Tomography synthetic aperture radar (TomoSAR) is a cutting-edge radar observation technique that has the ability to produce three-dimensional images and can effectively extract forest vertical structure parameters, including forest height, a key forest parameter closely related to forest biomass and carbon storage. However, the phase errors in the TomoSAR data are unavoidable due to the elements such as orbit errors, which can seriously affect the quality of tomographic imaging and thus affect the accuracy of forest parameter extraction. To address this issue, various methods have been proposed. Nevertheless, they still exhibit restrictions when addressing phase errors with complex trends. To solve such problem, a novel method was developed and implemented in this paper, which includes two steps and remove parts of the phase errors with different trends sequentially. First, a wavelet decomposition and polynomial fitting-based approach was applied to each track to remove the slowly but significantly spatially-varying part of the phase errors. Secondly, the modified autofocusing algorithm is proposed to correct the remaining phase errors, which adopted the two-dimensional image entropy as the optimization indicator, providing stronger robustness compared to traditional indicator. Furthermore, in order to overcome the initial value dependency of the traditional search method, the proposed autofocusing algorithm used the particle swarm algorithm as search engine. After the phase error correction, the forest height was extracted by identifying the upper and lower boundary of the forest from the corrected TomosAR profiles. Two P-band datasets obtained in north China are adopted to examine the proposed phase error correction method. Experimental results show that compared with traditional autofocusing algorithm, the proposed method can achieve higher quality tomographic imaging results. On the basis of TomoSAR imaging, higher precision forest height extraction is obtained based on the new method.

In this paper, we studied the multi-feature combination estimation approach of forest above ground biomass (AGB) using X-band InSAR and P-band PolInSAR data. We focus on a crucial step of the estimation process, which is selection of the optimal feature combination. Firstly, the feature pool was acquired using multi-frequency SAR data, which includes optimized features (forest height and polarimetric interferometric feature) and original features (polarimetric features, intensity features, and texture features). Then, using machine learning method to select the optimal feature combination. Finally, the forest AGB was estimated based on multiple types of the features combination. The experimental results showed that the combination of optimized features with original features has the highest accuracy in forest AGB estimation, followed by the combination using only optimized features. The accuracy of forest AGB estimation is lower for the feature combination that does not include optimized features.

The TomoSAR technique has been applied to forest aboveground biomass (forest AGB) estimation studies, but existing studies make insufficient use of the forest structure information detected by TomoSAR. In this paper, we proposed a forest AGB estimation method based on TomoSAR backscattered power distribution law. The method uses the TomoSAR vertical profiles calculated by the Beamforming spectral analysis algorithm to extract the backscattered power for fitting in order to obtain the power curve. Then the distribution law was summarized by analyzing the variation of backscattered power distribution at different forest AGB levels. Based on the distribution law of backscattered power, two new forest AGB estimation features, BPC-4 and GVPR-19, are proposed. After modeling and validation, the results show that the forest AGB estimation model built with BPC-4 and GVPR-19 as variables can have better accuracy compared to the models built with the features proposed in previous studies.

Accurate tree species information is a prerequisite for forest resource management. Combining light detection and ranging (LiDAR) and image data is one main method of tree species classification. Traditional machinelearningmethods rely on expert knowledge to calculatea large number of feature parameters.Deep learning technology can directly use the original image and pointclouddata to classify tree species. However, data with different patterns require the use of different types of deeplearningmethods. In this study, a multimodal deeplearningframework (TSCMDL) that fuses 2D and 3D features was constructed and then used to combine data from multiple sources for tree species classification. This framework uses an improved version of the PointMLP model as its backbone network and uses ResNet50 and PointMLP networks to extract the image features and pointcloudfeatures, respectively. The proposed framework was tested using UAV LiDAR data and RGB orthophotos. The results showed that the accuracy of the tree species classification using the TSCMDL framework was 98.52%, which was 4.02% higher than that based on pointcloudfeatures only. In addition, when the same hyperparameters were used for training the model, the efficiency of the model training was not significantly lower than for models based on pointcloudfeatures only. The proposed multimodal deeplearningframework extracts features directly from the original data and integrates them effectively, thus avoiding manual feature screening and achieving more accurate classification. The feature extraction network used in the TSCMDL framework can be replaced by other suitable frameworks and has strong application potential.

The objective is to estimate forest above-ground biomass (AGB) considering the vertical structures of tropical forest in Mondah test site. Ten DLR F-SAR P-band fully polarized images are utilized to reconstruct the vertical reflectivity through TomoSAR techniques, and then extract multi-feature parameters from vertical profiles of reflectivity, finally these parameters are used for forest AGB estimation with support vector regression (SVR) approach. The results show that the method we proposed for tropical forest AGB estimation is effective and accurate in our test site.

Quantifying the uncertainty of forest height estimation is crucial for accurate global carbon computation. Forest height have been estimated by random volume over ground (RVoG) model using polarimetric interferometry synthetic aperture radar (PolInSAR) data in recent two decades. By far, RVoG derived forest height uncertainty estimation has been limited to comparison against “true” validation data, which then lead to the impossible use of uncertainties analysis at a national, continental or global landscape. In this paper, Bayesian theorem were introduced to RVoG forest height estimation procedure and used to estimated the uncertainties of the estimated forest height. The results revealed the feasibility of Bayesian method in forest height uncertainties estimation.

