Joseph R. Buckley

Investigating RADARSAT-2 as a tool for monitoring grassland in western Canada

Native grasslands play an important role in ecosystem function, biodiversity, climate change, and economics, yet quantifiable estimates of the rate and location of native grassland change in western Canada are not readily available. To date, optical remote sensing has been explored for grassland mapping, but cloud cover limits the availability of timely data for the discrimination of improved, as opposed to native, grassland. In this study we investigated the utility of RADARSAT-2 polarimetric imagery to map native grassland, improved grassland, and agricultural crops. Fine quad-polarisation mode RADARSAT-2 data were acquired at two incidence angles over a test site in southern Alberta every 24 days from 1 April to 31 October 2009 and were processed using the Freeman–Durden decomposition. Double-bounce, volumetric, and surface scattering properties suggest that native grasslands can be distinguished from cultivated cropping, especially using a mid- to late-season image. However, discriminating native grasslands from improved grasslands was more difficult. Land cover classification of a single RADARSAT-2 image from July 2009 provided reasonable but slightly less accurate results compared with a single Landsat-5 Thematic Mapper image (Kappa value of 0.65 compared with 0.81).

Assessment of polarimetric SAR data for discrimination between wet versus dry soil moisture conditions

Radarsat-2 imagery from extreme dry versus wet conditions are compared in an effort to determine the value of using polarimetric synthetic aperture radar (SAR) data for improving estimation of fuel moisture in a chronosequence of Alaskan boreal black spruce ecosystems (recent burns, regenerating forests dominated by shrubs, open canopied forests, moderately dense forest cover). Results show strong distinction between wet and dry conditions for C-HH and C-LR polarized backscatter, and Freeman–Durden and van Zyl surface bounce decomposition parameters (35–65% change for all but the dense spruce site). These four SAR variables have high potential for evaluation of within site surface soil moisture, as well as for relative distinction between wet and dry conditions across sites for lower biomass and sparse canopy forested sites. However, for any given test site except the shrubby regrowth site, van Zyl volume, surface, and double bounce scattering all result in similar percentage increases from dry to wet soil condition. This indicates that for most of these test sites/cases moisture enhances the magnitude of the return for all scattering mechanisms evaluated. Thus, differences in scattering from the interaction of biomass, surface roughness, and moisture condition across sites remains an issue and backscatter due to surface roughness or biomass cannot be uniquely estimated. In contrast, the Cloude–Pottier C-band decomposition variables appear invariant to soil moisture, but may instead account for variations in ecosystem structure and biomass. Further investigation is needed, as results warrant future research focused on evaluation of multiple polarimetric parameters in algorithm development.

Enhanced classification of prairie landscapes using simulated RADARSAT-2 imagery

For the past decade, we have been using Landsat TM imagery to monitor changes in the area and distribution of ecosystems in a relatively undisturbed natural region in southwestern Manitoba, Canada. The 40 000 ha of Canadian Forces Base Shilo are protected from urbanization, commercialization, and agriculture, and are disturbed only by modest and controlled amounts of military training. A significant problem with the use of TM imagery for this study is that some of the more sensitive indicators of environmental change, such as the infestation of disturbed grassland by leafy spurge (Euphorbia esula), are very difficult to identify and delineate in the visible and near-IR bands. A fully polarized synthetic aperture radar that is sensitive to vegetation morphology may not have the same difficulty. C-band radars, such as that being constructed for RADARSAT-2, are known to have difficulty differentiating vegetation types in regions of dense foliage, but, in a prairie ecosystem such as the one under study here, even the densest foliage should be sparse enough to yield a unique polarization signature. To test this hypothesis, we have simulated RADARSAT-2 HH, HV, and VV images, using TM images of the area, polarization ratios extracted from the literature for vegetation types similar to those found in the area, and a RADARSAT-1 HH image of the area collected almost simultaneously with one of the TM images. Estimation of class separation by Bhattacharyya distance, first using only HH information and then using the full HH, HV, VV imagery, shows an increase of statistically separable classes and class groups when the fully polarized class characterization is used. Visual comparison of synthesized imagery from different years shows sufficient differences to imply that RADARSAT-2 imagery will likely be a useful tool to augment our continuing monitoring and assessment of this prairie ecosystem.

Monitoring grasslands with radarsat 2 quad-pol imagery

Radarsat 2 quad polarization imagery has been used to study the effectiveness of polarimetric radar to monitor the extent and health of prairie grasslands in southern Alberta, Canada. In this report of preliminary findings, the imagery is shown to be effective in the separation of cropped lands from rangelands, and in the separation of native grasslands and improved pastures. Classification was more accurate using Freeman-Durden decomposition parameters than using Cloude-Pottier parameters. Incidence angle differeces were noted and use of multiple angles in classification improved accuracy. In a second part of the study, it was shown that polarimetric imagery was capable of identifying weeds and brush growing in native rangland, and in separating different kinds of brush and weeds. Validated sample sets were too small to allow proper accuracy assessment, but a ‘performance metric’ showed that accuracy would be improved by use of multiple incidence angles, and by the use of Freeman-Durden or coherency matrix parameters.

