HH VV Research Articles

Synergistic Use of TanDEM-X and Landsat-8 Data for Crop-Type Classification and Monitoring

Classification of crop types using Earth Observation (EO) data is a challenging task. The challenge increases many folds when we have diverse crops within a resolution cell. In this regard, optical and Synthetic Aperture Radar (SAR) data provide complementary information to characterize a target. Therefore, we propose to leverage the synergy between multispectral and Synthetic Aperture Radar (SAR) data for crop classification. We aim to use the newly developed model-free three-component scattering power components to quantify changes in scattering mechanisms at different phenological stages. By incorporating interferometric coherence information, we consider the morphological characteristics of the crops that are not available with only polarimetric information. We also utilize the reflectance values from Landsat-8 spectral bands as complementary biochemical information of crops. The classification accuracy is enhanced by using these two pieces of information combined using a neural network-based architecture with an attention mechanism. We utilize the time series dual co-polarimetric (i.e., HH–VV) TanDEM-X SAR data and the multispectral Landsat-8 data acquired over an agricultural area in Seville, Spain. The use of the proposed attention mechanism for fusing SAR and optical data shows a significant improvement in classification accuracy by 6.0% to 9.0% as compared to the sole use of either the optical or SAR data. Besides, we also demonstrate that the utilization of single-pass interferometric coherence maps in the fusion framework enhances the overall classification accuracy by $\approx$ 3.0%. Therefore, the proposed synergistic approach will facilitate accurate and robust crop mapping with high-resolution EO data at larger scales.

A New Methodology for Rice Area Monitoring With COSMO-SkyMed HH–VV PingPong Mode SAR Data

In this paper, a novel approach is proposed to exploit a time series of COSMO-SkyMed (CSK) HH-VV SAR images to map rice fields and to estimate the sowing dates. The approach relies on multi-polarization features, i.e., the squared modulus of the HH and VV channels and the polarization ratio, extracted from CSK SAR scenes. The key step consists of extracting a rice training signature related to the multipolarization features. This signature allows estimating the sowing date that, at once, is used to refine the rice map obtained by the conventional interpretation of the CSK time series in terms of the scattering mechanisms of the different growing cycles. Experiments, carried out on a time series of 32 CSK images, collected from the Mekong Delta region, South Vietnam, confirm the soundness of the proposed methodology which is shown to provide results comparable to the ones obtained by a literature approach that exploits a similar dataset.

Pol-InSAR for Forest Biomass Estimation with the Transformation of the Polarization Basis

An improved method that uses polarimetric and Pol-InSAR information at the L-band and P-band was proposed in this study for estimating forest biomass. In the first phase, various polarimetric and polarimetric interferometry indicators were extracted via transformation of the polarization basis. In the second phase, the particle swarm optimization method is applied to identify optimal parameter values for biomass estimation. According to the results, the correlations between the polarimetric indicators and the biomass were stronger at the P-band compared to the L-band. Polarimetric indicators that involve HV and HH–VV show the maximum correlation with biomass prior to optimization. It is demonstrated that changing the polarization basis can significantly improve the correlations of estimators with the biomass, especially at the P-band. The globally optimal estimators involve the same scattering mechanisms (volume and double-bounce scattering), and the optimal polarization bases differ slightly among most of the cases, due to the adaptation to the forest natural geometry and/or the acquisition configuration. Regarding Pol-InSAR, several tree height retrieval estimators that are based on various assumptions have been analyzed under multiple polarization basis rotations. According to the results, the RVoG phase method, which represents the forest as a low-extinction structured volume, yielded the best accuracy, with R = 0.73 and RMSE = 4.30 m. The identification of the optimum polarization indicators via binary PSO could improve the biomass estimation accuracy by 2% and 6% at the P-band and L-band, respectively. It is demonstrated that such an approach can be employed to accurately estimate biomass via extrapolation of in situ measurements over an entire region.

