A new algorithm for forest height estimation based on the varied extinction random volume over ground (VERVoG) model using PolInSAR data

This paper proposes a new method for forest height estimation using single-baseline single frequency polarimetric synthetic aperture radar interferometry (PolInSAR) data. The new algorithm estimates the forest height based on the random volume over the ground with a volume temporal decorrelation (RVoG+VTD) model. We approach the problem using a four-stage geometrical method without the need for any prior information. In order to decrease the number of unknown parameters in the RVoG+VTD model, the mean extinction coefficient is estimated in an independent procedure. In this respect, the suggested algorithm estimates the mean extinction coefficient as a function of a geometrical index based on the signal penetration in the volume layer. As a result, the proposed four-stage algorithm can be used for forest height estimation using the repeat pass PolInSAR data, affected by temporal decorrelation, without the need for any auxiliary data. The suggested algorithm was applied to the PolInSAR data of the European Space Agency (ESA), BioSAR 2007 campaign. For the performance analysis of the proposed approach, repeat pass experimental SAR (ESAR) L-band data, acquired over the Remningstorp test site in Southern Sweden, is employed. The experimental result shows that the four-stage method estimates the volume height with an average root mean square error (RMSE) of 2.47 m against LiDAR heights. It presents a significant improvement of forest height accuracy, i.e., 5.42 m, compared to the three-stage method result, which ignores the temporal decorrelation effect.

Polarimetric synthetic aperture radar interferometry (PolInSAR) and light detection and ranging (LiDAR) have their own respective advantages and disadvantages in extracting large-scale forest height. In this letter, we present an advanced approach to obtain forest canopy height by combining these two strategies. More specifically, the novelty of the proposed method focuses on a dual-baseline selection from multibaseline PolInSAR data, which ensures the robust performance of the forest height inversion by effectively improving the estimation of volume-only coherence. The dual-baseline selection can be regarded as a supervised classification problem. We consider support vector machine (SVM) as an appropriate classifier, and a small amount of sparse LiDAR samples within the coverage of the PolInSAR data (less than 1%) are chosen to assist with the training of the dual-baseline combination classification, which can be met by the current spaceborne LiDAR missions. Finally, we validate the proposed approach by airborne P-band synthetic aperture radar (SAR) data acquired by the F-SAR system and LiDAR data acquired by the National Aeronautics and Space Administration (NASA) Land, Vegetation, and Ice Sensor (LVIS) during the 2016 AfriSAR campaign. The estimation accuracy of the proposed method [ $R^{2} = 0.73$ , root-mean-square error (RMSE) = 3.17 m] is 25.24% higher than that of the existing SVM fusion approach devoted to single-baseline selection ( $R^{2} = 0.59$ , RMSE = 4.24 m).

PolGAN: A deep-learning-based unsupervised forest height estimation based on the synergy of PolInSAR and LiDAR data

A new algorithm for forest height estimation based on dual polarimetric interferometric SAR data is presented in this study. The main objective is to consider the efficiency of the dual-polarization data compared to the full polarimetric images with respect to forest height retrieval. Accordingly, the forest height estimation based on the random volume over the ground model is examined using a geometrical procedure named the three-stage method. An exhaustive search polarization optimization technique is also applied to improve the results by employing the efficiency of all the polarization bases based on the four-dimensional lexicographic PolInSAR vector. The repeat-pass experimental SAR (ESAR) images, which include both L- and P-band full polarimetric data, are employed for the accuracy assessment of the dual PolInSAR data and the newly proposed method for forest height estimation. The experimental results on the L-band PolInSAR data show the ability of the dual PolInSAR data for forest height estimation with an average root mean square error (RMSE) of 4.97 m against Lidar data based on the conventional three-stage method. Additionally, the proposed method results in an accuracy of 2.95 m for forest height estimation, indicating its high potential for tree height retrieval.

