Robert N. Treuhaft

Vertical structure of vegetated land surfaces from interferometric and polarimetric radar

This paper describes the estimation of parameters characterizing the vertical structure of vegetated land surfaces, from combined interferometric and polarimetric radar data. Physical models expressing radar observations in terms of parameters describing vegetated land surfaces are the foundation for parameter estimation techniques. Defining a general complex cross correlation enables the unified development of models for interferometry and polarimetry, including polarimetric interferometry. Three simple physical models in this paper express this complex cross correlation in terms of vegetation parameters: (1) a randomly oriented volume, (2) a randomly oriented volume with a ground return, and (3) an oriented volume. For the first two models the parameters include vegetation height, extinction coefficient, underlying topography, and another parameter depending on ground electrical properties and roughness. For the oriented volume, additional parameters depend on the refractivity, extinction coefficients, and backscattering characteristics of waves propagating along eigenpolarizations of the vegetation volume. The above models show that the interferometric cross-correlation amplitude and the polarimetric {HHHH/VVVV} ratio both change by about 1% per meter of vegetation height change, for experimental conditions typical of airborne and spaceborne interferometric radars. These vertical-structure sensitivities prompt a parameter estimation demonstration with two-baseline TOPSAR interferometric and zero-baseline polarimetric data from the Boreal Ecosystem-Atmosphere Study (BOREAS) Southern Study Area in Prince Albert National Park, Saskatchewan, Canada. The demonstrations show the feasibility of measuring vegetation height to better than 4.2 m, underlying topography to better than 6.5 m, and the ratio of ground-to-volume power to better than 10%, using interferometry and polarimetry, coupled with parameter-constraining assumptions, concerning the degree of surface roughness. This paper suggests that single-baseline and multibaseline fully polarimetric interferometry have the potential to obviate the need for such assumptions, thereby making parameter estimation more robust, accurate, and realistic.

Vegetation characteristics and underlying topography from interferometric radar

This paper formulates and demonstrates methods for extracting vegetation characteristics and underlying ground surface topography from interferometric synthetic aperture radar (INSAR) data. The electromagnetic scattering and radar processing, which produce the INSAR observations, are modeled, vegetation and topographic parameters are identified for estimation, the parameter errors are assessed in terms of INSAR instrumental performance, and the parameter estimation is demonstrated on INSAR data and compared to ground truth. The fundamental observations from which vegetation and surface topographic parameters are estimated are (1) the cross-correlation amplitude, (2) the cross-correlation phase, and (3) the synthetic aperture radar (SAR) backscattered power. A calculation based on scattering from vegetation treated as a random medium, including the effects of refractivity and absorption in the vegetation, yields expressions for the complex cross correlation and backscattered power in terms of vegetation characteristics. These expressions lead to the identification of a minimal set of four parameters describing the vegetation and surface topography: (1) the vegetation layer depth, (2) the vegetation extinction coefficient (power loss per unit length), (3) a parameter involving the product of the average backscattering amplitude and scatterer number density, and (4) the height of the underlying ground surface. The accuracy of vegetation and ground surface parameters, as a function of INSAR observation accuracy, is evaluated for aircraft INSAR, which is characterized by a 2.5-m baseline, an altitude of about 8 km, and a wavelength of 5.6 cm. It is found that for ≈0.5% accuracy in the INSAR normalized cross-correlation amplitude and ≈5° accuracy in the interferometric phase, few-meter vegetation layer depths and ground surface heights can be determined from INSAR for many types of vegetation layers. With the same observational accuracies, extinction coefficients can be estimated at the 0.1-dB/m level. Because the number of parameters exceeds the number of observations for current INSAR data sets, external extinction coefficient data are used to demonstrate the estimation of the vegetation layer depth and ground surface height from INSAR data taken at the Bonanza Creek Experimental Forest in Alaska. This demonstration shows approximately 5-m average ground truth agreement for vegetation layer depths and ground-surface heights, with a clear dependence of error on stand height. These errors suggest refinements in INSAR data acquisition and analysis techniques which will potentially yield few-meter accuracies. The information in the INSAR parameters is applicable to a variety of ecological modeling issues including the successional modeling of forested ecosystems.

The structure of oriented vegetation from polarimetric interferometry

Polarimetric radar interferometry is much more sensitive to the distribution of oriented objects in a vegetated land surface than either polarimetry or interferometry alone. This paper shows that single-baseline polarimetric interferometry can be used to estimate the heights of oriented-vegetation volumes and underlying topography, while at least two baselines are needed for randomly oriented volumes. Single-baseline, calculated vegetation-height accuracies are in the range of 2-8 m for reasonable levels of vegetation orientation in forest canopies.