Separating the Influence of Vegetation Changes in Polarimetric Differential SAR Interferometry

Soil moisture and wet biomass changes between two noninstantaneous SAR observations markedly affect the displacement estimates obtainable with Differential Interferometric Synthetic Aperture Radar (DInSAR). The separation, the modeling of these influences besides their uncoupling from the displacement signal, and the atmospheric disturbances are still unsolved issues for several repeat-pass interferometric applications. This paper focuses on the separation of vegetation changes from the other phase contributions affecting repeat-pass measurements over vegetated areas. These phase terms mainly relate to changes in soil moisture, atmospheric delays, and surface deformation. The separation is achieved with a first-order scattering solution decomposing the observed HH and VV DInSAR phases in the sum of several phase terms. The latter mainly consider the changes in soil surface scattering and in the two-way propagation through a vertically oriented vegetation canopy. No assumption is made on the spatiotemporal evolution of the displacement and atmosphere. The overall approach is tested on a L-band data set acquired over an agricultural area. Upon calibration, the model allows for estimating changes in wet biomass based on the nonzero HH–VV DInSAR phase difference observed over several birefringent agricultural fields. The obtained biomass estimates provide then a correction for the effect of vegetation changes on the observed HH and VV DInSAR phases. Deprived of the vegetation contribution, the remainder phase terms can be more easily explored for further analyses, e.g., the estimation of soil moisture changes and/or surface movements in vertically oriented vegetated areas.

Evaluation of different orientation angle distributions within the X-Bragg scattering model for bare soil moisture estimation

ABSTRACTIn this paper, the applicability of three different orientation angle distributions of surface facets within the extended Bragg (X-Bragg) scattering model is investigated for estimation of soil moisture over bare surfaces using both Eigen-based and model-based polarimetric synthetic aperture radar (PolSAR) decomposition techniques. The three distributions considered for investigation in the X-Bragg model are uniform, half cosine, and the Lee distributions. In order to understand the sensitivity of the model using the three orientation angle distributions, key polarimetric parameters, such as scattering entropy (H), scattering anisotropy (A), scattering mechanism (α), cross-pol power (T33), linear T12 coherence (|γ(HH+VV)(HH–VV)|), are simulated and analysed for various widths of distributions. The analysis of the simulated polarimetric parameters show that the Lee distribution has a reduced roughness validity range compared with the uniform and half cosine distributions. DLR E-SAR L-band data from the AgriSAR’2006 campaign over the Demmin test site in Northern Germany are inverted for soil moisture over bare surfaces. The inverted soil moisture from the physics-based X-Bragg model is compared with in situ measured TDR (time domain reflectometry) soil moisture values. The inversion results using the Eigen-based decomposition reveal similar root mean square error (RMSE = 14 vol.%) and inversion rates for three distributions. The model-based decomposition inversion results obtained at various fixed widths of distributions reveal that the Lee distribution shows less RMSE of 8 vol.% and high inversion rates for moderate surface roughness (ks = 0.5) as compared with half cosine and uniform distributions.

Time series analysis of co-polarization phase difference (PPD) for winter field crops using polarimetric C-band SAR data

