Spaceborne L-band bistatic interferometric synthetic aperture radar (InSAR) is an advanced remote sensing technology used for forest height inversion. It enables the detection of forest vertical structures and avoids the effects of temporal decorrelation. However, forest height inversion using single-polarization L-band bistatic InSAR in mountainous areas still faces several challenges. First, multi-parameter scattering models cannot be directly resolved using single-polarization InSAR observations. Second, ground scattering in L-band InSAR significantly affects the accuracy of forest height inversion. Moreover, mountainous terrain alters the interaction between InSAR signals and forest scatterers, further increasing the inversion uncertainty. To address such challenges, this paper proposes a frequency-domain information enhancement adaptive volume coherence optimization (Ada-VolOpt) method using single-polarization LuTan-1 bistatic InSAR data. The proposed method expands the observation space of InSAR through time-frequency analysis. Subsequently, based on frequency-domain information enhancement and the random volume over ground (RVoG) model, an adaptive volume coherence optimization method is proposed to overcome the adverse effects of significant ground scattering on forest height inversion in mountainous areas. Finally, forest height inversion is performed using a slope-adaptive scattering model. The effectiveness of the proposed method was validated across three test sites in China. A total area of 63.11 thousand km2 (6.31 million hectares) was used to test the proposed method, resulting in reliable forest height products. The forest height is estimated with an accuracy of 5.04 m for tropical forests, 3.41 m for mixed forests, and 2.44 m for boreal forests, respectively. Compared to the method that ignores ground contributions, the proposed method improves the accuracy by 8%, 10%, and 33%, respectively. This study provides a comprehensive benchmark evaluation of the performance of large-scale forest height inversion using LuTan-1 bistatic InSAR data.

This article proposes an enhanced bistatic interferometric synthetic aperture radar random volume over ground (EBis-RVoG) model for modeling the coherence of L-band spaceborne bistatic interferometric synthetic aperture radar (InSAR) in forests. Simultaneously, a novel method for forest parameter robust estimation (RE) based on the EBis-RVoG model using single-polarization LuTan-1 bistatic InSAR images is proposed. The proposed EBis-RVoG model addresses the underdetermined problem in bistatic InSAR coherence modeling in forests, which arises from the independent double-bounce (DB) scattering contribution. The effects of DB scattering on the estimation of forest vertical structure parameters are studied. The proposed RE method expands the observation space of single-polarization InSAR through time-frequency analysis, and reduces the dependence on model parameter initial guesses by linearizing the model and constraining the solution space. The performances of the proposed EBis-RVoG model and the parameter RE method are validated using simulation datasets and LuTan-1 bistatic InSAR images from boreal forests in northern China. The underlying topography estimation achieves an accuracy better than 2 m, while the forest height estimation attains an accuracy exceeding 90%. This study provides a novel model and method for retrieving forest vertical structure parameters using LuTan-1, TanDEM-X, PIESAT-01, and future bistatic InSAR missions.

The soil freeze–thaw (F/T) cycle plays a critical role in understanding the water cycle, climate change, and land–atmosphere energy exchange. Although passive microwave remote sensing is widely used for large-scale F/T monitoring, the sensitivity of frozen soil microwave radiation to soil parameters remains insufficiently understood, limiting improvements in monitoring accuracy. To address this gap, this study employs the Sobol global sensitivity analysis (SA) method in combination with the seasonally frozen soil radiation transfer (SFS_DMRT) model to evaluate the sensitivity of brightness temperature (Tb) under different frequencies, incidence angles, polarizations, and their combinations to key soil parameters. The results reveal a frequency-dependent transition in the dominant mechanisms affecting Tb. Specifically, L-band Tb is sensitive to microwave emission from the bottom unfrozen layer, whereas, C- and X-band Tb are primarily sensitive to emission from the upper frozen layer. Ku-band Tb is influenced by both emission and volume scattering within the upper frozen layer, while Ka-band Tb is mainly affected by volume scattering. Beyond these frequency-dependent mechanisms, the analysis also reveals how specific microwave configurations can serve as indicators for different soil parameters. V-polarized Tb (TbV) exhibits sensitivity to frozen layer temperature. The polarization difference at C-band (TbV − TbH) is sensitive to frozen depth, especially within the top 30 cm. The polarization difference at L-band and the incidence angles difference of TbV between 55° and 40° (TbV55° − TbV40°) are sensitive to the effective dielectric constant of the frozen layer. The TbV ratio between C- and Ka-bands is sensitive to volume scattering effects. By systematically evaluating the sensitivity of microwave radiation to key soil parameters, the global SA identifies optimal passive microwave configurations for monitoring soil F/T in response to variations in soil temperature, frozen depth, dielectric properties, and volume scattering during F/T transitions, thereby providing valuable guidance for improving soil F/T monitoring algorithms and offering insights for the design of future satellite payload.

