246 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE~24, NO.2. MARCH 1986

Multipolarization Radar Images for Geologic Mapping and Vegetation Discrimination

DIANE L. EVANS, TOM G. FARR, MEMBER, IEEE, J. P. FORD, MEMBER, IEEE,

THOMAS W. THOMPSON, MEMBER, IEEE, AND C. L. WERNER

Abstract-The NASAl JPL airborne synthetic aperture radar system produces radar image data simultaneously in four linear polarizations (HH, VV, VH, HV) at 24.6-cm wavelength (L-band), with 10-m reso­lution, across a swath width of approximately 10 km. The signal data are recorded optically and digitally and annotated in each of the chan­nels to facilitate a completely automated digital correlation. Both stan­dard amplitude, and also phase difference images are produced in the correlation process. Individual polarization and range-dependent gain functions improve the effective dynamic range, but as yet do not per­mit absolute quantitative measurements of the scattering coefficients. However, comparison of the relative intensities of the different polar­izations in individual black-and-white and color composite images pro­vides discriminatory mapping information. In the Death Valley, Cali­fornia, area, rough surfaces of young alluvial deposits produce strong responses at all polarizations_ Smoother surfaces of older alluvial de­posits show significantly lower responses. Evaporite deposits of differ­ent types and moisture contents have distinct polarization signatures. In the Wind River Basin, Wyoming, sedimentary rock units show po­larization responses that relate to differences in weathering. Local in­tensity variations in like-polarization images result from topographic effects; strong cross-polarization responses denote the effects of vege­tation cover and, in some cases, possible scattering from the subsur­face. In the Savannah River Plant, South Carolina, forest cover char­acteristics are discriminated by polarization responses that reflect the density and structure of the canopy, and the presence or absence of standing water beneath the canopy. In each of the areas studied, mul­tiple polarization data allowed discrimination and mapping of unique characteristics of the surficial units.

I. INTRODUCTION

M UL TIPOLARIZED synthetic-aperture radar (SAR) images were acquired over numerous geological,

agricultural, and forested targets by the NASAl Jet Pro­pulsion Laboratory (JPL) airborne radar in August and September 1983 and again in March 1984. The data were acquired at L-band (wavelength of 24.6 cm) simulta­neously in four polarization states: horizontal transmit, horizontal receive (HH); horizontal transmit, vertical re­ceive (HV); vertical transmit, vertical receive (VV); and vertical transmit, horizontal receive (VH).

According to first-order theory for slightly rough sur­faces [1], like-polarized waves are most sensitive to the

Manuscript received June 14, 1984; revised July 10, 1985. The JPL Aircraft Radar Program and the NASAIARC CV-990 Aircraft Programs are supported by the NASA Office of Space Science and Applications. This work was performed in part by the JPL, California Institute of Technology, under contract with NASA.

The authors are with the Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA 91109.

IEEE Log Number 8407040.

-POLARIZED -- CROSS~POLARIZED

---VOLUME SCAffiR'lNG'-

1 I

o ]0 20 30 40 50 W 70 80 90

ANGLE OF INCIDENCE, (degrees)

Fig. 1. Relative contributions of different scattering processes to Iike~ and cross~polarized backscatter. Note that the relative cross~polarized level, normally much lower than like polarized, has been increased for clarity (from [3]).

spatial frequency corresponding to the Bragg resonant condition. For rougher surfaces, higher order theories predict a mixture of surface and subsurface scattering contributions [2]. Fig. I schematically shows the relative contributions of surface and subsurface volume scattering as shown in the Manual of Remote Sensing [3].

Cross-polarized returns result only from second-order effects involving surface multiple scatter and subsurface volume scatter. Most theories predict domination of cross­polarized returns by subsurface volume scattering [2], [4]. For rough surfaces with high dielectric constants, how­ever, surface multiple scattering may contribute signifi­cantly to the return. For low-loss targets such as dry al­luvium, the contribution from the subsurface probably dominates the cross-polarized return, and bulk properties such as number of rocks per cubic meter, porosity, and moisture content become the important target character­istics. Vegetation canopies are also strong volume scat­terers, and commonly exhibit a high cross-polarized com­ponent. Work is continuing on the evaluation of theories that predict the multipolarization radar backscatter re­sponse of natural targets. As calibrated data become more commonly available, quantitative determinations of ter­rain characteristics may be possible. At the present time, however, we are restricted to qualitative statements as in­dicated above.

Preliminary analysis of multiple polarization radar data shows that they are extremely useful for mapping surficial

0196-2892/86/0300-0246$01.00 © 1986 IEEE

EVANS et al.: RADAR IMAGES FOR MAPPING AND DISCRIMINATION 247

1- - - - - - - -AIRC-RA-FTI

I EQUIPMENT I DESK-TOP €il-

'

CV-990 COMPUTER RADAR NAVIGATION -------------+__e- ____________________ .,

,INFO LOGS

I I , I I I ,

I I

FORMATTER,! AID

CONVERTER

RADAR ELECTRONICS

L ____ _

DIGITAL TAPE

RECORDER I+--""""""~VAX DIGITAL

CORRELATOR

I POSTFLIGHT DATA PRODUCTION L- __ _

USER

_ __ --.J

Fig. 2. JPL aircraft SAR system overview. Aircraft operations acquire raw digital tapes and exposed signal film, which are processed to produce SAR images sometime after the aircraft flights.