Accurate and fine multiple tree species supervised classification based on few-shot learning has attracted close attention from researchers, because the sample collection is often hindered in forests. Prototypical networks (P-Nets), as a simple but efficient few-shot learning method, have significant advantages in forest tree species classification. Nevertheless, the overfitting phenomenon caused by the lack of training samples is still prevalent in few-shot classifiers, which brings challenges to training accurate classification models. In this study, we proposed a novel Proto-MaxUp (PM) framework to minimize the issue of overfitting from the perspective of data augmentation and a feature extraction backbone for tree species classification. Taking Gaofeng Forest Farm (GFF) in Nanning City, Guangxi Province, as the study area, nine tree species, cutting site, and road were classified. First, by analyzing the effects of a series of popular data augmentation methods and their combinations in different parts of the P-Net, several effective data augmentation pools were established. Then, the pools aforementioned were combined with PM to obtain the best classification performance. To verify the robustness and validity of the proposed strategy, we applied PM to the other four popular public hyperspectral datasets and achieved excellent results. Finally, this efficient data augmentation method was used in different feature extraction backbones. The results show that the classification accuracy was greatly improved with the optimal backbone (overall accuracy (OA) and Kappa, are 98.08% and 0.9789, respectively), and the difference between training accuracy and test accuracy is less than 2%. It is concluded that the accurate and fine classification for multiple tree species can be realized by the PM data augmentation strategy and backbone proposed in this article.

In this article, a multilevel model (MLM) for forest height inversion is introduced and investigated using X-band single-pass interferometric synthetic aperture radar (InSAR) coherence data. Compared with the two-level model (TLM), as a generalized model, the MLM is more in line with the characteristics of forest structure and the scattering mechanism for X-band data. Based on the MLM model, three simplified MLM models were derived: TLMm, SINC, and MTND. An improved method of calculating coherence considering the assumptions of the MLM model is proposed. At two test sites, airborne (CASMSAR) and spaceborne (TanDEM-X) X-band single-pass InSAR data and LiDAR H100 data were used to verify the proposed new approach. The results showed that the MTND model can obtain more reliable and accurate inversion results compared to the SINC model and the TLMm. With airborne InSAR data, the MTND model’s highest accuracy was 82.56%, and the root mean square error (RMSE) was 2.4 m. With spaceborne InSAR data, the highest inversion accuracy of the MTND model was 73.55%, with an RMSE of 4.92 m. The inversion accuracy of forest height can be effectively improved by ensuring that the theoretical models and calculations of coherence obey the same assumptions.

Forest canopy cover (CC) directly and indirectly influences various processes of forest ecosystems. Airborne light detection and ranging (LiDAR) can be used to characterize forest spatial structures and further obtain estimates of forest CC. However, nonreturn laser pulses from targets of interest impact the estimation accuracy of forest CC. The objective of this article was to develop a novel method of estimating the nonreturn laser pulses to improve the estimation accuracy of forest CC using LiDAR data. The improved models for estimating forest CC were developed by introducing the loss ratio of laser pulses and a CC coefficient into the original models. The forest CC reference data were collected and used to validate the forest CC estimates. The results show that the loss ratio for forested areas was much higher than that for open ground areas. The range between the sensor and a target was a crucial factor that caused the loss of returns. The relationship between the range and the loss ratio was nonlinear in both open ground and forested areas. Compared with the original models, the improved models combining the loss ratio and the CC coefficient statistically significantly increased the estimation accuracy of the forest CC. Moreover, the forest CC estimates from the canopy height model (CHM) were more accurate than those from the height normalized point cloud (NPC) data. In addition, the simplified models were more generalized than the other models. This article is novel and has great potential to improve mapping of forest CC.

Forest stand height is one of the most important tree measurement metrics in forest inventory. Airborne LiDAR is likely to be a preferred technology for acquiring high-quality forest stand height to support forest survey and management, but the cost is much higher if being used to monitor the changes of forest stand height in time series. In this study, we use ZIYUAN-3 (ZY-3) tri-stereo imagery to create photogrammetric digital surface models (DSM) by dense image matching methods and LiDAR data to derive the digital terrain models (DTM). Forest stand height can be derived from ZY3-lidar CHM which is produced by subtracting the lidar DTM from the ZY-3 photogrammetric DSM. The forest stand height map is validated by the systematic LiDAR CHM plot. The result shows that ZY-3 tri-stereo imagery can be used for forest stand height mapping.

In this paper, we studied the nonparametric estimation approach of wheat biomass using C-band PolSAR data. We focus on a crucial step of the estimation process, which is feature combination selection. Firstly, the original feature pool was acquired using PolSAR data. Then, using genetic algorithm (GA) as searching engine we selected the feature combination which has the best performance in the estimation model. Finally, the wheat biomass was estimated based on the features combination selected by GA. The experimental results showed that the selected feature combination by GA can outperform the original feature pool and the feature combination selected based on Pearson correlation coefficient (PCC) between feature and biomass.

WCM (water cloud model) developed in 1978 was proved useful for vegetation parameters estimation. However, its no sense of polarization dependence limited its application with quad-polarimetric data. In this study, we developed polarimetric water cloud model (PWCM) and applied to quad-polarimetric GF-3 data for winter wheat biomass estimation. (HH-VV)/HV, (HH+W)/HV, HH/VV and VOL (volume scattering from Freeman Durden decomposition)/(ODD (odd scattering or surface scattering)+DBL (double bounce scattering)) were used as surface volume ratio parameters and provided in the developed PWCM. The results showed the effectiveness of developed PWCM for crop biomass inversion. Among all the parameters, (HH-VV)/HV showed best performance for winter wheat biomass inversion. The RMSE is 157.58 g/m2, the relative accuracy is 63.88.