Accurate identification of eucalyptus plantation maturity, which is hierarchically defined as young forests, middle-aged forests, and mature forests, corresponding to different growth and physiological development stages of eucalyptus stands, is critical for forestry management and sustainable development. Currently, research on fine-grained semantic segmentation of different eucalyptus growth stages remains insufficient, and there is a lack of systematic evaluation of classical deep learning architectures and multimodal data for eucalyptus plantation maturity identification. This limits the application of remote sensing technology in forestry management, and constrains the understanding of growth dynamics in subtropical planted forests. Taking Gaofeng Forest Farm in Nanning, Guangxi, as the study area, this study analyzes the adaptability of four generations of deep learning segmentation architectures—U-Net (convolutional baseline), Trans-UNet (CNN-Transformer hybrid), Swin-UNet (pure Transformer), and Mamba-UNet (state-space model)—in the fine identification of eucalyptus plantation maturity based on spectral indices (SIs), C-band SAR data (S1), and multispectral data (S2). The results show the following: (1) There is no positive correlation between model complexity and recognition performance. Among all architectures, Mamba-UNet achieves the best performance, with a validation set mIoU of 79.10% and an F1-score of 88.26%. The performance ranking of the different architectures evaluated is Mamba-UNet > Swin-UNet > U-Net > Trans-UNet. (2) The S2+SI combination achieves the highest accuracy (mIoU 79.10%), outperforming S2+S1+SI (78.55%), S2+S1 (78.38%), and single S2 (77.72%), which indicates the strong correlation between spectral indices and eucalyptus physiological characteristics. The backscattering features of S1 are limited by canopy penetration in the subtropical rainforest, introducing redundancy and triggering negative fusion effects. (3) Independent verification with field survey points verifies the strong generalization ability of the proposed approach, with OA of 94.08%, mIoU of 85.62%, Precision of 93.00%, Recall of 91.37% and F1-score of 92.15%, which indicates that the methodological framework can realize the identification of eucalyptus maturity with high precision. (4) Eucalyptus accounts for 60.15% of the total area of Gaofeng Forest Farm. Within the eucalyptus stand age structure, young, middle-aged, and mature forests account for 20.67%, 13.62%, and 25.86%, respectively. The overall distribution exhibits a polarized pattern with high proportions of young and mature forests. The findings offer theoretical support and insights for dynamic monitoring of fast-growing plantations and refined management of stand development stages.

Limited sample collection in a complex forest stand is a serious challenge for automatic high-precision classification of multiple tree species. Few-shot learning (FSL) have significant advantages in handling small-sample problems, especially in transferring domain-invariant feature representations/meta-knowledge from the source domain (SD) to the target dataset (TD). However, domain shift, significant differences in ground categories between SD and TD, is commonly present in cross-domain FSL. A novel TSDCFSL framework is proposed, which decouples domain-invariant feature transfer from domain-specific feature extraction in a two-stage manner to solve the small-sample and domain shift problems simultaneously. Its core contribution is to align the global and related subdomain distributions of SD and TD via local maximum mean discrepancy in the pretraining, thereby enhancing the domain-invariant feature representation, which is beneficial to extracting domain-specific features using FSL supervised contrastive learning, and data augmentation in the fine-tuning to improve the discriminability of TD samples. Five hyperspectral datasets, including three SDs and two TDs are adopted to conduct thorough experiments. Taking the XiongAn as SD, the training set contains 5 samples per class on each TD, the overall accuracy (OA) of the optimal model on GaoFeng-A and GaoFeng-B is 90.74% and 90.02%, which is better than the SD of Chikusei and HanChuan. As the training set increases to 20 samples per class, the OA reaches the highest 98.51% and 95.26% on GaoFeng-A and GaoFeng-B, respectively. The outstanding results indicate that our TSDCFSL has great potential in training high-precision tree species classification models.

