Science Service System
You are here : Home : Proposals_Summary

Summary of Proposal LAN2194

TitleDetermining Disease Risk for Canola Using TerraSAR-X and RADARSAT-2
Investigator McNairn, Heather - Canadian Federal Government, Agriculture and Agri-Food Canada
Team Member
Manager Powers, Jarrett - Government of Canada, Agriculture and Agri-Food Canada
Physical Scientist Merzouki, Amine - Government of Canada, Agriculture and Agri-Food Canada
Research Scientist Shang, Jiali - Government of Canada, Agriculture and Agri-Food Canada
Physical Scientist Pacheco, Anna - Government of Canada, Agriculture and Agri-Food Canada
SummaryCanola crops are susceptible to a range of diseases which can significantly impact production. Given the acreages of canola planted in Canada, early assessment of risk is paramount to assist with optimal application of fungicides and insecticides. Risk is linked closely with farming practices, meteorological conditions, soil moisture and crop phenology. This proposal will acquire temporally dense repeat passes of TerraSAR-X to determine whether this sensor is capable of detecting the onset of flowering and seed development for canola. These are stages critical for the timing of chemical applications to minimize yield losses. Preliminary results from a limited data set collected during the SMAPVEX12 experiment demonstrated that the X-Band VV/VH ratio fell when canola began to flower, and dropped again with the onset of podding. The project team proposes to collect a larger data set for the 2014 growing season to validate these results and test how robust the change in this ratio is in terms of indicatingthe onset of these growth stages. Twenty-two VV-VH stripmap images are being requested beginning at canola seeding and concluding at harvest. The team will report on the statistical sensitivity of the change in this ratio, will produce sample map products, plans to communicate these results to the agriculture community, and will publish results in the scientific literature. Agriculture and Agri-Food Canada has a long term research site at SMAPVEX and thus funding is in place from an internal project ($151K CDN/year) to support remote sensing activities at this site.
Final ReportReference: Anna Pacheco, Heather McNairn, Yifeng Li, George Lampropoulos and Jarrett Powers. “Using RADARSAT-2 and TerraSAR-X Satellite Data for the Identificaiton of Canola Crop Phenology”, SPIE Remote Sensing 2016, 26-20 September 2016, Edinburgh, Scotland. Introduction: Crop phenology is an important variable in crop monitoring and management. Combined with other factors such as persistent wet conditions, some crops are particularly vulnerable to the development of crop disease at specific crop development stages. For example, canola is most vulnerable to infestation of sclerotinia when petals emerge. Given variances in planting dates and growing conditions, growth stage must be determined at the field level. Monitoring favorable conditions to the development of crop disease, including crop growth stage, can be challenging given the temporally varying conditions. In Canada, the challenge in crop disease risk assessment is exacerbated by the requirement to monitor over large spatial extents. The objective of this study is to further determine which frequency, X- or C-band, is optimal to identify canola flowering, and which polarimetric parameters could potentially be integrated in an approach to assess crop disease risk related to the flowering of canola. Study Site and Data: Image and ground data were collected by Agriculture and Agri-Food Canada (AAFC) in the agriculture area of Carman and Elk Creek in the province of Manitoba. Annual crops that are typically grown in this area include cereals, canola, corn, soybeans and edible beans. In total, 21 canola fields were identified and surveyed throughout the 2014 growing season. The BBCH-scale was used to identify the phenological development stages of these canola fields. Inflorescence emergence will typically begin 50 days following the seeding date of canola and flowering will generally occur 10 days after that. In 2014, the appearance of flower buds was noted on the majority of the fields by June 26 (stage 50 to 59 on the BBCH-scale) and by July 4, all but two fields were within the flowering growth stage (stage 60 to 69 on the BBCH-scale). Ground and image data were collected on several dates. Seven C-band RADARSAT-2 (5.405 GHz) and thirteen X-band TerraSAR-X (9.6 GHz) images were programmed throughout the growing season. RADARSAT-2 data were acquired in wide fine quad polarization mode whereas the TerraSAR-X data were acquired in StripMap dual-polarization mode using the VH and VV channel combination. For RADARSAT-2, acquisitions were collected in beam mode FQ8W and FQ15W. The TerraSAR-X data were collected in StripFar-004 and StripNear-013 with incidence angles ranging from 23.701 to 25.404 and from 41.735 to 42.933, respectively. Methods: Wide fine quad RADARSAT-2 images were acquired in Single Look Complex (SLC) format at a nominal spatial resolution of 8-m. TerraSAR-X data were acquired in Single Look Slant Range Complex (SSC) format at a spatial resolution of 6.6-m. SLC and SSC data were first calibrated to sigma naught and a speckle reduction filter was applied on the data. Specifically, a polarimetric Gaussian-BoxCar speckle filter was used on a 5-by-5 pixel sliding window. Various polarimetric parameters were then calculated from the data. Each of these polarimetric image products was ortho-rectified and various statistics (mean and standard deviation) were extracted for each canola field surveyed within this study. An inverted buffer of ~30-m was applied on all field polygons during data extraction to avoid any signal contamination from neighboring fields. Various polarimetric parameters were computed from the RADARSAT-2 and TerraSAR-X imagery. The following polarimetric parameters were computed for both the C- and X-band data: entropy, anisotropy, and alpha angle. Polarimetric variables including the conformity coefficient, the scatter predominance and diversity, the degree of purity and depolarisation index were also calculated for the RADARSAT-2 data. Moreover, reflectivity ratios and differential reflectivity values were computed on the data for all possible channel combinations, yielding 3 (HH/HV; HH/VV; HV/VV) and 1 (VH/VV) image products for RADARSAT-2 and TerraSAR-X, respectively. The entropy, anisotropy, and alpha angle parameters are derived from the Cloude-Pottier decomposition. The differential