![CSK (blue crosses) and TDX (red squares) InSAR coherence versus GSV [m 3 /ha] for spruce stands (slopes below 4 °).](https://www.researchgate.net/profile/Christian-Thiel-8/publication/261209241/figure/fig1/AS:631950154928131@1527680066641/CSK-blue-crosses-and-TDX-red-squares-InSAR-coherence-versus-GSV-m-3-ha-for-spruce_Q320.jpg)
![CSK (blue crosses) and TDX (red squares) InSAR coherence versus GSV [m 3 /ha] for spruce stands (slopes below 4 °).](https://www.researchgate.net/profile/Christian-Thiel-8/publication/261209241/figure/fig2/AS:631950154936320@1527680066639/CSK-blue-crosses-and-TDX-red-squares-InSAR-coherence-versus-GSV-m-3-ha-for-spruce_Q320.jpg)
Content uploaded by Christian Thiel
Author content
All content in this area was uploaded by Christian Thiel on May 30, 2018
Content may be subject to copyright.
A preview of the PDF is not available
Cosmo-SkyMed backscatter intensity and interferometric coherence signatures over Germany's low mountain range forested areas
Abstract and Figures
In this paper, we present some investigations based on X-band interferometric coherence for retrieving forest Growing Stock Volume (GSV). Exanimating first temporal decorrelation, a comparison indicated total (γ=0.1-0.2), (γ=0.1-0.4) and low (γ=0.4-0.9) temporal decorrelation for TSX (11 days repeat pass), CSK (1 day repeat pass) and TDX (single pass) respectively. After studying also the spatial decorrelation, it could be pointed out that the increase of the perpendicular baseline enhanced the sensitivity to the volume of the forest canopies. CSK and TDX were further investigated in order to highlight their sensitivity to GSV. Volume decorrelation seemed to be dominant, as the coherence decreased with increasing vegetation. However, TDX seemed more suitable than CSK to retrieve GSV as it showed higher correlation (r2TDX=0.64, r2CSK=0.13).
Figures - uploaded by Christian Thiel
Author content
All figure content in this area was uploaded by Christian Thiel
Content may be subject to copyright.
... The potential of C-band multitemporal interferometric coherence observations for land-cover mapping became evident after the launch of ERS-1 in 1991. A large number of studies confirmed the high potential of repeat-pass coherence magnitude calculated from multitemporal InSAR data for biomass estimation, forest type classification, and clear-cut detection [30]- [32]. Authors in [31], [33]- [36] showed that C-band repeat-pass coherence acquired at short-term intervals shows a decline with increasing biomass. ...
There is a need for mapping of forest areas with young stands under regeneration in Norway, as a basis for conducting tending, or precommercial thinning (PCT), whenever necessary. The main objective of this paper to show the potential of multitemporal Sentinel-1 (S-1) and Sentinel-2 (S-2) data for characterization and detection of forest stands under regeneration. We identify the most powerful radar and optical features for discrimination of forest stands under regeneration versus other forest stands. A number of optical and radar features derived from multitemporal S-1 and S-2 data were used for the class separability and cross-correlation analysis. The analysis was performed on forest resource maps consisting of the forest development classes and age in two study sites from southeastern Norway. Important features were used to train the classical random forest (RF) classification algorithm. A comparative study of performance of the algorithm was used in three cases: I) using only S-1 features, II) using only S-2 optical bands, and III) using combination of S-1 and S-2 features. RF classification results pointed to increased class discrimination when using S-1 and S-2 data in relation to S-1 or S-2 data only. The study shows that forest stands under regeneration in the height interval for PCT can be detected with a detection rate of 91% and F-1 score of 73.2% in case III as most accurate, while tree density and broadleaf fraction could be estimated with coefficient of determination (R 2) of about 0.70 and 0.80, respectively.
... In this respect, Synthetic Aperture Radar (SAR) has proven to be a suitable instrument for forest investigations. The SAR capability of estimating several forest parameters, such as forest density, tree height, and forest biomass, was demonstrated in several studies (e.g., [2][3][4][5][6][7][8][9][10][11]). The SAR capability of observing forest parameters depends on the operating frequency, that is, at microwaves, the sensitivity of low frequencies (i.e., P-or L-band) to forest biomass was largely proven [12][13][14]. ...
