Search

The assessment of forest structural parameters is crucial for understanding carbon storage, habitat suitability, and timber stock. However, the labor-intensive and expensive nature of field measurements, coupled with inadequate sample sizes for large-scale modeling, poses challenges. To address the forest structure parameters in the Western Tianshan Mountains, this study used UAV-LiDAR to gather extensive sample data. This approach was enhanced by integrating Sentinel satellite and topographic data and using a Bayesian-Random Forest model to estimate forest canopy height, average height, density, and aboveground biomass (AGB). Validation against independent LiDAR-derived samples confirmed the model's high accuracy, with coefficients of determination (R2) and root mean square errors (RMSE) indicating strong predictive performance (R2 = 0.63, RMSE = 5.06 m for canopy height; R2 = 0.64, RMSE = 2.88 m for average height; R2 = 0.68, RMSE = 62.84 for density; and R2 = 0.59, RMSE = 29.71 Mg/ha for AGB). Notably, the crucial factors include DEM, Sentinel-1 (VH and VV backscatter in dB), and Sentinel-2 (B6, B8A, and B11 bands). These factors contribute significantly to the modeling of forest structure. This technology aims to expedite and economize forest surveys while augmenting the range of forest parameters, especially in remote and rugged terrains. Using a wealth of UAV-LiDAR data, this outcome surpasses its counterparts' by providing essential insights for exploring climate change effects on Central Asian forests, facilitating precise carbon stock quantification, and enhancing knowledge of forest ecosystems. [ABSTRACT FROM AUTHOR]

Many forest cover maps have been generated by using optical and/or microwave images, but these forest cover maps have large area and spatial discrepancies. To date, few studies have assessed forest cover maps in terms of two biophysical parameters used in forest definition: canopy height and canopy coverage. We generated annual forest cover maps from 2007 to 2010 and evergreen forest cover maps from 2000 to 2021 in the Brazilian Amazon using the images from the Phased Array type L-band Synthetic Aperture Radar and the time series images from the Moderate Resolution Imaging Spectroradiometer, using the forest definition of the Food and Agriculture Organization of the United Nations (>5 m tree height and >10 % canopy coverage) as the reference. We used the canopy height and canopy coverage datasets from the Geoscience Laser Altimeter System during 2003–2007 to assess annual forest cover maps from 2007 to 2010 and annual evergreen forest cover maps from 2003 to 2007, and the results show high accuracy of these forest cover and evergreen forest cover maps. These annual forest cover maps and annual evergreen forest cover maps provide data support for the analyses of the causes, processes, and consequences of forest cover changes in the Brazilian Amazon (10.6084/m9.figshare.21445626; Qin and Xiao, 2022a; 10.6084/m9.figshare.21445590; Qin and Xiao, 2022b). [ABSTRACT FROM AUTHOR]

Mangrove forests, mostly found in the intertidal zone, are among the highest-productivity ecosystems and have great ecological and economic value. The accurate mapping of mangrove forests is essential for the scientific management and restoration of mangrove ecosystems. However, it is still challenging to perform the rapid and accurate information mapping of mangrove forests due to the complexity of mangrove forests themselves and their environments. Utilizing multi-source remote sensing data is an effective approach to address this challenge. Feature extraction and selection, as well as the selection of classification models, are crucial for accurate mangrove mapping using multi-source remote sensing data. This study constructs multi-source feature sets based on optical (Sentinel-2) and SAR (synthetic aperture radar) (C-band: Sentinel-1; L-band: ALOS-2) remote sensing data, aiming to compare the impact of three feature selection methods (RFS, random forest; ERT, extremely randomized tree; MIC, maximal information coefficient) and four machine learning algorithms (DT, decision tree; RF, random forest; XGBoost, extreme gradient boosting; LightGBM, light gradient-boosting machine) on classification accuracy, identify sensitive feature variables that contribute to mangrove mapping, and formulate a classification framework for accurately recognizing mangrove forests. The experimental results demonstrated that using the feature combination selected via the ERT method could obtain higher accuracy with fewer features compared to other methods. Among the feature combinations, the visible bands, shortwave infrared bands, and the vegetation indices constructed from these bands contributed the greatest to the classification accuracy. The classification performance of optical data was significantly better than SAR data in terms of data sources. The combination of optical and SAR data could improve the accuracy of mangrove mapping to a certain extent (0.33% to 4.67%), which is essential for the research of mangrove mapping in a larger area. The XGBoost classification model performed optimally in mangrove mapping, with the highest overall accuracy of 95.00% among all the classification models. The results of the study show that combining optical and SAR remote sensing data with the ERT feature selection method and XGBoost classification model has great potential for accurate mangrove mapping at a regional scale, which is important for mangrove restoration and protection and provides a reliable database for mangrove scientific management. [ABSTRACT FROM AUTHOR]

Forest aboveground biomass (AGB) retrieval using synthetic aperture radar (SAR) backscatter has received extensive attention. The water cloud model (WCM), because of its simplicity and physical significance, has been one of the most commonly used models for estimating forest AGB using SAR backscatter. Nevertheless, forest AGB estimation using the WCM is usually based on simplified assumptions and empirical fitting, leading to results that tend to overestimate or underestimate. Moreover, the physical connection between the model and the polarimetric synthetic aperture radar (PolSAR) is not established, which leads to the limitation of the inversion scale. In this paper, based on the fully polarimetric SAR data from the Advanced Land Observing Satellite-2 (ALOS-2) Phased Array-type L-band Synthetic Aperture Radar (PALSAR-2), the relative contributions of the three major scattering mechanisms were first analyzed in a hilly area of southern China. On this basis, the traditional WCM was extended by considering the secondary scattering mechanism. Then, to establish the direct relationship between the vegetation scattering mechanism and forest AGB, a new relationship equation between the PolSAR decomposition model and the improved water cloud model (I-WCM) was constructed without the help of external data. Finally, a nonlinear iterative method was used to estimate the forest AGB. The results show that volume scattering is the dominant mechanism, accounting for more than 60%. Double-bounce scattering accounts for the smallest fraction, but still about 10%, which means that the contribution of the double-bounce scattering component is not negligible in forested areas because of the strong penetration capability of the long-wave SAR. The modified method provides a correlation coefficient R2 of 0.665 and a root mean square error (RMSE) of 21.902, which is an improvement of 36.42% compared to the traditional fitting method. Moreover, it enables the extraction of forest parameters at the pix scale using PolSAR data without the need for low-resolution external data and is thus helpful for high-resolution mapping of forest AGB. [ABSTRACT FROM AUTHOR]