Forest Attributes from Radar Interferometric Structure and Its Fusion with Optical Remote Sensing

Abstract The possibility of global, three-dimensional remote sensing of forest structure with interferometric synthetic aperture radar (InSAR) bears on important forest ecological processes, particularly the carbon cycle. InSAR supplements two-dimensional remote sensing with information in the vertical dimension. Its strengths in potential for global coverage complement those of lidar (light detecting and ranging), which has the potential for high-accuracy vertical profiles over small areas. InSAR derives its sensitivity to forest vertical structure from the differences in signals received by two, spatially separate radar receivers. Estimation of parameters describing vertical structure requires multiple-polarization, multiple-frequency, or multiple-baseline InSAR. Combining InSAR with complementary remote sensing techniques, such as hyperspectral optical imaging and lidar, can enhance vertical-structure estimates and consequent biophysical quantities of importance to ecologists, such as biomass. Future InSAR experiments will supplement recent airborne and spaceborne demonstrations, and together with inputs from ecologists regarding structure, they will suggest designs for future spaceborne strategies for measuring global vegetation structure.

2‐cm GPS altimetry over Crater Lake

Dierences in electromagnetic path delay, be- tween direct Global Positioning System (GPS) signals and those reflected from the surface of Crater Lake, have led to lake surface height estimates with 2-cm precision in 1 sec- ond. This is the rst high-precision altimetric demonstra- tion with GPS from sucient altitude (480 m) to probe fundamental experimental errors, which bear on future air- and spaceborne passive GPS altimetry. It also serves as the rst demonstration of a new approach to altimetric remote sensing in the coastal region, an area that is poorly mea- sured by conventional radar altimetry. Time-series analy- sessuggest thattroposphericandthermalnoisefluctuations dominate the altimetric error in this experiment. Estimat- ing the dierential delay from several simultaneously visi- ble satellites may enable tropospheric error estimation and correction. Thermal noise on the reflected signal will be reduced with fully polarimetric observations and larger an- tenna apertures.

View angle effects on canopy reflectance and spectral mixture analysis of coniferous forests using AVIRIS

The dependence of vegetation reflectance on sun and sensor geometry can potentially provide information on canopy properties, but also may be a source of unmodelled systematic error in single-angle remote sensing measurements. In this study, we investigated the angular variability of reflectance measurements from the NASA Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), and the resulting impact on spectral mixture analysis (SMA) using both full-range (400-2500 nm) and shortwave-infrared wavelengths (2080-2280 nm; AutoSWIR ). The study was conducted in coniferous forests in Central Oregon using five AVIRIS overpasses to generate multiple view angle measurements. Canopy reflectance was highly anisotropic, with the strength of the angular signal controlled by species type, canopy cover and soil reflectance. Canopy cover estimates from full-range SMA averaged only slight decreases (∼6% relative) toward the retro-solar direction for 16 field plots in the study region. AutoSWIR was even less influenced by view angle, exhibiting changes only for large differences in view angle. In addition, AutoSWIR s ability to accommodate endmember variability led to stronger agreement with field cover values than full-range SMA. The results suggest that while view angle can significantly affect reflectance measurements from AVIRIS, the consequent variability in vegetation cover estimates from SMA and AutoSWIR is low.

The calculated performance of forest structure and biomass estimates from interferometric radar

Abstract Vertical structure and biomass are key characteristics of the forest random medium. This paper calculates the power and interferometric synthetic aperture radar (InSAR) sensitivity to tree height and vegetation density as it manifests in extinction, using a homogeneous, random-volume model of the forest medium, and accounting for speckle and thermal noise. Tree height and extinction are both related to biomass within the context of this simple model. Signal and noise calculations show that InSAR coherence and phase are more sensitive than radar power to structure and biomass to 10% variations in structure parameters over a wide range of medium to high density forests. For example, for extinctions of 0.2 db m−1 and other parameters as noted in the text, the sensitivity of InSAR coherence to 10% changes in tree height exceeds observation errors for trees shorter than about 37 m, as opposed to 14 m for radar power. InSAR phase sensitivity to 10% structural and associated biomass changes exceeds observation errors for all tree heights. Only for the very lowest densities, for extinctions less than 0.15 db m−1, does radar power sensitivity to density exceed that of InSAR; even at low densities, InSAR sensitivity to tree height outperforms that of radar power in these calculations.

