Search

The interactivism model introduces a dynamic approach to language, communication and cognition. In this work, we explore this fundamental theory in the context of dialogue modelling for spoken dialogue systems (SDS). To extend such a theoretical framework, we present a set of design principles which adhere to central psycholinguistic and communication theories to achieve interactivism in SDS. From these, key ideas are linked to constitute the basis of our proposed design principles.

The Seawinds scatterometer on the advanced Earth observing satellite-2 (ADEOS-2) will determine surface wind vectors by measuring the radar cross section. Multiple measurements will be made at different points in a wind-vector cell. When dense clouds and rain are present, the signal will be attenuated, thereby giving erroneous results for the wind. This report describes algorithms to use with the advanced mechanically scanned radiometer (AMSR) scanning radiometer on ADEOS-2 to correct for the attenuation. One can determine attenuation from a radiometer measurement based on the excess brightness temperature measured. This is the difference between the total measured brightness temperature and the contribution from surface emission. A major problem that the algorithm must address is determining the surface contribution. Two basic approaches were developed for this, one using the scattering coefficient measured along with the brightness temperature, and the other using the brightness temperature alone. For both methods, best results will occur if the wind from the preceding wind-vector cell can be used as an input to the algorithm. In the method based on the scattering coefficient, we need the wind direction from the preceding cell. In the method using brightness temperature alone, we need the wind speed from the preceding cell. If neither is available, the algorithm can work, but the corrections will be less accurate. Both correction methods require iterative solutions. Simulations show that the algorithms make significant improvements in the measured scattering coefficient and thus is the retrieved wind vector. For stratiform rains, the errors without correction can be quite large, so the correction makes a major improvement. For systems of separated convective cells, the initial error is smaller and the correction, although about the same percentage, has a smaller effect.

Teh SIR-C/X-sar radar observations of rain storms are the first multi-polarization and multi-frequency observations of percipitation from space. In addition to numerous, often dramatic images of severe weather systems obtained by forming a synthetic aperture in the usual side-looking attitude, several data takes were performed while the radar antennas were parallel to the ground and the radar beams were pointing at nadir. These oppoprtunities coincided with the passage of the Shuttle over Tropical Cyclone Odille in the southern Indian Ocean during the first flight, and over Typhoon Seth in the Western Pacific during the second flight. The resulting observations, or more apropriately, the resulting measurements, demonstrate for the first time the capability of a spaceborne multi-frequency multi-polarization microwave radar system to quantify precipitation rates, to detect hydrometer phase, and to classify rain type.

Most models of radar backscatter from the sea ignore sea spikes, the nonlinear effects that result in large excursions above the local mean signal. They are strong enough to change the mean scattering level significantly, and they may cause streaks in synthetic aperture radar (SAR) ocean images. A threshold based on short signal excursions above the local mean is used to identify spikes. With this method, spikes can be found in regions of low signal level. An autoregressive spectral method is used to identify the locations of the spikes on the dominant waves. Sample results from Ka-band measurements in the North Sea made during the SAXON-FPN experiment are given. It is shown that spikes can occur anywhere on the dominant wave, although they are most prevalent on the front face.

Measurements of radar backscatter from an artificial tree were made in the laboratory at 6 and 10 GHz at horizontal incidence. The system had a resolution cylinder 18 cm in diameter and 11 cm long, so about half of the tree (30.5 cm high) was within the beam. The mean returned power from the target with leaves was always higher than that from the target without leaves, as expected. The signal faded with an exponential distribution when the artificial tree was rotated about its trunk, with no angular trend apparent. When the leaves were present, tilted polarizations gave results favoring an angle corresponding with that of the branches. When leaves were absent, the trunk of the tree dominated the radar backscatter at all polarizations.

Results of a feasibility study for a geostationary rain radar are presented. A 2-cm wavelength radar with a 15 to 20 mm antenna will be useful for general scale meteorology. The transmitter power of 500 W with a pulse compression ratio of 200 will provide adequate signal-to-noise ratio for a rain rate of 1 mm/hour. Various problems associated with a geostationary radar and solutions are also discussed.

