Canada-wide 1-km Water Fraction From National Topographic Data Base Maps Summary: This product is a raster coverage representing the fraction of area within each 1km grid cell over Canada's land mass covered by waterbodies, as mapped within the National Topographic Data Base v3.1. In NTDB, waterbody is defined as area covered by open water belonging to one of five categories (section 1.5.1). The product involves negligible errors due to aggregation and data format conversion in comparison with the NTDB digital base maps used in its generation. As such, the errors between mapped and actual water fractions are primarily due to differences between the NTDB source maps and the actual waterbody coverage in a grid cell. Since water area fluctuates seasonally and interanually one should use caution in applying this product for local mapping and modeling purposes. At a minimum, one should verify that the survey dates of the source NTDB maps are appropriate. Furthermore, some consideration should be given to seasonal shifts in waterbody extent. The majority of NTDB source maps were surveyed within the growing season (post snow-melt) and should therefore be relatively unbiased during the growing season in the absence of drought or severe precipitation events. At regional and national scales we expect that seasonal (weather-induced) fluctuations in water extent are the major uncertainty in water fractions. Table of Contents * 1 Data Set Overview * 2 Investigator(s) * 3 Theory of Measurements * 4 Equipment * 5 Data Acquisition Methods * 6 Observations * 7 Data Description * 8 Data Organization * 9 Data Manipulations * 10 Errors * 11 Notes * 12 Application of the Data Set * 13 Future Modifications and Plans * 14 Software * 15 Data Access * 16 Output Products and Availability * 17 References * 18 Glossary of Terms * 19 List of Acronyms * 20 Document Information 1. Data Set Overview 1.1 Data Set Identification ABC3 Canada-wide 1-km AVHRR composite maps Citation : Fernandes, R.A., Pavlic, G., Chen, W., and Fraser, R. 2001. Canada- wide 1-km water fraction derived from National Topographic Data Base maps. Natural Resources Canada. 1.2 Data Set Introduction This dataset represents the only raster representation of water fraction coverage across Canada that we are aware of. Care has been taken to ensure traceability to the NTDB source data. 1.3 Objective/Purpose There are three major objectives for generating this product: 1. Providing information of water fraction coverage as input to the processing, analysis and interpretation of satellite data. 2. Providing a visualization of regional and national water fraction coverage. 3. Providing initial estimates of water fraction coverage for numerical simulation models independent of topology information available from hydrographic networks. There may be other applications we have not envisioned. The spatial resolution of the raster grid can be easily improved, limited only by copyright restrictions when the resolution allows representation of base map features indistinguishable from the vector data layers. 1.4 Summary of Parameters Proportion of grid cell covered by water (0 to 100%) plus a no-data value. 1.5 Discussion 1.5.1 Introduction Nearly 25% of the world's fresh water is found in Canada. Canada does not currently provide an accurate spatial or statistical summary of fresh water covered areas. Statistics Canada reports the total area of waterbodies per province and nationally based on the same NTDB sources used in this product. Total area Land Freshwater % of total area of fresh water km² Canada 9,984,670 9,093,507 891,163 100.0 Newfoundland 405,212 373,872 31,340 4.1 Prince Edward Island 5,660 5,660 – 0.1 Nova Scotia 55,284 53,338 1,946 0.6 New Brunswick 72,908 71,450 1,458 0.7 Quebec 1,542,056 1,365,128 176,928 15.4 Ontario 1,076,395 917,741 158,654 10.8 Manitoba 647,797 553,556 94,241 6.5 Saskatchewan 651,036 591,670 59,366 6.5 Alberta 661,848 642,317 19,531 6.6 British Columbia 944,735 925,186 19,549 9.5 Yukon 482,443 474,391 8,052 4.8 Northwest Territories 1,346,106 1,183,085 163,021 13.5 Nunavut 2,093,190 1,936,113 157,077 21.0 – nil or zero Source: Natural Resources Canada, GeoAccess Division. The Canadian forest inventory reports water-covered areas within forest regions aggregated to 10x10km square grid cells based on provincial data bases. Their total estimate is on the order of 77x10^6 km^2. The difference in comparison to the Statistics Canada estimate of 89x10^6km^2 reflects (i) the use of provincial rather than uniform NTDB data sources and (ii) the omission of non-forested regions. NRCan has recently produced a vector hydrographic coverage of Canada using the NTDB for reference. However, this coverage emphasizes compatibility with the VMAP0 global vector database. As such, modifications to areal coverage are performed to ensure a reasonable network topology