reflectivity ratios are calculated using the ratio of the reflected horizontal and vertical power returns. Other polarimetric parameters were generated from the RADARSAT-2 data including the conformity coefficient, the scatter predominance and diversity, the degree of purity and the depolarisation index. Some of these parameters are closely related to each other, and to the Cloude-Pottier decomposition parameters. Results: Canola fields were surveyed in 2014 to assess their phenological stages at the time of coincident satellite RADARSAT-2 and TerraSAR-X overpasses. Cloude-Pottier decomposition parameters, i.e. entropy, anisotropy and alpha angle were calculated for both C- and X-band data. The RADARSAT-2 and TerraSAR-X entropy and alpha angle demonstrate similar sensitivity to the inflorescence emergence and flowering stages in the canola. A sharp increase can be noted on both the entropy and alpha angle as flower buds emerge. These parameters continue to increase during the canola flowering stage but appears to saturate at full flowering according to the BBCH scale, i.e. once 50% of the flowers have opened. This is an interesting observation given that effective fungicide applications in canola need to occur prior to 50% flowering. The increase in entropy is due to the change from one dominant scattering to an equal contribution from all three scattering mechanisms (surface, volume, and double bounce) whereas the increase in alpha angle reflects the transition between single-bounce and volume scattering. The X-band anisotropy appears to be more responsive than the anisotropy from the C-band SAR data where values only slightly decrease for the latter. The anisotropy responds to the relative strengths of the second and third scattering mechanisms, and these are thus more sensitive to higher SAR frequencies. Reflectivity ratios and differential reflectivity values were calculated for the various RADARSAT-2 channel combinations (HH/HV, HH/VV and HV/VV). There is a significant decrease of the HH/HV ratio and a notable increase of the HV/VV ratio when the canola flower buds appear. A similar observation can be made on the differential reflectivity where the backscatter responses are most significant with the HH/HV and HV/VV combinations during the canola inflorescence emergence and flowering stages. A significant drop in intensity can be noted with the HH/HV differential reflectivity (about 4) whereas a clear increase in intensity is observed from the HV/VV combination (about 4). This decrease and increase is related to the change in volume scattering which becomes greater from the continuation of the stem elongation and the increase in leaf area from the leaves and buds during the inflorescence emergence. A significant drop is observed in the TerraSAR-X VH/VV response at the emergence of the flower buds, which is continued throughout the flowering stage. Based on these results, it appears that any of the reflectivity ratios and differential reflectivity values that incorporate an HV response, either it be in X- or C-band, are sensitive to canola budding and flowering. Several other polarimetric parameters were generated from RADARSAT-2 including the conformity coefficient, the scatter predominance and diversity, the degree of purity and the depolarisation index. All parameters are responsive to the inflorescence emergence and the first phase of the canola flowering. Overall, the conformity coefficient, the degree of purity and the depolarisation index demonstrate the greatest potential at identifying when the flower buds begin to emerge. The conformity coefficient decreases from about 0.18 on average to about -0.02 on average indicating the presence of volume scattering derived from the emergence of flower buds and its eventual flowering. The conformity coefficient then continues to either decrease or increase slightly midway into the final stage of the flowering period. The degree of purity and the depolarization index measure somewhat the same element of the polarimetric signal. All canola fields present a similar behavior for both parameters throughout the season with the exception of two fields. Growth stages for these fields are within the average range, and the ground data did not flag any particular differences with these fields. Discussion and Conclusion: It is well understood that microwave SAR data are sensitive to changes in canopy structure. Several RADARSAT-2 and TerraSAR-X polarimetric parameters were derived in this study over 21 canola fields throughout the growing season, resulting in polarimetric temporal responses over these fields for each of the parameters. Specifically, seven fully polarimetric RADARSAT-2 images and thirteen dual-polarimetric TerraSAR-X images were acquired throughout the 2014 growing season, spanning June 17 to August 31.This analysis has helped to determine which of the polarimetric parameters have the greatest potential to be incorporated into a methodology for the assessment of risk related to the development of crop disease in canola. Both the entropy and alpha angle derived from RADARSAT-2 and TerraSAR-X were sensitive to the emergence of flower buds and the flowering in canola, demonstrating that either C- or X-band can be used for this application, hence providing greater opportunities for satellite data acquisitions which is particularly important when crop growth is so temporally dynamic. Other RADARSAT-2 polarimetric parameters such as the conformity coefficient, the degree of purity and the depolarisation index were also useful in identifying the change in canola crop phenology. These parameters are easier to calculate than the alpha angle and entropy, and thus easier to implement on large polarimetric datasets. This could be especially useful when data processes require efficiency and streamlining, which is generally the case for the implementation of any operational mapping activity. Finally, reflectivity ratios and differential reflectivity values were computed on the data for all possible channel combinations, yielding 3 (HH/HV; HH/VV; HV/VV) and 1 (VH/VV) image products for RADARSAT-2 and TerraSAR-X, respectively. All reflectivity ratios and differential reflectivity values that incorporated a cross-polarized response were sensitive to the change in crop phenology during the pre-flowering and flowering stages in canola. The calculation of these ratios would also be fairly simple to implement within a methodology for risk assessment in canola crop disease.

Back to list of proposals

© DLR 2004-2016