Dan G. Blumberg

Classification using ASTER data and SVM algorithms;: The case study of Beer Sheva, Israel

Abstract New sensors and new technologies for remotely sensed data acquisition have proven useful for mapping urban environments. In this paper, a new dataset from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) spaceborne sensor was used with support vector machine (SVM)-based algorithms for classification processing. A case study of Beer-Sheva, Israel demonstrates that ASTER data are suitable for urban studies. The classification results also show that the approach based on SVM has high performance in convergence, speed, and accuracy of training and classifying. Field validation shows that the classification is reliable for urban studies with high classification precision and little confusion (88.6% average overall accuracy and 9.6% average omission error for 15-m resolution image classification vs. 89.9% average overall accuracy and 11.2% average omission error for 30-m resolution image classification). For the 30-m multispectral data, there are only five classes vs. six for the 15-m visible spectrum data. Despite the additional classes retrieved from the visible data, the short-wave infrared (SWIR) data provides the accuracy in differentiating some of the confused classes.

Elongation and migration of sand dunes

Two distinct processes are known to act on dynamic dunes, the process of migration by erosion on the windward side and deposition on the lee side, typical for transverse dunes, and the process of elongation typical for linear dunes. These two processes are determined by wind direction relative to the dune alignment. This article reviews the assertion that linear dunes experience lateral displacement in addition to elongation. Fieldwork on vegetated linear dunes (VLDs) and GIS work on seif dunes indicates no lateral migration for these dunes. Linear dunes can shift laterally only when a slip face, formed on the lee side, reaches the plinth of the dune. The winds from both sides of the seif dune are never symmetric; usually winds from one direction are more dominant and effective. The outcome is the formation of peaks and saddles along the dune. The strongest winds create a slip face on the lee side of the peak segments of the dune, oblique to the dune alignment, which reaches the base of the dune and displaces the peak downwind along the dune alignment. The internal structure of the seif dune is formed mostly by this dominant wind direction and gives the impression that the dune has shifted laterally. On the other hand, there are cases in which the wind directions relative to dune alignment fall between those of transverse and seif dunes. In such cases, both processes act on the dune, which subsequently experiences migration as well as elongation.

Applications of spaceborne radar laboratory data to the study of aeolian processes

Aerodynamic roughness (z0) is an important parameter in studies of sand and dust transport, as well as atmospheric circulation models. Aerodynamic roughness is a function of the size and spacing of surface roughness elements and is typically determined at point locations in the field from wind velocity profiles. Because field measurements require complex logistics, z0 values have been obtained for very few localities. If radar can be used to map z0, estimates can be obtained for large areas. In addition, because aerodynamic roughness can change in response to surface processes (e.g., flooding of alluvial surfaces), radar remote sensing could obtain new measurements on short timescales. Both z0 and the radar backscatter coefficient σ0 are dependent on topographic roughness at the submeter scale, and correlation between these two parameters was developed based on radar data obtained from aircraft (AIRSAR). The Spaceborne Radar Laboratory (SRL) afforded the opportunity to test the correlation for data obtained from orbit. SRL data for sites in Death Valley, California; Lunar Lake, Nevada; and Gobabeb, Namibia, were correlated with wind data and compared with previous radar z0 relations. Correlations between σ0 and z0 for L band (λ=24 cm) HV (H, vertically and V, vertically polarized modes) L band HH, and C band (λ=5.6 cm) HV compare favorably with previous studies. Based on these results, maps of z0 values were derived from SRL data for each site, demonstrating the potential to map z0 for large vegetation-free areas from orbit using radar systems.

Remote sensing of desert dune forms by polarimetric Synthetic Aperture Radar (SAR)

Abstract This report describes remote sensing experiments using highly capable polarimetric synthetic aperture radar (SAR). In preparation for the launch of the SIR-C/X-SAR mission, NASA’s JPL airborne polarimetric SAR (AIRSAR) was deployed over many locations worldwide. The report describes SAR observations of desert dune fields in North America, Bolivia, and Australia acquired by AIRSAR and in Namibia acquired by SIR-C/X-SAR. The use of polarimetric and multiwavelength SAR is reviewed and comments on the advantages of the various combinations are noted. SAR data are very useful in mapping dunes because of the ability to control illumination parameters such as look angle, wavelength, and polarization. Furthermore, the unique responses of SAR to surface roughness are found very useful in mapping dune forms. For most sites, the longer wavelengths, P- and L-band at λ=68 cm and 24 cm, are advantageous over the shorter (C-band at λ=5.6 cm) providing better contrast. Overall, copolarized channels (HH and VV) are found best for mapping the dunes and dune types. When trying to discriminate active from inactive dunes cross-polarized channels provide essential information on vegetation.

Assessing the Martian surface distribution of aeolian sand using a Mars general circulation model

A sand transport model using Whites [1979] sand flux equation and the Mars general circulation model [Pollack et al., 1990] was developed to understand the erosional sources, transport pathways, and depositional sinks of windblown sand on Mars. An initially uniform distribution of sand (4 mm over the entire surface) is regionally transported based on wind stress, saltation threshold, and percentage of topographic trapping. Results are consistent with the observed polar and Hellespontus dunes and Christensens [1986] modeled block size distribution, but only for an extremely low saltation threshold (0.024 N/m 2 ). Low thresholds generally result in transport of sand-sized particles originating in the northern mid latitudes to the north pole, and transport from the northern lower latitudes to the southern hemisphere. Our results indicate that the polar dune fields could form in 50,000 years, consistent with the active polar dunes and lack of longitudinal dunes observed on the surface of Mars.