In this study, a simple semi-empirical model based on random volume over ground model framework was used to estimate forest canopy height through single-pass TanDEM-X interferometric coherence data. Many studies have showed that SINC model could have good performance for forest canopy height inversion in relatively flat areas, this paper is aimed at assessing whether this method works well over complex terrain areas. The result showed that the inversion accuracy is influenced by local incidence angle and should be considered when mapping forest canopy height using SINC model.

Synthetic Aperture Radar (SAR) is able to image the Earth's surface in all weather conditions, regardless of whether it is day or night. It can provide high quality data for large-scale mapping. In recent years, the deep learning method, which has robust performance, can automatically extract hierarchic features of images with reduced manual participation, and is beneficial to the rapid implementation of large-scale mapping. Based on the theory of deep learning, we propose an end-to-end framework for the dense, pixel-wise classification of GF-3 dual-pol SAR imagery with convolutional highway unit network (U-net). U-net consists of a stack of learned convolutional highway unit that extract hierarchical contextual image features, and is a popular form of deep learning networks. A series of experiments show that the networks consider a large amount of context to provide fine-grained classification maps. The overall classification accuracy is 85.0%, Kappa is 0.803. The results are good enough to meet the needs of the mapping application.

In this study, we tested the performance of averaged Stokes parameters in rape biomass inversion over Radarsat-2 data. Five time series Radarsat-2 data were acquired and they covered the whole rape growth season. At each growth stage, 16 Stokes and its child parameters were extracted to analyze their sensitivity to rape biomass. When it comes to biomass inversion, random forest (RF) algorithm were chosen for the rape biomass inversion. The results showed that the total scattered intensity g0 has highest coherence with rape biomass during the whole growth cycle. R2 between them was 0.704. Other parameters including degree of polarization m, the degree of depolarization (1- m), the degree of linear polarization (Pl) and the linear polarization ratio (Uc) showed secondary higher coherence with R2 between 0.3 to 0.4. These parameters also showed higher importance for the parameter importance analysis with RF method. RF method showed potential performance for rape biomass inversion with R2 of 0.585 and RMSE of 71.86 g/m2.

In this paper, we studied the synergetic estimation approach of forest above ground biomass (AGB) based on the multidimensional SAR data (dual antenna X-band InSAR and P-band PolSAR data) that acquired by air-borne CASMSAR system of China. Firstly, the high-resolution DSM data of the experimental area was acquired from the X-InSAR data. Then, based on the filtered DSM, the terrain correction of P-PolSAR data and X-InSAR coherence was completed. Finally, forest AGB was estimated based on the characteristics of multi-dimensional SAR that after the terrain correction. The experimental results showed that the combined multi-dimensional SAR features can obtain higher estimation accuracy than the single-dimensional SAR features. Compared to only using P-PolSAR features and X-InSAR coherence feature, the accuracy of combined estimation approach was improved by 6.4% and 5.1%, respectively.

In this paper, we focus on the topographic effects of InSAR interferometric land use (ILU) images and a corresponding terrain correction method is proposed. Firstly, Effective scattering area correction and angular variation effect correction are performed for intensity information. Secondly, for the coherence information, a novel difference equation method of terrain correction based on the SINC volume decorrelation model is proposed. The Tandem-X/Terrasar-X InSAR data and SRTM DEM data are used to illustrate the method. The results showed that terrain effects not only existed in the intensity image, but also existed obviously in the coherence image and can been effectively removed by the proposed method. Finally, the interpretability of the ILU imagery treated in this manner is improved in comparison to the uncorrected case.

In this work, we present a strategy for obtaining the dynamics of forest above-ground biomass (AGB) at a fine spatial and temporal resolution. Our strategy rests on the assumption that combining estimates of both AGB and carbon fluxes results in a more accurate accounting for biomass than considering only one of the terms since the cumulative carbon flux should be consistent with AGB increments. We estimated forest AGB dynamics by combining two types of models driven by field, remote sensing, and auxiliary data. The strategy was successfully applied to the Qilian Mountains, a cold arid region located in northwest China. In the first step, we improved the efficiency of existing non-parametric methods for estimating forest AGB. We applied the Random Forest (RF) model in order to pre-select the most relevant remotely sensed features in Landsat Thematic Mapper 5 (TM) and ASTER GDEM V2 products (GDEM). These features were further used to construct an optimal configuration for the k-Nearest Neighbor (k-NN). Validation using forest measurements from 159 plots and the leave-one-out (LOO) method indicated that the optimal k-NN configuration yielded satisfactory performance (R2= 0.70 and RMSE = 24.52 tones ha-1). Hence, the k-NN configuration was used to generate a regional forest AGB basis map for 2009. In the second step, we obtained one seasonal cycles (2011) of carbon fluxes using the MODIS MOD_17 GPP (MOD_17) model that was driven by meteorological fields of a numerical weather prediction model (WRF) and calibrated to Eddy Covariance (EC) flux tower data. The calibrated model for 2010_well predicted GPP for 2011 (R2= 0.88 and RMSE = 5.02 gC m-28d-1). In the third step, we calibrated the ecological process model (Biome-BioGeochemical Cycles (Biome-BGC)) to above GPP estimates (for 2011) for 30 representative forest plots over an ecological gradient in order to simulate AGB changes over time. The Biome-BGC outputs of GPP and net ecosystem exchange (NEE) were validated against EC data (R2= 0.75 and RMSE = 1. 27 gC m-2d-1for GPP, and R2= 0.61 and RMSE = 1.17 gC m-2d-1for NEE). We used Biome-BGC to produce a longer time series for net primary productivity (NPP), which, after conversion into AGB increments, were compared to dendrochronological measurements (R2= 0.73 and RMSE = 46.65 g m-2year-1). The calibrated Biome-BGC model provided estimates of forest carbon fluxes that were converted into interannual AGB increments according to site-calibrated coefficients. With combination of these increments with the AGB map of 2009, the modeling of forest AGB dynamics was accomplished.