Satellite-based forest alert systems are an important tool for ecosystem monitoring, planning conservation, and increasing public awareness of forest cover change. Continuous monitoring in tropical regions, such as those experiencing pronounced monsoon seasons, can be complicated by spatially extensive and persistent cloud cover. One solution is to use Synthetic Aperture Radar (SAR) imagery acquired by the European Space Agency's Sentinel-1A and B satellites. The Sentinel 1A and B satellites acquire C-band radar data that penetrates cloud cover and can be acquired during the day or night. One challenge associated with operational use of radar imagery is that the speckle associated with the backscatter values can complicate traditional pixel-based analysis approaches. A potential solution is to use deep learning semantic segmentation models that can capture predictive features that are more robust to pixel-level noise. In this analysis, we present a prototype SAR-based forest alert system that utilizes deep learning classifiers, deployed using the Google Earth Engine cloud computing platform, to identify forest cover change with near real-time classification over two Cambodian wildlife sanctuaries. By leveraging a pre-existing forest cover change dataset derived from multispectral Landsat imagery, we present a method for efficiently developing a SAR-based semantic segmentation dataset. In practice, the proposed framework achieved good performance comparable to an existing forest alert system while offering more flexibility and ease of development from an operational standpoint. [ABSTRACT FROM AUTHOR]

Deep learning (DL) models are gaining popularity in forest variable prediction using Earth observation (EO) images. However, in practical forest inventories, reference datasets are often represented by plot- or stand-level measurements, while high-quality representative wall-to-wall reference data for end-to-end training of DL models are rarely available. Transfer learning facilitates expansion of the use of deep learning models into areas with sub-optimal training data by allowing pretraining of the model in areas where high-quality teaching data are available. In this study, we perform a "model transfer" (or domain adaptation) of a pretrained DL model into a target area using plot-level measurements and compare performance versus other machine learning models. We use an earlier developed UNet based model (SeUNet) to demonstrate the approach on two distinct taiga sites with varying forest structure and composition. The examined SeUNet model uses multi-source EO data to predict forest height. Here, EO data are represented by a combination of Copernicus Sentinel-1 C-band SAR and Sentinel-2 multispectral images, ALOS-2 PALSAR-2 SAR mosaics and TanDEM-X bistatic interferometric radar data. The training study site is located in Finnish Lapland, while the target site is located in Southern Finland. By leveraging transfer learning, the SeUNet prediction achieved root mean squared error (RMSE) of 2.70 m and R2 of 0.882, considerably more accurate than traditional benchmark methods. We expect such forest-specific DL model transfer can be suitable also for other forest variables and other EO data sources that are sensitive to forest structure. [ABSTRACT FROM AUTHOR]

Change detection methods based on Earth Observations are increasingly used to monitor rainforest in the intertropical band. Until recently, deforestation monitoring was mainly based on remotely sensed optical images which often face limitations in humid tropical areas due to frequent cloud coverage. This leads to late detections of disturbance events. Since the launch of Sentinel-1 acquiring images with a revisit time of 12 days and a spatial resolution of 5 × 20 m in Brazil, Synthetic Aperture Radar (SAR) images have been increasingly used to monitor deforestation. In this study, we propose a multitemporal version of the change detection method we previously applied to timeseries of Sentinel-1 SAR images, to monitor deforestation/degradation in the Congo rainforest. Our approach is based on a Cumulative Sum (CuSum) method combined with a spatial recombination of Critical Thresholds (CuSum cross-T c). The newly developed multitemporal CuSum method (ReCuSum) was applied to a time-series of 82 dual polarization (VH, VV) Ground Range Detected (GRD) Sentinel-1 images acquired in the Para State, in the Brazilian Amazonia, from 29/09/2016 to 01/07/2019. The ReCuSum method consists of iteratively applying the CuSum cross-T c to monitor multiple changes in a time-series by splitting the time-series at each date of detected change and by independently iterating over the time periods resulting from the splits. The number of changes in the time-series was then analysed according to the vegetation type (Forest, non-forest vegetation) determined by visual inspection of optical Sentinel-2 image and PlanetScope monthly mosaic. This showed a difference between non-forest vegetation and forested areas. A threshold based on the number of changes (T nbc) was then developed to differentiate forest from non-forest disturbances. The ability to monitor non-forest vegetation was analysed: the CuSum cross-T c detected up to 90% of the total non-forest vegetation area over the study region in the past period. After removing past disturbances and past non-forest vegetation, then removing the pixels covered with non-forest vegetations based on T nbc , the application of the ReCuSum led to a precision of 81%, a recall of 68%, a kappa coefficient of 0.72 and a F1- score of 0.74 in forest disturbance monitoring. According to these results, ReCuSum applied to Sentinel-1 time-series of images can be used for efficient forest disturbance monitoring and for generating a forest / non-forest map after the application of newly developed post-processing steps. Sentinel-1 imagery can be used for both Forest / Non-forest mapping and for forest disturbance detection. ReCuSum was released as an open-source GIT project available at: https://forgemia.inra.fr/bertrand.ygorra/cusum-deforestation_monitoring. [ABSTRACT FROM AUTHOR]

Polarimetric Synthetic Aperture Radar (PolSAR) images with dual polarization modes have great potential to map forest stock volume (FSV) by excellent penetration capabilities and distinct microwave scattering processes. However, the response of these SAR data to FSV is still uncertain in the non-growing season. To further interpret the response of FSV to different dual polarization SAR images, three types of dual polarization SAR images (GF-3, Sentinel-1, and ALOS-2) were initially acquired in coniferous planted forest in the non-growing season. Then, sensitivity between FSV and all alternative features extracted from each type of SAR image was analyzed to express the response of FSV to dual polarization SAR images with bands and polarization modes in the non-growing season in deciduous (Larch) and evergreen (Chinese pine) forests. Finally, mapped FSV using single and combined dual polarization images were derived by optimal feature sets and four machine learning models, respectively. The combined effects were also analyzed to clarify the difference of bands and polarization modes in deciduous and evergreen forests in the non-growing season. The results demonstrated that the backscattering energy from different sensors is significantly different in Chinese pine, and the difference is gradually reduced in Larch forests. It is also implied that the polarization mode is more important than penetration capability in mapping forest FSV in deciduous forest in the non-growing season. By comparing the accuracy of mapped FSV using single and combined images, combined images have more capability to improve the accuracy and reliability of mapped FSV. Meanwhile, it is confirmed that compensation effects with bands and polarization modes not only have great potential to delay the saturation phenomenon, but also have the capability to reduce errors caused by overestimation. [ABSTRACT FROM AUTHOR]