In this paper, we proposed a forest aboveground biomass (AGB) estimation approach exploiting few-look averaged interferograms to construct forest vertical structure profile and extract forest structural features. Firstly, we used X-band InSAR data to obtain the initial Digital Surface Model (DSM) and removed topographic trends using the low-pass filtering technology. Then, the forest vertical structure profile was constructed using the InSAR-derived surface height within specified range. Finally, the forest AGB was estimated by combining the extracted features reflecting both forest height and density from the vertical structure profile. The experimental results showed that the estimation method of extracting forest structural parameters from vertical structural profiles improved the accuracy of forest AGB estimation compared with the conventional estimation method, which depends solely on forest height.

Tomography synthetic aperture radar (TomoSAR) is a cutting-edge radar observation technique that has the ability to produce 3-D 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 elements such as orbit errors, which can seriously affect the quality of tomographic imaging and 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 a problem, a novel method was developed and implemented in this article, which includes two steps and removes parts of the phase errors with different trends sequentially. First, a wavelet decomposition and polynomial fitting-based approach were applied to each track to remove the slowly but significantly spatially varying part of the phase errors. Second, the modified autofocusing (MA) algorithm is proposed to correct the remaining phase errors, which adopted the 2-D image entropy as the optimization indicator, providing stronger robustness compared with the 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 a search engine. After the phase error correction, the forest height was extracted by identifying upper and lower boundaries 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 the 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.

The past five years are the five years when big model and general model of Artificial Intelligence(AI)are gradually integrated into people's daily work and life,and the five years when remote sensing+AI technology develops rapidly in the fields of land cover type identification,change detection,etc.It is also the first five-year for the implementation of the national strategy of"ecological civilization"and"beautiful China".Summarizing the progress made in the research,development and application of forestry and grassland remote sensing technology in these five years is of great significance for the country to formulate the development plan of forestry and grassland remote sensing in the future. The paper summarizes the main progress of the forestry and grassland remote sensing research and development in China in the past five years into four research directions,namely,change detection and classification of forest and grassland cover types,quantitative inversion/estimation of forest parameters by remote sensing,and that of grassland vegetation and early warning and monitoring of forest and grassland disasters.From a general point of view,the research on forestry and grassland remote sensing technology shows a rapid development trend from traditional shallow machine learning to deep learning,and from"data"-driven to"data+mechanism"-double-driven direction,and the deep learning method develops quickly and deeply in change detection and classification,but not in quantitative parameter inversion/estimation.The production technology of large-scale forest and grassland thematic products,such as global and national products,has also been developed rapidly. An analysis of the integration of remote sensing technology into existing technical standards and technical programs for forestry and grassland resources and ecological monitoring,disaster early warning monitoring and monitoring of nature reserves shows that forest and grassland cover type change detection/monitoring and classification technologies have been widely and deeply applied to various resource supervision and disaster early warning and monitoring operations in the forestry and grassland industry,but the degree of operational application of quantitative inversion/estimation technologies of forest and grassland quality parameter is still very low. In view of the challenges in promoting the comprehensive and in-depth application of forestry and grassland remote sensing technology,it is suggested that the forestry and grassland industry should vigorously integrate the"space-air-ground"multi-source earth observation resources,comprehensively apply remote sensing,artificial intelligence(AI),statistical inference and other cutting-edge technologies to build a"space-air-ground"integrated monitoring technology system,and greatly strengthen the investment in scientific research,technology exchange and talent exchange and cultivation.