ABSTRACTThe utility of time series polarimetric C-band data for vegetation state monitoring was explored to understand the mechanism of growth and phenology for important winter crops in India. Parameters investigated were HH–VV phase difference (co-polarization phase difference, PPD), amplitude ratio, and polarization indices. Data were acquired during the entire growth phase categorized as early, mid/peak vegetative, and post-vegetative /flowering phase. The trend emerging in this study showed a shift in the phase difference distribution for agricultural areas relating to the growth rate for various crops. The time series data set revealed that the PPD is a function of frequency and was directly affected by crop type (planophile or erectophile), vigour, structure, and crop biophysical parameters, particularly biomass. The behaviour of crop biomass with PPD responded differentially across crop architectures and vigour classes. Co-polarization index was found to be a good measure for discrimination in early growth stages while cross-polarization index suited in advanced vegetative stages where geometrical orientation was uniform. The PPD captured the change in frequency distribution resulting in a peaked distribution at sowing changing to smooth, well-spread frequency distribution as the peak vegetation stage approaches. This histogram nature is observed to be gradual for high-biomass crops and peaked in case of the low-biomass crops. It is indicative of the rate of growth; a low peaked normal curve indicates faster growth rate and resulting in high biomass. The amplitude ratio in the later phase of growth as on the third date is similarly altered as in the VV returns from the crop. Intuition of the rate of growth and plant vigour is obtained from the temporal PPD pattern. The current study shows that while phase differences and amplitude ratio carry little information content on a single resolution cell basis, their spatial distribution over a wider time span can be used to derive quantitative relationships between SAR response and crop condition. The synergy of information involving the above parameters were used to derive useful information on the vegetation.

Investigation into Different Polarimetric Features for Sea Ice Classification Using X-Band Synthetic Aperture Radar

Satellite-borne synthetic aperture radar has proven to be a valuable tool for sea ice monitoring for more than two decades. In this study, we examine the performance of an automated sea ice classification algorithm based on polarimetric TerraSAR-X images. In the first step of our approach, we extract 12 polarimetric features from HH–VV dualpol StripMap images. In a second step, we train an artificial neural network, and then, feed the feature vectors into the trained neural network to classify each pixel into an ice type. The first part of our analysis addresses the predictive value of different subsets of features for our classification process (by means of measuring mutual information). Some polarimetric features such as polarimetric span and geometric intensity are proven to be more useful than eigenvalue decomposition based features. The classification is based on and validated by in situ data acquired during the N-ICE2015 field campaign. The results on a TerraSAR-X dataset indicate a high reliability of a neural network classifier based on polarimetric features. Performance speed and accuracy promise applicability for near real-time operational use.

Effects of Inundated Vegetation on X-Band HH–VV Backscatter and Phase Difference

TerraSAR-X HH and HH-VV polarization data of a wetland in Estonia are used to study the effects of HH-VV phase difference on detecting inundated vegetation. Patches with different tree species composition are selected from two images representing flooded and unflooded conditions. Contribution of the double bounce mechanism is studied via phase difference and backscatter analysis. An increase in both HH and HH-VV backscatter was observed for all study cases in flooded conditions, as compared to unflooded periods. Over the different patches, HH backscatter increased from 2.3 dB to 8.2 dB, while for HH-VV - from 3 dB to 9.8 dB. The HH-VV channel allowed for a better separation between flooded and un flooded forest than HH, with the difference between the HH-VV and HH backscatter increase in inundated conditions ranging from 0.2 dB for coniferous stands with a tree height of less than 10 m to 1.6 dB for deciduous stands with a tree height of less than 10 m. A larger separation on both channels was in general observed in deciduous stands as the leaf-off season allowed for a better penetration depth. A considerable phase shift between HH and VV channels ranging from 10° to 28° was observed due to inundation, suggesting a strong contribution from the double bounce mechanism. For unflooded patches, the difference in backscatter between the two acquisition dates ranged from -0.8 dB to 1 dB for HH, and from 0 dB to 2 dB for HH-VV channel. Phase shift for unflooded patches varied from 5° to 8°.