of hydrographic elements. Furthermore, the water coverage area is checked with the NTDB and is not equivalent to NTDB water covered areas. Therefore, this raster product is the current estimate of water fraction coverage across Canada which is traceable to the NTDB v3.0. The product will be revised to reflect updates to the NTDB. This raster product contains the value of the total area of each 1km grid cell (according to a specific map projection) covered by waterbodies. Waterbodies are operationally defined in the NTDB as: i. Generic/Unknown - a region covered with water at time of survey; ii. Intermittent/slough 0 a waterbody normally dry at some time of the year; iii. Other - other regions designated as waterbodies; iv. Irrigation canal - a manmade waterbody; v. Flooded area - permanently flooded region. The only other NTDB categories that may contain water at the surface are 'tundra ponds' and 'wells'. We have not included these categories in the water fraction product. 'Tundra ponds' are classified as wetlands and will be released with a separate wetland product. 1.5.2 Methodology 1. All NTDB v3.0 1:50,000 and 1:250,000 scale map sheets were acquired in vector data format. 2. The water body layer was extracted. This consisted of polygon features with topology. 3. Each vector map was reprojected to the output projection. 4. A raster-to-vector conversion was performed at 5m grid resolution. This resolution was selected based on error analysis that showeed that this resolution produced a negligible bias between the raster- and vector- based estimates of waterbody polygon area within 1km square regions. 5. A raster overlay of the 1km grid and the 5m grid was performed and the waterbody area of all 5m grid cells completely within the 1km cell were added. Partial cells were included if they crossed the southern or eastern edges of the grid cell. As such, there is a slight random bias in the mapped area, although the total area of Canada is accurate. 6. The aggregated sum of waterbody area was converted into a proportion by dividing by 10^6 m^2. 7. The final water fraction map was generated by tiling 1km estimates from 1:50,000 map sheets. Regions not entirely covered by 1:50,000 map sheets were filled in with estimated from 1:250,000 map sheets. 1.5.3 Discussion of the Resulting Water fraction Map The water fraction map is intended for application for regional modelling of environmental systems. There will be differences between the water fraction in an a 1km pixel and the actual water fraction on the surface at any point in time. There are two sources of error in the final map: i. Errors between the NTDB and reality. ii. Errors between areas represented by 1km grid cells versus the total area of NTDB waterbody polygons in that cell. i. Errors between the NTDB and reality can fall under three categories: a. Generalisation. The exentent to which spatial detail is simplified during the mapping process. b. The maximum spatial displacement between a feature on the map and in reality. c. Thematic Accuracy: The difference in feature labels on the map and in reality (i.e. water body coverage will vary with climate). iii. Errors between the NTDB and the 1km raster grid areas are due to three sources: a. Numerical rounding errors b. Raster to vector conversion error c. Raster aggregation 1.5.4 Recommendations This raster product is intended as a traceable representation of the NTDB waterbody areas. We emphasize that, given the dynamic nature of water area with both seasonal and inter-annual changes, this coverage should be checked with other data sources (field survey or interpretation of current satellite imagery) to ensure it is thematically valid for a region and the time period of interest. 1.6 Related Data Sets Geomatics Canada (1996), National Topographic Data Base, Standards and Specifications, Edition 3.1, Minister of Supply and Services Canada. Rhian Evans, (2000) National Atlas Framework Data – Release Notes http://geogratis.cgdi.gc.ca/framework/Introduction%20Framework%20- %20English.htm, GeoAccess Division, Canada Centre for Remote Sensing. Statics Canada, (2001), Land and Freshwater Area, Table, http://www.statcan.ca/english/Pgdb/Land/Geography/phys01.htm 2. Investigator(s) 2.1 Investigator(s) Name and Title Richard Fernandes Research Scientist Environmental Monitoring Section Applications Division Canada Centre for Remote Sensing richard.fernandes@ccrs.nrcan.gc.ca Wenjun Chen Research Scientist Environmental Monitoring Section Applications Division Canada Centre for Remote Sensing wenjun.chen@ccrs.nrcan.gc.ca 2.2 Title of Investigation correct: 1km Gridded Water Fraction Cover Map of Canada from NTDB 2.3 Contact Information Richard Fernandes Canada Centre for Remote Sensing 588 Booth Street, 4th Floor Ottawa, Ontario K1A 0Y7 Canada (613) 947-1202 (613) 947-1383 (fax) richard.fernandes@ccrs.nrcan.gc.ca 3. Theory of Measurements The NTDB maps waterbody features by visual interpretation based on the following survey sources: Aerial Photography; Reproduction Material (paper maps); MSS Landsat Imagery; TM Landsat Imagery; Spot XS Imagery; Spot PAN Imagery; GPS Data; Digital Vector Data from other sources. Features are guaranteed to be mapped with less than 5% total (commission + omission) error if they are larger than a specified dimensional threshold (see section 4.2.1.1 ). Features may be also mapped if they fall below this threshold. Specifications for temporal extent of survey periods are not provided but are assumed to correspond to the growing season post snow-melt peak runoff periods. 