TABLE I JPL L-BAND RADAR PARAMETERS

Parameter ValuE'

Radar freq uency

Wavelength

Pulse length

Bandwidth

Peak power

Antenna azimuth beamwidth

Antenn~ range beamwidth

Antenna beam center gain

Nominal altitude

Nominal velocity

Nominal pulse repetition frequency

Number of looks

Dynamic range

Azimuth ambiguities

Receiver noise figure

1215 MHz

24.6 em

4.9 )lS

19.3 MHz

4 kW

12 dB

6.0 to 12.0 kID

zoo to 250 m/ s

600 to 800 pps (Dual pol.) 1200 to 1600 pps (Quad pol.)

2 optical 4 digital

12 dB optical 22 dB digital

-20 dB optical -30 dB digital

8 dB

deposits, mapping sedimentary rocks, and mapping a for­est canopy.

II. DATA ACQUISITION

The NASA/JPL L-band aircraft SAR is described by Thompson [5]. A block diagram for the radar is shown in Fig. 2, and a list of the radar operating parameters is pre­sented in Table I. The radar is normally installed on the NASAl Ames Research Center (ARC) CV -990 research aircraft. Most of the radar electronics, including dual re­ceivers and the traveling wave tube transmitter, are 10-

cated in the CV -990 aft baggage compartment. Dual an­tennas, one horizontally polarized and the other vertically polarized, are mounted on a baggage door on the aircraft's starboard side. The antennas have an 18° beamwidth along-track (i.e., parallel to the aircraft's velocity vector) and 75 0 beamwidth in a plane perpendicular to the air­craft's velocity. Radar echoes from ground targets are re­ceived by the antenna, amplified, and heterodyned to video frequencies. These video signals are recorded on both optical and digital recorders. In order to facilitate the processing, the radar pulse repetition frequency (prO is varied with aircraft ground speed, which is derived from the aircraft's inertial navigation system. A desk-top com­puter located in the passenger compartment of the CV-990 aircraft is used to control the radar operation.

In-flight operations produce two forms of raw data-an optical recorder film and a high-density digital tape (HOOT). Following the flights, optical survey images are produced using optical correlator techniques such as those described by Kozma et al. [6]. Digital data are recorded in-flight by digitally sampling the heterodyned video sig­nal at a sample rate of 40 MHz.

The JPL SAR has the capability of simultaneously col­lecting linear like-polarized (HH and VV) and cross­polarized (HV and VH) backscatter data. The transmitter alternately drives the horizontally and vertically polarized antennas while dual receivers simultaneously record the like-polarized and cross-polarized echo signals. In this manner, all possible combinations of linear polarization are recorded. The polarizations, which are interleaved on the high-density digital tape (HOOT), are decoded by the computer during the correlation process. The resulting multiple polarization images are perfectly registered. Since the four polarization channels are recorded essentially

248 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-24, NO.2, MARCH 1986

simultaneously, phase differences between the differently polarized returns also can be formed into images. Work is being pursued on the utility of images depicting the phase difference between the VV and HH channels,

Data acquired in different polarization modes have dif­ferent swath widths. In the quad polarization mode, the optical and digital coverage is 50 to 60 fJ-S in range, which represents ground coverage from nadir to approximately 10 km to the right of the aircraft ground track. In the dual polarization mode, the range coverage is doubled. In this mode, transmitted pulses are either vertically or horizon­tally polarized and both polarizations are recorded. Thus, the data are either HH and HV or VV and VH. Typical flight altitudes are 6 to 12 km. The resulting far-range angles of incidence are near 45 a and 63 0 for the "quad pol" and "dual pol" images, respectively. Typical re­cording times are a few minutes, yielding up to several kilometers of along-track coverage.

III. DIGITAL DATA PROCESSING

Digital data acquired by the JPL SAR are annotated with the aircraft altitude, attitude and position, date and time of day, transmitter power, receiver gains, and polariza­tion mode. This annotation makes it possible to com­pletely automate the correlation process. The processing is performed on a VAX 111780 with a FPS AP-120B array processor and takes approximately 5 h per four-polariza­tion image set.

A. Imaging Algorithm

A synthetic aperture radar return can be modeled as the two-dimensional convolution of the projected scene back­scatter and point target response [7]. The projection is onto a curved plane defined by the direction the sensor is travelling and the range vector pointing from the radar to the imaged point (Fig. 3). The convolving functions are the transmitted waveform expressed in spatial coordi­nates, and the two-dimensional azimuthal chirp [8].

