Jack F. Paris

Radar remote sensing of forest and wetland ecosystems in the Central American tropics

Abstract We analyzed airborne synthetic aperture radar (AIRSAR) imagery of forest, wetland, and agricultural ecosystems in northern Belize, Central America. Our analyses are based upon four biophysical indices derived from the fully polarimetric SAR data: the volume scattering index (VSI), canopy structure index (CSI), biomass index (BMI), calculated from the backscatter magnitude data, and the interaction type index (ITI), calculated from the backscatter phase data. We developed a four-level landscape hierarchy based upon clustering analyses of the 12 index parameters (four indices each for P, L, and C band) from two test site images. Statistical analyses were used to examine the relative importance of the 12 parameters for discriminating ecosystem characteristics at various landscape scales. We found that ITI was the most important index (primarily C band = CITI) for level, vegetated terrain at all levels of the hierarchy. BMI was most important for differentiating between vegetated and nonvegetated areas and between sloping and level terrain. These findings indicate that upper canopy spatial characteristics and flooding in marshlands (reflected in the CITI) are more important than biomass in differentiating many tropical ecosystems with radar data. The relative importance of the indices varied with vegetation type; for example, PVSI was the most important for distinguishing between upland forests and regrowth, and PCSI was the most important for differentiating swamp forest types. Finally, we evaluated the potential of present and future spaceborne SARs for tropical ecosystem studies based on our results. Most of these SARs are single channel systems and will provide limited capability for characterizing biomass and structure of tropical vegetation. This is especially true for C band systems, which produce data similar to our CBMI parameter, which was one of the least important in our analyses. The SIR-C/X-SAR and proposed EOS SAR are future spaceborne multifrequency fully polarimetric SAR systems, and they will provide a significant contribution to tropical ecosystem studies.

Detecting seasonal flooding cycles in marshes of the Yucatan Peninsula with SIR-C polarimetric radar imagery☆

Abstract Polarimetric L- and C-band radar imagery from the shuttle imaging radar-C (SIR-C) were acquired over wetlands of the Yucatan Peninsula during the dry (April) and wet (October) seasons of 1994. Field surveys during the flights recorded biophysical data and water depth in 11 marsh sites containing communities of three principal emergent macrophytes: Cladium jamaicense, Typha domingensis, and Eleocharis cellulosa. The only major seasonal change was in flooding. Seasonal changes in polarimetric backscatter magnitude (HH, VV, and CS=(HV+VH)/2) and phase [βH-V phase differenceβ=PD) were extracted for a stable evergreen mangrove forest calibration site, which confirmed that the absolute calibration of the Yucatan imagery exceeded the SIR-C system calibration. We estimate that seasonal changes of ⩾2dB in backscatter magnitude and ⩾10° in phase (PD) are significant in our data. Seasonal changes in L- and C-band magnitude and phase were extracted from the 11 marshes, and significant changes above the calibration limit were noted. Increased flooding in the marshes was detected by: 1) an increase in backscatter magnitude in marshes with tall, dense cover; 2) a decrease in backscatter magnitude in marshes with short, sparse cover, and 3) an increase in PD in all types of marshes. Magnitude increases result from an increase in double-bounce interactions between the emergent vegetation and water surface, whereas decreases result from an increase in forward scattering off the open water. Average PD values increase owing to an absolute or relative increase in double- compared with single-bounce interaction. Changes from dry or partially flooded to completely flooded, as well as increases in water depth, could be detected by most of the polarimetric parameters, but changes from dry to partially flooded could not. C-band PD (CPD) was the radar parameter most sensitive to flooding. CPD changed significantly for all eleven marshes, followed by L-band PD (LPD) and LVV (nine marshes) and LHH, LCS, and CVV (seven marshes). CHH detected significant changes in five marshes but produced changes of ±1.8–1.9 dB (just below our estimated calibration limit) in four others. An evaluation of current spaceborne radars indicates that a combination of the European Remote Sensing Satellite (ERS-1,2) and Radarsat radars could detect seasonal flooding in a wide variety of marsh ecosystems, excluding partial flooding and flooding in small patches of short, sparse vegetation.

Distinguishing high and low anopheline-producing rice fields using remote sensing and GIS technologies

Abstract Worldwide, 140 million ha are devoted to rice cultivation, mostly in developing countries of the tropics and subtropics where malaria still constitutes a serious human health problem. Because rice fields are flood-irrigated on a semi-permanent basis during each growing season, they provide an ideal breeding habitat for a number of potential mosquito vectors of malaria. One of these vectors, Anopheles freeborni , is distributed throughout nearly 240 000 ha of irrigated rice in northern and central California, and may serve as a model for the study of rice field mosquito population dynamics using spectral and spatial information. Analysis of field data revealed that rice fields with rapid early season vegetation canopy development, located near livestock pastures (i.e. bloodmeal sources), had greater mosquito larval populations than fields with more slowly developing vegetation canopies located further from pastures. Remote sensing reflectance measurements of early season rice canopy development and geographic information system (GIS) measurements of distance to livestock pasture were combined to distinguish between high and low mosquito-producing rice fields. These distinctions were made with 90% accuracy nearly 2 months before anopheline larval populations peaked.