Copernicus Sentinel-1 images are widely used for forest mapping and predicting forest growing stock volume (GSV) due to their accessibility. However, certain important aspects related to the use of Sentinel-1 time series have not been thoroughly explored in the literature. These include the impact of image time series length on prediction accuracy, the optimal feature selection approaches, and the best prediction methods. In this study, we conduct an in-depth exploration of the potential of long time series of Sentinel-1 SAR data to predict forest GSV and evaluate the temporal dynamics of the predictions using extensive reference data. Our boreal coniferous forests study site is located near the Hyytiälä forest station in central Finland and covers an area of 2500 km2 with nearly 17,000 stands. We considered several prediction approaches and fine-tuned them to predict GSV in various evaluation scenarios. Our analyses used 96 Sentinel-1 images acquired over three years. Different approaches for aggregating SAR images and choosing feature (predictor) variables were evaluated. Our results demonstrate a considerable decrease in the root mean squared errors (RMSEs) of GSV predictions as the number of images increases. While prediction accuracy using individual Sentinel-1 images varied from 85 to 91 m3/ha RMSE, prediction accuracy with combined images decreased to 75.6 m3/ha. Feature extraction and dimension reduction techniques facilitated the achievement of near-optimal prediction accuracy using only 8–10 images. Examined methods included radiometric contrast, mutual information, improved k-Nearest Neighbors, random forests selection, Lasso, and Wrapper approaches. Lasso was the most optimal, with RMSE reaching 77.1 m3/ha. Finally, we found that using assemblages of eight consecutive images resulted in the greatest accuracy in predicting GSV when initial acquisitions started between September and January. [ABSTRACT FROM AUTHOR]

Accurate estimation of forest height over a large area is beneficial to reduce the uncertainty of forest carbon sink estimation, which is of great significance to the terrestrial carbon cycle, global climate change, forest resource management, and forest-related scientific research. Forest height inversion using polarimetric interferometry synthetic aperture radar (PolInSAR) data through Random volume over ground (RVoG) models has demonstrated great potential for large-area forest height mapping. However, the wavelength and baseline length used for the PolInSAR data acquisition plays an important role during the forest height estimation procedure. In this paper, X–, C–, L–, and P–band PolInSAR datasets with four different baseline lengths were simulated and applied to explore the effects of wavelength and baseline length on forest height inversion using RVoG models. Hierarchical Bayesian models developed with a likelihood function of RVoG model were developed for estimated results uncertainty quantification and decrease. Then a similar procedure was applied in the L– and P–band airborne PolInSAR datasets with three different baselines for each band. The results showed that (1) Wavelength showed obvious effects on forest height inversion results with the RVoG model. For the simulated PolInSAR datasets, the L– and P–bands performed better than the X– and C–bands. The best performance was obtained at the P–band with a baseline combination of 10 × 4 m with an absolute error of 0.05 m and an accuracy of 97%. For the airborne PolInSAR datasets, an L–band with the longest baseline of 24 m in this study showed the best performance with R2 = 0.64, RMSE = 3.32 m, and Acc. = 77.78%. (2) It is crucial to select suitable baseline lengths to obtain accurate forest height estimation results. In the four baseline combinations of simulated PolInSAR datasets, the baseline combination of 10 × 4 m both at the L– and P–bands performed best than other baseline combinations. While for the airborne PolInSAR datasets, the longest baseline in three different baselines obtained the highest accuracy at both L– and P–bands. (3) Bayesian framework is useful for estimation results uncertainty quantification and decrease. The uncertainties related to wavelength and baseline length. The uncertainties were reduced obviously at longer wavelengths and suitable baselines. [ABSTRACT FROM AUTHOR]

Nowadays, polarimetric interferometric synthetic aperture radar (PolInSAR) has attracted increasing attention for the simultaneous acquisition of scattering and terrain information. The scattering mechanism vectors corresponding to the optimal coherences have shown great effect in land cover classification and forest mapping. To extract scattering mechanism vectors, different optimization algorithms are proposed but always accompanied with spatial discontinuities. This letter proposes a method to optimize scattering mechanism vectors based on nonlocal estimations. A pixel‐by‐pixel optimization model is established by calculating patch‐based nonlocal similarity, which effectively reduces the uncertainty of scattering mechanism vector estimation while maintaining high resolution. The phase recovery experiment is conducted on simulated data and the root mean square errors (RMSEs) of three algorithms are compared. The proposed method performs better with a RMSE of 0.22 in the whole image, against 0.25 and 1.27 for other algorithms. The recovery results in the target region (RMSE 0.11 against 0.21 and 0.99) and linear edge region (RMSE 0.26 against 0.49 and 1.52) also indicate strong ability in detail preservation. Then the observed ESAR data is utilized for further assessment. It demonstrates that the proposed method has a significant improvement to reduce singularities and distinguish artificial buildings. [ABSTRACT FROM AUTHOR]

Unaffected by cloud cover and solar illumination, synthetic aperture radar (SAR) images have great capability to map forest growing stem volume (GSV) in complex biophysical environments. Up to now, c-band dual-polarization Gaofen-3 (GF-3) SAR images, acquired by the first Chinese civilian satellite equipped with multi-polarized modes, are rarely applied in mapping forest GSV. To evaluate the capability of dual-polarization GF-3 SAR images in mapping forest GSV, several proposed derived features were initially extracted by mathematical operations and applied to obtain optimal feature sets by different feature sorting methods and feature selection methods. Then, the maps of GSV in an evergreen coniferous forest were inverted by various machine learning algorithms and stacking ensemble learning methods with different strategies. The results implied that backscattering coefficients and partially proposed derived features showed high sensitivity to the forest GSV, and the saturation phenomenon also obviously occurred once the forest GSV was larger than 300 m3/ha. Furthermore, the results showed that the accuracy of the mapped GSV was significantly improved using the stacking ensemble learning methods. Using various optimal feature sets and base models (MLR, KNN, SVM, and RF), the rRMSE values mainly ranged from 30% to 40%. After using the stacking ensemble learning methods, the values of rRMSE ranged from 16.71% to 20.51%. This confirmed that dual-polarization GF-3 images have great potential to map forest GSV in evergreen coniferous forests. [ABSTRACT FROM AUTHOR]

Mapping tropical forest aboveground biomass (AGB) is important for quantifying emissions from land use change and evaluating climate mitigation strategies but remains a challenging problem for remote sensing observations. Here, we evaluate the capability of mapping AGB across a dense tropical forest using tomographic Synthetic Aperture Radar (TomoSAR) measurements at P-band frequency that will be available from the European Space Agency's BIOMASS mission in 2024. To retrieve AGB, we compare three different TomoSAR reconstruction algorithms, back-projection (BP), Capon, and MUltiple SIgnal Classification (MUSIC), and validate AGB estimation from models using TomoSAR variables: backscattered power at 30 m height, forest height (FH), backscatter power metric (Q), and their combination. TropiSAR airborne campaign data in French Guiana, inventory plots, and airborne LiDAR measurements are used as reference data to develop models and calculate the AGB estimation uncertainty. We used univariate and multivariate regression models to estimate AGB at 4-ha grid cells, the nominal resolution of the BIOMASS mission. Our results show that the BP-based variables produced better AGB estimates compared to their counterparts, suggesting a more straightforward TomoSAR processing for the mission. The tomographic FH and AGB estimation have an average relative uncertainty of less than 10% with negligible systematic error across the entire biomass range (~ 200–500 Mg ha−1). We show that the backscattered power at 30 m height at HV polarization is the best single measurement to estimate AGB with significantly better accuracy than the LiDAR height metrics, and combining it with FH improved the accuracy of AGB estimation to less than 7% of the mean. Our study implies that using multiple information from P-band TomoSAR data from the BIOMASS mission provides a new capability to map tropical forest biomass and its changes accurately. [ABSTRACT FROM AUTHOR]