Recently, multi-source remote sensing data and their derived features such as vegetation indices, texture metrics have been frequently applied to quantitatively estimate forest above-ground biomass (AGB). However, it is still challenging to efficiently select the optimal features for modeling the forest AGB. In this study, a fast, efficient and automatic method has been proposed, called as k-nearest neighbor with fast iterative features selection (KNN-FIFS). This method iteratively pre-select the optimal features which determined by the minimum root mean square error (RMSE) between the forest field data and the k-nearest neighbor (k-NN) estimates based on the leave-one-out (LOO) cross-validation. By use of KNN-FIFS and multisource data, including Landsat-8 OLI (operational land imager) and its vegetation indices, texture metrics, HV polarization of P-band Synthetic Aperture Radar (SAR) data (PHV), and forest inventory data, were applied to estimate forest AGB over Genhe forest reserve located in Inner Mongolia, China. Afterwards, the model behaviors between KNN-FIFS and stepwise multiple linear regression (SMLR) methods were compared, which showed that the KNN-FIFS method (R2= 0.77 and RMSE = 22.74 t·ha-1) was superior to the SMLR method (R2= 0.53 and RMSE = 32.37 t·ha-1).

Estimating and measuring building height has become one of the significant factors in urban planning, legal and illegal construction inspection, urban disaster warning and assessing, as well as providing initial mapping data for creating three dimensional (3D) digital city models. In this paper we examine the feasibility of extracting building height information using computer vision algorithms with Structure from Motion procedures (SfM) from overlapping airborne images in urban environment. 3D land surface models can be generated from airborne images, and then DTM is subtracted from the DSM to form the nDSM layer. Using object-based image analysis, building height can be differentiated from bush, trees, and others by rulesets of spectral features, geometric features and contextual information. The accuracy of the building height based on orthorectified images and nDSM from airborne imagery was at a similar level to those based on airborne LiDAR data from the same study area.

This study investigated forest changes over Jinhe town of Inner Mongolia among 2001, 2004, 2008, 2013 and 2015 by use of multi-temporal Landsat Thematic Mapper-5 (TM), Chinese high resolution satellite data, Gaofen-1 (GF-1) remote sensing and forest inventory data. Vegetation fractional coverage (VFC) was extracted by the dimidiate pixel model. Forest above-ground biomass (AGB) was estimated by k-nearest neighbor (k-NN) with fast iterative features selection (KNN-FIFS). The results showed that, due to the serious forest fire disaster which taken plaihquce in 2003, the local VFC declined in the high level (60%2, accounting for 17.90% of the total area. Further analysis for the years after the fire showed that the forest recovered well, with the high level areas increasing to 2977.99 km2in 2015. The total forest AGB was increased from 1,792,900 tons in 2008 to 1,903,000 tons in 2013, with an average annual growth of 300,200 tons.

In this study, 5 compact polarimetric (CP) data were simulated from 5 consecutive fully polarimetric images, which covered the whole rape growth period. Four groups including 25 CP parameters were extracted from each CP image at different rape growth stages, respectively. With the analysis of the relationships between CP parameters and LAI, g3in stokes parameter group, Ucin stokes child parameters group, σRRin backscattering parameters group and PVαin decomposition parameters group were chose to inverse rape LAI. The results showed that, among these four CP parameters, PVαperformed best for rape LAI inversion, the determination coefficient R2 between inversion results and the ground truth data is 0.917, root mean square error (RMSE) is 0.36.

Two extraction models applied in estimating forest height and above-ground biomass (AGB) were developed using the X-band Interferometric Synthetic Aperture Radar (InSAR) data, which was acquired from the China airborne SAR system in 2013 covering part of forested area in northeast China. The models using multi-passes InSAR data for the estimations of forest height and AGB were introduced respectively with detailed ground true data derived from both forest plot data and airborne Light Detection And Ranging (LiDAR) data. The estimation results indicated that it is feasible to estimate forest height using X-band InSAR data without the external DEM. Moreover, the model using multi-passes InSAR data could descript the forest structure characteristics and can be used to estimate forest AGB faithfully.

Forest cover will be changed by natural disaster, deforest and other activities. Remote sensing has the ability to monitor the change of forest cover in large area. However, in order to obtain the change information of forest cover, the method for getting the change information from remote sensing images is needed. China has launched GF-1 satellite in 2013, the 16m spatial resolution multi-spectral data of it should be suitable for forest cover detection applications. In this paper, I describe a set of procedures that automate forest cover change detection using a pair of GF-1 images. The proposed method was tailored to work with high spatial resolution images acquired over forest in large area. To achieve a high level of automation, automatic optimum threshold selection algorithm were applied in the processing steps. In this study, an overall change recognition accuracy of 83% has been achieved.