The transmitted waveform for the JPL SAR is a linear FM chirp signal with a bandwidth of 19.3 MHz. Match­filtering of the radar return with the azimuthal chirp cor­relates the image in the range dimension and determines the range impulse response. The transmitter waveform has been weighted with a Hamming window to reduce the range sidelobe level, with a commensurate reduction in range resolution. The unwindowed range resolution is given by

c 2')' = 7.9 m

where c is the speed of light and')' is the chirp rate. Azimuth correlation of the range-compressed data re­

quires matched-filtering of the phase history of each pixel as recorded by the sensor as it travels along the flight path. For spaceborne SAR data, azimuth correlation is a two­dimensional filtering operation, to account for range mi­gration [9], [10]. For aircraft SAR, however, the effective

Fig. 3. Aircraft SAR imaging geometry. The aircraft moves with velocity V along X and images a(x, r) at incidence angle e.

SLANT RANGE SAMPLI NG----j INTERVAL I

r(xll

Fig. 4. Curve of equal range in the (x, r) plane demonstrating range cur­vature.

azimuth window can be made narrow enough, either physically or by digital filtering of the raw data, so that the range migration remains within a slant range sampling interval (Fig. 4).

Azimuthal resolution is determined by the length of the synthetic aperture and the distance the sensor travels while a target is within the effective azimuth antenna pattern. The azimuth resolution (\) is given by

r(O) " r(O) " b = -- = -- = 10.98 m x 2L 2N&

where reO) is the perpendicular range to the scatterer, " is the radar wavelength, L the synthetic aperture length, & the azimuth sampling interval, and N is the number of samples in the azimuth reference function. The small size of the JPL L-band antenna requires that the azimuth sam­pling interval be small. Rather than processing the ac­quired data to full resolution, the data are decimated in azimuth by a digital bandpass filter so that the correlated image resolution is similar in range and azimuth. Speckle noise in the SAR image is reduced through summation of high-resolution images ("looks") obtained by bandpass­filtering of the complex-valued correlator output.

B. Data Format

Digital image parameters are summarized in Table II. Output from the correlation process is in two forms: real­valued data files, which preserve the full dynamic range of the image; or 8-bit pixel format. The 8-bit images have a linear scaling applied so that the average magnitude is

EVANS et al.: RADAR IMAGES FOR MAPPING AND DISCRIMINATION

TABLE II RADAR IMAGE PARAMETERS

Parameter Value

Azimuth pixel spacing 11 meters

Azimuth resolution 13 meters

Number of looks

Slant range pixel spacing 7.5 meters

Slant range resolution 7.9 meters

Ground range pixel spacing, @1O deg. 43 meters

@30 deg. 15 meters

@50 deg. 9.8 meters

Look angle range 0-57 deg. Quad pol.

0-68 deg. Dual pol.

0-75 deg. Single pol.

set to the middle of the display dynamic range. The black­and-white images presented in the following section rep­resent the standard output product from the JPL digital correlator. They have had only a Gaussian contrast stretch applied before the images were printed. Since the multiple polarization images are automatically registered, images acquired in each of three polarization states can be en­coded as the red, green, and blue planes of color com­posite images. In this way, differences in the polarization response of different targets can be viewed simulta­neously.

There are several constraints to consider when compar­ing image intensities from different polarizations and in­cidence angles. Individual polarization and range-depen­dent gain functions have been applied to the receivers to improve their effective dynamic range. The correlator does not compensate for these gain variations at present. The gain functions have been chosen such that surfaces of equal roughness have similar brightness for most inci­dence angles across the swath. Other systematic errors, including range and azimuth antenna patterns, have not yet been removed from the data. However, the gain vari­ations do not change rapidly, especially for incidence an­gles greater than 30 0 • In this regime, it is valid to make relative comparisons of intensities for different polariza­tions and incidence angles. Although absolute quantita­tive measurements of scattering coefficients are not avail­able as yet, the multipolarization data have been shown to be extremely useful in the analyses described below. In addition to color composite images, "polarization sig­natures" of some units have been extracted from the dig­ital data based on mean DN values in extended regions.

IV. GEOLOGIC MAPPING AND VEGETATION

DISCRIMINATION

A. Mapping Suificial Deposits

Radar images can furnish valuable data for the mapping of surficial deposits of differing age, lithology, and chem-

249

ical compOSItIon. Surfaces such as alluvial fans, pedi­ments, and playas are common in desert areas and contain information related to the geologic conditions in the pres­ent and recent past.

Death Valley, in eastern California, has been used as a test site for remote-sensing technique development for a number of years because of the well-exposed examples of common geologic surfaces. Multipolarized radar images were obtained August 30, 1983, over the central part of Death Valley. HH, VV, VH images, and a color com­posite of the three polarizations are shown in Fig. 5(a)­(d), respectively. "Polarization signatures" were ex­tracted for several units in the scene by finding the mean DN of an extended area within each unit. These signatures are displayed in Fig. 6. The area of the images is char­acterized by a variety of Quaternary alluvial gravels, eva­porite deposits and Tertiary alluvial, lacustrine, and vol­canic deposits (Fig. 5(e)). Furnace Creek fan centered at C3, is the most prominent feature in the images. The bright point-returns at the apex of the fan are from Fur­nace Creek Inn and the square feature at B3 is the resort complex of Furnace Creek Ranch. The active part of the fan consists of its upper half at C3, above the dark band. This part of the fan consists of gravels washed down by Furnace Creek. The dark band represents the point at which the particles on the fan become too small, less than about one-tenth the radar wavelength, to scatter the L­band radar waves back to the sensor [11]. Below the dark band (A3, B4-D4, E3), Furnace Creek fan is composed of flood plain deposits and carbonate evaporites. Mes­quite trees preferentially occupy drainages on the flood plain deposits as shown in Fig. 7. Along with the vege­tation at the Ranch, these trees are the only significant vegetation in the scene. Their relatively strong response in the cross-polarized channel (Fig. 5(c)) is a result of multiple scattering from the leaves and the limbs of the trees. This is expressed as a blue to purple color in these areas in Fig. 5(d), but also can be seen in curve a of Fig. 6, which illustrates the relative polarization response of vegetation on Furnace Creek fan.