This study aimed at evaluating the potential of machine learning (ML) for estimating forest biomass from polarimetric Synthetic Aperture Radar (SAR) data. Retrieval algorithms based on two different machine-learning methods, namely Artificial Neural Networks (ANNs) and Supported Vector Regressions (SVRs), were implemented and validated using the airborne polarimetric SAR data derived from the AfriSAR, BioSAR, and TropiSAR campaigns. These datasets, composed of polarimetric airborne SAR data at P-band and corresponding biomass values from in situ and LiDAR measurements, were made available by the European Space Agency (ESA) in the framework of the Biomass Retrieval Algorithm Inter-Comparison Exercise (BRIX). The sensitivity of the SAR measurements at all polarizations to the target biomass was evaluated on the entire set of data from all the campaigns, and separately on the dataset of each campaign. Based on the results of the sensitivity analysis, the retrieval was attempted by implementing general algorithms, using the entire dataset, and specific algorithms, using data of each campaign. Algorithm inputs are the SAR data and the corresponding local incidence angles, and output is the estimated biomass. To allow the comparison, both ANN and SVR were trained using the same subset of data, composed of 50% of the available dataset, and validated on the remaining part of the dataset. The validation of the algorithms demonstrated that both machine-learning methods were able to estimate the forest biomass with comparable accuracies. In detail, the validation of the general ANN algorithm resulted in a correlation coefficient R = 0.88, RMSE = 60 t/ha, and negligible BIAS, while the specific ANN for data obtained R from 0.78 to 0.94 and RMSE between 15 and 50 t/ha, depending on the dataset. Similarly, the general SVR was able to estimate the target parameter with R = 0.84, RMSE = 69 t/ha, and BIAS negligible, while the specific algorithms obtained 0.22 ≤ R ≤ 0.92 and 19 ≤ RMSE ≤ 70 (t/ha). The study also pointed out that the computational cost is similar for both methods. In this respect, the training is the only time-demanding part, while applying the trained algorithm to the validation set or to any other dataset occurs in near real time. As a final step of the study, the ANN and SVR algorithms were applied to the available SAR images for obtaining biomass maps from the available SAR images.
... At wavelengths shorter than 10 cm, i.e., 5.6 cm (C-band) and 3 cm (X-band), the backscatter is mainly due to leaves or needles and twigs of the upper canopy. This makes it possible to obtain information on the upper layer of vegetation cover, allowing the identification of forest/non-forest areas [5][6][7][8][9][10][11][12]. ...
In this paper, multifrequency synthetic aperture radar (SAR) images from ALOS/PALSAR, ENVISAT/ASAR and Cosmo-SkyMed sensors were studied for forest classification in a test area in Central Italy (San Rossore), where detailed in-situ measurements were available. A preliminary discrimination of the main land cover classes and forest types was carried out by exploiting the synergy among L-, C- and X-bands and different polarizations. SAR data were preliminarily inspected to assess the capabilities of discriminating forest from non-forest and separating broadleaf from coniferous forests. The temporal average backscattering coefficient (σ°) was computed for each sensor-polarization pair and labeled on a pixel basis according to the reference map. Several classification methods based on the machine learning framework were applied and validated considering different features, in order to highlight the contribution of bands and polarizations, as well as to assess the classifiers’ performance. The experimental results indicate that the different surface types are best identified by using all bands, followed by joint L- and X-bands. In the former case, the best overall average accuracy (83.1%) is achieved by random forest classification. Finally, the classification maps on class edges are discussed to highlight the misclassification errors.
... Moreover, the ability of SAR data in estimating several parameters including forest density, tree height and above ground biomass or stem volume was confirmed (e.g. Le Toan et al., 1989;Ulaby and Dobson, 1989;Wang et al. 1994 andFerrazzoli et al., 1997;Paloscia et al., 1999;Kasischke et al., 1997;Ackermann et al., 2012;Kaasalainen et al., 2015;Baghdadi et al., 2015). ...