Forest height estimation by means of compact PolInSAR data

ABSTRACTMonitoring the earth’s biosphere is an essential task to understand the global dynamics of ecosystems, biodiversity, and management aspects. Forests, as a natural resource, have an important role to control the climate changes and the carbon cycle. For this reason, biomass and consequently forest height are known as the key information for monitoring the forest and its underlying surface. Several studies have shown that Synthetic Aperture Radar (SAR) imaging systems can provide an appropriate solution to estimate the biomass and the forest height. In this framework, Polarimetric SAR Interferometry (PolInSAR) technique is an effective tool for forest height estimation, due to its sensitivity to location and vertical distribution of the forest structural components. From one point of view, the employed methods are either based on model-based decomposition techniques or inversion models. In this paper, a method based on the combination of two categories has been proposed. Indeed, introducing a new way of combining the two categories for forest height estimation is the novel contribution of this study. The main motivation is to find directly and simultaneity the volume only and ground only complex coherences using the PolInSAR decomposition technique without the need to any a priori information for improving the forest height estimation procedure in the inversion models such as Random Volume over Ground (RVoG) model. The efficiency of the proposed approach was demonstrated by the E-SAR L-band single baseline PolInSAR data over the Remningstorp test site, in southern Sweden. Moreover, Light Detection and Ranging (LiDAR) data were used to evaluate the results. The experimental results showed that the proposed method improved the forest height estimation by 6.86 m.

This paper investigates the potentials and limitations of a simple dual-baseline PolInSAR (DBPI) method for forest height inversion. This DBPI method follows the classical three-stage inversion method’s idea used in single baseline PolInSAR (SBPI) inversion, but it avoids the assumption of the smallest ground-to-volume amplitude ratio (GVR) by employing an additional baseline to constrain the inversion procedure. In this paper, we present for the first time an assessment of such a method on real PolInSAR data over boreal forest. Additionally, we propose an improvement on the original DBPI method by incorporating the sloped random volume over ground (S-RVoG) model in order to reduce the range terrain slope effect. Therefore, a digital elevation model (DEM) is needed to provide the slope information in the proposed method. Three scenes of P-band airborne PolInSAR data acquired by E-SAR and light detection and ranging (LIDAR) data available in the BioSAR2008 campaign are employed for testing purposes. The performance of the SBPI, DBPI, and modified DBPI methods is compared. The results show that the DBPI method extracts forest heights with an average root mean square error (RMSE) of 4.72 m against LIDAR heights for trees of 18 m height on average. It presents a significant improvement of forest height accuracy over the SBPI method (with a stand-level mean improvement of 42.86%). Concerning the modified DBPI method, it consistently improves the accuracy of forest height inversion over sloped areas. This improvement reaches a stand-level mean of 21.72% improvement (with a mean RMSE of 4.63 m) for slopes greater than 10°.

Forest height is one of the most important forest biophysical parameters influencing light competition, stand productivity and carbon sequestration. This thesis considered investigating the existing and new methods for estimating forest height. The specific focus of the research project concerned the improved estimation of forest height (h) over tropical forest from both single- and multi-source remotely sensed datasets. The potential applications of single-baseline (B) PolInSAR, fusion of multi-baseline PolInSAR and LiDAR, and a combined use of polarimetric SAR and LiDAR were assessed. The single-baseline PolInSAR demonstrated the limitation of the model for estimation of forest height over a heterogeneous forest because of the lack of appropriate values of the vertical wavenumber. For the smallest baseline (B = 20m), height is overestimated (~ 5m) for short stand (h < 10m); and for the largest baseline (B = 120m), height is underestimated (~ 30m) for tall stand (20 < h < 60m). As conventional PolInSAR, small baselines are suitable for estimation of larger stand heights, whereas large baselines are for shorter stands. Therefore, for an improved result over a heterogeneous forest with different height ranges (0-60m), a multi-baseline merging approach was introduced based on a fusion of PolInSAR with LiDAR. The results demonstrated improvement (r2 = 0.81, RMSE = 7.1m) in comparison to PolInSAR alone (r2 = 0.67, RMSE = 9.2m). Whilst this fusion indicated improvement, the model performance is limited by the availability of the PolInSAR data. Therefore, a second data fusion approach was introduced using polarimetric SAR and LiDAR. The obtained results are significant (r2 = 0.70, RMSE = 10m) as the technique relies on polarimetric measurements and not interferometric. Overall, the research results demonstrate the capabilities of SAR and LiDAR data fusions in estimation of tropical forest height, adding value to both scientific understanding and management of forest ecosystems.

This paper addresses an algorithm for forest height estimation using single frequency and single baseline dual polarization radar interferometry data. The proposed method is based on a physical two layer volume over ground model and is represented by using polarimetric synthetic aperture radar interferometry (PolInSAR) technique. The presented algorithm provides the opportunity to take advantages of the dual polarimetric data, i.e, better spatial resolution and wider swath width, in comparison with the full polarimetric data, in forest height estimation application. In this research, a polarimetric optimization method is utilized to choose the optimum volume polarization basis in order to improve the results of the three-stage inversion algorithm. For the performance analysis of the proposed approach, the L-band ESAR data of the European Space Agency from BioSAR 2007 campaign (ESA) which is acquired over the Remningstorp test site in southern Sweden, is employed. The experimental result shows the dual PolInSAR HH/HV data capability in the forest height estimation without decreasing the accuracy of the result compared with the full polarimetric data. The suggested method leads to the average root mean square error (RMSE) of 4.39 m and the determination of coefficient of 0.66 in the forest height estimation in 15 predetermined stands in comparison with the LiDAR reference heights.

