Catherine Champagne

The Contribution of ALOS PALSAR Multipolarization and Polarimetric Data to Crop Classification

Mapping and monitoring changes in the distribution of cropland provide information that aids sustainable approaches to agriculture and supports early warning of threats to global and regional food security. This paper tested the capability of Phased Array type L-band Synthetic Aperture Radar (SAR) (PALSAR) multipolarization and polarimetric data for crop classification. L-band results were compared with those achieved with a C-band SAR data set (ASAR and RADARSAT-1), an integrated C- and L-band data set, and a multitemporal optical data set. Using all L-band linear polarizations, corn, soybeans, cereals, and hay-pasture were classified to an overall accuracy of 70%. A more temporally rich C-band data set provided an accuracy of 80%. Larger biomass crops were well classified using the PALSAR data. C-band data were needed to accurately classify low biomass crops. With a multifrequency data set, an overall accuracy of 88.7% was reached, and many individual crops were classified to accuracies better than 90%. These results were competitive with the overall accuracy achieved using three Landsat images (88.0%). L-band parameters derived from three decomposition approaches (Cloude-Pottier, Freeman-Durden, and Krogager) produced superior crop classification accuracies relative to those achieved using the linear polarizations. Using the Krogager decomposition parameters from all three PALSAR acquisitions, an overall accuracy of 77.2% was achieved. The results reported in this paper emphasize the value of polarimetric, as well as multifrequency SAR, data for crop classification. With such a diverse capability, a SAR-only approach to crop classification becomes increasingly viable.

Validation of a hyperspectral curve-fitting model for the estimation of plant water content of agricultural canopies

Abstract The estimation of plant water content is essential to the integration of remote sensing into precision agriculture. Hyperspectral models developed to estimate plant water content have had limited application under field conditions and have not been rigorously validated. A physical model using a spectrum matching technique was applied to hyperspectral data to directly calculate the canopy equivalent water thickness (EWT) using a look-up table approach. The objective of this study was to test the validity of this algorithm using plant water content information collected under field conditions and to relate this to the needs of precision agriculture. Image data were acquired over two experimental test sites in Canada, near Clinton, Ontario and Indian Head, Saskatchewan, using the Probe-1 airborne hyperspectral sensor. Plant biomass samples were collected simultaneously from plots spanning fourteen fields of various crop types (wheat, canola, corn, beans, and peas). The model was validated against EWT estimated from biomass samples, as well as more conventional measures of crop water status. The model accurately predicts water content in the range found with all crop types pooled together, with an index of agreement ( D ) of 0.92 and a root mean squared error (RMSE) of 26.8% of the average. On an individual crop basis, the model proved to be a poor predictor for wheat crops (RMSE=69.9%). When wheat fields were removed from the overall analysis, the RMSE was 17.9% and the D was 0.87. While the model provided a reasonably accurate prediction of EWT for broadleaf crops like beans, corn, canola, and peas ( D =0.88, 0.69, 0.88, and 0.84, respectively), the error margin in the prediction was too large for to precisely detect within-crop variation for the low variability found in corn and bean crops in this study. EWT is related to plant biomass and leaf area index (LAI), both quantities of interest to precision agriculture.

Evaluation of soil moisture derived from passive microwave remote sensing over agricultural sites in Canada using ground-based soil moisture monitoring networks

Passive microwave soil moisture datasets can be used as an input to provide an integrated assessment of climate variability as it relates to agricultural production. The objective of this research was to examine three passive microwave derived soil moisture datasets over multiple growing seasons in contrasting Canadian agricultural environments. Absolute and relative soil moisture was evaluated from two globally available datasets from the Advanced Microwave Scanning Radiometer for EOS (AMSR-E) sensor using different retrieval algorithms, as well as relative soil wetness at a weekly scale from the Special Sensor Microwave/Imager (SSM/I) sensor. At a daily scale, the Land Parameter Retrieval Model (LPRM) provides a better estimate of surface soil moisture conditions than the National Snow and Ice Data Center (NSIDC) dataset, with root mean squared errors ranging from 5 to 10% for LPRM and 12 to 18% for NSIDC soil moisture when a temporal smoothing is applied to the dataset. Both datasets provided better estimates of soil moisture over the temperate site near Elora, Ontario than the prairie site near Davidson, Saskatchewan. The LPRM dataset tends to overestimate soil moisture conditions at both sites, where the NSIDC dataset tends to underestimate absolute soil moisture. These differences in retrieval methods were independent of radiometric frequency used. At weekly scales, the LPRM dataset provides a better relative estimate of wetness conditions when compared to the NSIDC and the Basist Wetness Index (BWI) from SSM/I data, but the SSM/I dataset did provide a reasonably good relative indicator of moisture conditions. The high variability in accuracy of soil moisture estimation related to retrieval algorithms indicates that consistency is needed in these datasets if they are to be integrated in long term studies for yield estimation or data assimilation.