In this paper, a non-local spatio-temporal fusion filtering method for Polarimetric SAR data is proposed. The methodology consists of two primary stages: initially, a combination of temporal dimension distance weighting and a ratio method is employed to enhance the image signal-to-noise ratio. Subsequently, this is complemented by the integration of the spatial dimension with non-local means filtering. Notably, unlike conventional filtering techniques designed for single-polarized data, the proposed method is grounded in the polarization covariance matrix. The principal merit of this approach is its ability to preserve intricate image details while effectively reducing noise. To substantiate the efficacy of the proposed method, Gaofen-3 QPSI (C-band, quad-polarization) time-series data were utilized for validation. The results demonstrate that the proposed method outperforms traditional filtering methods significantly in preserving image detail information.

Accurate assessment of canopy density is crucial for forest resource management, ecological monitoring, and carbon cycle research. Currently, high-resolution remote sensing imagery is extensively utilized for canopy density estimation through statistical analyses based on spectral and structural features, as well as machine learning models. Among these, deep learning models like U-Net have gained widespread application in recent years due to their automatic feature extraction capabilities, enabling canopy density estimation through single spectral and spatial features. However, these approaches often exhibit limitations in estimation accuracy and adaptability when confronted with the spectral and structural diversity of different tree species. Additionally, existing studies seldom integrate tree species information as auxiliary features into the canopy density estimation process, resulting in reduced reliability and accuracy of estimates in heterogeneous forest environments. To address these challenges, advanced multi-task learning frameworks are essential, integrating tree species information into the feature extraction process and simultaneously optimizing canopy density estimation. The Swin Transformer employs a hierarchical structure and self-attention mechanism to effectively capture both local and global image features, thereby enhancing the model's understanding of complex spatial structures. Leveraging these capabilities, this study develops a multi-task, tree species-driven canopy density estimation method that simultaneously extracts tree species information and integrates it as auxiliary input to improve canopy density predictions. The study was conducted in Wangyedian Forest Farm, Inner Mongolia (500 km2) utilizing Sentinel-2 satellite imagery. Compared to UNet, the proposed method reduced Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) by approximately 4% and 8%, respectively, while increasing the coefficient of determination (R) and F1 score by around 6% and 4%. In the tree species-driven canopy density estimation task, incorporating tree species information decreased RMSE for Pinus tabuliformis, Larix, broadleaf forests, and mixed forests by approximately 3%, 5%, 6%, and 2%. Error analysis revealed that the proposed method had smaller estimation errors for Pinus tabuliformis and mixed forests, while errors for Larix and broadleaf forests were slightly higher due to spectral feature overlaps with other tree species.

Airborne LiDAR (ALS) data have been extensively utilized for aboveground biomass (AGB) estimation; however, the high acquisition costs make it challenging to attain wall-to-wall estimation across large regions. Some studies have leveraged ALS data as intermediate variables to amplify sample sizes, thereby reducing costs and enhancing sample representativeness and model accuracy, but the cost issue remains in larger-scale estimations. Satellite LiDAR data, offering a broader dataset that can be acquired quickly with lower costs, can serve as an alternative intermediate variable for sample expansion. In this study, we employed a three-stage up-scaling approach to estimate forest AGB and introduced a method for quantifying estimation uncertainty. Based on the established three-stage general-hierarchical-model-based estimation inference (3sGHMB), an RK-3sGHMB inference method is proposed to make use of the regression-kriging (RK) method, and then it is compared with conventional model-based inference (CMB), general hierarchical model-based inference (GHMB), and improved general hierarchical model-based inference (RK-GHMB) to estimate forest AGB and uncertainty at both the pixel and forest farm levels. This study was carried out by integrating plot data, sampled ALS data, wall-to-wall Sentinel-2A data, and airborne P-SAR data. The results show that the accuracy of CMB (Radj2 = 0.37, RMSE = 33.95 t/ha, EA = 63.28%) is lower than that of GHMB (Radj2 = 0.38, RMSE = 33.72 t/ha, EA = 63.53%), while it is higher than that of 3sGHMB (Radj2 = 0.27, RMSE = 36.58 t/ha, EA = 60.43%). Notably, RK-GHMB (Radj2 = 0.60, RMSE= 27.07 t/ha, EA = 70.72%) and RK-3sGHMB (Radj2 = 0.55, RMSE = 28.55 t/ha, EA = 69.13%) demonstrate significant accuracy enhancements compared to GHMB and 3sGHMB. For population AGB estimation, the precision of the proposed RK-3sGHMB (p = 94.44%) is the highest, providing that there are sufficient sample sizes in the third stage, followed by RK-GHMB (p = 93.32%) with sufficient sample sizes in the second stage, GHMB (p = 90.88%), 3sGHMB (p = 88.91%), and CMB (p = 87.96%). Further analysis reveals that the three-stage model, considering spatial correlation at the third stage, can improve estimation accuracy, but the prerequisite is that the sample size in the third stage must be sufficient. For large-scale estimation, the RK-3sGHMB model proposed herein offers certain advantages.