For forest with complex structure, the vertical structure backscatter is influenced by a combination of factors, including the frequencies of the radar waves and the forest biophysical parameters (i.e., density, species). The backscatter-induced error is thus a critical element in limiting the accuracy of polarimetric synthetic aperture radar interferometry (Pol-InSAR) forest height inversion based on a single model. It is, therefore, necessary to select the optimal backscatter profile from the multiple possible solutions in each pixel of the test site. In this letter, the impacts of backscatter in forest height estimation based on the models of random volume over ground (RVoG) ( $\sigma >0$ ), RVoG ( $\sigma ), and Gaussian vertical backscatter (GVB) were investigated in the complex plane, and then with the combined use of Pol-InSAR and light detection and ranging (LiDAR), a random forest (RF) classifier is trained to obtain the optimal backscatter function and Pol-InSAR forest height from the results based on the different models in each pixel. The proposed method was tested with single- and multi baseline Pol-InSAR data in the P-band, and the root-mean-square errors (RMSEs) of the proposed approach were 2.85 and 2.69 m, respectively, which represented average improvements of 20.6% and 17.7% over the optimal single-model inversion.

A slope three-layer scattering model (STSM) for retrieving forest height in mountain forest region using L-band polarimetric synthetic aperture radar interferometry (PolInSAR) data is proposed in this paper. The proposed model separates the vertical structure of forest into three layers: canopy, tree trunk, and ground layer, which account for the effect of topography for forest height calculation in a sloping forest area. Compared to the conventional two-layer random volume over ground model, the STSM improves substantial for modeling of actual mountain forest, allowing better understanding of microwave scattering process in sloping forest area. The STSM not only enables the accuracy improvement of the forest height estimation in sloping forest area but also provides the potential to isolate more accurately the direct scattering, double-bounce ground trunk interaction, and volume contribution, which usually cannot be achieved in the previous forest height estimation methods. The STSM performance is evaluated with simulated data from PolSARProSim software and ALOS/PALSAR L-band spaceborne PolInSAR data over the Kalimantan areas, Indonesia. The experimental results indicate that forest height could be effectively extracted by the proposed STSM.

Monitoring forest height is crucial to determine the structure and biodiversity of forest ecosystems. However, detailed spatial patterns of forest height from 30 m resolution remotely sensed data are currently unavailable. In this study, we present a new method for mapping forest height by combining spaceborne Light Detection and Ranging (LiDAR) with imagery from multiple remote sensing sources, including the Landsat 5 Thematic Mapper (TM), the Phased Array L-band Synthetic Aperture Radars (PALSAR), and topographic data. The nationwide forest heights agree well with results obtained from 525 independent forest height field measurements, yielding correlation coefficient, root mean square error (RMSE), and mean absolute error (MAE) values of 0.92, 4.31 m, and 3.87 m, respectively. Forest heights derived from remotely sensed data range from 1.41 m to 38.94 m, with an average forest height of 16.08 ± 3.34 m. Mean forest heights differ only slightly among different forest types. In natural forests, conifer forests have the greatest mean forest heights, whereas in plantations, bamboo forests have the greatest mean forest heights. Important predictors for modeling forest height using the random forest regression tree method include slope, surface reflectance of red bands and HV backscatter. The uncertainty caused by the uneven distribution of Geoscience Laser Altimeter System (GLAS) footprints is estimated to be 0.64 m. After integrating PALSAR data into the model, the uncertainty associated with forest height estimation was reduced by 4.58%. Our finer-resolution forest height could serve as a benchmark to estimate forest carbon storage and would greatly contribute to better understanding the roles of ecological engineering projects in China.