Simulation results show that a scatterometer design using two pencil beams scanning at different incidence angles measures the near-surface oceanic winds from a satellite better under most conditions than previous designs. The return signals from the ocean surface are much stronger than those from the fan beams used previously. Performance on a polar-orbiting satellite is compared with that of a fan beam spaceborne scatterometer. A wider and continuous swath is covered. The improvement in performance is higher at low wind speeds, so it is particularly suitable for measuring the low-mean-speed tropical wind fields. Performance on a low altitude tropic-orbiting platform such as the Space Station is also shown.

A fine resolution polarimetric FM SAR was developed for optimization of polarimetric SARs and interpretation of SAR data via controlled experiments with surface-base sensors. The system is designed for collecting polarimetric data at 5.3 and 10 GHz over incidence angles from 0 to 60 deg. Features of the system include broad bandwidth to obtain fine range resolution, phase stabilization and linearization loop circuitry, and digital signal processing capability. The system is used in a research program to collect polarimetric backscatter data from artificial sea ice research and design trade-offs, laboratory and field evaluation, as well as results from experiments on artificial sea ice are presented.

The sources of scattering in vegetation and other surfaces and objects were studied. A special radar, SOURCESCAT, that could resolve a cylindrical volume 18 cm in diameter and 11 cm long was built. This system provided the first really fine-resolution measurements of radar backscatter from vegetation. The measurements showed that many of the assumptions used previously in modeling vegetation backscatter were false. Vegetation studied included various field crops, prairie grass, and various trees. Major differences were found in the roles of leaves, branches, stems, and trunks for different species. An artificial tree was studied in the laboratory using the systems. The most significant findings were that the average radar volume scattering coefficient is independent of azimuth, and that slanting of the polarization vector can give useful information not available with ordinary vertical and horizontal polarization. A model for scattering from a single leaf was developed. This model, for the first time, took into account the presence of veins in leaves. The pattern of scatter from a leaf was shown quite different from that for which the veins are ignored. A list of publications and presentations resulting from this project are attached.

A very-fine-range resolution Frequency Modulated-Continuous Wave radar, operating at X-band, was used to evaluate the primary backscattering contributors to the radar signal from a tall-grass prairie. Defoliation was used to help measure relative backscatter from various components of the canopy with both linear polarizations and at 30 deg and 50 deg incidence angles. Variation of the average returned power with incidence angle was greatest for the natural prairie at horizontal polarization and was the least for vertical polarization. Vertical polarization seemed to be more sensitive to the upper portions of the plants, whereas horizontal polarization was to the lower portions. Attenuation for horizontal polarization was only half that for vertical polarization. Dielectric measurements of the grass and some of the canopy components were conducted as well.

The radar backscattering coefficient of simulated first year sea ice was measured at 5.2, 9.6, 13.6, and 16.6 GHz, using a dual-antenna FM-CW scatterometer. The measurements were corrected for antenna separation and beamwidth, to obtain true values of scattering coefficient. Above 15 deg, snow-covered and rough gray ice have similar angular responses, while the decrease in scattering coefficient with incidence angle is more rapid for smoother surfaces. True scattering coefficients and surface roughness data estimated with an antenna-beamwidth correction based on a physical-optics model give values close to those measured.

A short-range very-fine-resolution FM-CW radar scatterometer has been used to identify the primary contributors to 10-GHz radar backscatter from pine, pin oak, American sycamore and sugar maple trees, and from creeping juniper shrubs. This system provided a range resolution of 11 cm and gave a 16-cm diameter illumination area at the target range of about 4 m. For a pine tree, the needles caused the strongest backscatter as well as the strongest attenuation in the radar signal. Cones, although insignificant contributors to the total backscatter, were more important for backscattering than for attenuation. For the rest of the trees, leaves were the strongest cause of backscattering and attenuation. However, in the absence of leaves, the petioles, small twigs, and branches gave relatively strong backscatter. For American sycamore and sugar maple trees, the fruits did not affect the total backscatter unless they were packed in clusters. For creeping juniper the backscattered energy and attenuation in the radar signal were mainly due to the top two layers of the evergreen scales. The contribution of the tree trunks was not determined.