The sensitivity of RADARSAT-2 polarimetric SAR data to corn and soybean leaf area index

In this study, quadrature-polarization (quad-pol) RADARSAT-2 data at steep (25o) and shallow (40o) incidence angles were acquired during the 2008 season, imaging 13 corn and soybean fields. The leaf area index (LAI) was derived from optical imagery, and volumetric soil moisture was measured coincident with each overpass. Many synthetic aperture radar (SAR) parameters were significantly correlated with derived corn and soybean LAI. The highest correlations were observed for parameters sensitive to volume scattering (HV, LL, and RR backscatter, pedestal height, and the Freeman–Durden volume-scattering parameter) at the steeper angle. For corn, the minimum correlation coefficient was 0.95. For soybeans, the coefficients were between 0.83 and 0.86. Sensitivity to LAI was lost late in the season, when the derived LAI exceeded 3.0 m2m−2. The derived LAI and the measured soil moisture were used to model several radar parameters (HV backscatter, pedestal height, and the Freeman–Durden volume-scattering parameter) using the water-cloud model. Early in the season, the SAR response was primarily affected by the vegetation, but soil moisture was also an important contributor. When the derived LAI exceeded 1, soil-moisture contributions became minimal. The water-cloud model adequately simulated SAR responses as the canopy developed and LAI increased, demonstrating the potential of polarimetric SAR data for monitoring indicators of crop productivity.

Preprocessing of EO-1 Hyperion data

A procedure for processing hyperspectral data acquired with Hyperion has been developed with an aim to correct for sensor artifacts and atmospheric and geometric effects. Advances in preprocessing of hyperspectral remote sensing data have enabled more accurate atmospheric correction and have led to the development of new information extraction techniques in the areas of agriculture, forestry, geosciences, and environmental monitoring. These processing and analysis tools have been incorporated into Imaging Spectrometer Data Analysis Systems (ISDAS), a software package developed at the Canada Centre for Remote Sensing (CCRS). The procedure, as applied for Hyperion data, begins with geometric corrections to the short-wave infrared (SWIR) component to register the SWIR and visible near-infrared (VNIR) data spatially. This is followed by the removal of stripes and pixel (column) dropouts and noise reduction, using recently developed automated software tools. The data cube is subsequently analyzed using keystone and spectral smile detection software to characterize these distortions. Included in the smile detection procedure is an optional gain and offset correction technique. The radiance data are converted to reflectance using a MODTRAN-based atmospheric correction procedure. Only at this point are the data corrected for smile effects. Any artifacts still remaining after these corrections are removed by post-processing.

TerraSAR-X and RADARSAT-2 for crop classification and acreage estimation

This research outlines a preliminary assessment of the use of TerraSAR-X data for classifying agricultural crop land in Canada. X-Band data were able to identify crops (pasture-forage, soybeans, corn and wheat) to accuracies of 95% once a post-classification filter was applied. These accuracies were achieved using six TerraSAR-X images from 2008 and a decision-tree classification algorithm. Acquisitions began only mid-season and consequently a second full season TerraSAR-X data set is being collected in 2009. C-Band classification accuracies were about 10% lower in comparison. These results clearly demonstrate the potential of X-Band data for crop identification.

Monitoring Agricultural Risk in Canada Using L-Band Passive Microwave Soil Moisture from SMOS

AbstractSoil moisture from Soil Moisture Ocean Salinity (SMOS) passive microwave satellite data was assessed as an information source for identifying regions experiencing climate-related agricultural risk for a period from 2010 to 2013. Both absolute soil moisture and soil moisture anomalies compared to a 4-yr SMOS satellite baseline were used in the assessment. The 4-yr operational period of SMOS was wetter than the 30-yr climate normal in many locations, particularly in the late summer for most regions and in the spring for the province of Manitoba. This leads to a somewhat unrepresentative baseline that skews anomaly measures at different parts of the growing season. SMOS soil moisture does, however, show a clear trend where extremes are present, with drier-than-average conditions during periods that drought and dry soil risks were identified and wetter-than-average conditions when flooding and excess moisture were present. Areas where extreme weather events caused crop losses were identifiable using S...