Continuous and extensive monitoring of forest height is essential for estimating forest above-ground biomass and predicting the ability of forests to absorb CO2. In particular, forest height at the national scale is an important indicator reflecting the national forestry economic construction, environmental governance, and ecological balance. However, the lack of inventory data restricts large-scale monitoring of forest height to some extent. Conducting manual surveys of forest height for large-scale areas would be labor-intensive and time-consuming. The successful launch of the new generation of spaceborne light detection and ranging (LiDAR) (The Ice, Cloud, and Land Elevation Satellite-2/the Advanced Topographic Laser Altimeter System, ICESat-2/ATLAS) has brought new opportunities for national-scale forestry resource surveys. This paper explores a method to survey national forest canopy height from the new generation of ICESat-2/ATLAS data. In view of the sparse sampling and little overlap between repeated spaceborne LiDAR data, a strategy for assessing the overall change of canopy height for large scales is provided. Some spatially continuous ancillary data were used to assist ICESat-2/ATLAS data to generate a wall-to-wall (spatially continuous) forest canopy height map in China by using the machine learning approach and then quantifying the analysis of forest canopy height in various provinces. The results show that there is a good correlation between the model forest height and the verification data, with a root mean squared error (RMSE) of 3.30 m and a coefficient of determination (R2) of 0.87. This indicates that the method for retrieving national forest canopy height is reliable. There are some limitations in areas with lower vegetation coverage or complex topography which need additional filtering or terrain correction to achieve higher accuracy in measuring forest canopy height. Our analysis suggests that ICESat-2/ATLAS data can achieve the retrieval of national forest height at an overall level, and it would be feasible to use ICESAT-2/ATLAS products to estimate forest canopy height change for large-scale areas.

Canopy height serves as a good indicator of forest carbon content. Remote sensing-based direct estimations of canopy height are usually based on Light Detection and Ranging (LiDAR) or Synthetic Aperture Radar (SAR) interferometric data. LiDAR data is scarcely available for the Indian tropics, while Interferometric SAR data from commercial satellites are costly. High temporal decorrelation makes freely available Sentinel-1 interferometric data mostly unsuitable for tropical forests. Alternatively, other remote sensing and biophysical parameters have shown good correlation with forest canopy height. The study objective was to establish and validate a methodology by which forest canopy height can be estimated from SAR and optical remote sensing data using machine learning models i.e., Random Forest (RF) and Symbolic Regression (SR). Here, we analysed the potential of Sentinel-1 interferometric coherence and Sentinel-2 biophysical parameters to propose a new method for estimating canopy height in the study site of the Bhitarkanika wildlife sanctuary, which has mangrove forests. The results showed that interferometric coherence, and biophysical variables (Leaf Area Index (LAI) and Fraction of Vegetation Cover (FVC)) have reasonable correlation with canopy height. The RF model showed a Root Mean Squared Error (RMSE) of 1.57 m and R2 value of 0.60 between observed and predicted canopy heights; whereas, the SR model through genetic programming demonstrated better RMSE and R2 values of 1.48 and 0.62 m, respectively. The SR also established an interpretable model, which is not possible via any other machine learning algorithms. The FVC was found to be an essential variable for predicting forest canopy height. The canopy height maps correlated with ICESat-2 estimated canopy height, albeit modestly. The study demonstrated the effectiveness of Sentinel series data and the machine learning models in predicting canopy height. Therefore, in the absence of commercial and rare data sources, the methodology demonstrated here offers a plausible alternative for forest canopy height estimation.

Polarimetric Synthetic Aperture Radar Interferometry (Pol-InSAR) based forest height estimation for ecosystem monitoring and management has been developing rapidly in recent years. Spaceborne Pol-InSAR systems with long temporal baselines of several days always lead to severe temporal decorrelation, which can cause a forest height overestimation error. However, most forest height estimation studies have not considered the change in dielectric property as a factor that may cause temporal decorrelation with a long temporal baseline. Therefore, it is necessary to propose a new method that considers dielectric fluctuations and random motions of scattering elements to compensate for the temporal decorrelation effect. The lack of ground truth for forest canopy also needs a solution. Unsupervised methods could be a solution because they do not require the use of true values of tree heights as the ground truth to calculate their estimation accuracies. This paper aims to present an unsupervised forest height estimation method called Dielectric Fluctuation and Random Motion over Ground (DF-RMoG) to improve accuracy by considering the dielectric fluctuations and random motions. Its performance is investigated using Advanced Land Observing Satellite (ALOS)-1 Pol-InSAR data acquired over a German forest site with temporal intervals of 46 and 92 days. The authors analyze the relationship between forest height and different parameters with DF-RMoG and conventional models. Compared with conventional models, the proposed DF-RMoG model significantly reduces the overestimation error due to temporal decorrelation in forest height estimation according to its lowest average forest height.