To the south of Furnace Creek fan, at F-G2, are several smaller fans made up of young gravels. These have the same dark band at their bases, but are darker overall than the Furnace Creek fan. This is because their streams are draining areas of less resistant rock types that weather to pebbles less than 2 cm in size. To the right of these fans, at H3, is an outcrop of the Pliocene Funeral fanglomerate [12]. This ancient alluvial deposit contains volcanic boul­ders up to 1 m in diameter. The outcrop has a distinct bright pink signature in the color composite indicating strong returns at all polarizations, with a slightly higher relative return at HH. Fig. 6 shows the polarization sig­nature of the Funeral fanglomerate graphically. The rel­atively high cross-polarized return may be the result of multiple scattering among the closely packed boulders. These results indicate that multipolarization radar images may be used to discriminate rock types by their erosional characteristics in an alluvial fan environment.

250

(a)

(c)

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-24, NO.2, MARCH 1986

~km

(e)

EXPLANATION

D ROUGH AND fROOEOSAlT

D SMOOTH SALT

O CARBONATE ZONEI SALINE FACIES

O CARBONATE ZONE SAND FACIES

• FLOODPLAIN DEPOSITS

[J YOUNG FAN GRAVELS

m'i INTERMEDIATE Lill FAN GRAVELS

§ OLD FAN GRAVELS

I§ FUNERAL f§I FANGLOMERATE

~ UNDIVIDED ~ PRE PLEISTOCENE ROCKS

~km

(b)

~km (Ai lUtlll')

(d)

Fig. 5. Multipolarization images of Death Valley, CA, centered on 36°20'N, 116°50'W. Images acquired August 30, 1983. (a) Like polarization (HH). (b) Like polarization (VV). (c) Cross polarization (VH). (d) Color composite (HH = red , VV = green, VH = blue). (e) Geologic map of Furnace Creek area of Death Valley simplified from Hunt and Mabey [12) . Also shown are locations of field photographs (Figs. 7-9).

EVANS el at.: RADAR IMAGES FOR MAPPING AND DISCRIMINATION

220

LHH"RED LVH -BLUE LVV -GREEN

Fig. 6. Polarization signatures composed of mean DN values for rectan­gular areas of units in Death Valley . Curve a: vegetation on Furnace Creek fan . Curve b: Funeral fanglomerate . Curve c: smooth salt unit. Curve d: rough salt unit.

Fig. 7. Low-altitude air photograph of Furnace Creek fan showing mes­quite trees occupying drainages at lower right. Area shown is at D3- E3 in Fig. 5.

Another alluvial fan at Tucki Wash is at the lower left of the images (A7-D7). This fan illustrates the differences in radar returns from alluvial surfaces of differing age. Three gravel units have been mapped here on the basis of age (Fig. 5(e» [12]. The two youngest represent active and recently abandoned washes. Both are relatively bright at all polarizations because they are very rough at deci­meter to meter scales. The older surfaces, however, are dark in all of the polarizations. These surfaces are com­posed of the interlocking mosaics of pebbles that form desert pavement, shown in Fig. 8. Their smooth surfaces result in specular reflection away from the radar antenna. Work is ongoing to evaluate the consistency of age effects on multi polarization radar images of alluvial surfaces in arid areas.

Between the alluvial fans lies the lowest part of Death Valley where salts have accumulated from the evapora­tion of lakes that once occupied the valley. These deposits are zoned such that the center of the valley is composed of halite (NaCl) surrounded by less soluble carbonates and sulfates. The dark area in the center of the valley is the seasonally wet flood plains which are very smooth and often covered with a thin veneer of salt. The halite de­posits have been divided into rough and smooth facies [12]

251

Fig . 8. Field photograph of smooth desert pavement surface represented by dark areas at A /B7 in Fig . 5. Scale is 15 cm long.