Forest is an important component of terrestrial ecosystems, so it is necessary to estimate the forest aboveground biomass (AGB) accurately in order to reduce the uncertainty of the carbon stock in forest ecosystem. In this study, a fast, efficient and automatic method has been proposed, called as Automatic KNN(AKNN). Using the Landsat 8 OLI data, airborne polmetric SAR (PolSAR) data, the AKNN has been applied to estimate the forest AGB and stem volume at two levels, the pixel and the subcompartment levels, respectively. The results showed that the power of AKNN in quantitative retrieval at two levels(R2=0.75, RMSE=23.43t/ha; R2=0.52 and RMSE=21.86m3/ha, respectively).

Forest above ground biomass (AGB) retrieval by TomoSAR technology has been preliminary studied over the last two years. Recent experiments have demonstrated that the backscattered power in HV channel at 30-m layer have strong correlation with the forest AGB, and be successfully used to build the estimation model. However, much progress remains necessary to make the best of the vertical distribution of the backscattered power. In this work, we proposed a multivariate regression equation to refine the AGB estimation model. The experiments were carried out on TropiSAR campaign data over the site of Paracou, French Guiana. As expected, the results demonstrate the relevance of the proposed method.

The goal of this paper is to investigate the influence of temporal decorrelation on InSAR / Pol-INSAR and T-SAR in boreal forest. The P-, L-band Pol-InSAR data collected in campaign BioSAR 2008 was used in our study. The impact of temporal decorrelation on InSAR / Pol-InSAR and T-SAR is reflected by coherences, phases and vertical backscattering power. Markov model is applied to describe the quantitative impact of time decorrelation, and correlation, time decorrelation constant is identified by GA. And the influence of temporal decorrelation on T-SAR is also related with coherences and phases. The backscattering power represents more ambiguous with longer time interval than that with shorter time interval for single baseline.

In this paper a method was demonstrated that GF-1 multi-spectral and MODIS multi-temporal NDVI data were integrated for forest land cover classification. The test site is located in the central of the Xiaoxing'anling region in Heilongjiang province where covered the area of one scene of GF-1 image. The random forests algorithm was adopted to select the best features automatically which contains spectral, texture and shape features from GF-1 multi-spectral data and phenological features from multi-temporal MODIS NDVI data. A decision tree was used to supervise the classification result. Experimental results show that the overall classification accuracy and Kappa coefficient of the developed method combing multi-sources data can reach 89.46% and 0.874 respectively, with significant improvement compared with that using either GF-1 multi-spectral data or MODIS NDVI time series data alone, especially for the classification of evergreen forest.

A strategy of data assimilation using the refined remote sensing product for the process-based model (Biome-BGC) in order to improve the simulated carbon fluxes was proposed. Firstly, we applied the optimized the remote-sensing-based MODIS MOD_17 GPP (MOD_17) model to calibrate the process-based Biome-BGC model. This incorporation strategy for the parameterization of Biome-BGC has been proved to be more reliable in carbon fluxes' simulations. The calibrated Biome-BGC model agreed better with the Eddy covariance (EC) measurements (R2=0.87, RMSE=1.583 gC/m2/d) than the original model (R2=0.72, RMSE=2.419 gC/m2/d). Afterwards, two years (8-day during 2003 and 2004) Global LAnd Surface Satellite (GLASS) LAI products were applied to test the data assimilation procedure for the calibrated Biome-BGC using Ensemble Kalman Filter (EnKF). The results indicated that simulated LAI through assimilation agreed better with GLASS LAI, and the carbon fluxes are hoped to improve by further adaption of the filter.

Plant biomass is an important parameter for crop management and yield estimation. The potential of compact polarimetric (CP) synthetic aperture radar (SAR) data in estimating biomass of oilseed rape crop (Brassica napus L.) is investigated in this study. Five CP SAR imagery was simulated using five fully polarimetric Radarsat-2 data, and the dynamic evolution of polarimetric features, relying on different polarimetric decomposition methods (m-χ, m-δ, and Freeman-Durden), with the crop growth, was compared. It was found that the Dbl indicator, by the m-χ decomposition method, can reflect well the dynamic growth of canola. Therefore, a method of monitoring fresh and dry biomass of canola was put forward. The result showed that the root mean square error (RMSE) was 56.5g/m2, 448.2g/m2, and the relative error (RE) was 23.9%, 25.0% for fresh and dry biomass, respectively. In addition, the precision of the model will be affected when the crop becomes mature since its vegetation water content declines. The results were also compared with those of the fully polarization SAR. It revealed that the performance of CP SAR on rapeseed monitoring can achieve the level of fully polarization SAR, considering the advantages of CP SAR, such as wider coverage and less data volume etc. It revealed that the polarization information was necessary in quantitatively monitoring of broad leaf crops, such as rapeseed, and CP SAR has a great potential in crop monitoring.

The radiometric quality of polarimetric SAR (PolSAR) image is affected by terrain undulations due to 1) the variation of effective scattering area, 2) the variation of scattering mechanisms, and 3) the variation of polarization states. This paper proposed a three-stage terrain correction method of PolSAR data for the impact of above three aspects. PALSAR-2 PolSAR data and LiDAR forest AGB data are used to illustrate the method. The results showed that terrain effects can been effectively removed and the correlation between forest biomass and backscatter coefficient are improved after three-stage terrain correction. The HV polarization has the best correlation with the forest AGB (R = 0.75) and the correlation coefficient was increased about 0.3 compared with uncorrected case.