The extraction of forest information from SAR images is particularly complex in Mediterranean areas, since they are characterized by high spatial fragmentation and heterogeneity. We have investigated the use of multi-frequency SAR data from different sensors (ALOS/PALSAR and ENVISAT/ASAR) for estimating forest biomass in two test areas in Central Italy (San Rossore and Molise), where detailed in-situ measurements and Airborne Laser Scanning (ALS) data were available. The study focused on the estimation of growing stock volume (GS, in m³/ha) by using an inversion algorithm based on artificial neural networks (ANN). The ANN algorithm was first appropriately trained using the available GS estimates obtained from ALS data. The potential of this algorithm was then improved through the innovative use of a simulated dataset, generated by a forward electromagnetic model based on the Radiative Transfer Theory (RTT). The algorithm is able to merge SAR data at L and C bands for predicting GS in diversified Mediterranean environments. The performed analyses indicated that GS was correctly estimated by integrating information from L and C bands on both test areas, with the following statistics: R > 0.97 and RMSE = 28.5 m³/ha for the independent test, and R = 0.86 and RMSE ≈ 77 m³/ha for the final independent validation, the latter performed on the forest stands of both areas not included in the ALS acquisitions and where conventional measurements were available. The research then illustrates the potential of using the obtained GS estimates from SAR data to drive the simulations of forest net primary production (NPP). This experiment produced spatially explicit estimates of GS current annual increments that are slightly less accurate than those obtained from ground observations (R = 0.75 and RMSE ≈ 1.5 m³/ha/year).
... Extension of forests, their thermal state (frozen/thawed), biomass density, tree height and woody volume can be derived from the analysis of the backscattering data produced by synthetic aperture radar (SAR) systems [e.g. Eriksson et al., 2005;Santoro et al., 2008;Ackermann et al., 2012], especially after the launch of new sensors characterized by high spatial resolution and very frequent revisit time. The most suitable frequency for estimating forest biomass (generally expressed as the Woody Volume, WV, in m 3 /ha) is P band due to the longer wavelength and consequent higher penetration in dense canopies. ...
This work aims at investigating the potential of L (ALOS/PALSAR) and C (ENVISAT/ ASAR) band SAR images in forest biomass monitoring and setting up a retrieval algorithm, based on Artificial Neural Networks (ANN), for estimating the Woody Volume (WV, in m 3 /ha) from combined satellite acquisitions. The investigation was carried out on two test areas in central Italy, where ground WV measurements were available. An innovative retrieval algorithm based on ANN was developed for estimating WV from L and C bands SAR data. The novelty consists of an accurate training of the ANN with several thousands of data, which allowed the implementation of a very robust algorithm. The RMSE values found on San Rossore area were ≅40 m 3 /ha (L band data only), and 25-30 m 3 /ha (L with C band). On Molise, by using combined data at L and C bands, RMSE<30m 3 / ha was obtained.
The application of remote sensing techniques in the retrieval of forest stand proprieties, especially biomass and stem volume, has been intensively investigated during the last few years. New methods using SAR and Optical systems have been pointed out. However, these instruments present intrinsic limitations which prevent operational biomass retrieval while using them separately. In order to overcome SAR and Optical system restrictions, with respect to biomass mapping, the present study will assess the potential of fusion of these two different sources of information. The Thuringian Forest located in the center of Germany has been chosen for the investigations. An extensive set of L-Band data (ALOS PALSAR) and X-Band data (TerraSAR-X) has been acquired in various sensor configurations (acquisition modes, polarizations, incident angles). High resolution optical data (RapidEye and Kompsat2) has been also acquired and a forest inventory with a great deal of forest stands has been delivered. The study will first concentrate on the pre-processing of the data. A data analysis will then be carried out and preliminary results will be discussed with a view to the fusion of SAR and Optical information. The fusion of SAR and Optical systems has the potential to give new possibilities in terms of temporal and spatial transferability. In this context, it will be expected to increase the biomass map accuracy, its spatial extension and its updated frequency.