Unaffected by cloud cover and solar illumination, synthetic aperture radar (SAR) images coupled with quad-polarimetric techniques have significant potential for mapping forest aboveground biomass (AGB) in the mountains of southern China. To improve the accuracy of mapping forest AGB, it is necessary to accurately interpret and evaluate the sensitivity of polarimetric features related to polarimetric response in complex forests. In this study, several rotated polarimetric features were extracted from L-band quad-polarimetric ALOS PALSAR-2 images based on uniform polarimetric matrix rotation theory. In addition, the sensitivity of rotated polarimetric features with forest parameters was evaluated by the Pearson correlation coefficient, sensitivity index (SI), and saturation levels. Ultimately, the forest AGB was mapped with various combinatorial feature sets by a proposed feature selection method based on the sensitivity index. The results illustrated that rotated polarimetric features extracted from the rotational domain have higher sensitivity with various forest parameters and higher saturation levels for mapping forests than other traditional features. After using the proposed feature selection method and combinatorial feature sets, the rRMSE of mapped forest AGB ranged from 22.5% to 33.9% for two acquired images, and the best result was obtained from the combination of three types of polarimetric features (BC + C4 + Ro). It is also confirmed that different types of features extracted from quad-polarimetric SAR images have better compensation effects and the accuracy of mapped forest AGB is significantly improved. [ABSTRACT FROM AUTHOR]

Annual forest maps at a high spatial resolution are necessary for forest management and conservation. Large uncertainties remain among the existing forest maps, because of different forest definitions, satellite datasets, in-situ training datasets, and mapping algorithms. In this study, we generated annual forest maps and evergreen forest maps at a 30-m resolution in the Contiguous United States (CONUS) during 2015-2017 by integrating microwave data (Phased Array type L-band Synthetic Aperture Radar (PALSAR-2)) and optical data (Landsat) using Knowledgebased algorithms. The resultant PALSAR-2/Landsat-based forest maps (PL-Forest) were compared with five major forest datasets in the CONUS: (1) the Landsat tree canopy cover from Global Forest Watch datasets (GFW-Forest), 2) the Landsat Vegetation Continuous Field datasets (Landsat VCF-Forest), (3) the National Land Cover Database 2016 (NLCD-Forest), (4) the Japan Aerospace Exploration Agency (JAXA) forest maps (JAXA-Forest), and (5) the Forest Inventory and Analysis (FIA) data from the USDA Forest Service (FIA-Forest). The forest structure data (tree canopy height and canopy coverage) derived from the lidar observations of the Geoscience Laser Altimetry System GLAS) onboard of NASA's Ice, Cloud, and land Elevation Satellite (ICESat-1) were used to assess the five forest datasets derived from satellite images. Using the forest definition by the Food and Agricultural Organization (FAO) of the United Nations, more forest pixels from the PL-Forest maps meet the FAO's forest definitions than the GFW-, Landsat VCF-, and JAXA-Forest datasets. Forest area estimates from the PL-Forest were close to those from the FIAForest statistics but higher than the GFW-Forest, NLCD-Forest and lower than the Landsat VCF-Forest, which highlights the potential of using both PL-Forest and FIA-Forest datasets to support the FAO's Global Forest Resources Assessment. Furthermore, the PL-based annual evergreen forest maps (PL-Evergreen Forest) showed reasonableconsistency with the NLCD product. Together with our previous work in South America and monsoon Asia, this study further demonstrates the potential of integrating PALSAR and Landsat images for developing annual forest maps and forest-type maps at high spatial resolution across the scales from region to the globe, which could be used to support FAO Global Forest Resources Assessments. The PL-Forest and PL-Evergreen Forest datasets are publicly available at https://doi.org/10.6084/m9.figshare.21270261 (Wang et al., 2022). [ABSTRACT FROM AUTHOR]

Many forest cover maps have been generated by using optical and/or microwave images and various forest definitions, but these forest cover maps have large discrepancies. Both forest definition and validation data used for accuracy assessment of forest cover maps are often considered as two of the major factors for the discrepancy among these forest cover maps. To date, few studies have assessed forest cover maps in terms of two biophysical parameters used in forest definition: (1) tree canopy height and (2) canopy coverage. We generated annual forest cover maps from 2007 to 2010 and evergreen forest cover maps from 2000 to 2021 in the Brazilian Amazon using the images from the Phased Array type L-band Synthetic Aperture Radar and Moderate Resolution Imaging Spectroradiometer, and the forest definition of the Food and Agriculture Organization (FAO) of the United Nations (>5-m tree height and >10% canopy coverage). The canopy height and coverage datasets from the Geoscience Laser Altimeter System during 2003–2007 were used to assess annual forest cover maps from 2007 to 2010 and evergreen forest cover maps from 2003 to 2007, and the results show high accuracy of these forest and evergreen forest cover maps in the Brazilian Amazon. [ABSTRACT FROM AUTHOR]

The interferometric synthetic aperture radar (InSAR) technique based on time–frequency (TF) analysis has great potential for mapping the forest canopy height model (CHM) at regional and global scales, as it benefits from the additional InSAR observations provided by the sublook decomposition. Meanwhile, due to the wider swath and higher spatial resolution of single-polarization data, InSAR has a higher observation efficiency in comparison with PolInSAR. However, the accuracy of the CHM inversion obtained by the TF-InSAR method is attenuated by its inaccurate coherent scattering modeling and uncertain parameter calculation. Hence, a new approach for CHM estimation based on single-baseline InSAR data and sublook decomposition is proposed in this study. With its derivation of the coherent scattering modeling based on the scattering matrix of sublook observations, a time–frequency based random volume over ground (TF-RVoG) model is proposed to describe the relationship between the sublook coherence and the forest biophysical parameters. Then, a modified three-stage method based on the TF-RVoG model is used for CHM retrieval. Finally, the two-dimensional (2-D) ambiguous error of pure volume coherence caused by residual ground scattering and temporal decorrelation is alleviated in the complex unit circle. The performance of the proposed method was tested with airborne L-band E-SAR data at the Krycklan test site in Northern Sweden. Results show that the modified three-stage method provides a root-mean-square error (RMSE) of 5.61 m using InSAR and 14.3% improvement over the PolInSAR technique with respect to the classical three-stage inversion result. An inversion accuracy of RMSE = 2.54 m is obtained when the spatial heterogeneity of CHM is considered using the proposed method, demonstrating a noticeable improvement of 32.8% compared with results from the existing method which introduces the fixed temporal decorrelation factor. [ABSTRACT FROM AUTHOR]