The objective of this paper is to derive DEM and DHM in tropical forests based on the technology of multi-baseline InSAR tomography, focusing on the requirement of the minimum number of flight tracks. The experiments were carried out on P-band HH polarization airborne multi-baseline InSAR data over the site of Paracou, French Guiana, during the European Space Agency campaign TropiSAR 2009. Tomographic processing was carried out by Capon spectral estimation technique with three-track observations, subsequently two elevation values with respect to the relative maximum peaks were retrieved representing DEM and DSM respectively. As expected, the retrieved DEM and DHM are found agreements with LiDAR measurements.

Estimating forest above-ground biomass (AGB) and monitoring its variation are relevant for sustainable forest management, monitoring global change, carbon accounting, particularly for the Qilian Mountains (QMs), a water resource protection zone. In this work, the results of above-ground biomass (AGB) estimates from Landsat Thematic Mapper 5 (TM) images and field data from the fragmented landscape of the upper reaches of the Heihe River Basin (HRB), located in the Qilian Mountains of Gansu province in northwest China, are presented. An optimized k-Nearest Neighbor (k-NN) method was determined by varying both the mathematical formulation of the algorithm and remote sensing data input which resulted in 3,000 different model configurations. Following the sun-canopy-sensor plus C (SCS+C) topographic correction, performance of the optimized k-NN method was satisfied (R2=0.59, RMSE=24.92 ton/ha) which indicated that the optimized k-NN is capable of operational applications of forest AGB estimates in regions where only a few inventory data are available. Afterwards, the calibrated BIOME-BGC was applied to simulate the carbon fluxes over QMs forests with satisfactory accuracy. Finally, the dynamic analysis and modeling of forest AGB was conducted based on the remotely sensed estimation of forest AGB and the annual forest AGB increment from the ecological process model.

Forest biomass reflects the ecological succession and human disturbance of the forest, and can fully embody the quality of forest ecosystem environment. The Qilian Mountain forest reserve at upper reaches of the Heihe River Basin was selected for the study. Landsat Thematic Mapper 5 (TM) images were selected as the source data, which were rectified by SCS + C terrain radiometric correction. Forest above-ground biomass was estimated using k-nearest neighbor (k-NN) method and support vector regression (SVR) method, respectively. The results show that spectral information of remote sensing image was recovered by the sun-canopy-sensor plus the C (SCS+C) terrain correction which can effectively improve the estimation accuracy of the models regardless of k-NN or SVR. The optimal k-NN method (R2=0.54, RMSE=26.62ton/ha) performs better than the optimal SVR method (R2=0.51, RMSE=27.45ton/ha).

The aim of this paper is to analyze the feasibility of hemi-boreal forest above ground biomass (AGB) estimation based on Tomo-SAR technique. Repeat-path multi-baseline P-band Pol-InSAR data collected during March and May, 2007 in Remningstorp test site is used. The result shows that the correlation coefficient (R) of P-band HH backscattering coefficient reaches 0.87 with the in-situ forest biomass. The R of P-band VV backscattering power at 5m and 10m is 0.71, 0.72 with the in-situ forest biomass, respectively.

An approach was used to incorporate the forest carbon flux for Qilian Mountains by ecological-process-based model (Biome-BGC), and remote-sensing-based model (MODIS-PSN). The calibration phase, aiming at setting the ecophysiological parameters to effectively simulate the daily GPP behavior of the Qilian Mountains, was proceeded by adjusting the 8 day GPP outputs obtained from Biome-BGC using the optimized MODIS-PSN algorithm and the observations. The results showed that the optimized MODIS-PSN could describe the GPP behavior faithfully comparing to the eddy covariance-observed GPPs, with R2=0.77, RMSE=6.219gC/m2/8d. After validation, the calibrated Biome-BGC has been proved to estimate the performances of daily GPP behavior effectively comparing to the eddy covariance-observed GPPs, especially in summer and winter (with R2= 0.76, RMSE= 1.3115gC/m2/d), which illustrated that the combination of Biome-BGC and optimized MODIS-PSN could express the carbon fluxes well over the Qilian Mountains.

In this paper we analyze the effects of polarization, environmental conditions and forest structure upon the backscatter response of forested stands. This analysis is based upon a time series of ALOS-1 PALSAR images acquired over our study site in Xunke County, Heilongjiang Province, China. Based on six scenes, we analyzed the polarization and environment conditions on the forest stands. Backscatter coefficients of HV channel had a greater dynamic range than HH channel. HV channel was less influenced by weather and wind speed conditions. Our observations found canopy density greatly influenced the forest stand backscatter. Backscatter coefficient showed weak correlations to canopy density, mean tree height and mean diameter at breast height (DBH). Correlations were much stronger when the forest stands were grouped with canopy density or mean tree height. Before grouping the highest correlation coefficient between backscatter coefficients and tree height was 0.377, the value for HV image acquired on August 07, 2007. After grouping these forest stands by canopy density, the correlation was 0.95 to tree height. We also analyzed the correlations with two different tree species, and obtained similar results.