Tropical-Forest Biomass Estimation at X-Band From the Spaceborne TanDEM-X Interferometer

This letter reports the sensitivity of X-band interferometric synthetic aperture radar (InSAR) data from the first dual-spacecraft radar interferometer, TanDEM-X, to variations in tropical-forest aboveground biomass (AGB). It also reports the first tropical-forest AGB estimates from TanDEM-X data. Tropical forests account for about 50% of the worlds forested biomass and play critical roles in the control of atmospheric carbon dioxide by emission through deforestation and uptake through forest growth. The TanDEM-X InSAR data used in this analysis were taken over the Tapajós National Forest, Pará, Brazil, where field measurements from 30 stands were acquired. The magnitude of the InSAR normalized complex correlation, which is called coherence, decreases by about 25% as AGB increases from 2 to 430 Mg-ha-1, suggesting more vertically distributed return-power profiles with increasing biomass. Comparison of InSAR coherences to those of small-spot (15 cm) lidar suggests that lidar penetrates deeper into the canopies than InSAR. Modeling InSAR profiles from InSAR coherence and lidar profiles yields an estimate of 0.29 dB/m for the X-band extinction coefficient relative to that of lidar. Forest AGB estimated from InSAR observations on 0.25-ha plots shows RMS scatters about the field-estimated AGB between 52 and 62 Mg-ha-1, which is between 29% and 35% of the average AGB of 179 Mg-ha-1, depending on the data analysis mode. The sensitivity and biomass-estimation performance suggest the potential of TanDEM-X observations to contribute to global tropical-forest biomass monitoring.

Subpixel canopy cover estimation of coniferous forests in Oregon using SWIR imaging spectrometry

The percent cover of vegetation canopies is an important variable for many land-surface biophysical and biogeochemical models and serves as a useful measure of land cover change. Remote sensing methods to estimate the subpixel fraction of vegetation canopies with spectral mixture analysis (SMA) require knowledge of the reflectance properties of major land cover units, called endmembers. However, variability in endmember reflectance across space and time has limited the interpretation and general applicability of SMA approaches. In this study, a subpixel vegetation cover of coniferous forests in Oregon, United States, was successfully estimated by employing shortwave infrared reflectance measurements (SWIR2 region, 2080–2280 nm) collected by the NASA Airborne Visible Infrared Imaging Spectrometer (AVIRIS). The approach presented here, referred to as AutoSWIR [Asner and Lobell, 2000], was originally developed for semiarid and arid environments and exploits the low SWIR2 variability of materials found in most ecosystems. SWIR2 field spectra from Oregon were compared with spectra from an arid systems database, revealing significant differences only for soil reflectance. However, SWIR2 variability remained low, as indicated by field spectra and principal component analysis, and AutoSWIR was then modified to use coniferous forest spectra collected in Oregon. Subsequent high spatial resolution estimates of forest canopy cover agreed well with estimates from low-altitude air photos (rms = 3%), demonstrating the successful extension of AutoSWIR to a coniferous forest ecosystem. The generality of AutoSWIR facilitates accurate estimates of vegetation cover that can be automatically retrieved from SWIR2 spectral measurements collected by forthcoming spaceborne imaging spectrometers such as NASAs New Millenium Program EO-1 Hyperion. These estimates can then be used to characterize landscape heterogeneity important for land-surface, atmospheric, and biogeochemical research.

Stem volume of tropical forests from polarimetric radar

In this study, we investigated the potential of polarimetric synthetic aperture radar (PolSAR) data for the estimation of stem volume in tropical forests. We used calibrated L-band, high incidence angle data from the airborne system SAR-R99B, acquired over an experimental area in the Tapajós National Forest, Pará, Brazil. To evaluate the potential of PolSAR data for this application we used regression analysis, in which first-order models were fit to predict stem volume per hectare, as determined from field measurements. Unlike previous studies in tropical forests, the set of potential explanatory variables included a series of PolSAR attributes based on phase information, in addition to power measurements. Model selection techniques based on coefficient of determination (R 2) and mean square error (MSE) identified several useful subsets of explanatory variables for stem volume estimation, including backscattering coefficient in HH polarization, cross-polarized ratio, HH-VV phase difference, polarimetric coherence, and the volume scatter component of the Freeman decomposition. Evaluation of the selected models indicated that PolSAR data can be used to quantify stem volume in the study site with a root mean square error (RMSE) of about 20–29 m3 ha−1, corresponding to 8–12% of the mean stem volume. External validation using independent data showed average prediction errors of less than 14%. Saturation effects in measured versus modelled volume were not observed up to volumes of 308 m3 ha−1 (biomasses of ∼357 Mg ha−1). However, no formal assessment of saturation was possible due to limitations of the volume range of the dataset.