In effort to improve design of Earthquake-resistant structures, mathematical study undertaken to simulate interactions among soil, foundation, and superstructure during various kinds of vibrational excitation. System modeled as three lumped masses connected vertically by springs, with lowest mass connected to horizontal vibrator (representing ground) through springs and dashpot. Behavior of springs described by elastic or elastoplastic force/deformation relationships. Relationships used to approximate nonlinear system behavior and soil/foundation-interface behavior.

Aspects of volume scattering and emission theory are discussed, taking into account a weakly scattering medium, the Born approximation, first-order renormalization, the radiative transfer method, and the matrix-doubling method. Other topics explored are related to scatterometers and probing systems, the passive microwave sensing of the atmosphere, the passive microwave sensing of the ocean, the passive microwave sensing of land, the active microwave sensing of land, and radar remote sensing applications. Attention is given to inversion techniques, atmospheric attenuation and emission, a temperature profile retrieval from ground-based observations, mapping rainfall rates, the apparent temperature of the sea, the emission behavior of bare soil surfaces, the emission behavior of vegetation canopies, the emission behavior of snow, wind-vector radar scatterometry, radar measurements of sea ice, and the back-scattering behavior of cultural vegetation canopies.

Short-range FM scatterometers with single antennas are plagued by interference set up by feedthrough and internal reflections. Dual-antenna systems have much lower internal interference, but there are problems associated with pointing the antennas at the same spot. This note quantifies these problems for Gaussian-shaped beams. The use of antennas with beamwidth ratios of up to 5:1 is shown to improve performance significantly over that obtained with identical beamwidths. For an angle between antenna centers (as viewed from the beam intersection) that is three times the beamwidth of the narrower antenna, the usable spread of range for equal beamwidths is only about 1.85:1, while for an angular ratio of 5, the usable spread of range is greater than 20:1.

High-resolution backscatter measurements were made at X-band, with vertical polarization, and incidence angles of 30 and 50 deg, for winter wheat, soybeans, and corn during the summer of 1984. Plants were observed both under natural conditions and with partial-to-complete defoliation. The resolution volume of the measurement system is better than 20 x 20 x 11.25 cu cm within the entire range of measurements. Results are presented as a radar image, which also provides information on the spatial distribution of the (attenuated) scattering inside the canopy observed under natural conditions. Measured results indicate that: (1) for the full-grown wheat plants, heads are the dominant scatterers at the early heading stage and the ground is the dominant scatterer at the later growth stages; (2) for the soybeans, returns are mostly from the upper 30-cm portion of the plants, and leaves and stems dominate the radar signal at 30 and 50 deg incidence angles, respectively; (3) for the corn plants, leaves are the single dominant scatterers at both angles of incidence; and (4) ground returns are relatively unimportant as they are highly attenuated by the vegetation cover.

The objectives of this research are to identify the primary contributors to 10 GHz radar backscatter from various natural and man-made surfaces and objects, and to use this information in developing new and better models for the scatter. When the true sources are known for the scattering that leads to variation in intensity on radar images, the images (and sets of them) may be interpreted more meaningfully in terms of the variation of parameters of interest for science or application. For example, better interpretation of vegetation images may be possible for yield forecasting and stress detection.