Fig. 9. Field photograph of rough salt unit. Unit is composed of silt and sodium chloride that have been eroded into a surface with microrelief up to 0 .5 m. This unit is best seen at H5 in Fig. 5.

which are most distinct in the cross-polarized image at H5 and 04, respectively. They are blue-white and greenish­orange in the color composite, and curves c and d of Fig. 6 also show the differences in their polarization signa­tures. The distribution of these signatures closely matches that of the geologic map (Fig. 5(e». The relatively high VH return from the rough salt unit is probably the result of multiple scattering from the extemely rough surface shown in Fig. 9. The polarization response of the smooth salt unit, when compared to that of the rough salt unit, is unusual in that it has much lower VH return, and slightly lower HH and VV returns. The difference between the like and cross-polarized returns of the two units may be due to high moisture conditions causing the smooth salt unit to have an anomalously high dielectric constant. The margins of the valley floor are the areas where ground­water from the alluvial fans usually surfaces. Thus, these are commonly wet or marshy. This may be the cause of the green color of the smooth salt unit at 04 and around the base of the alluvial fans. However, because the im­ages are uncalibrated, it is not possible to quantify this effect. Similar signatures are observed for marshy areas of the Savannah River Swamp discussed later.

252 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-24, NO.2, MARCH 1986

B. Mapping Sedimentary Rocks

One of the main goals of geologic mapping in a sedi­mentary basin is to delineate changes in rock type that relate to changes in paleoenvironment. Models of depo­sitional history are not only useful for exploration for minerals, and oil and gas, but are important for determi­nation of landslide potential and identification of potential aquifers for ground water.

As part of a larger study headed by H. Lang (JPL) to evaluate the utility of remote-sensing measurements for mapping in a sedimentary basin environment, multipolar­ized airborne SAR images were acquired over a portion of the Wind River Basin in central Wyoming. These im­ages, which are shown in Fig. 10, cover the Deadman Butte area on the west flank of the Casper Arch. The strati­graphic sequence exposed in this area ranges from Trias­sic to Late Cretaceous in age and includes limestones, siltstones, shales, and sandstones. The oldest unit ex­posed in the area covered by this image is the Sundance Formation at Al-7. The Sundance is made up of nonre­sistant and resistant members labeled ISs and IS, in Fig. 10( e). The nonresistant member is a shale unit and ap­pears dark in all polarizations because it weathers to out­crops that are smooth on the scale of the radar wave­length. The polarization signature of this unit (Fig. 11), however, is very different from the other shale units in this area owing to compositional differences. Specifically, this shale unit contains more organic material and clay than the Mowry and Thermopolis Shales described below.

The resistant member of the Sundance Formation ap­pears fairly bright and slightly reddish in the color com­posite at A6 and A 7. This strong response is caused by the very blocky nature of this unit. The slight pinkish tone results from a slightly high cross-polarized return (Fig. 11). This is most likely caused by the vegetation cover that this unit supports (Fig. 12). The dark parallel lines that can be seen in the Sundance Formation at A 7 are small horsts and grabens that have resulted from normal faulting [13]. These structures expose a smoother unit with a lower backscatter than the Sundance.

The Morrison and Cloverly Formations are both made up of interbedded sandstones and shales and are difficult to separate in this area [13]. However, variations within these formations are easily visible on the color composite SAR image. For example, the slightly pink shade in the color composite at B6 corresponds to an area of resistant sandstone that is similar in outcrop morphology and vegetation cover to the resistant Sundance unit. The sim­ilarity between these units is also seen in their polariza­tion signatures shown in Fig. 11.

The Thermopolis Shale overlies the Cloverly Forma­tion. This unit is easily mappable where drainages are ac­centuated (e.g., B/C4, 5, E7). The dissected morphology and the low radar backscatter from this unit make this readily interpretable as a nonresistant, easily eroded unit (Fig. 13). In addition, the polarization signature of the Thermopolis Shale is very similar to that of the overlying

Mowry Shale (Fig. 11). These two units are very similar in composition and are both quartz-rich.

The youngest unit exposed in the area covered by this image is the Frontier Formation (E-H, 1-5). This unit has a relatively higher cross-polarized than like-polarized re­turn which results in a red tone in the color composite (Fig. 1O(d». However, unlike the Cloverly and Sundance units, based on airphoto analysis and field work, this rel­atively high cross-polarized return cannot be related to vegetation cover since only sparse, dry grasses are found in this area (Fig. 14). Thus, it may be possible that the cross-polarized return is dominated by scattering from the subsurface.

A basal member of the Frontier Formation is also sep­arable on the color composite. Owing to the similarity in polarization signature between this unit and the Thermo­polis and Mowry Shales (Fig. 11), one would expect this basal unit to be compositionally more similar to those units than to the remainder of the Frontier Formation.

C. Mapping in a Forested Environment

The Savannah River Plant (SRP) site covers some 780 km 2 of the upper coastal plain in western South Carolina. The site has level to gently rolling topography, with low relief, and a substrate of gently dipping Tertiary and Cre­taceous sediments. It is mostly forest covered. The natu­ral composition of the forest is closely governed by the availability of moisture to the trees, and by the extent and duration of flooding. The forest habitats range from very dry sandy hilltops to continuously flooded swamp [14]. The range of habitats is divided into zones that are char­acterized by a community of tree species.

Simultaneous like-polarized (HH and VV) and cross­polarized (HV and VH) airborne SAR coverage of the SRP site was acquired in March 1984. Polarization effects should be mainly related to the forest cover, as geologic and topographic variations in the area are relatively neg­ligible. The like-polarized images are shown in Fig. 15(a) and (b). The cross-polarized image is shown in Fig. 15(c). The area of interest in this study extends from A3 to H3, through H7 and A 7 on the image grids. The radar re­sponses in the HH, VV, and VH polarizations are com­bined to form a color composite image (Fig. 15(d». This false color image enhances the relative inputs of the three polarizations, and reveals colors that are related to major differences in the forest cover. Polarization signatures of some units also have been displayed graphically (Fig. 16).

