Application of Multi-Frequency and Multi-Temporal Polarimetric SAR for Operational Crop Inventories in Canada

Heather McNairn(1), Jiali Shang(1), Catherine Champagne(1) and Xianfeng Jiao(1)

(1) Agriculture and Agri-Food Canada, 960 Carling Ave., Ottawa, Ontario K1A OC6, Canada


As a federal government department, Agriculture and Agri-Food Canada (AAFC) is responsible for delivering agricultural programs that promote economic and environmental sustainability. Annual crop inventories, if reliably and consistently delivered, could support these programs. To meet this operational requirement, AAFC carried out a multi-year (2004 – 2007), multi-sensor (optical and radar), and multi-site (five provinces: Ontario, Saskatchewan, Alberta, Manitoba, P.E.I.) research activity to develop robust methodologies to inventory crops across Canada. Results revealed that multi-temporal (2 to 3 scenes acquired at different growth stages) optical data are ideal for crop classification. Overall accuracies of at least 85% were achieved, and most major crops (corn, wheat, soybean) were also classified to this accuracy. The timing of these acquisitions, however, was critical. Optical data acquired later in the growing season provided the best overall classification accuracy. The critical importance of late-season optical data in successful crop classification presents a number of operational challenges. Late season optical data may not be available due to clouds, jeopardizing delivery of an accurate crop inventory. The other problem is that dependency on late season imagery will not permit early season estimates of crop acreages. Radar image acquisitions late in the season can be helpful in mitigating this dependency on late season optical data. Results have demonstrated that when only one optical image is available, the addition of two ASAR images acquired in VV/VH polarization provides acceptable accuracies.

Further reducing dependency on optical imagery remains of interest. With ever-increasing access to multi-frequency, multi-polarization and polarimetric space-borne SAR, radar imagery is expected to play a greater role in delivering crop information. Since 2006, AAFC has acquired ALOS PALSAR data (quad-pol and fine-beam dual-pol) to evaluate the contribution of L-band SAR for crop identification. Combining L-band (ALOS) and C-band data (ASAR and RADARSAT-1) produced improved classification accuracies using conventional classifiers (Maximum Likelihood, decision tree). Decomposition techniques were also evaluated and integration of this information further improved classification accuracies. In addition to the results derived from the 2006 and 2007 ALOS data, this paper will also report on preliminary results from 2008 growing season, with the integration of X-, C- and L-band polarimetric data acquired from ALOS, TerraSAR-X, RADARSAT-2. Differential canopy penetrations associated with multi-frequency dataset will likely permit improved crop discrimination. Exploitation of these multi-dimensional SAR data will help bring the agricultural community closer to operational monitoring with SAR sensors, and may permit the delivery of earlier crop acreage estimates.


Symposium presentation


  Higher level                 Last modified: 07.05.06