4. Equipment 4.1 Sensor/Instrument Description Not applicable. 4.1.1 Collection Environment NTDB is provided as ARC/INFO interchange files in GRS1980 UTM projection. 4.1.2 Source/Platform Not applicable. 4.1.3 Source/Platform Mission Objectives The objective of the NTDB is to provide topographic information over the entire Canadian landmass in a consistent manner. 4.1.4 Key Variables Polygon features corresponding to waterbodies. 4.1.5 Principles of Operation The NTDB is based on a manual survey and image interpretation. The water fraction raster database was produced by GIS overlay procedures on a PC workstation. 4.1.6 Sensor/Instrument Measurement Geometry Not applicable. 4.1.7 Manufacturer of Sensor/Instrument The NTDB is produced by Natural Resources Canada, Government of Canada. 4.2 Calibration Polygon coverages in ARC/INFO interchange format with full topology under UTM (GRS1980) projection. 4.2.1.1 Tolerance None provided for geometric tolerances of the databases. See Section 10 Sources of Error. 4.2.1 Specifications Not applicable. 4.2.2 Frequency of Calibration Vintage of features and survey checks of the NTDB vary from feature to feature. Users interested in this information should check the NTDB source data. 6. Observations 6.1 Data Notes None. 6.2 Field Notes None. 7. Data Description 7.1 Spatial Characteristics 7.1.1 Spatial Coverage Canadian terrestrial region. 7.1.2 Spatial Coverage Map The North American Datum of 1983 (NAD83) corner coordinates of the Canada-wide map are: Latitude Longitude Corner Location 1) 66.912698° or 66°54'45.7" N -177.263306° or -177°15'47.9"W UL (Northwest) 2) 62.428531° or 62°25'42.7" N -9.997006° or -09°59'49.2"W UR (Northeast) 3) 36.218551° or 36°13'06.8" N -122.904216° or -122°54'15.2"W LL (Southwest) 4) 34.301567° or 34°18'05.6" N -62.547110° or -62°32'49.6"W LR (Southeast) 7.1.3 Spatial Resolution 1km grid cells (nominal size, due to use of conformal rather than equal area projection). 7.1.4 Projection The established grid system is based on the ellipsoidal version of the Lambert Conformal Conic (LCC) projection as defined within the NAD83 datum and WGS84 ellipsoid. The origin of the grid is at 95° W, with the central meridian at 95° W, latitude of origin at 0° N and the standard parallels are set to 49° N and 77° N (Snyder, 1987). 7.1.5 Grid Description The composite maps are projected into the LCC projection (section 7.1.4) at a grid cell size of 1-km square per pixel. The grid size is 5700 lines and 4800 pixels. 7.2 Temporal Characteristics 7.2.1 Temporal Coverage Spatial features corresponding to waterbodies have survey dates extending back to 1949. Users should check the NTDB source data for the exact survey date for their area of interest. 7.2.2 Temporal Coverage Map None. 7.2.3 Temporal Resolution Waterbody features were mapped post snow-melt during the growing season. It is expected that water fraction will fluctuate seasonally and inter-annually. Users should perform field or remote (satellite/air photo) verification of the general level of fluctuation for their region on interest. 7.3 Data Characteristics 7.3.1 Parameter/Variable Final data products include: Water fraction coverage proportion scaled from 1 (0%) to 10000 (100%). 'No data' is represented by 10000 although this will either represent large lakes or oceans not mapped within the NTDB framework OR land regions not contained within Canada. An associated Canada mask is being produced to allow distinction between these two cases. 7.3.2 Variable Description/Definition Water fraction = proportion of planimetric area of each 1km grid cell covered by waterbody polygons from the NTDB source data. 7.3.3 Unit of Measurement Proportion of planimertic area relative to local datum. 7.3.4 Data Source National Topographic Data Base, Natural Resources Canada, Government of Canada. 7.3.5 Scaled Data Range Data is represented in 16 bit unsigned (Intel) format. A value of 1 corresponds to 0% water fraction coverage. A value of 10000 corresponds to 100% water body coverage. Intermediate values correspond to: %coverage = value/10000 'No data' is represented by a 0. 