Spruce budworm (Choristoneura fumiferana; SBW) outbreaks in the northeastern USA and Canada are recurring phenomena leading to large-scale mortality of spruce (Picea sp.) and balsam fir (Abies balsamea (L.) Mill.) forests as susceptibility to SBW is primarily determined by the availability of host species and their maturity. Our study examined several satellite remote sensing (Sentinel-1 C-band synthetic aperture radar (SAR), PALSAR L-band SAR, and Sentinel-2 multispectral) and site variables over space and time to develop a method to produce large-scale SBW stand impact types and susceptibility maps in Maine, USA. We used two machine-learning algorithms (Random Forest, RF; Multi-Layer Perceptron, MLP) to map SBW host species where RF produced better results than MLP. Our best model with site (elevation and aspect) and Sentinel-2 data attained an overall accuracy (OA) of 83.4%. However, the addition of SAR variables did not improve the model further. Combining host species data with age data retrieved from Land Change Monitoring, Assessment, and Projection (LCMAP) products, we demonstrated that SBW susceptibility map (based on stand impact types) could be produced with an OA of 88.3%. The fine spatial resolution (20 m) maps derived from our study provide reliable products for landscape-level SBW interventions in the region. [ABSTRACT FROM AUTHOR]

Several generic methods have recently been developed for change detection in heterogeneous remote sensing data, such as images from synthetic aperture radar (SAR) and multispectral radiometers. However, these are not well-suited to detect weak signatures of certain disturbances of ecological systems. To resolve this problem we propose a new approach based on image-to-image translation and one-class classification (OCC). We aim to map forest mortality caused by an outbreak of geometrid moths in a sparsely forested forest-tundra ecotone using multisource satellite images. The images preceding and following the event are collected by Landsat-5 and RADARSAT-2, respectively. Using a recent deep learning method for change-aware image translation, we compute difference images in both satellites' respective domains. These differences are stacked with the original pre- and post-event images and passed to an OCC trained on a small sample from the targeted change class. The classifier produces a credible map of the complex pattern of forest mortality. [ABSTRACT FROM AUTHOR]

A timely and spatially high-resolution detection of drought-affected forest stands is important to assess and deal with the increasing risk of forest fires. In this paper, we present how multitemporal Sentinel-1 synthetic aperture radar (SAR) data can be used to detect drought-affected and fire-endangered forest stands in a spatially and temporally high resolution. Existing approaches for Sentinel-1 based drought detection currently do not allow to deal simultaneously with all disturbing influences of signal noise, topography and visibility geometry on the radar signal or do not produce pixel-based high-resolution drought detection maps of forest stands. Using a novel Sentinel-1 Radar Drought Index (RDI) based on temporal and spatial averaging strategies for speckle noise reduction, we present an efficient methodology to create a spatially explicit detection map of drought-affected forest stands for the year 2020 at the Donnersberg study area in Rhineland-Palatinate, Germany, keeping the Sentinel-1 maximum spatial resolution of 10 m × 10 m. The RDI showed significant (p < 0.05) drought influence for south, south-west and west-oriented slopes. Comparable spatial patterns of drought-affected forest stands are shown for the years 2018, 2019 and with a weaker intensity for 2021. In addition, the assessment for summer 2020 could also be reproduced with weekly repetition, but spatially coarser resolution and some limitations in the quality of the resulting maps. Nevertheless, the mean RDI values of temporally high-resolution drought detection maps are highly correlated (R2 = 0.9678) with the increasing monthly mean temperatures in 2020. In summary, this study demonstrates that Sentinel-1 data can play an important role for the timely detection of drought-affected and fire-prone forest areas, since availability of observations does not depend on cloud cover or time of day. [ABSTRACT FROM AUTHOR]

Time series of SAR imagery combined with reference ground data can be suitable for producing forest inventories. Copernicus Sentinel-1 imagery is particularly interesting for forest mapping because of its free availability to data users; however, temporal dependencies within SAR time series that can potentially improve mapping accuracy are rarely explored. In this study, we introduce a novel semi-supervised Long Short-Term Memory (LSTM) model, CrsHelix-LSTM, and demonstrate its utility for predicting forest tree height using time series of Sentinel-1 images. The model brings three important modifications to the conventional LSTM model. Firstly, it uses a Helix-Elapse (HE) projection to capture the relationship between forest temporal patterns and Sentinel-1 time series, when time intervals between datatakes are irregular. A skip-link based LSTM block is introduced and a novel backbone network, Helix-LSTM, is proposed to retrieve temporal features at different receptive scales. Finally, a novel semisupervised strategy, Cross-Pseudo Regression, is employed to achieve better model performance when reference training data are limited. CrsHelix-LSTM model is demonstrated over a representative boreal forest site located in Central Finland. A time series of 96 Sentinel-1 images are used in the study. The developed model is compared with basic LSTM model, attention-based bidirectional LSTM and several other established regression approaches used in forest variable mapping, demonstrating consistent improvement of forest height prediction accuracy. At best, the achieved accuracy of forest height mapping was 28.3% relative root mean squared error (rRMSE) for pixel-level predictions and 18.0% rRMSE on stand level. We expect that the developed model can also be used for modeling relationships between other forest variables and satellite image time series. [ABSTRACT FROM AUTHOR]

Accurately assessing forest structures at a national level is essential for promoting sustainable development, managing forests effectively, and fulfilling commitments under the Paris Agreement to combat climate change. In Japan, the traditional method of estimating forest carbon stock through National Forest Inventories (NFI) has certain limitations. The limited scope of forest survey data samples, coupled with the extended duration of surveys, typically spanning five years, induces significant biases in the carbon stock estimation. Achieving comprehensive forest mapping through the integration of remote sensing data and ground-referenced data for validation remains challenging due to the limited number and representativeness of referenced data, particularly at large scales. To address these issues, we collected LiDAR data from 17 sites across Japan between 2016 and 2022, covering an area of 4185.50 km2. This effort resulted in 212,067 squared plots, each with a standard size of 40 m × 40 m, which were used to develop regression models for forest structures. We utilized time-series analysis of Sentinel-2 data from 2015 to 2023 and Sentinel-2 data from different seasons in 2020. These datasets were integrated with 10 m resolution ALOS PALSAR-2 Synthetic Aperture Radar (SAR) data from JAXA and topographic data from the Shuttle Radar Topography Mission (SRTM), combined with machine learning algorithms to map forest structures throughout Japan. We initially produced a forest height map with a spatial resolution of 10 m and achieved a Relative Root Mean Square Error (rRMSE) of 21.58%. Subsequently, the estimated forest height was used as an additional predictor in machine learning models to estimate forest volume and aboveground biomass (AGB), achieving rRMSE values of 43.67% and 44.63%, respectively. Based on the predicted AGB, we calculated that the total aboveground carbon stock in Japanese forests is 1440.26 Teragrams (Tg) with an uncertainty of ±564.58 Tg. Our carbon stock estimation results outperformed those from ESA CCI, GEDI L4B, and the Ecosystem Demography Model-LiDAR (ED-LiDAR), representing the lowest rRMSE (18.07%) compared with the airborne LiDAR data. Overall, using time-series data from Sentinel-2 improved the estimation of forest structures. This study presents the first 10 m resolution, national-scale estimation of forest structures (height, volume, AGB), and forest carbon stock for the entirety of Japan. It introduces a new framework for estimating forest structures and carbon stock that supports forest management and sustainable development goals. • Mapping of multiple forest structures and carbon stock at 10-m resolution throughout Japan • Generating uncertainty maps of forest structures and carbon stock at 10-m resolution • Forest carbon stocks predictions outperformed those from other global products. [ABSTRACT FROM AUTHOR]