Environmental change detection in prairie landscapes with simulated Radarsat 2 imagery

For the past decade, we have been using Landsat TM imagery to monitor changes in the area and distribution of ecosystems in a relatively undisturbed natural region in southwestern Manitoba, Canada. The 40,000 hectares of Canadian Forces Base Shilo are protected from urbanisation, commercialisation and agriculture, and are disturbed only by modest and controlled amounts of military training. A significant problem with the use of TM imagery for this study is that some of the more sensitive indicators of environmental change, such as the infestation of disturbed grassland by leafy spurge, are very difficult to identify and delineate in the visible and near-IR bands. A fully polarized SAR that is sensitive to vegetation morphology may not have the same difficulty. C band radars, such as that being constructed for Radarsat 2, are known to have difficulty differentiating vegetation types in regions of dense foliage, but, in a prairie ecosystem such as the one under study here, even the densest foliage should be sparse enough to yield a unique polarisation signature. In order to test this hypothesis we have simulated Radarsat 2 HH, HV and VV images, using several TM images of the area, polarisation ratios extracted from the literature for vegetation types similar to those found in the area, and a Radarsat 1 HH image of the area collected almost simultaneously with one of the TM images. Comparison of synthesized imagery from different years shows sufficient differences to imply that Radarsat 2 imagery will likely be a useful tool for continuing our prairie ecosystem monitoring and assessment.

Can geometric optics fully describe radar images of the sea surface at grazing incidence

It is possible to estimate the significant wave height of the ocean surface using shadowing statistics computed from grazing incidence radar imagery. The theoretical basis for this estimation is a geometric optics view of a gaussian surface. In practice it is necessary to apply an empirical correction based on the intensity of the radar backscatter in order to produce a wide range of verifiable wave heights. To explore the reasons for this correction, we have conducted a one-dimensional Monte-Carlo simulation of the sea surface at a wide range of significant wave heights, and have observed it with a ray trace program from several source heights and configurations. We conclude that, although the finite amplitude effects simulated in this experiment partially explain the results observed at sea, some electromagnetic interaction unaccounted for in geometric optics must also be responsible for the result. We suggest that increased local incidence angle is a possible source of the difference between observations and our simulation.

Use of an adapted marine radar for the short-range detection and tracking of small birds in flight

As part of a project to develop an inexpensive songbird tracking and identification system, we have modified the antenna system of a standard marine radar to estimate the three dimensional position of point targets. Although the radar formally gives only two coordinate parameters, we show how the Incorporation of other information about the targets allows estimation of the full three dimensional position. We present two independent methods for this estimation. We have used this system in several field experiments at Prince Edward Point, Picton County, Ontario, Canada, where we have tracked various targets including a C-130 Hercules aircraft (RCS = 11 m/sup 2/) and 10-40g birds (RCS = 7-22 cm/sup 2/). Target locations were verified using an array of acoustic receivers to provide an independent measure of location.

Use of Radarsat-2 polarimetric SAR images for fuel moisture mapping in the Kruger National Park, South Africa

Fully polarimetric Radarsat-2 imagery from wet and dry conditions over the South African Lowveld is compared to assess its value for fuel moisture mapping. Imagery was acquired at two different dates, in May (end of summer, wet) and in August (mid of winter, dry). Sample plots were classified into two broad Lowveld site types (herbaceous-dominated and shrub and tree-dominated). Linear and circular polarized backscatters, polarimetric discriminators and polarimetric decomposition parameters were computed to find suitable parameters for fuel moisture estimation. The results show a significant distinction between wet and dry conditions for C-HH, C-HV, C-RR, and C-LL, all Freeman-Durden and van Zyl decomposition parameters and some polarimetric discriminators (dmin, Prmax, Prmin, Smax, Smin). In almost all cases the normalized difference between wet and dry condition is lower for the shrub and tree-dominated sites. The Freeman-Durden double bounce scattering decomposition parameter performs best in both site types.

Does modis sea surface temperature accurately represent the temperature of the dynamically significant surface layer of the ocean

This paper describes a comparison between ocean skin temperature as estimated by MODIS 11μm radiation in both day and night overpasses, and near-surface measurements made by Argo buoys co-located in time and space with MODIS pixels located in two 20° × 20° regions in the eastern Pacific Ocean, for the period 2003-2010. On average, Argo temperatures were warmer than MODIS temperatures, by about 0.3°C on average in the daytime, and 0.8°C at night. There was significant month-to-month variation in these relationships, but little interannual variability. In general, the measurements are not directly interchangeable within established limits of uncertainty for SST measurements, but are when the mean monthly differences are taken into account. The mean differences themselves provide useful information on the relationship between ocean skin temperature and that a few metres below the surface.

Comparing RADARSAT 2 and Terrasar-X quad-pol SAR imagery of grasslands

Quad-pol imagery from both Terrasar-X and RADARSAT-2 were acquired over the grasslands of southern Alberta, Canada, in the early spring of 2010. Dates and incidence angles were as closely matched as was possible. Both sets of imagery showed qualitative agreement in spatial structure, but the Terrasar-X imagery was significantly less well defined with a much higher noise floor. The range of pixel values due to speckle noise, even after modest filtering, was large enough to mask any significant relationship between the two sources of imagery on a pixel-by-pixel basis.