The use of spaceborne synthetic aperture radar (SAR) systems to estimate stem volume and biomass in boreal forests has shown some promising results, but with saturation of the radar backscatter at relatively low stem volumes and limited accuracy of stem volume estimation. These limitations have motivated evaluation of more advanced methods, such as interferometry. The results presented in this study show that ERS interferometry, under favourable conditions, may be used to estimate stem volume at stand level with saturation level and accuracy useful for operational forestry management planning in boreal forests. Five interferograms were analysed, covering a test site located in the central part of Sweden with stem volume in the range of 0-305 m3 ha-1. The best interferogram showed a linear relationship between stem volume and coherence with a root mean square error (RMSE) of approximately 26 m3 ha-1, corresponding to 20% of the average stem volume, throughout the range of stem volume. No saturation was observed up to the maximum stem volume. However, the sensitivity of coherence to stem volume varied considerably between the interferograms. Finally, four SPOT XS images were evaluated and compared with the stem volume estimations obtained from the interferograms, resulting in a relative RMSE of about 24% of the stem volume, for the best case. The estimation of stem volume using coherence data was found to be better than optical data for stem volumes exceeding about 110 m3 ha-1. The statistical analysis was performed using linear regression models with cross-validation.
SAR sensors are being evaluated for application to global carbon
modeling as a means of acquiring biomass data for vegetated ecosystems.
While good correlations have been reported linking SAR backscatter to
biomass, biomass saturation limits, or the point where backscatter no
longer scales with biomass, have also been noted. This study examines
two SAR and biomass data sets of forests with different canopy
architectures for commonalties regarding backscatter/biomass saturation.
The SAR data were collected using the same instrument and incidence
angles. The effect of the saturation limits on the usefulness of SAR for
making global biomass inventories was assessed by comparing the biomass
limits to a global vegetation biomass map. The areal extent and total
global biomass represented by ecosystems that exceed the biomass
saturation level for each band are estimated
The authors have developed a SAR and an interferometric processor
suitable for all current SAR systems. The main modules of the processor
are the Modular SAR Processor (MSP), the Interferometric SAR Processor
(ISP), and the Differential Interferometry and Geocoding Module
(DIFF&GEO). The potential and limitations of the selected algorithms
and processing techniques are discussed. Typical results are shown
SAR surveys from separate passes show relative shifts of the
ground wavenumber spectra that depend on the local slope and the
off-nadir angle. The authors discuss the exploitation of this spectral
shift for different applications: 1) generation of “low
noise” interferograms benefiting phase unwrapping, 2) generation
of quick-look interferograms, 3) decorrelation reduction by means of
tunable SAR systems (TINSAR), 4) range resolution enhancement, and 5)
the combination of SAR data gathered by different platforms (airborne
and satellite) for a “long-time coherence” study
A radar interferometric technique for topographic mapping of
surfaces, implemented utilizing a single synthetic aperture radar (SAR)
system in a nearly repeating orbit, is discussed. The authors
characterize the various sources contributing to the echo correlation
statistics, and isolate the term which most closely describes surficial
change. They then examine the application of this approach to
topographic mapping of vegetated surfaces which may be expected to
possess varying backscatter over time. It is found that there is
decorrelation increasing with time but that digital terrain model
generation remains feasible. The authors present such a map of a
forested area in Oregon which also includes some nearly unvegetated lava
flows. Such a technique could provide a global digital terrain map
The authors present the results of an experiment defined to
demonstrate the use of radar to retrieve forest biomass. The SAR data
were acquired by the NASA/JPL SAR over the Landes pine forest during the
1989 MAESTRO-1 campaign. The SAR data, after calibration, were analyzed
together with ground data collected on forest stands from a young stage
(eight years) to a mature stage (46 years). The dynamic range of the
radar backscatter intensity from forest was found to be greatest at
P -band and decreased with increasing frequencies.
Cross-polarized backscatter intensity yielded the best sensitivities to
variations of forest biomass. L -band data confirmed past
results on good correlation with forest parameters. The most striking
observation was the strong correlation of P -band backscatter
intensity to forest biomass
Potential of fusion of SAR and Optical Satellite imagery for biomass estimation in temperate forested areas
- Schmullius
Schmullius, "Potential of fusion of SAR and Optical
Satellite imagery for biomass estimation in temperate
forested areas," in Proceedings of the ESA Living Planet
Symposium, Bergen, 2010.