Detecting subcanopy invasive plant species in tropical rainforest by integrating optical and microwave (InSAR/PolInSAR) remote sensing data, and a decision tree algorithm

In this paper, we propose a decision tree algorithm to characterize spatial extent and spectral features of invasive plant species (i.e., guava, Madagascar cardamom, and Molucca raspberry) in tropical rainforests by integrating datasets from passive and active remote sensing sensors. The decision tree algorithm is based on a number of input variables including matching score and infeasibility images from Mixture Tuned Matched Filtering (MTMF), land-cover maps, tree height information derived from high resolution stereo imagery, polarimetric feature images, Radar Forest Degradation Index (RFDI), polarimetric and InSAR coherence and phase difference images. Spatial distributions of the study organisms are mapped using pixel-based Winner-Takes-All (WTA) algorithm, object oriented feature extraction, spectral unmixing, and compared with the newly developed decision tree approach. Our results show that the InSAR phase difference and PolInSAR HH–VV coherence images of L-band PALSAR data are the most important variables following the MTMF outputs in mapping subcanopy invasive plant species in tropical rainforest. We also show that the three types of invasive plants alone occupy about 17.6% of the Betampona Nature Reserve (BNR) while mixed forest, shrubland and grassland areas are summed to 11.9% of the reserve. This work presents the first systematic attempt to evaluate forest degradation, habitat quality and invasive plant statistics in the BNR, and provides significant insights as to management strategies for the control of invasive plants and conversation in the reserve.

Urban change detection with polarimetric Advanced Land Observing Satellite phased array type L-band synthetic aperture radar data: a case study of Tai’an, China

Change detection in Tai’an city of eastern China using a pair of qual-polarimetric Advanced Land Observing Satellite phased array type L-band synthetic aperture radar (ALOS PALSAR) data was studied. The procedures consisted of polarimetric features extraction, optimal polarimetric feature group selection, supervised classification, and result accuracy assessment. Feature extraction from PALSAR data was performed first, and then the polarimetric features were categorized into several groups. Polarimetric optimum index factor (POIF) and distance factor (DF) were selected to measure and evaluate the suitability of each feature group for urban change detection. The best group of features was identified including linear polarization correlation coefficient ( ρ HH–VV ), right–left (R–L) circular polarization correlation coefficient ( ρ RR–LL ), total power (TP), and cross-polarization isolation (XPI). Afterward, four difference images of the identified features extracted from the two PALSAR data were derived, respectively. Then, the random forest (RF) classifier was employed to perform a supervised classification of the four difference images. Three classes were quantified, including no-change, change from undeveloped area to developed area, and vice versa. The overall accuracy of change detection was about 84% and Cohen’s Kappa coefficient was 0.71. Consequently, satisfactory outcomes were obtained in the application of the polarimetric ALOS PALSAR data of moderate resolution in detecting urban land use and land cover type changes.

Ground-based C-band tomographic profiling of a conifer forest stand

We provide a demonstration of the new tomographic profiling (TP) technique, here applied to forestry for the first time. The portable ground-based synthetic aperture radar (GB-SAR) system was used to capture profiles of the vertical polarimetric backscattering patterns through a ∼7 m tall stand of Norway spruce trees. The TP scheme collects data as for normal SAR imaging, but with the antennae aligned in the along-track direction. Adaptive post-processing meant that each TP scan simultaneously captured along-track image transects over the incidence angle range 0°–60°. An important feature of the derived image products is that incidence angle is constant at every point within an image. The measured HH–VV height backscatter profiles were very similar, whilst the cross-/co-polarization ratio showed very little variation with height through the stand. Backscattering profiles showed closest agreement with the branch biomass distribution through the canopy, rather than with trunk or branch + trunk biomasses. Equivalent interferometric tree heights were estimated from the centre of mass of the backscatter-height distribution, which displayed increasing height with increasing incidence angle. There was no significant vertical separation between the cross- and co-polarization returns.