With the upcoming spaceborne synthetic aperture radar (SAR) missions (BIOMASS, LuTan-1, NISAR, and TanDEM-L), it will become possible to extract vegetation height at a global scale by utilizing spaceborne low-frequency (L- and P-band) polarimetric synthetic aperture radar interferometry (PolInSAR) data. However, in the context of single-baseline parameter retrieval by the random volume over ground (RVoG) model, three main error sources that affect the inversion accuracy should be carefully considered, i.e., the ground scattering contribution, the spatial baseline configuration, and the temporal decorrelation (the main part of non-volume decorrelation). To make the estimation more reliable, several kinds of multibaseline PolInSAR inversion methods have been proposed over the past few years and have achieved improved inversion performances. Dual-baseline inversion effectively avoids the ambiguity of the ground contribution, whereas the performance is highly dependent on the appropriate combination of two spatial baselines as well as the mitigation of the non-volume decorrelation. In this study, we conducted in-depth research into the effect of these influencing factors on dual-baseline inversion, aiming to provide effective guidance on dual-baseline combination selection among multibaseline data. Accordingly, a novel multibaseline inversion scheme (MBLFPI) suitable for low-frequency PolInSAR data is proposed in this paper. The significant advantage of the new method is that the three aforementioned error sources can be taken into account simultaneously, without relying on external data. The proposed scheme was validated using L- and P-band SAR data acquired by the DLR's E-SAR/F-SAR and ONERA's SETHI systems, as well as corresponding light detection and ranging (LiDAR) data collected during the BioSAR-2008, AfriSAR-2016, and TropiSAR-2009 campaigns. A series of experiments was performed to evaluate the applicability and generalizability of the proposed method. The results showed that this innovative scheme produced forest height maps with a root-mean-square error (RMSE) of 2.37 m (R2 = 0.88) and 3.13–4.43 m (R2 = 0.35–0.94) in the L-band and P-band scenarios (boreal and tropical forest), respectively, indicating a significant improvement over the three conventional multibaseline methods and pure dual-baseline inversion. The comprehensive analysis provided in this paper should assist with and provide strong support for SAR system and mission design, and the proposed scheme could be considered a promising way for future spaceborne missions to invert vegetation parameters at a global scale. • A generalizable MB method is proposed for global forest mapping by BIOMASS and LT-1. • Three general issues in low-frequency PolInSAR inversion are simultaneously solved. • B1 is more sensitive than B2 in dual-baseline PolInSAR geometric inversion method. • An error model of dual-baseline forest height inversion is constructed. • Findings help for system design of spaceborne low-frequency PolInSAR missions. [ABSTRACT FROM AUTHOR]

Terrain-induced variations of radar backscatter represent an important limiting factor of many Synthetic Aperture Radar (SAR)-based applications. Radiometric terrain flattening (RTF) is a well-established method that minimizes these variations in SAR imagery. To fully understand the implications of SAR RTF, validation of its impact on the derived products is needed. In this study, we quantified the influence of the RTF on a forest mapping and classification algorithm over Austria, and compared the classification results for the conventional sigma naught and radiometrically terrain-corrected gamma backscatter. The overall accuracy for forest/non-forest mapping and forest type classification improved by 2% and 4%, respectively, over the whole of Austria, with improvements of up to 16% and 20%, respectively, in regions with strong topography. [ABSTRACT FROM AUTHOR]

The TanDEM-X synthetic aperture radar (SAR) system allows for the recording of bistatic interferometric SAR (InSAR) acquisitions, which provide additional information to the common amplitude images acquired by monostatic SAR systems. More concretely, the volume decorrelation factor, which can be derived from the bistatic interferometric coherence, is a reliable indicator of the presence of vegetation and it was used as main input feature for the generation of the global TanDEM-X forest/non-forest map, by means of a clustering algorithm. In this work, we investigate the capabilities of deep Convolutional Neural Networks (CNNs) for mapping tropical forests at large-scale using TanDEM-X InSAR data. For this purpose, we rely on a U-Net architecture, which takes as input a set of feature maps selected on the basis of previous preparatory works. Moreover, we design an ad hoc training strategy, aimed at developing a robust model for global mapping purposes, which has to properly manage the large variety of different acquisition geometries characterizing the TanDEM-X global data set. In addition to detecting forest/non-forest areas, the CNN has also been trained to detect water surfaces, which are typically characterized by low values of coherence. By applying the proposed method on single TanDEM-X images, we achieved a significant performance improvement with respect to the baseline clustering approach, with an average F-score increase of 0.13. We then applied such a model for mapping the entire Amazon rainforest, as well as the other tropical forests in Central Africa and South-East Asia, in order to test its robustness and generalization capabilities, and we observed that forests are typically well detected as contour closed regions and that water classification is reliable, too. Finally, the generated maps show a great potential for mapping temporal changes occurring over forested areas and can be used for generating large-scale maps of deforestation. [ABSTRACT FROM AUTHOR]