Most notable are the relatively bright returns in each polarization from the Savannah River swamp area (A3 and H3 and A4 to H4). The swamp supports a dominant cy­press-tupelo community, in a standing-water environ­ment (Fig. 17). Reflections from the tree and water sur­faces provide strong responses in each polarization. The combined input is strikingly displayed in pale yellow tones on the color composite image (Fig. 15(d». Fig. 16 shows in graphical form the relatively high returns with a slight deficiency in VH.

Returns from the Pen Branch delta (D4-E4) are not

EVANS el al.: RADAR IMAG ES FOR MAPPING AND DISCRIMINATION

(a)

• ILLUMINATION ~km

(c)

(e)

EXPLANATION

I(I~ UPI'ER FkONTIER FORMATION

Kil LOWER FRONTIER FORMATION

11m MOWRY FORMATION

lit THERMOPOLIS FORMATION

K.Im CLOVEI'll'!' AND MORRISON FORMATIONS

J., NON RESISTANT MEMBER OF SUNDANCE FORMATION

.... , RESI$TANTMEMBER OF SUNDANCE FORMATION

253

o lkm ~

~km (Al lt.fUII Il

Fig. 10. Multipolarization images of Deadman Butte, WY , centered on 43 °22 'N 106 °58 'W. Images acquired September I , 1983. (a) Like polarization (HH). (b) Like polarization (VV). (c) Cross polarization (VH). (d) Color composite (HH = green, VV = blue , VH = red). (e) Geologic sketch map of Deadman Butte area simplified from [13]. Locations of field photographs (Figs . 12- 14) are also shown.

254 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-24, NO.2, MARCH 1986

z c

220 ,-------,----------,

180

~==-----v---c:::::..-....,=_~ JS r Kim

60L-______ L-_____ ~

LHH • GREEN LVH • RED LVV • BLUE

Fig. II. Polarization signatures of units in Wind River Basin. Explanation is the same as Fig. ID(e).

Fig. 12. Field photograph taken looking south to the resistant member of the Sundance Formation capping the butte. Location shown in Fig. ID(e).

Fig. 13. Field photograph of the Thermopolis Shale. Location shown in Fig. lD(e).

readily discriminated from the responses of the adjacent swamp on the like-polarized images, but combination with the cross-polarized image yields a distinct yellow tone on the color composite image. Curve b of the polarization signatures in Fig. 16 also shows the distinctive VH re-

Fig. 14. Field photograph of the basal Frontier Formation. Picture loca­tion shown in Fig. ID(e).

sponse of the Pen Branch delta. The delta is an area of nonpersistent emergent marsh. Much of the vegetation cover has been killed by the elevated temperature of ef­fluent waters that are drained via Pen Branch to the swamp.

Islands in the swamp, and levees along the Savannah River (e.g., B4-C4, E4-H4, and parallel to the river in A3-H3) are elevated very slightly above the level of the swamp. The small variation in elevation strongly affects the kinds of trees present, and permits a different com­munity that includes species of maple, sweetgum, and pine. This community produces relatively stronger re­sponses in VH and HH polarizations resulting in distinc­tive blue and reddish-blue tones on the color composite image (Fig. IS(d)).

The area north and east of the swamp (AS and HS through H7 to A 7) has an extensive cover of managed pine plantations. An example of a pine stand where the pines are predominantly longleaf species is shown in Fig. 18. Relatively bright responses in the VH polarization from this area produce predominantly blue tones on the color composite image. Local areas of stronger response in the HH polarization (e.g., at E6, E7, G7) show more reddish tones on the color composite image. Fig. 16 graphically shows the comparison between these two sig­natures. Virtually the only difference is in the HH re­sponse. Preliminary field observations in the area, and comparison of the radar images with corresponding color­infrared images that were acquired on the same mission indicate that the relatively low HH response (blue) cor­responds to areas of dense pine canopy, and relatively stronger HH response (reddish tones) corresponds to open pine plantations with grassy understory. This indicates that the type and intensity of polarization response varies with the structure of the canopy. More detailed quantita­tive relations between polarization response and under­story, relative height (age) and density of the pine stands, and contrasting responses of mixed hardwoods are topics that are currently under investigation in conjunction with D. Wickland (JPL), and R. Sharitz (Savannah River Ecol­ogy Laboratory).

EVANS el al.: RA uA R IMAGES FOR MAPPING AND DISCRIMINATION

(a)

(C)

o 1km ~

1\;I'-lUIH

o tkm ~

(e)

~ (b)

(d)

~km {.\/IUUlln

o 1km ~ (A/U.fU THl

255

Fig. 15 . Multipolarization images of Savannah River swamp , SC, centered on 33 °08 ' N, 81 °43'W . Images acquired March I , 1984; (a) Like polarization (HH) , (b) Like polarization (VV), (c) Cross polarization (VH), (d) Color composite (HH = red, VV = green, VH = blue). (e) Sketch map of Savannah River Plant area showing location of major features and field photographs (Figs. 17-19).