Synthetic aperture radar (SAR) simulation of 3-D forest canopies is a powerful tool for studying the interaction between radar and forest, for testing new applications, and for devising inversion algorithms of forest structures. SAR raw-signal generation is frequently used in point-target simulation but is rarely used in 3-D forest simulation. The existing simulators directly produce SAR images based on an impulse response function (IRF) without involving raw-signal generation and various nonideal factors. In this paper, a novel simulator to produce SAR images of 3-D forest canopies is proposed. It incorporates a SAR raw-signal generation process taking account of various nonideal factors such as trajectory deviation of radar platforms and complexity of natural environments, which is more faithful to realistic remote sensing systems. Furthermore, an approach to speed up the raw-signal generation is put forward based on the equivalent scattering model consisting of a few virtual scatterers with specially calculated positions and backscattering matrices. Thus, the raw signals received from the entire forest canopy can be equivalent to those from virtual scatterers in the case of tiny slant-range errors. The error sensitivity of equivalent conditions is analyzed, and the optimum selection of equivalent parameters is derived considering the compromise between precision and efficiency. The results of simulation and forest height inversion demonstrate the feasibility and potential utilities of the proposed simulator.

This study explores retrieval of wall-to-wall forest above-ground biomass (AGB) over Jilin province in Northeast China, using the optimized non-parametric k-NN method, the 7thNational Forest Inventory (NFI) data, and the remote sensing data: Landsat-TM/ETM+ images. For pixel-based validation, the estimated result was compared to the NFI data by leave-one-out process and R2= 0.40 and RMSE = 54.29 tons/hm2. For county-scale validation, the result was verified by the intensive forest sub-compartment data of eight county and R2= 0.80 and RMSE = 34.26 tons/hm2.

A campaign consists of in-situ measurement, airborne Lidar data and spaceborne data fusion was designed and implemented in Daxinganling of Northeastern China. The purpose of the campaign is to collect field and multi-sensor data to produce spatial explicit map of forest aboveground biomass for the entire region. This paper presents preliminary results of ABG mapping using field data, airborne lidar, ICESat GLAS, MODIS and ENVISAT MERIS data. The AGB map from multi-sensor fusion had a good performance with 8.1% difference after comparison with the estimated AGB in the Daxinganling Ecological Research Station.

Compared with traditional remote sensing, multiangular observation provides 3-D structural information of a forest through different directional observations. The MGeoSAIL model, suitable for multiangular observations, was developed based on the single-angle model GeoSAIL. The MGeoSAIL model combines the geometric-optic model with the radiation transfer model and has the advantages of both models. Thus, it is more accurate and feasible. The geometric-optic model calculates the amount of shadowed and illuminated components within a forest scene, while the radiation transfer model [Scattering by Arbitrarily Inclined Leaves (SAIL)] calculates the reflectance and transmittance of tree crowns. The uniform index is introduced to characterize the relationship quantitatively between tree distribution pattern and the bidirectional reflectance distribution function (BRDF). The simulation results show that the MGeoSAIL model could simulate the “hot” spot in red and near-infrared bands, as well as the “bowl” shape in the near-infrared band. The relationship between the uniform index and BRDF is negatively exponential. Finally, the look-up table was calculated using the MGeoSAIL model, and leaf area index (LAI) was inversed from compact high-resolution imaging spectrometry data. The results compared well with the measured LAI in Changbai Mountain area, China.

In order to improve the land cover classification accuracy for SAR image, Support Vector Machine (SVM), which has wide applicability is used on the land cover classification of POLSAR image in this paper. The study site is located in Tahe County, Heilongjiang Province, China, and two scenes of quad-polarization Radarsat-2 SAR images were acquired. the land cover classification of single-temporal POLSAR image by SVM, and multi-temporal POLSAR image by SVM and maximum likelihood classification (MLC) is studied separately. Then all the classification results are evaluated. Some conclusions can be got according to the analysis of all results and accuracy: Firstly, it is difficult to distinguish the different types of vegetation for the similar scattering among them in July. However, water, whose scattering characteristic is simplex, can be distinguished from others easily. Scondly, in October, the scattering characteristics among forest, shrub, grass, crop are different, therefore it is easy to distinguish vegetation because of their one from others in this period. But for water, with reduced in winter, the river width narrows, compared with it in summer, water classification accuracy is lower in this period. Thirdly, joint July and October SAR data for classification, can offset espective their own disadvantages. and improve overall accuracy. And the last one, With the characteristics that different probability density distribution, small sample, non-linear and so on, SVM shows the wide applicability.

The main objective of this study was to investigate the potential of using Support Vector Machines (SVM) and Random forest (RF) to estimate forest above ground biomass (FAGB) by using multi-source remote sensing data. To do so, we introduced a basic flow of SVM to estimate FAGB from multisource remote sensing data. RF method was adept at identifying relevant features having main effects in multisource remote sensing data. Results show that: (i) In the stage of feature selection, the Random Forest model provide better results compared to the typical F-scores method. (ii) The optimal SVM model, based on the selection of features clearly demonstrate that the estimation accuracy increased by feature selection algorithm. (iii) Compared to the optimal KNN, BPNN and RBFNN model, the optimal SVM algorithm provided more accurate and robust result on the considered case.

The objective of this study is to estimate forest above ground biomass (AGB) of the whole areas covered with forest basing on Polarization coherence tomography (PCT) and polarimetric interferometric segmentation. Two scenes of DLR E-SAR L band quad-polarization images were acquired in Traunstein test site. Results show that there is no saturation using this method even when the biomass ups to 500tons/ha and parameters extracted from PCT is easier to implement in practice.

