be found in Guindon et al (2018). The change detection is based on reflectance-corrected yearly summer (July and August) Landsat mosaics from 1984 to 2015 created from individual scenes developed from USGS reflectance products (Masek et al, 2006; Vermote et al, 2006). Briefly, the change detection method uses a six-year temporal signature centered on the disturbance year to identify fire, harvest and no change. The signatures were derived from visually-interpreted disturbance or no-change polygons that were used to fit a decision tree model. The method detects about 91% of the areas harvested and 85% of the areas burned across Canada’s forests over the study period, but overestimates areas disturbed in the two initial and mostly in the two final years of the 1985 to 2015 time series. This is caused by the absence of appropriate pre-disturbance and post-disturbance data for the model-based detection and attribution. Disturbance coverage in those four years should therefore be used with caution. As in Guindon et al (2014), the method was designed to minimize commission errors and has a disturbance class attribution success rate of about 98%. The attribution success rate of disturbance year for fire is of about 69% for the exact year and of about 99% when attribution to the following year is also considered as a success. This common one-year lag is mostly due to the use of mid-summer Landsat mosaics for the analysis that will cause spring and fall events of the same year to be attributed to successive years. For example, a fire that occurred in the fall of 2004 (after July and August), will be detected and attributed to 2005, while for a fire that occurred in the spring of 2004 will be detected and attributed to 2004. The presence of clouds and shadows or image availability causes 10% of missing data annually and therefore can too delay the capture of events. The data provides uniform spatial and temporal information on fire and harvest across all provinces and territories of Canada and is intended for strategic-level analysis. Since no attention was given to other minor disturbances such as mining, road or flooding, the product should not be used for their identification. Finally, calibration datasets were developed for only three major forest pests (mountain pine beetle, eastern spruce budworm and forest tent caterpillar), but were folded within the “no-change” class in order to minimize commission errors for fire and harvest . Less common pests for which validation datasets are hard to develop were not considered and therefore could in rare circumstances generate false fire events. Considering that area having two (3.3%) to three disturbances (less than 1%) events are not common, only the most recent disturbance is provided, overlapping older disturbances in these rare case. ## Please cite this dataset as: Guindon, L., P. Villemaire, R. St-Amant, P.Y. Bernier, A. Beaudoin, F. Caron, M. Bonucelli and H. Dorion. 2017. Canada Landsat Disturbance (CanLaD): a Canada-wide Landsat-based 30-m resolution product of fire and harvest detection and attribution since 1984. https://doi.org/10.23687/add1346b-f632-4eb9-a83d-a662b38655ad ## Scientific article citation: The creation, validation and limitations of the CanLaD product are described in the Supplementary Material file associated with the following article: Guindon, L.; Bernier, P.Y.; Gauthier, S.; Stinson, G.; Villemaire, P.; Beaudoin, A. 2018. Missing forest cover gains in boreal forests explained. Ecosphere, 9 (1) Article e02094. doi:10.1002/ecs2.2094. ## Cited references: Masek, J.G., Vermote, E.F., Saleous N.E., Wolfe, R., Hall, F.G., Huemmrich, K.F., Gao, F., Kutler, J., and Lim, T-K. (2006). A Landsat surface reflectance dataset for North America, 1990–2000. IEEE Geoscience and Remote Sensing Letters 3(1):68-72. http://dx.doi.org/10.1109/LGRS.2005.857030. Vermote, E., Justice, C., Claverie, M., & Franch, B. (2016). Preliminary analysis of the performance of the Landsat 8/OLI land surface reflectance product. Remote Sensing of Environment. http://dx.doi.org/10.1016/j.rse.2016.04.008.

McElhanney Consulting Services Ltd (MCSL) has performed a LiDAR and Imagery survey in southern Saskatchewan. The acquisition was completed between the 16th and 25th of October, 2009. The survey consisted of approximately 790 square kilometers of coverage. While collecting the LiDAR data, we also acquired aerial photo in RGB and NIR modes consisting of 1649 frames each.