Calibration of the vertical pattern of the antennas for the SEASAT scatterometer was accomplished using the nearly-uniform radar return from the Amazon rain forest. A similar calibration will be attempted for the SIR-B antenna. Thick calibration is important to establish the radiometric calibration across the swath of the SIR-B, and the developed methodology will provide an important tool in the evaluation of future spaceborne imaging radars. This calibration was made by the very-wide-beam SEASAT scatterometer antennas because at 14.65 GHz the scattering coefficient of the rain forest is almost independent of angle of incidence. It is expected that the variation in scattering coefficient for the rain forest across the relatively narrow vertical beam of the SIR-B will be very small; even at L band the forest should be essentially impenetrable for radar signals, the volume scatter from the treetops will predominate as at higher frequencies. The basic research elements include: (1) examination of SIR-B images over the rain forest to establish the variability of the scattering coefficient at finer resolutions than that of the SEASAT scatterometer; (2) analysis of the variability of SIR-B data detected prior to processing for either azimuth compression or; possibly, range compression so that averages over relatively large footprints can be used; (3) processing of data of the form of (2) using algorithms that can recover the vertical pattern of the antenna.

During the last decade, the utility of spaceborne microwave remote sensing systems for ocean windspeed measurement, ocean wave imaging and sea ice studies was demonstrated. Development of large space antennas offers some interesting possibilities for rain rate measurements, ocean and ice studies, and radar imaging. The joint use of active and passive sensors using the 15 m antenna for ocean, ice, and soil moisture studies; rain rate measurements; and radar imaging is considered. Verification of the frequency agile rain radar concept with Shuttle offers the possibility of much needed rain rate statistics over the ocean.

The present article is concerned with a helicopter-borne spectrometer (Heloscat), which has been developed to permit high-quality scattering measurements from a mobile platform at remote sites. The term 'spectrometer' referes to a class of scatterometers. The term 'scatterometer' is employed to denote a specialized radar for measuring scattering coefficients as a function of angle. A spectrometer, on the other hand, is a scatterometer which can measure backscatter at several frequencies. The Heloscat system is discussed, taking into account two antennas, RF hardware, and an externally mounted pendulum for angle encoding. A dual-antenna configuration is used for cross-polarized measurements, while a single-antenna system is used for like-polarized measurements. Attention is also given to oscillator characteristics, efficient data handling, and aspects of calibration.

The variability of the mesoscale wind fields near coastlines which can be caused by mountains that shadow offshore wind and by valleys that enhance them. These wind, provide relatively fixed patterns that must be considered in the development of algorithms for future spaceborne scatterometer systems; mesoscale variability over the offshore regions is random and must be averaged out for forecasting yet nearshore fixed patterns are treated differently. Before the patterns of interest can be defined quantitatively, the scattering response of the ocean to winds at the L-band frequency and SIR-B angles of incidence must be developed from the SIR-B data. Patterns can be analyzed on the images both in regions selected for high probability of the occurrence of suitable patterns, and in other regions where the patterns are observed. The patterns are analyzed for topographic effects and the distance to sea over which these effects cause variations in the oceanic wind patterns. The results are interpreted in terms of quantitative description of the processes involved and in for need of modifications of future scatterometer algorithms.

A series of measurements of the active microwave properties of sea ice under fall growing conditions was conducted. Ice in the inland waters of Mould Bay, Crozier Channel, and intrepid inlet and ice in the Arctic Ocean near Hardinge Bay was investigated. Active microwave data were acquired using a helicopter borne scatterometer. Results show that multiyear ice frozen in grey or first year ice is easily detected under cold fall conditions. Multiyear ice returns were dynamic due to response to two of its scene constituents. Floe boundaries between thick and thin ice are well defined. Multiyear pressure ridge returns are similar in level to background ice returns. Backscatter from homogeneous first year ice is seen to be primarily due to surface scattering. Operation at 9.6 GHz is more sensitive to the detailed changes in scene roughness, while operation at 5.6 GHz seems to track roughness changes less ably.