The primary aim of this study was to extract DEM using Polarimetric SAR Interferometry (Pol-InSAR) ALOS/PALSAR data. We make use of three Pol-InSAR coherence optimization methods, including Singular Value Decomposition, Numerical Radius and Phase Diversity ways, to extract DEM with processing of filtering, unwrapped phase, base-line estimation and so on. At the end, compare optimized results with the single polarimetric interferometry such as HH, HV and VV. It has been observed that optimization ways can reduce the interferometric noise, reduce the phase unwrapping residuals, and improve the precision of the DEM extraction. Meanwhile, Numerical Radius approach has a better result than Singular Value Decomposition and Phase Diversity ways; Interferogram with integrated Contoured Median and Goldstein two-step filter method can further improve and optimize the quality of the interferogram, reduce residual mostly, improve the precision of the DEM, but if two-step filter window setting big will affect the area of non-water-covers while improve the water-covers area.

Extraction of vegetation water content and soil moisture from microwave observations requires development of a high fidelity scattering model. A number of factors associated with the vegetation canopy and with the underlying bare soil should be taken into account. In this paper, we propose an electromagnetic scattering model for a corn canopy which includes the coherent effect due to the corn structure and takes advantage of recently advanced scattering models for dielectric cylinder of finite length and for rough surface.

In this paper, a set of polarimetric eigenvalue and eigenvector based parameters, e.g. entropy and anisotropy, are investigated for forest application. The correlation terms of the eigenvectors, μ1and μ2, are found to be better for forest mapping in both summer and winter using Radarsat-2 quad-polarimetric space borne SAR data. These are used to automatically identify forest class pixels from the volume scattering category of a Freeman-Durden Wishart unsupervised segmentation map. The algorithm scheme was developed and implemented using fully polarimetric Radarsat-2 SAR (PolSAR) data acquired in July and October and the validity was evaluated using the ground reference data created from SPOT5 K-clustering classification map.

The objective of this paper is to assess the use of ALOS PALSAR FBD data to map rice growing areas. Image enhancement in backscattering in rice fields as a result of Bragg resonance scattering was found only at HH polarization since double-bounce scattering is a prerequisite to Bragg resonance scattering for radar backscatter from bunches of rice plants. A rice mapping method using HV images was developed and applied to Haian test site. Validation showed that rice mapping using L-band SAR is promising when cross-polarized data are available to cope with the Bragg resonance scattering effects.

This paper investigates to what extent interferometric coherence optimization in radar polarimetry improves the performance of forest height inversion method using POLInSAR measured data. Based on repeat pass E-SAR data and the corresponding ground measured forest stand heights, several available forest height inversion methods are validated and compared together with coherence optimization algorithms such as the iteration for the maximization of the magnitude difference (BF-mag) coherence optimization algorithm and phase diversity (PD) coherence optimization algorithm. The results show that coherence phase optimization can improve the performance of the tree height retrieval method based on coherence phase information, but can not improve the performance of the retrieval method based on coherence amplitude information alone. Furthermore, an integrated inversion method, which combines coherence phase with coherence amplitude information and includes corresponding polarization coherence optimization and compensation of non-volume scattering decorrelation, is proposed and discussed.

With the arisen spaceborne multi-parameter Synthetic Aperture Radar (SAR) systems, such as Envisat ASAR, TerraSAR-X, ALOS PALSAR, and RADARSAT-2, the interest of crop mapping has been increasing. The present study compares the capabilities of the multi-parameter SAR in discriminating the main crop types by object-based classification in Haian county of Jiangsu province, South China. Two kinds of information, SAR intensity based and SAR statistical properties based are used for Maximum Likelihood Classification (MLC) and Minimum Distance Classification (MDC) respectively. The results show that, the L-band SAR can uniquely identify mulberry from dry-land crops, such as maize and vegetable and C-band SAR has some advantages in mapping rice. Specifically, the polarimetric RADARASAT-2 data can identify the rice with accuracy about 75% ~ 80% which is similar as the result from X-band TerraSAR-X Spotlight data but higher than that from C-band dual-polarization Envisat ASAR data. Nevertheless, both of X- and C-band can hardly separate the mulberry from the other dry-land crops.

As the land degradation is a complex process, influenced also by climatic and human-induced factors, its understanding and mapping requires a methodology based on Earth Observation integrated with ancillary data such as socio- economic and, optionally, meteorological data. For this reason, the service, which has been defined, implemented, and validated in close cooperation with End Users, is based both on scaleable indexes - derived from Earth Observation data - and on indicators relevant to the influence of human and animal pressure on natural resources. The ultimate goal is the generation of vulnerability maps to provide to decision makers for prevention purposes.

Average height of forest stand is an important parameter related to forest ecological function and carbon cycling. Estimation of stand height using remote sensing data becomes possible due to the new technologies. Interferometry synthetic aperture radar (InSAR) technology uses the phase information of radar returns to get three-dimensional information of the earth surface. The existence of trees in a radar pixel modifies the phase information, so InSAR data has been studied for its potential of tree height estimation. In this study, the C and L-band InSAR data acquired during STR-C/X SAR Mission (1994) were used to estimate average tree height of forest stands in the Daxinganling forest region in Northeast China. The DGPS is used to locate each forest sample plot and the tree heights were measured in 1999. Tree heights in 1994 were deduced from the 1999 field measurements and the tree growth table. The results from C and L bands were compared. Following conclusions were drawn from the study: (i) Stand height information can be acquired from the digital surface model (DSM) generated from InSAR data, (ii) If the forest stand is very uniform in terms of tree height and density, the C-band InSAR data gives better result than L band. Otherwise, the results from L-band InSAR data are more stable, (iii) Multi-temporal InSAR technology may be used to monitor stand height increment.