256 IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING. VOL. GE-24. NO . 2. MARCH 1986

250 ,--------,--------"l

210

z o

~ 170

~ f~~------~~C--

90'---_____ -'-_____ ---'

LHH"RED LVH"BLUE LVV "GREEN

Fig. 16. Polarization signatures of units in the Savannah River Plant area. Curve a: Savannah River Swamp. Curve b: Pen Branch Delta. Curve c: dense pine forest. Curve d: open pine plantations with grassy understory .

Fig. 17. Field photograph of cypress-tupelo community in standing water. taken at D4 . in Fig . 15.

Fig. 18. Field photograph of longleaf pine and some slash pine in man­aged plantation . taken at E6/F6 in Fig. IS . This area appears blue in the color composite (Fig. 15(d)) .

A mixed hardwood forest is found on the moist soils that are associated with small streams and old floodplains. When the radar images were acquired in March 1984, the hardwood trees were not leafed out and in some areas the soils were saturated. This results in a mottled pattern of

Fig. 19. Field photograph of clearcut taken at D6 in Fig. IS.

bright HH and VV responses, mostly in the area from A5 to D5 and A6 to D6. The pattern is virtually imperceptible on the VH image (Fig. 15(c». This produces reddish-yel­low tones on the color composite image (Fig. 15(d». Cross-polarized returns are significantly lower than in the swamp areas described above.

Clearcuts such as the one shown in Fig. 19, and open water produce dark tones in all polarizations. Areas of pine regrowth in the clearcuts yield relatively higher HH returns. Such areas appear in dark reddish tones on the color composite image (e.g., at C7 and G6). The margins of the clearcuts that face toward the radar illumination show stronger responses in the HH polarization (Fig. 15(a». This may be due to reflections between the trunks of the pines and adjacent clearcut surfaces. Some small ponds (Carolina bays) that are located mostly in the area from A 7 to G7 on the images have a concentric vegetation pattern. Where the ponds are surrounded by trees, the sides of the trees that face the radar illumination produce very bright responses in both HH and VV polarizations probably for the same reason as the forest-bounded clear­cuts.

V. CONCLUSIONS

Images from portions of Death Valley, California, Wind River Basin, Wyoming, and Savannah River Plant, South Carolina, show that multiple polarization data aid in map­ping surficial deposits, mapping sedimentary rocks, and mapping in forested environments, respectively.

Multiple polarization radar data for Death Valley pro­vided a valuable tool for the study of alluvial surfaces of different ages and lithologies, and of evaporite deposits of different types and moisture contents. In the Wind River Basin area, multiple polarization radar data were used to discriminate sedimentary rocks based on surface rough­ness, vegetation cover, and possibly compositional infor­mation provided in the images. In the Savannah River Plant area, brightness in the multipolarization images was found to be qualitatively related to tree size, distribution and density, and to presence or absence of standing water beneath the canopy. Further investigations are directed to

EVANS et 0/ .: RADAR IMAGES FOR MAPPING AND DISCRIMINATION

relations between polarization responses, soil moisture, and understory.

These results are particulary relevant to the Shuttle Im­aging Radar-C currently scheduled for 1989, which will have multiple polarization capability.

ACKNOWLEDGMENT

The Aircraft Radar Group at JPL, headed by W. Brown, was responsible for acquiring the radar images presented here. The Radar Systems Science and Engineering Group at JPL, headed by D. Held, was responsible for the digital processing of the data. We gratefully acknowledge their contribution to this work. H. Lang, of the Geology Group at JPL, provided one of the ground photos of the Dead­man Butte area.

We thank C. Elachi and D. Held for their helpful re­views of this manuscript, and the NASAl ARC Medium Altitude Missions Branch for their support of our efforts via the excellent operation of the NASAl ARC CV -990 Airborne Laboratory, Galileo II.

Note: On July 17, 1985, the NASAIARC CV-990 caught fire on takeoff from March AFB, CA. The crew escaped without injury, but the aircraft, with the JPL ra­dar on board, was completely destroyed. The radar is being rebuilt with the same parameters, and plans call for it to be mounted on a replacement aircraft by mid-1987.

REFERENCES

[I] G. R. Valenzuela, "Depolarization of EM waves by slightly rough surfaces," IEEE Trans. Antennas Propagat ., vol. AP-15, pp. 552-557, 1967.

[2] A. 1. Blanchard and 1. W. Rouse , Jr. , "Depolarization of electro­magnetic waves scattered from an inhomogeneous half space bounded by a rough surface," Radio Sci., vol. 15, pp . 773-779,1980.

[3] R. N. Colwell, Ed., Manual of Remote Sensing. American Society of Photogrammetry, 1983.

[4] A. K. Fung and H. 1. Eom, "Note on the Kirchoff rough surface solution in backscattering," Radio Sci., vol. 16, pp. 299-302, 1981.