Accurate and effective mapping of forest aboveground biomass (AGB) in heterogeneous mountainous regions is a huge challenge but an urgent demand for resource managements and carbon storage monitoring. Conventional studies have related the plot-measured or LiDAR-based biomass to remote sensing data using pixel-based approaches. The object-based relationship between AGB and multi-source data from LiDAR, multi-frequency radar, and optical sensors were insufficiently studied. It deserves the further exploration that maps forest AGB using the object-based approach and combines LiDAR data with multi-sensor images, which has the smaller uncertainty of positional discrepancy and local heterogeneity, in heterogeneous mountainous regions. To address the improvement of mapping accuracy, satellite LiDAR data from GEDI and ICEsat-2, and images of ALOS-2 yearly mosaic L band SAR (Synthetic Aperture Radar), Sentinel-1 C band SAR, Sentinel-2 MSI, and ALOS-1 DSM were combined for pixel- and object-based forest AGB mapping in a vital heterogeneous mountainous forest. For the object-based approach, optimized objects during a multiresolution segmentation were acquired by the ESP (Estimation of the Scale Parameter) tool, and suitable predictors were selected using an algorithm named VSURF (Variable Selection Using Random Forests). The LiDAR variables at the footprint-level were extracted to connect field plots to the multi-sensor objects as a linear bridge. It was shown that forests' AGB values varied by elevations with a mean value of 142.58 Mg/ha, ranging from 12.61 to 514.28 Mg/ha. The north slope with the lowest elevation (<1100 m) had the largest mean AGB, while the smallest mean AGB was located in the south slope with the altitude above 2000 m. Using independent validation samples, it was indicated by the accuracy comparison that the object-based approach performed better on the precision with relative improvement based on root-mean-square errors (RIRMSE) of 4.46%. The object-based approach also selected more optimized predictors and markedly decreased the prediction time than the pixel-based analysis. Canopy cover and height explained forest AGB with their effects on biomass varying according to the elevation. The elevation from DSM and variables involved in red-edge bands from MSI were the most contributive predictors in heterogeneous temperate forests. This study is a pioneering exploration of object-based AGB mapping by combining satellite data from LiDAR, MSI, and SAR, which offers an improved methodology for regional carbon mapping in the heterogeneous mountainous forests. [ABSTRACT FROM AUTHOR]

The monitoring of land cover and land use patterns is pivotal in the joint effort to fight deforestation in the Amazon and study its relation to climate change effects with respect to anthropogenic activities. Most of the region, typically monitored with optical sensors, is hidden by a persistent cloud cover for most of the wet season. The necessity for a consistent and reliable deforestation warning system based on cloud-independent radar data is therefore of particular interest. In this paper, we investigated the potential of combining deep learning with Sentinel-1 (S-1) Interferometric Synthetic Aperture Radar (InSAR) short time series (STS), covering only 24 d of acquisitions, to map endangered areas in the Amazon Basin. To this end, we implemented a U-Net-like convolutional neural network (CNN) for multi-layer semantic segmentation, trained from scratch with different sets of input features to evaluate the viability of the proposed approach for different operating conditions. As input features, we relied on both multi-temporal backscatter and interferometric coherences at different temporal baselines. We provide a detailed performance benchmark of the different configurations, also considering the current state-of-the-art approaches based on S-1 STS and shallow learners. Our findings showed that, by exploiting the powerful learning capabilities of CNNs, we outperformed the STS-based approaches published in the literature and significantly reduced the computational load. Indeed, when considering the entire stack of Sentinel-1 data at a 6 d revisit time, we were able to maintain the overall accuracy and F1-score well above 90% and reduce the computational time by more than 50% with respect to state-of-the-art approaches, by avoiding the generation of handcrafted feature maps. Moreover, we achieved satisfactory results even when only S-1 InSAR acquisitions with a revisit time of 12 d or more were used, setting the groundwork for an effective and fast monitoring of tropical forests at a global scale. [ABSTRACT FROM AUTHOR]

Sentinel-1 interferometric time-series allow for the accurate retrieval of the target’s temporal decorrelation and, therefore, the inversion of land cover information and its temporal monitoring. This paper describes the development of an observation scenario for monitoring monthly deforestation over the Amazon rainforest, which relies on the use of radar for overcoming the physical limitations of optical sensors caused by the presence of cloud coverage. Specifically, we implement a classification scheme that exploits multi-temporal SAR features, such as backscatter, spatial textures, and interferometric parameters, to map forested areas. Distinct forest maps are generated for consecutive months and further processed to detect deforestation phenomena and map clear-cuts evolution. The obtained results are validated by selecting cloud-free Sentinel-2 multispectral data on the selected area and acquired during the same observation time.

The TanDEM-X synthetic aperture radar (SAR) system allows for the recording of bistatic interferometric SAR (InSAR) acquisitions, which provide additional information to the common amplitude images acquired by monostatic SAR systems. More concretely, the volume decorrelation factor, which can be derived from the bistatic interferometric coherence, is a reliable indicator of the presence of vegetation and it was used as main input feature for the generation of the global TanDEM-X forest/non-forest map, by means of a clustering algorithm. In this work, we investigate the capabilities of deep Convolutional Neural Networks (CNNs) for mapping tropical forests at large-scale using TanDEM-X InSAR data. For this purpose, we rely on a U-Net architecture, which takes as input a set of feature maps selected on the basis of previous preparatory works. Moreover, we design an ad hoc training strategy, aimed at developing a robust model for global mapping purposes, which has to properly manage the large variety of different acquisition geometries characterizing the TanDEM-X global data set. In addition to detecting forest/non-forest areas, the CNN has also been trained to detect water surfaces, which are typically characterized by low values of coherence. By applying the proposed method on single TanDEM-X images, we achieved a significant performance improvement with respect to the baseline clustering approach, with an average F-score increase of 0.13. We then applied such a model for mapping the entire Amazon rainforest, as well as the other tropical forests in Central Africa and South-East Asia, in order to test its robustness and generalization capabilities, and we observed that forests are typically well detected as contour closed regions and that water classification is reliable, too. Finally, the generated maps show a great potential for mapping temporal changes occurring over forested areas and can be used for generating large-scale maps of deforestation.

Temperate mixed forest ecosystems are composed of various tree functional types (TFTs) that differ in canopy structure, phenology, and physiological response to climate change. An accurate characterization of the composition of these TFTs is important for quantifying land surface carbon, energy, and water cycling, as well as process-based simulation of forest dynamics. However, because the pixel size of satellite imagery is usually larger than temperate tree crowns, it is challenging to untangle the significant pixel-wise signal mixture of TFT across mixed forest regions. Spectral Mixture Analysis (SMA) has been widely used to derive the sub-pixel fractional composition of TFT from satellite imagery, but accounting for the broad spectral variability within TFTs across space and time remains a challenge. Synthetic aperture radar (SAR) can indicate biomass mixture information, but it has not been fully exploited for deriving subpixel TFT composition. To improve TFT composition mapping in mixed forest regions, we developed a Fisher-transformation-based Spectral and Radar Time-series Mixture Analysis (F-SRTMA) framework on Google Earth Engine. The F-SRTMA framework aims to address the space-time TFT variability of satellite signatures based on two modified modules: (1) the use of spectral and radar data with spatial and temporal information, and (2) feature optimization based on Fisher Discriminant Analysis (FDA). We tested the F-SRTMA at three representative temperate mixed landscapes located in the northeastern United States, where time-series Sentinel-1 and -2 data were used to calibrate our F-SRTMA approach. Airborne hyperspectral and LiDAR-derived canopy height data were used to generate ground reference TFT fraction maps for validation. The results demonstrate that (1) compared to the spectral time-series model, the synergy of spectral and radar time-series features yielded higher accuracy at the local sites (r2 = 0.649 vs. 0.680); (2) optimized feature based on FDA significantly minimized the within-TFT variability while maximizing the between-TFT variability, which further improved model generalizability across different landscapes, yielding the highest accuracy with cross-site r2 increasing from 0.634 to 0.715 and RMSE decreasing from 0.207 to 0.164. Collectively, these results suggest that F-SRTMA can be an accurate and generalizable approach for sub-pixel fraction mapping across temperate mixed landscapes, with the potential to be applied to other mixed forest ecosystems. • A sentinel-1 and -2 based unmixing method enhanced tree functional types mapping. • Spectral and radar time-series integration improved forest composition estimation. • Fisher-transformed features enhanced model generalizability across ecoregion. [ABSTRACT FROM AUTHOR]