Soil moisture (water-content) assessment by an airborne scatterometer: the Chernobyl disaster area and the Negev desert.

Abstract Soil water content is an important component that influences meso- and microscale processes. The agricultural capacity of a site is directly affected by soil water content and it is especially important in arid regions. In temperate and humid regions, soil water content is important in determining flood risks. Environmentally, soil water content will influence the risk of carrying pollutants through the soil. For these reasons, a scatterometer was developed as a remote sensor for mapping soil water content. The scatterometer is frequency modulated using a continuous wave. This scatterometer operates at nadir with a wide antenna diagram of 10°. Measurements were conducted in two environments with different implications. The first was in the Chernobyl nuclear disaster area and the second in the Negev desert. Results show a good correlation between soil water content and the amplitude of the returned signal measured by the scatterometer. Thus, the scatterometer provides an efficient tool for mapping soil water content. The long wavelength (P-band) of λ=68 cm makes this scatterometer more sensitive to soil water content and less affected by surface roughness than scatterometers operating at shorter wavelengths.

Spectral properties and hydraulic conductance of soil crusts formed by raindrop impact

In this study, a controlled spectral investigation of the properties of structural crust, using infiltration rate as an indicator, was carried out, on a grumusol (typic chromoxerets) soil. The scope of this paper is to investigate the spectral relationship between structural crust and the infiltration rate and some related properties of a grumusol soil. Structural crust is a thin compact layer generated on the soils surface after rainstorm events due to raindrop energy. The crust, which is a result of physical segregation and rearrangement of soil particles, affects some of the soils physical properties, such as infiltration, run-off and soil erosion. A set of soil samples was subjected in the laboratory to increased levels of cumulative energy from a simulated rainstorm, while measuring the infiltration rate. The result was a set of soil samples with increased levels of crusting and correspondingly reduced infiltration rates. After drying the soil samples, their spectral parameters across the SWIR region (1.2 w m-2.4 w m) were studied in the laboratory. Empirical relations were found between the infiltration rate and reflectance values (either in raw or in first derivative domains). A spectral ratio manipulation, using a fog type rain treatment (characterized as having no rain energy) as reference, confirmed that, within the spectrum, changes in both albedo and absorption enabled the crust to be detected using reflectance radiation. This is basically because fine (mostly clay) mineral enrichment processes occurred at the surface of the crusted soil. These minerals (montmorillonite, calcite) are spectrally active across the selected spectral region. Although further study is required to fully account for other soils, rainstorm energies and water qualities, this technique suggests great potential as a tool for rapid, non-destructive, in situ assessment of crusted soil properties.

Subsurface microwave remote sensing of soil-water content: field studies in the Negev Desert and optical modelling

The population increase in the Middle East and the respective decrease of water resources necessitate innovative methods for utilization and monitoring of water resources. Development of remote sensing tools can pave the way for remote, rapid mapping of soil-water content, control of excessive irrigation, and prevention of water waste. This paper describes a series of experiments conducted in the Negev Desert that were aimed at developing such tools for monitoring soil-water content. The use of visible near infrared and microwave techniques seems suitable. All provide good correlation with soil-water content measured on the ground. However, the microwave techniques presented here using a P-band scatterometer and ERS-2 SAR seem the most promising. Finally the possibility of optical simulation of the microwave processes is presented in an effort to improve the physical basis for empirical studies. A method of fabrication of optical samples that model soils with different water content and different surface roughness is developed, and a system for measuring backscattered signals is designed. It is shown that the reflectivity of a layered medium is a non-monotonic function of the water content. The effect of the surface roughness on the reflection from a strong buried reflector is being studied.

Potential Transport of Windblown Sand: Influence of Surface Roughness and Assessment with Radar Data

The transport of windblown sand is controlled by many factors, including wind regime and sediment supply. Surface roughness at the sub-meter scale is also important because it influences both the threshold conditions for particle entrainment and the flux of sand once it is set into motion. In general, increases in surface roughness result in higher threshold speeds for particle movement and decreases in sand fluxes. Aerodynamic roughness (z 0) is the aeolian parameter related to surface roughness and is defined as the height above some mean level at which average wind speed is zero. Values of z 0 are derived from wind measurements through the boundary layer, but few z 0 values have been obtained over natural surfaces because of the expense and limitations of making such measurements. Rather, remote sensing using radar systems has the potential for addressing this problem. In this investigation, we derived z 0 values for a wide variety of surfaces in the southwestern United States and obtained radar data for these sites in P-band (wavelength = 68 cm), L-band (wavelength = 24 cm) and C-band (wavelength = 5.6 cm). We show that there are good correlations among z 0, the RMS height of the surface, and the radar backscatter coefficient, σ0, with the best correlation for L-band HV polarized radar data. This study shows the potential for mapping large regions with radar in order to derive aerodynamic roughness values, which in turn can be used in predictive models of sand transport.

Evaluating subpixel target detection algorithms in hyperspectral imagery

Our goal in this work is to demonstrate that detectors behave differently for different images and targets and to propose a novel approach to proper detector selection. To choose the algorithm, we analyze image statistics, the target signature, and the targets physical size, but we do not need any type of ground truth. We demonstrate our ability to evaluate detectors and find the best settings for their free parameters by comparing our results using the following stochastic algorithms for target detection: the constrained energy minimization (CEM), generalized likelihood ratio test (GLRT), and adaptive coherence estimator (ACE) algorithms. We test our concepts by using the dataset and scoring methodology of the Rochester Institute of Technology (RIT) Target Detection Blind Test project. The results show that our concept correctly ranks algorithms for the particular images and targets including in the RIT dataset.