A scanning pencil-beam scatterometer for ocean windvector determination has potential advantages over the fan-beam systems used and proposed heretofore. The pencil beam permits use of lower transmitter power, and at the same time allows concurrent use of the reflector by a radiometer to correct for atmospheric attenuation and other radiometers for other purposes. The use of dual beams based on the same scanning reflector permits four looks at each cell on the surface, thereby improving accuracy and allowing alias removal. Simulation results for a spaceborne dual-beam scanning scatterometer with a 1-watt radiated power at an orbital altitude of 900 km is described. Two novel algorithms for removing the aliases in the windvector are described, in addition to an adaptation of the conventional maximum likelihood algorithm. The new algorithms are more effective at alias removal than the conventional one. Measurement errors for the wind speed, assuming perfect alias removal, were found to be less than 10%.

Fine-resolution radar backscattering measurements were proposed to determine the backscattering sources in various vegetation canopies and surface targets. The results were then used to improve the existing theoretical models of terrain scattering, and also to enhance understanding of the radar signal observed by an imaging radar over a vegetated area. Various experiments were performed on targets such as corn, milo, soybeans, grass, asphalt pavements, soil and concrete walkways. Due to the lack of available references on measurements of this type, the obtained results will be used primarily as a foundation or future experiments. The constituent backscattering characteristics of the vegetation canopies was also examined.

The focused parabolic antenna has far field pattern characteristics in the radiating near field region. Therefore, it can provide fine resolutions in the across range dimensions. The technique of focusing the parabolic antenna is discussed and applied to a 2-1/2 foot parabolic antenna at X-band. The results of the pattern measurements at various ranges from 2.8 m to 5 m are provided.

The ability to measure the radar scattering coefficient of ice with a helicopter or surface spectrometer was extended into the 4-8 GHz spectral region. The scattering coefficient was measured at Mould Bay, N.W.T., over a frequency range from 4 to 18 GHz for both summer and fall conditions. Scatter from fresh water ice in the St. Lawrence River and from numerous seasonal sea-ice types along the coast of Newfoundland were also measured. The C-band (near 5 GHz) scattering cross section for different types of ice shows poorer contrast than the scattering coefficient at higher frequencies, but better contrast than the negligible value found at L-band (1.5 GHz). At frequencies above 4 GHz the contrast in scattering coefficient between the different ice types is much less in summer than in other seasons; at most times of year the scattering is much stronger from multiyear than from other ice types, but in early summer it is actually slightly weaker than that from first year ice.

The behavior of radiometer corrections for the scatterometer is investigated by simulating simple situations using footprint sizes comparable with those used in the SEASAT-1 experiment and also actual footprints and rain rates from a hurricane observed by the SEASAT-1 system. The effects on correction due to attenuation and wind speed gradients are examined independently and jointly. It is shown that the error in the wind-speed estimate can be as large as 200% at higher wind speeds. The worst error occurs when the scatterometer footprint overlaps two or more radiometer footprints and the attenuation in the scatterometer footprint differs greatly from those in parts of the radiometer footprints. This problem could be overcome by using a true radiometer-scatterometer system having identical coincident footprints comparable in size with typical rain cells.

The radar response from the Amazon rain forest was studied to determine the suitability of this region for use as a standard target to calibrate a scatterometer like that proposed for the National Oceanic Satellite System (NOSS). Backscattering observations made by the SEASAT Scatterometer System (SASS) showed the Amazon rain forest to be a homogeneous, azimuthally-isotropic, radar target which was insensitive to polarization. The variation with angle of incidence was adequately modeled as scattering coefficient (dB) = a theta b with typical values for the incidence-angle coefficient from 0.07 to 0.15 dB/deg. A small diurnal effect occurs, with measurements at sunrise being 0.5 dB to 1 dB higher than the rest of the day. Maximum-likelihood estimation algorithms presented here permit determination of relative bias and true pointing angle for each beam. Specific implementation of these algorithms for the proposed NOSS scatterometer system is also discussed.