A bootstrap method for assessing classification accuracy and confidence for agricultural land use mapping in Canada

Abstract Land cover and land use classifications from remote sensing are increasingly becoming institutionalized framework data sets for monitoring environmental change. As such, the need for robust statements of classification accuracy is critical. This paper describes a method to estimate confidence in classification model accuracy using a bootstrap approach. Using this method, it was found that classification accuracy and confidence, while closely related, can be used in complementary ways to provide additional information on map accuracy and define groups of classes and to inform the future reference sampling strategies. Overall classification accuracy increases with an increase in the number of fields surveyed, where the width of classification confidence bounds decreases. Individual class accuracies and confidence were non-linearly related to the number of fields surveyed. Results indicate that some classes can be estimated accurately and confidently with fewer numbers of samples, whereas others require larger reference data sets to achieve satisfactory results. This approach is an improvement over other approaches for estimating class accuracy and confidence as it uses repetitive sampling to produce a more realistic estimate of the range in classification accuracy and confidence that can be obtained with different reference data inputs.

Contribution of Multi-Frequency, Multi-Sensor, and Multi-Temporal Radar Data to Operational Annual Crop Mapping

Information on agricultural land use (crop inventory) is needed by various organizations on an annual basis. To meet this operational requirement, Agriculture and Agri-Food Canada (AAFC) has carried out a multi-year (2004 - 2007), multi-sensor (Landsat TM, SPOT, RADARSAT-1, ASAR), and multi-site (five provinces: Ontario, Saskatchewan, Alberta, Manitoba, P.E.I.) research activity to develop a robust methodology to inventory crops across Canadas large and diverse agricultural landscapes. Results clearly demonstrated that multi-temporal satellite data can successfully classify crops for a variety of cropping systems across Canada. Overall accuracies of at least 85% were achieved. When available, multi-temporal (2 to 3 scenes acquired at different growth stages) optical data are ideal for crop classification. However due to cloud and haze interference, good optical data are not always obtainable. A SAR-optical combination offers a good alternative. This research has found that when only one optical image is available, the addition of two ASAR images acquired in VV/VH polarization will provide acceptable accuracies. Of particular interest is the observation that with the incorporation of radar, crop inventories can be delivered earlier in the growing season.

Spatial Variability Mapping of Crop Residue Using Hyperion (EO-1) Hyperspectral Data

Soil management practices that maintain crop residue cover and reduce tillage improve soil structure, increase organic matter content in the soil, positively influence water infiltration, evaporation and soil temperature, and play an important role in fixing CO2 in the soil. Consequently, good residue management practices on agricultural land have many positive impacts on soil quality, crop production quality and decrease the rate of soil erosion. Several studies have been undertaken to develop and test methods to derive information on crop residue cover and soil tillage using empirical and semi-empirical methods in combination with remote sensing data. However, these methods are generally not sufficiently rigorous and accurate for characterizing the spatial variability of crop residue cover in agricultural fields. The goal of this research is to investigate the potential of hyperspectral Hyperion (Earth Observing-1, EO-1) data and constrained linear spectral mixture analysis (CLSMA) for percent crop residue cover estimation and mapping. Hyperion data were acquired together with ground-reference measurements for validation purposes at the beginning of the agricultural season (prior to spring crop planting) in Saskatchewan (Canada). At this time, only bare soil and crop residue were present with no crop cover development. In order to extract the crop residue fraction, the images were preprocessed, and then unmixed considering the entire spectral range (427 nm–2355 nm) and the pure spectra (endmember). The results showed that the correlation between ground-reference measurements and extracted fractions from the Hyperion data using CLMSA showed that the model was overall a very good predictor for crop residue percent cover (index of agreement (D) of 0.94, coefficient of determination (R2) of 0.73 and root mean square error (RMSE) of 8.7%) and soil percent cover (D of 0.91, R2 of 0.68 and RMSE of 10.3%). This performance of Hyperion is mainly due to the spectral band characteristics, especially the availability of contiguous narrow bands in the short-wave infrared (SWIR) region, which is sensitive to the residue (lignin and cellulose absorption features).