Dynamic surface water maps of Canada from 1984 to 2019 Landsat satellite imagery

Abstract

Knowledge of the location of surface water in time and space is critical to inform policy on the environment, wildlife, and human welfare. Dynamic surface water maps have been created at continental to global scales from medium and coarse resolution satellite image archives, most notably by Pekel et al. (2016), who mapped global water dynamics from 1984 to 2015 Landsat data. Their occurrence layer, depicting the percentage of time that water was observed in each 30 m resolution cell, has been relied upon heavily by Natural Resources Canada's (NRCan) Emergency Geomatics Service (EGS) to map surface water during flood events. To generate dynamic surface water maps that are optimized for Canada's unique geography, the fully automated EGS surface water mapping methodology was applied in a cloud environment to the 1984–2019 Landsat archive over Canada. National-scale surface water maps and derived inundation frequency akin to Pekel's occurrence, as well as inter-annual wetting and drying trends calculated using per-pixel logistic regression, were produced to form the complete dataset. Separate comparisons of our frequency layer with Pekel's occurrence layer and Canada's water base layer from the National Hydrographic Network (NHN) were conducted. The comparison with Pekel's occurrence showed that our product contains a large number of unique water objects, the majority of which are correct when assessed against Google Earth (GE) imagery. Comparison with the NHN indicated that the NHN contained a large number of permanent water objects that were mapped as ephemeral water objects in our product, with the majority of these being verified in GE as true ephemeral water objects such as floodplains and wetlands. Wetting trends were found to be more than five times greater than drying trends across Canada, with notable wetting in the Prairie Pothole region and Low Arctic verified with examples of statistically significant wetting and drying features. The dataset will enhance EGS flood and river ice mapping operations, provide information on floodplain location and extent, and give insight into the effects of climate change on surface water availability.