Complete and accurate burned area map data are needed to document spatial and temporal patterns of fires, to quantify their drivers, and to assess the impacts on human and natural systems. To achieve the the purpose of identifying burned area accurately and efficiency from remote sensing images, a lightweight deep learning model is proposed based on Deeplab V3+, which employs the combination of attention mechanism and deep transitive transfer learning (DTTL) strategy. The lightweight MobileNet V2 network integrated with Convolutional Block Attention Module (CBAM) is designed as the backbone network to replace the traditional time-consuming Xception of Deeplab V3+. The attention mechanism is introduced to enhance the recognition ability of the proposed deep learning model, and the deep transitive transfer learning strategy is adopted to solve the problem of incorrect identification of the burned area and discontinuous edge details caused by insufficient sample size during the extraction process. For the process of DTTL, the improved Deeplab V3 + network was first pre-trained on ImageNet. Sequentially, WorldView-2 and the Sentinel-2 dataset were employed to train the proposed network based on the ImageNet pre-trained weights. Experiments were conducted to extract burned area from remote sensing images based on the trained model, and the results show that the proposed methodology can improve extraction accuracy with OA of 92.97% and Kappa of 0.819, which is higher than the comparative methods, and it can reduce the training time at the same time. We applied this methodology to identify the burned area in Western Attica region of Greece, and a satisfactory result was achieved with. OA of 93.58% and Kappa of 0.8265. This study demonstrates the effectiveness of the improved Deeplab V3 + in identifying forest burned area. which can provide valuable information for forest protection and monitoring.

Currently, it is very important to accurately estimate growing stock volumes; it is crucial for quantitatively assessing forest growth and formulating forest management plans. It is convenient and quick to use the Structure from Motion (SfM) algorithm in computer vision to obtain 3D point cloud data from captured highly overlapped stereo photogrammetry images, while the optimal algorithm for estimating growing stock volume varies across different data sources and forest types. In this study, the performance of UAV stereo photogrammetry (USP) in estimating the growing stock volume (GSV) using three machine learning algorithms for a coniferous plantation in Northern China was explored, as well as the impact of point density on GSV estimation. The three machine learning algorithms used were random forest (RF), K-nearest neighbor (KNN), and support vector machine (SVM). The results showed that USP could accurately estimate the GSV with R2 = 0.76–0.81, RMSE = 30.11–35.46, and rRMSE = 14.34%–16.78%. Among the three machine learning algorithms, the SVM showed the best results, followed by RF. In addition, the influence of point density on the estimation accuracy for the USP dataset was minimal in terms of R2, RMSE, and rRMSE. Meanwhile, the estimation accuracies of the SVM became stable with a point density of 0.8 pts/m2 for the USP data. This study evidences that the low-density point cloud data derived from USP may be a good alternative for UAV Laser Scanning (ULS) to estimate the growing stock volume of coniferous plantations in Northern China.