The effect of attenuation in precipitation regions of the sea, which must be considered in order to measure the radar backscatter from the ocean with sufficient accuracy to allow determination of the wind vector, can be ascertained from a knowledge of the brightness temperature observed by a microwave radiometer such as the Seasat multifrequency scanning radiometer. Two algorithms relating radiometric measurements and attenuation, and thereby correcting measured scattering coefficient values, were compared with wind vectors reported by surface observers and those determined by the Seasat scatterometer measurements with and without correction for atmospheric attenuation. Although the attenuation correction yields some improvements, it is constrained by both radiometer footprint differences and different scan patterns.

The fundamental principles of radar backscattering measurements are presented, including measurement statistics, Doppler and pulse discrimination techniques, and associated ambiguity functions. The operation of real and synthetic aperture sidelooking airborne radar systems is described, along with the internal and external calibration techniques employed in scattering measurements. Attention is given to the physical mechanisms responsible for the scattering emission behavior of homogeneous and inhomogeneous media, through a discussion of surface roughness, dielectric properties and inhomogeneity, and penetration depth. Simple semiempirical models are presented. Theoretical models involving greater mathematical sophistication are also given for extended ocean and bare soil surfaces, and the more general case of a vegetation canopy over a rough surface.

A brief survey is given of some fundamental physical concepts of optimal polarization characteristics of a transmission path or scatter ensemble of hydrometers. It is argued that, based on this optimization concept, definite advances in remote atmospheric sensing are to be expected. Basic properties of Kennaugh's optimal polarization theory are identified.

The radar response from the Amazon rain forest was studied to determine the suitability of this region for use as a standard target to calibrate a scatterometer like that proposed for the National Ocean Satellite System (NOSS). Backscattering observations made by the SEASAT-1 scatterometer system show the Amazon rain forest to be a homogeneous, azimuthally-isotropic, radar target which is insensitive to polarization. The variation with angle of incidence may be adequately modeled as sigma deg (dB) = alpha theta + beta with typical values for the incidence-angle coefficient from 0.07 dB deg to 0.15 dB/deg. A small diurnal effect occurs, with measurements at sunrise being 0.5 dB to 1 dB higher than the rest of the day. Maximum likelihood estimation algorithms are presented which permit determination of relative bias and true pointing angle for each beam. Specific implementation of these algorithms for the proposed NOSS scatterometer system is also discussed.

The spaceborne radar panel considered how radar could be used to measure precipitation from satellites. The emphasis was on how radar could be used with radiometry (at microwave, visible (VIS), and infrared (IR) wavelengths) to reduce the uncertainties of measuring precipitation with radiometry alone. In addition, the fundamental electromagnetic interactions involved in the measurements were discussed to determine the key work areas for research and development to produce effective instruments. Various approaches to implementing radar systems on satellites were considered for both shared and dedicated instruments. Finally, a research and development strategy was proposed for establishing the parametric relations and retrieval algorithms required for extracting precipitation information from the radar and associated radiometric data.

To provide useful precipitation measurements from space, two requirements must be met: adequate spatial and temporal sampling of the storm and sufficient accuracy in the estimate of precipitation intensity. Although presently no single instrument or method completely satisfies both requirements, the visible/IR, microwave radiometer and radar methods can be used in a complementary manner. Visible/IR instruments provide good temporal sampling and rain area depiction, but recourse must be made to microwave measurements for quantitative rainfall estimates. The inadequacy of microwave radiometer measurements over land suggests, in turn, the use of radar. Several recently developed attenuating-wavelength radar methods are discussed in terms of their accuracy, dynamic range and system implementation. Traditionally, the requirements of high resolution and adequate dynamic range led to fairly costly and complex radar systems. Some simplications and cost reduction can be made; however, by using K-band wavelengths which have the advantages of greater sensitivity at the low rain rates and higher resolution capabilities. Several recently proposed methods of this kind are reviewed in terms of accuracy and system implementation. Finally, an adaptive-pointing multi-sensor instrument is described that would exploit certain advantages of the IR, radiometric and radar methods.