7.4 Sample Data Record Image representation of the data base is included below: 8. Data Organization 8.1 Data Granularity Single raster file: canadawf.bil 8.2 Data Format(s) Raw binary 16 bit unsigned (Intel) raster. 9. Data Manipulations The methods here are repeated form Section 1.5.2 Methodology for convenience. All NTDB v3.0 1:50,000 and 1:250,000 scale map sheets were acquired in vector data format. The waterbody layer was extracted. This consisted of polygon features with topology. Each vector map was reprojected to the output projection. A raster-to-vector conversion was performed at 5m grid resolution. This resolution was selected based on error analysis that suggested it produced negligible bias between raster and vector based estimates of waterbody polygon area over 1km square regions. A raster overlay of the 1km grid and 5m grid was performed and the waterbody area of all 5m grid cells completely within the 1km cell were added. Partial cells were included if they crossed the southern or eastern edges of the grid cell. As such, there is a slight random bias in mapped area although the total area of Canada is accurate. The aggregated sum of waterbody area was converted into a proportion by dividing by 10^6 m^2. The final waterbody map was generated by tiling 1km estimates from 1:50,000 map sheets. Regions not entirely covered by 1:50,000 map sheets were filled in with estimated from 1:250,000 map sheets. Data is output in a binary raster image with 16bit unsigned data. 9.1 Formulae Refer to section 1.5.2 for processing steps. 10. Errors 10.1 Sources of Error There are two sources of error in the final map: i. Errors between the NTDB and reality. ii. Errors between areas represented within 1km grid cells versus the total area of NTDB waterbody polygons in that cell. i. Errors between the NTDB and reality can fall under three categories: a. Generalisation. The NTDB provides specifications for the minimum size of a waterbody represented as waterbody features as a function of map scale in units of meters: Category 1:50,000 1:250,000 Range> 50* 250** Width> 25 125 Length> 500*** 1500*** * The guaranteed size for intermittent/slough and flooded areas are 100m and 250m respectively at 1:50,000 scale; ** The guaranteed size for intermittent/slough and flooded areas are 375m and 1250m respectively at 1:250,000 scale; *** The guaranteed length values apply only to the attribute 'Irrigation canal'. The range corresponds to the length of the sides of a square corresponding to the superfices used for representing a waterbody. The width is the shorter dimension. The length is the dimension perpendicular to the width. b. Locational Accuracy The NTDB aims at three levels of locational planimetric accuracy expressed as a circular map accuracy (the size of a circle that will encompass the true feature location with respect to the mapped location): Urban areas : 10m or better; Rural areas: 25m or better; Isolated areas: 125m or better. c. Thematic Accuracy The NTDB specifies a total error of omission and classification of less than 5%. All spatial features include a 'data validation date' field giving the month and year of validation. This field gives an idea of the temporal representativeness of the waterbody feature. Furthermore, the field 'type of data source' is also available identifying the survey method. This source may provide some idea of uncertainties inherent in a survey method (e.g., Landsat imagery). iii. Errors between the NTDB and the 1km raster grid areas are due to three sources: a. Numerical rounding errors This source is negligible until final raster representation as double precision floating point computations are performed. The final map is represented as a proportion scaled from 0 to 32767 (100%). b. Raster to vector conversion error A 5m grid cell size is applied for raster to vector conversion. This size will not miss features since the 5m size is well below the criteria for including a feature in the 1:50,000 scale maps. However, the gridding procedure will result in errors that will increase as features become more convoluted. An error analysis (provided below) suggests the difference between 5m gridded areas and the polygon area is well below the 5% classification error. c. Raster aggregation Aggregation of 5m cells within a 1km region is performed using a lattice grid specified by the NRCan LCC Projection Grid (see Map Projection below). As such, 5m cells will straddle the boundary of 1km grid cells. The current aggregation includes all 5m grid cells straddling the southern and eastern boundaries but not the northern and western boundaries. This ensures that tiling a number of adjacent 1km grid cells will reduce total mapped area errors to only the error at the boundary of the tiled region. 