Information about the distribution of coniferous forests holds significance for enhancing forestry efficiency and making informed policy decisions. Accurately identifying and mapping coniferous forests can expedite the achievement of Sustainable Development Goal (SDG) 15, aimed at managing forests sustainably, combating desertification, halting and reversing land degradation, and halting biodiversity loss. However, traditional methods employed to identify and map coniferous forests are costly and labor-intensive, particularly in dealing with large-scale regions. Consequently, a methodological framework is proposed to identify coniferous forests in northwestern Liaoning, China, in which there are semi-arid and barren environment areas. This framework leverages a multi-classifier fusion algorithm that combines deep learning (U2-Net and Resnet-50) and shallow learning (support vector machines and random forests) methods deployed in the Google Earth Engine. Freely available remote sensing images are integrated from multiple sources, including Gaofen-1 and Sentinel-1, to enhance the accuracy and reliability of the results. The overall accuracy of the coniferous forest identification results reached 97.6%, highlighting the effectiveness of the proposed methodology. Further calculations were conducted to determine the area of coniferous forests in each administrative region of northwestern Liaoning. It was found that the total area of coniferous forests in the study area is about 6013.67 km2, accounting for 9.59% of northwestern Liaoning. The proposed framework has the potential to offer timely and accurate information on coniferous forests and holds promise for informed decision making and the sustainable development of ecological environment.

Research investigating the estimation ability of forest stock volume combining multiband polarimetric SAR(PolSAR)has hardly been explored,particularly the complementarity between long wavelengths,such as P-band and other shorter wavelengths.This study takes cold temperate coniferous forests in Inner Mongolia as the research object.Having available a multiband stack of airborne P-,L-,S-,C-,and X-band PolSAR data acquired by the high-resolution airborne multidimensional space joint-observation SAR(MSJosSAR)system,the aim is to analyze systematically the response and sensitivity of polarimetric characteristics in different bands to forest stock and evaluate the performance of forest stock retrieval using single and multiband PolSAR data. First,geocoding and terrain radiometric correction were performed on multiband PolSAR data,and then a polarimetric feature set containing backscatter intensity and polarization decomposition components was extracted.Second,on the basis of the water cloud model and correlation coefficient,the response law and sensitivity of polarimetric characteristics in different bands to forest stock was analyzed.Finally,machine learning algorithms were used to perform feature selection and modeling,and the ability of each band and jointly with multiband to estimate forest stock was evaluated. The response of backscatter intensity in different bands to forest stock shows a similar upward trend,but the saturation point varies depending on wavelength and polarimetric channel.Among them,the saturation point for the P-band is higher than 160 m3/ha,whereas it does not exceed 110 m3/ha for the other bands.In addition,the correlation between forest stock and the P-band,L/S-band,and C/X-band decreases in order,with values above 0.6,between 0.3 and 0.4,and below 0.3,respectively.When forest stock was estimated on the basis of a single band,the accuracy of the P-band was 73.79％,and the accuracy of other bands did not exceed 60％.When multiband joint estimation was used,the estimation accuracy of L-or S-band and P-band joint estimation was approximately 2％higher than using P-band alone.The contribution of adding the C-or X-band to the accuracy improvement was minimal.The best estimation performance was achieved through the combination of all bands with an accuracy of 77.25％. Considering various indicators,such as signal dynamic range,saturation point,and correlation,the P-band exhibits the highest sensitivity to forest stock,followed by the L/S-band,and the C/X-band,which is the least sensitive.Therefore,when estimating forest stock using PolSAR data,the P-band should be the first choice.Additionally,when using multiband joint estimation,the combination of P-and L-or S-band should be preferred.In recent years,long-wavelength SAR satellites are being vigorously developed from China and overseas,e.g.,China's LT-1 satellite is already in orbit,ESA BIOMASS and NASA-ISRO NISAR missions are about to be launched,and China's civil P-band SAR satellite has also entered the preliminary research stage.The above long-wavelength SAR satellites will greatly enhance the estimation ability of regional forest stock in our country and provide strong support for the refined and scientific management of forest resources.

Accurate estimation of forest aboveground biomass (AGB) is crucial for research on terrestrial carbon cycling and global climate change. In this study, we introduce an improved approach for estimating forest AGB combining P-band and X-band interferometric synthetic aperture radar (InSAR) data. Forest AGB was estimated by combining unbiased forest height and volume backscatter intensity. For forest height, a multilayer model and subaperture decomposition technology were used to remove the penetration bias of the X-band and reduce the effects of forest scatterers on the extraction of a pure understory terrain phase based on P-band, respectively. For volume backscatter intensity, a ground cancellation algorithm based on P-band InSAR was used to eliminate ground scattering contributions unrelated to forest AGB. The proposed method was validated using airborne P-band InSAR data and spaceborne X-band InSAR data gathered over the study area on the Saihanba Forest Farm in Hebei, China. The unbiased forest height and volume backscatter intensity had stronger correlations with forest AGB than estimates derived from unimproved features. The proposed method returned high-precision estimates of forest AGB with an accuracy of 83.73%, an improvement of 8.80% over an estimate derived from unoptimized features. Additionally, AGB estimates combined with forest height and backscatter intensity were greater than those based on a single feature, with the contribution of the former is greater than that of the latter.