Spaceborne imaging radar-C (SIR-C) observations of groundwater discharge and wetlands associated with the Chicxulub impact crater, northwestern Yucatan Peninsula, Mexico

Analyses of spaceborne imaging radar-C (SIR-C) data and field data from the northwestern Yucatan Peninsula, Mexico, demonstrate that spaceborne multifrequency polarimetric radars are excellent tools for characterizing patterns of wetland flooding. Seasonal flooding can be detected in most types of forest and marsh in the radar backscatter magnitude and phase data of both L and C band. Field observations made in the wet and dry seasons concurrent with the space missions and chemical analyses of floodwaters confirm that flooding is the product of discharge from the Yucatan aquifer, which consists of a fresh-water lens floating on seawater. This discharge controls the distribution of wetlands. Therefore, vegetation and flooding patterns, mapped with SIR-C imagery, provide valuable information on the hydrogeology of the region. Radar-image maps of wetlands and flooding indicate that there are three major zones of groundwater discharge that correlate with structures of the buried Chicxulub crater—zone 1 with the peak ring, zone 2 with the crater rim, and zone 3 with the exterior ring. Zone 1 has sulfate-poor discharge, unlike the sulfate-rich discharge in zones 2 and 3. The highest discharge is in zone 3, where the buried crater is closest to the surface. This groundwater-discharge pattern can be explained by tidal pumping of fresh water to the surface through high permeability zones developed in the Tertiary carbonates overlying crater faults and escarpments.

Inclusion of a simple multiple scattering model into a microwave canopy backscatter model

Abstract A simple multiple scattering model has been incorporated into a microwave canopy backscatter model for forest stands with continuous or discontinuous tree canopies. The multiple scattering model was empirically derived using available calculated multiple scattering values and Monte Carlo simulation. All orders of scattering within canopies beyond single scattering were assumed to be isotropic. Multiple scattering was divided evenly among HH, HV, VH, and VV polarizations. The corresponding single scattering term was polarization-sensitive. The effect of the multiple scattering term on modeled canopy backscatter was less at long wavelengths than at short wavelengths. At a given wavelength, the multiple scattering term affected copolarized scattering less than cross-polarized scattering. These predictions were consistent with calibrated SAR observations and with our understanding of microwave scattering in forested environment. Including multiple scattering effects improved the agreement between modeled and measured canopy backscatter particularly for cross-polarized backscatter at short wavelengths.

A Simple Algorithm To Correct Single-scattering Model Outputs For Multiple-scattering Effects

When EM radiation interacts with discrete elements in a volumescattering medium (e.g., with leaves in a forest canopy), some of the radiant energy is scattered upward. To aid remotely-sensed data analysis, a model may be used to simulate the interactions between the elements of the medium and the radiant energy. Usually, such models include only single scattering (SS). In reality, significant multiple scattering (MS) may be involved. Rigorous mathematical algorithms for modeling MS processes are usually very complex and very time consuming. In this paper, I present a simple algorithm, which uses SS model parameters, to estimate the strength of the MS contribution. THE PROBLEM CONSIDERED For short-wave optical and active-microwave situations, the spectral radiance, Lh, coming from a volume-scattering medium is the result of interactions between spectral irradiance, Eh, and discrete elements in the medium. Often, Eh comes primarily from a source having a small solid angle (e.g., the sun or an active-microwave device). In this paper, I am considering the case of volume-scattering media having uniform properties within the volume. Let N (m-3) be the element-number density of the medium. For an element, the spectral power absorbed or scattered is proportional to its cross section for absorption, Qa (m2), or for scattering, Qs (m2), respectively. The sum, Q (m2) = Q + Qs, represents the attenuation properties of the element The probability that the outcome of the interaction is scattering (not absorption) is o0 (the single-scattering albedo), where oo = Qs I Q. To account for transmittance, t, along a path length, Ad (m), one can use Beers Law,

Progressive Transformation: A Simple, Customized Approach for Dimensional Reduction and Scalar Parameter Generation for Remotely-Sensed Data Analysis

The extraction of measures of certain properties of objects from multichannel data is a basic process in analytical remote sensing. One approach is the use of standard rotational transformations, e.g., those designed for use in agricultural areas for “soil brightness,” (vegetative) “greenness,” (vegetative) “yellowness,” (plant or soil) “wetness,” etc. from Landsat Multispectral Scanner System data [l] or Landsat Thematic Mapper data [2]. While standard rotational coefficients may work well in many cases, a need exists for a simple approach which will produce a customized set of transformation coefficients which will yield scalar measures of certain physical properties in particular situations. This paper presents, in a tutorial fashion, a technique called the Progressive Transformation for achieving this goal. Key Words: Information extraction, customized transformations, multichannel remote-sensing.

Simultaneous Image Data Acquired By A Multispectral Scanner And A Polarimetric, Multifrequency Synthetic Aperture Radar Over A Wetland Site In Colusa County, California: A Comparative Analysis

In studies of cultural wetlands (e.g., rice) or natural wetlands (e.g., marshes), scientists may be interested in using remotely-sensed data to map certain wetland features. In our studies, we considered visible and infrared scanners (VIRS) and synthetic aperture radar (SAR). In support of our investigation, NASA acquired, from aircraft platforms ,VIRS data [in the Thematic Mapper (TM) bands] and full-polarimetric, multifrequency SAR data over a wetland area in Colusa County, California, on the same day and at nearly the same time. The SAR measured the backscattering properties of the surface at three frequencies, C-band (6-cm wavelength), L-band (24 cm), and P-band (68 cm), and included full polarization properties (backscatteringmagnitudes and phase-angle differences). The two data sets were acquired at mid-day on May 27, 1988, over rice fields and natural wetlands (e.g., bullrushes) managed for duck hunting and other wildlife. The date was early in the rice growing season after fields had been flooded and rice had emerged in some fields. These data presented a unique opportunity for us to compare broad-band VIRS data to full polarimetric, multifrequency SAR data.