Wind Speed Estimation Using Polarimetric RADARSAT-2 Images: Finding the Best Polarization and Polarization Ratio

As the number of operational wind scatterometers is getting smaller, other sources of spaceborne sensors are now included in global wind mapping. One of the prominent sensors is the Synthetic Aperture Radar (SAR). Besides serving as a generic scatterometer, SAR systems are the only type of radar systems that can provide sub-km resolution sea surface wind data and offers near shore mapping capability. This unique feature is important for assessing the offshore wind resources. As an important source of renewable energy, offshore wind farms are growing rapidly. Furthermore, recent research shows that the cross-polarization radar backscatter does not seem to saturate in high winds, and provides an excellent supplement for scatterometer wind sensing in storm conditions. The saturation issues of co-polarization radar returns have so far made it difficult to resolve wind speeds beyond roughly 20 m/s, or even less for lower incidence angles. The scope of this paper is to show the potential of RADARSAT-2's polarimetric modes for wind speed retrieval. RADARSAT-2 is the first operational fully polarimetric (HH VV HV VH) C-band satellite. Standard Quad-pol images have been collected in the St. Lawrence Gulf and compared against the Mont-Louis buoy and QuikSCAT scatterometer data. Co-polarization wind speeds were computed with CMOD-5 algorithms. A few polarization ratios were tested to determine the most suitable one for RADARSAT-2's HH polarization mode. For Cross-polarization, two different models were compared. Cross-polarization gives excellent results when wind exceeds 5 m/s. In general, SAR wind retrieval is suitable for resolution of 400 m.

Pine forest investigation using high resolution P-band Pol-InSAR data

A Pol-InSAR dataset acquired at P-band over a maritime pine forest (Nezer forest of Landes, France) has been investigated and compared with tree structural ground measurements in order to identify scatterers associated with elementary scattering mechanisms in the canopy layer. Since this managed forest is constituted by different height stands, it allows to investigate mechanism change in the vertical direction during the forest growth. It has been observed that VV channel was the most elevated, which can be accounted for a maximum attenuation in the vertical direction and partially for the presence of sub-vertical branches in the top canopy layer, and that this channel was the most sensitive to the tree density in reason of an important ground reflectivity. The most pure double-bounce scattering mechanism component is received in HH–VV channel where the corresponding phase center reaches the ground level for large Diameter at Breast Height. It is also supposed that HH channel, which receives double bounce backscatter, was affected by the sub-horizontal primary branch development, which induces an elevation of the phase center over oldest trees while tree density diminution enhances backscatter from the ground level. Finally, the crosspolar channel appeared the less perturbed by the tree density than VV channel since it receives predominantly the canopy contribution.

AIRSAR Studies of Woody Shrub Density in Semiarid Rangeland: Jornada del Muerto, New Mexico

This study evaluates the use of polarimetric Air borne S ynthetic A perture R adar (AIRSAR) data to assess woody shrub density in a semiarid site where the vegetation consists primarily of varied mixtures of herbaceous vegetation and shrubs. AIRSAR data and field observations of vegetation cover and growth form-composition were obtained for 59 sites in the Jornada del Muerto plain in southern New Mexico. Radar signature measures examined were C-, L- and P-band backscattering coefficients (σ 0) for HH, HV and VV polarizations, ratios of σ 0HH and σ 0HV to σ 0VV, and the HH–VV polarization phase difference and correlation coefficient. The most effective measure for estimation of shrub density was L-band σ 0HV, which distinguished among shrub density classes with no misclassification. Sensitivity of this measure to small amounts of shrub cover was indicated by successful separation of sites with <1% shrub cover from sites with 1–5% cover. Separability of shrub density classes was generally least for C-band signature measures. A distinctive radar signature was exhibited by dense stands of Yucca elata, a semitreelike plant with uniformly thick (≈10 cm diameter) fibrous stems. Yucca sites were distinguished from others by their high P-band σ 0HV relative to L-band σ 0HV. The results are largely explained by the greater sensitivity of longer wavelengths to larger canopy structural elements. L-band σ 0HV and other measures responsive to canopy volume scattering were more strongly related to shrub than to herbaceous plant cover because woody shrub canopies have numerous stems of the intermediate size to which L-band is most sensitive, whereas stems of this size are mostly lacking in herbaceous canopies. The uniform-diameter stems of yucca have larger dimensions to which P-band is more sensitive than L-band.