Fine-grained identification of forest types and tree species represents a critical aspect of forest resource inventory and monitoring. The use of airborne hyperspectral remote sensing imagery stands out for its ability to finely differentiate among tree species, leveraging its exceptional spatial resolution and rich spectral details. However, this approach is limited by several challenges (e.g., high spectral correlation and information redundancy). In accordance, the adoption of a lightweight deep learning approach in the form of a few-shot learning model can effectively resolve the challenges of multi-forest tree species classification. Therefore, integrating a data dimensionality reduction algorithm with a few-shot classification model presents a promising avenue for resolving the fine-grained classification of forest tree species. In this study, we propose the innovative classification framework FAST 3D-CNN P-Net. This framework utilizes CNN for band selection, enhances the fine-grained identification process in hyperspectral data, and integrates an optimized FAST 3D-CNN into the P-Net classifier (a few-shot classifier). First, a CNN-based band selection method is employed to learn the nonlinear dependencies between spectral bands, assign weights to rank the bands, and reconstruct the global spectral information using the most informative bands. It then constructs a novel classification model, designated FAST 3D-CNN P-Net, through the integration of an optimal 3D-CNN with a prototypical network. To enhance classification performance, the FAST 3D-CNN P-Net utilizes reconstructed hyperspectral images derived from the band selection results as input. The effectiveness of the proposed framework was assessed with the airborne GFF dataset and the widely accessible medium-resolution hyperspectral datasets, Indian Pines (IP) and Kennedy Space Center (KSC). The overall classification accuracy reached 98.33 % for the GFF dataset and 97.21 % and 99.43 % for the IP and KSC, respectively, exhibiting performance superiority compared to the standalone 3D-CNN classification network. This classification framework demonstrates efficiency in selecting a subset of hyperspectral bands with minimal redundancy, empowering the rapid and accurate classification and mapping of tree species in complicated, multi-species forest stands, even with a limited quantity of labeled samples.

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. Index Terms-Forest biomass estimation, Feature

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.

Polarimetric calibration is essential for the pre-processing of Polarimetric Synthetic Aperture Radar (PolSAR) data because it effectively mitigates polarimetric distortions in the measured PolSAR data. Traditional methods of polarimetric calibration employ man-made calibrators that offer high accuracy. However, the frequency of calibration is often limited due to the labor-intensive and time-consuming nature of deploying such calibrators. Some polarimetric calibration methods based on distributed targets in nature enable more frequent calibration. Nevertheless, these methods are constrained by the availability of specific distributed targets with known polarimetric properties for estimating parameters related to co-polarization channel imbalance (co-pol-imba) parameters. If distributed targets are not appropriately selected or suitable targets are absent within the image scene, the accuracy of calibration will be compromised. To address this challenge, this paper introduces the idea of cross-calibration, which uses calibrated PolSAR data to determine the real polarimetric property of distributed targets and cross-estimates co-pol-imba parameters for uncalibrated PolSAR data. Furthermore, considering the disparities in imaging geometry between calibrated and uncalibrated PolSAR data, the Cross-Co-Polarization Ratio (CCPR) coefficients are proposed to select stable distributed targets in both the calibrated PolSAR data and the uncalibrated PolSAR data. This approach not only takes into account the impact of co-pol-imba on feature extraction but can also be applied in a broader range of scenes. The real data experiments on the Uninhabited Aerial Vehicle Synthetic Aperture Radar (UAVSAR) data show that the estimated results of the cross-calibration method for co-pol-imba are comparable to the polarimetric calibration method based on man-made calibrators, with an amplitude difference of 0.17 dB and a phase difference of 0.69°.