[5] T. W. Thompson, "A user's manual for the NASA/JPL synthetic aperture radar and the NASA/JPL L- and C-band scatterometers, " JPL Pub. 83-38 , 1983.

[6] A. Kozma, E. M. Leith, and N. G. Massey , "Ti lted plane optical processor," Appl. Opl., vol. II , pp. 1766-1777, 1972.

[7] C. Wu, "Modeling and a correlation algorithm for spaceborne SAR signals," IEEE Trans. Aerosp. Electron . Syst., vol. AES-18, 563-575, 1982.

[8] M. Jin and C. Wu , "SAR correlation technique-A modified inter­pretation algorithm," in Proc. IGARSS , 1983.

[9] K. Tomiyasu, "Tutorial review of synthetic aperture radar (SAR) with applications to imaging the ocean surface," Proc. IEEE, vol. 66, pp. 563- 583 , 1978.

[10] C. Elachi, T. Bicknell, R. L. Jordan, and C. Wu, "Spaceborne syn­thetic aperture radars: Applications, techniques , and technology , " Proc. IEEE, vol. 70, pp. 1174-1209, 1982.

[II] G. G. Schaber, G. L. Berlin, and W. E. Brown, Jr., "Variations in surface roughness within Death Valley, California: Geologic evalu­ation of 25-cm-wavelength radar images ," Geol. Soc. Amer. Bull. , vol. 87, pp. 29-41 , 1976.

[12] C. B. Hunt and D. R. Mabey, "Stratigraphy and structure, Death Valley, California," U.S. Geol. Surv. Prof. Paper 494-A , 1966.

[13] T. C. Woodward, "Geology of Deadman Butte Area, Natrona County, Wyoming," Bull. Amer. Assoc. Petroleum Geol., vol. 41, pp. 212-262, 1957 .

[14] T. M. Langley, and W. L. Marter, "The Savannah River Plant site," E.!. Du Pont DeNemours and Co. , Savannah River Lab., Aiken, SC, DP-1323 , 1973 .

257

Diane L. Evans received the A.B. degree in ge­ology from Occidental College, Los Angeles, CA, in 1976 and the M.S. and Ph.D. degrees in 1978 and 1981 , respectively , in geological sciences from the University of Washington , Seattle.

She is currently a Member of the Technical Staff in the Radar Sciences Group and the Assis­tant Program Manager for the Land Processes Program at the Jet Propulsion Laboratory . She is the Principal Investigator for a study of new tech­niques for quantitative analysis of SAR images and

a Collaborator for a Shuttle Imaging Radar-B (SIR-B) study of the quanti­tative use of multi-incidence angle SAR for geologic mapping. She was part of the operations team during the SIR-B mission in October 1984, and is presently serving on the RADARSAT Science Working Group and as the Experiment Scientist for SDR-C.

* Tom G. Farr (M'84) received the B.S. degree in 1974 and the M.S . degree in 1976, both in geol­ogy from the California Institute of Technology, Pasadena, and the Ph .D. degree in 1981 in geol­ogy from the University of Washington, Seattie.

Throughout his Masters and Ph.D. work, he was employed by the Jet Propulsion Lab, at first working on airborne radar data of arctic sea ice and later on Seasat and SIR-A images of geologic targets. He is currently an Investigator on the SIR­B experiment, which has involved the develop­

ment of new techniques for the extraction of quantitative geologic infor­mation from multi-incidence angle radar images. His current research in­terests include the improvement of electromagnetic scattering models for rough geologic surfaces and subsurfaces, and the use of multiparameter radar systems to quantitatively predict surface and subsurface propelties on Earth and other planets.

Dr. Farr is a member of the SIR-C Science Working Group in geology.

* J. P. Ford (S'78-M'82) received the B.Sc. de­gree with honors in geology from the University of London, England, in 1959 and the Ph.D degree in geology from The Ohio State University, Co­lumbus, in 1965.

He has worked at the Jet Propulsion Labora­tory , California Institute of Technology , Pasa­dena, since 1977, as a Senior Resident Research Associate (1977-1979), Membcr of Technical

!I Staff (1979-1981), Technical Group Leader (1981-1982), and Supervisor (1982 to the pres­

ent) in the Radar Sciences Group, Earth and Space Sciences Division. His research interests are currently directed in the area of radar remote sensing for geologic mapping, with particular application in forested environments.

* Thomas W. Thompson (S'58-M'80) was born in Canton, OH, on May 25, 1936. He received the B.S . degree from Case Institute of Technology in 1958, the M.E. degree from Yale University in 1959, and the Ph.D. degree from Cornell Univer­sity in 1966.

He was one of the first experimenters to use the 430-MHz radar at the Arecibo Observatory. From 1964 through 1969, he made extensive radar ob­servations of the Moon. From 1970 through 1976, he was a coinvestigator on the Apollo Lunar

Sounder and was a principal investigator on one of the first lunar data syn­thesis programs. From 1973 through 1981, he participated in the Seasat program and also served as Science Operations Coordinator on the Voyager mission. Since 1983 he has coordinated the NASA/JPL Aircraft SAR pro­gram.

* C. L. Werner, photograph and biography not available at the time of pub­lication.