10.2 Quality Assessment The relative root mean square difference between vector polygon and raster gridded waterbody proportions for 1km pixels in sampled NTDB map sheets was always below 3%. The sampled map sheets were chosen to represent worst case scenarios, i.e. many small lakes and rivers with convoluted boundaries or isolated rivers. The total water fraction error between 1km gridded waterbody estimates and the polygon based area of the map sheet was quantified for all 1;250,000 map sheets. The mean bias for 1:250,000 map sheets was 0.1287% of the area over the map sheet (standard deviation of 0.1539%). The total water fraction error between 1km gridded waterbody estimates and the polygon based area of the map sheet was quantified for 1081 of the 1;50,000 map sheets used. The mean bias of a 1:50,000 map sheets was -2.08% (standard deviation of 4.9%). The larger errors for the 1:50,000 map sheets is due to the small amount of area covered by water in some 1:50,000 map sheets (southern Canada), combined with larger errors in regions with small features. For example, removing the 1:50,000 map sheets corresponding to the 1% smallest waterbody coverage reduces the mean bias to -1.94% (standard deviation of 2.68%). 11.2 Known Problems with the Data The major problem with the data is the uncertainty in temporal changes in waterbody coverage, especially small ponds/lakes and rivers. This problem exists in the source data. A secondary problem is the error due to generalization in the source data. Comparison of 1:50,000 and 1:250,000 scale maps indicates small features are ot represented with 1:250,000 scale NTDB maps. We can expect the waterbody area to be underestimated in regions where only 1:250,000 scale maps are available (northern Canada). This underestimate results in a small bias over an entire 1:250,000 map sheet (<2% in the worst cases seen) since the generalization used at 1:250,000 scale induces random errors that often mask bias when there are small lakes. Differences between the 1km raster grid and the source NTDB polygon water fraction areas are negligible in comparison to errors in the source data. Users may require improved spatial resolution if they are performing spatial analysis on the raster data set. Improvement can be achieved by re-gridding the 5m intermediate grid data. However, a valid hydrographic network would require extensive editing. Other NRCan products are available for hydrographic network analysis, and may be used in conjunction with this raster product to adjust area fractions. Users of ARCMAP software should make use of the waterfractiontif.zip archive rather than the waterfraction.zip archive to avoid artifacts in coding 16 bit unsigned data when importation into ARCMAP. 13. Future Modifications and Plans A 100m resolution coverage will be available after the next NTDB revision (based on Landsat TM data). 14. Software Not applicable. 15. Data Access 15.1 Contact Information Richard Fernandes Canada Centre for Remote Sensing 588 Booth Street, 4th Floor Ottawa, Ontario K1A 0Y7 Canada (613) 947-1265 (613) 947-1383 (fax) richard.fernandes@ccrs.nrcan.gc.ca 15.2 Data Center Identification Canada Centre for Remote Sensing, Natural Resources Canada 15.3 Procedures for Obtaining Data E-mail or written request to contact person. 15.4 Data Center Status/Plans It is planned to make this product available through NRCan's GeoConnections website. 16. Output Products and Availability 16.1 Tape Products 16.2 Film Products None. 16.3 Other Digital Products 17. References 17.1 URLs Cited in the Text 17.2 Platform/Sensor/Instrument/Data Processing Documentation 17.3 Journal Articles and Study Reports Pavlic, G. Fernandes, R.A., Chen, W. and Fraser, R. 2001. A raster water fraction coverage of Canada. To be presented at Canadian Institue for Geomatics Conference, Pttawa, 2002. Snyder, J.P. 1987. Map Projections - A Working Manual. USGS Professional Paper 1395. 17.3 Archive/DBMS Usage Documentation None. 18. Glossary of Terms AEAC - Albers Equal Area Conic ASCII - American Standard Code for Information Interchange BSQ - Band Sequential CCRS - Canada Centre for Remote Sensing CCT - Computer Compatible Tape CD-ROM - Compact Disk-Read-Only Memory CPIDS - Calibration Parameter Input Data Set DAAC - Distributed Active Archive Center DAT - Digital Archive Tape DN - Digital Number LCC - Lambert Conformal Conic NAD83 - North American Datum of 1983 NTDB National Topographic Data Base URL - Uniform Resource Locator UTM - Universal Transverse Mercator 19. List of Acronyms 20. Document Information 20.1 Document Revision Date Written: 20-Oct-2001 Last Updated: 20- Oct-2001 20.2 Document Review Date(s) Internal Review: 08-Nov-2001 External Review: 20.3 Document ID Waterfraction_metadata.doc (MS Word 2000) 20.4 Citation Fernandes, R.A., Pavlic, G., Chen, W., and Fraser, R. 2001. Canada-wide 1-km water fraction derived from National Topographic Data Base maps. Natural Resources Canada. 20.5 Document Curator Richard Fernandes rfernand@ccrs.nrcan.gc.ca 12