ISO 19131 AAFC Annual Crop Inventory – Data ...

URL: http://130.179.67.140/dataset/f2f37ad5-0afa-4fc4-885d-6a6e9f8b6dc8/resource/04a7532d-a7b0-416d-939a-fb950ca8012e/download/iso19131aafcannualcropinventorydataproductspecifications.pdf

Agriculture and Agri-Food Canada (AAFC) has been moving towards the development of an operational software system for mapping the crop types of individual fields using satellite observations. Successful crop identification relies on image acquisitions from multiple sensors during key crop phenological stages (reproduction, seed development and senescence). Multi-temporal optical data are the primary data source for crop classification because the NIR/SWIR channels are vital to crop classification. Over a growing season, at least three optical images are required to successfully identify crops. To the optical data, dual-polarization RADARSAT-2 data is added. In 2009 and 2010 the ScanSAR mode, with its large swath (300 km) and moderate resolution (50 m), was used as it fits the agricultural landscape of the Prairie Provinces. From 2011-present, the finer resolution of the Wide mode (30 m) is used as it is better suited to narrower fields. Annual crop insurance data are the most accurate, detailed and complete sources of information for crop types in Canada. As such, AAFC cooperates with provincial crop insurance agencies to use their data for the training and validation of satellite data analysis. For provinces where insurance data cannot be accessed, ground-truth information is provided by point observations from AAFC staff or other provincial sources. Each year, AAFC staff collects tens of thousands of points identifying crops across the country. Both these point sources are combined and used as training or reference sites. In addition, the AAFC Land Cover for Agricultural Regions of Canada, circa 2000 map is used to define the agricultural extent under each annual crop inventory. This map is updated to the current year through an automated process. Focusing on the Prairie Provinces in 2009 and 2010, a Decision Tree (DT) based methodology was applied using optical (Landsat-5, AWiFS, DMC) and radar (RADARSAT-2) imagery. The final map had a spatial resolution of 56m. For the 2011 and 2012 growing season, this activity was extended to all the other provinces (except Newfoundland) in support of a national crop inventory. The final spatial resolution was increased to 30m, to aid in differentiating the smaller fields in the rest of Canada. For 2012, the lack of affordable optical data forced AAFC to rely mostly on RADARSAT-2 data. In 2013, this activity expanded to include Newfoundland for the first time, and used Landsat-8 as its sole-source of optical imagery. RADARSAT-2 continued to be the source of radar imagery. This combination of optical and radar imagery was again used over the entire agricultural extent of Canada for 2014. At present, this approach can consistently deliver a crop inventory that meets the overall target accuracy of at least 85% at a final spatial resolution of 30m. Note: At the national scale, the crop type legend is not homogeneous. In some provinces, such as Alberta, Saskatchewan and Quebec, we have been able to divide the cereal crops in to sub-categories (Barley, Oats, Wheat, etc.). For other provinces, the cereals class may not have been subdivided. The lack of training sites and, in some cases, the limited availability of spectral data does not allow for the differentiation of cereals into sub-categories with sufficient precision. This results in class discontinuities between provinces.

This resource can not be previewed at the moment. Click here for more information.

Download resource

Additional Information

Field Value
Last updated December 15, 2015
Created December 15, 2015
Format PDF
License Other (Attribution)
can be previewed1
createdover 1 year ago
formatPDF
id04a7532d-a7b0-416d-939a-fb950ca8012e
on same domain1
position5
resource group id2271c96e-d38b-40b8-abd3-67483660c039
revision id4a7d7023-c8e9-48be-bd0c-990095fb585d
stateactive
url typeupload