The monitoring of land cover and land use patterns is pivotal in the joint effort to fight deforestation in the Amazon and study its relation to climate change effects with respect to anthropogenic activities. Most of the region, typically monitored with optical sensors, is hidden by a persistent cloud cover for most of the wet season. The necessity for a consistent and reliable deforestation warning system based on cloud-independent radar data is therefore of particular interest. In this paper, we investigated the potential of combining deep learning with Sentinel-1 (S-1) Interferometric Synthetic Aperture Radar (InSAR) short time series (STS), covering only 24 d of acquisitions, to map endangered areas in the Amazon Basin. To this end, we implemented a U-Net-like convolutional neural network (CNN) for multi-layer semantic segmentation, trained from scratch with different sets of input features to evaluate the viability of the proposed approach for different operating conditions. As input features, we relied on both multi-temporal backscatter and interferometric coherences at different temporal baselines. We provide a detailed performance benchmark of the different configurations, also considering the current state-of-the-art approaches based on S-1 STS and shallow learners. Our findings showed that, by exploiting the powerful learning capabilities of CNNs, we outperformed the STS-based approaches published in the literature and significantly reduced the computational load. Indeed, when considering the entire stack of Sentinel-1 data at a 6 d revisit time, we were able to maintain the overall accuracy and F1-score well above 90% and reduce the computational time by more than 50% with respect to state-of-the-art approaches, by avoiding the generation of handcrafted feature maps. Moreover, we achieved satisfactory results even when only S-1 InSAR acquisitions with a revisit time of 12 d or more were used, setting the groundwork for an effective and fast monitoring of tropical forests at a global scale.

The successful launch and operation of the Sentinel satellite platform has provided access to freely available remotely sensed data useful for commercial forest species discrimination. Sentinel – 1 (S1) with a synthetic aperture radar (SAR) sensor and Sentinel – 2 (S2) multi-spectral sensor with additional and strategically positioned bands offer great potential for providing reliable information for discriminating and mapping commercial forest species. In this study, we sought to determine the value of S1 and S2 data characteristics in discriminating and mapping commercial forest species. Using linear discriminant analysis (LDA) algorithm, S2 multi-spectral imagery showed an overall classification accuracy of 84% (kappa = 0.81), with bands such as the red-edge (703.9–740.2 nm), narrow near infrared (835.1–864.8 nm), and short wave infrared (1613.7–2202.4 nm) particularly influential in discriminating individual forest species stands. When Sentinel 2's spectral wavebands were fused with Sentinel 1's (SAR) VV and VH polarimetric modes, overall classification accuracies improved to 87% (kappa = 0.83) and 88% (kappa = 0.85), respectively. These findings demonstrate the value of combining Sentinel's multispectral and SAR structural information characteristics in improving commercial forest species discrimination. These, in addition to the sensors free availability, higher spatial resolution and larger swath width, offer unprecedented opportunities for improved local and large scale commercial forest species discrimination and mapping. [ABSTRACT FROM AUTHOR]

The use of freely accessible Sentinel-1 synthetic aperture radar (S-1 SAR) and Sentinel-2 multispectral instrument (S-2 MSI) data are currently a feasible way of mapping forest aboveground biomass (AGB) over large areas. Despite the extensive mapping of forest AGB by remote sensing, how to effectively combine different sensors data, selecting the proper statistical modelling method, and variable screening are still poorly understood. This paper presents a framework for Sentinel-based AGB estimation through the use of four variable screening techniques, namely, genetic algorithm (GA), least absolute shrinkage and selection operator (LASSO), boruta, and removal-based; and three statistical modelling methods including multiple linear regression (MLR), artificial neural network (ANN), and random forest (RF). These methods were examined across two forested sites with different levels of human activity (highly-degraded: HD and minorly-degraded: MD) in Zagros coppice oak forests. We used backscatter coefficients and texture metrics of S-1 SAR, in addition to raw bands, simple band ratios, vegetation indices and texture metrics of S-2 MSI to extract 53 explanatory variables. Our findings showed that GA and removal-based variable screening techniques outperformed the others and RF and MLR statistical modelling methods obtained the best results. The use of different variable screening techniques and modelling methods demonstrated higher accuracies of AGB estimates for the MD site compared to the HD site. S-2 MSI (Coefficient of determination (R2) = 0.76 and Root Mean Square Error (RMSE) = 35.08% of the mean) provided a higher accuracy than S-1 SAR (R2 = 0.42 and RMSE = 52.94% of the mean). Overall, the study demonstrates encouraging results in the retrieval and predictive mapping of the AGB of Zagros forests using the freely accessible Sentinel imagery. [ABSTRACT FROM AUTHOR]

Polarimetric synthetic aperture radar (SAR) data are inevitably contaminated by polarization crosstalk and channel imbalance, which propagate to the error of final remote sensing product. To ensure the successful estimation of forest heights from forthcoming polarimetric SAR interferometry (PolinSAR) campaigns, a critical study on the polarimetric system requirements of PolinSAR for forest height mapping must be carried out. This paper establishes an analytical model for forest height estimation error including dependences on polarimetric system parameters including crosstalk, channel imbalance, and system noise. Simulation analyses are conducted on the real airborne SAR data acquired by the E-SAR system to validate the proposed theoretical error dependence model. We demonstrate that the established error model can be used not only by the system designers as a guidance for setting the polarimetric system requirements of PolinSAR for forest height mapping, but also by the data analyst to correct for systematic bias in the forest height inversion. [ABSTRACT FROM AUTHOR]

Abstract Advanced Land Observing Satellite (ALOS) Phased Arrayed L-band Synthetic Aperture Radar (PALSAR) HH and HV polarization data were used previously to produce annual, global 25 m forest maps between 2007 and 2010, and the latest global forest maps of 2015 and 2016 were produced by using the ALOS-2 PALSAR-2 data. However, annual 25 m spatial resolution forest maps during 2011–2014 are missing because of the gap in operation between ALOS and ALOS-2, preventing the construction of a continuous, fine resolution time-series dataset on the world's forests. In contrast, the MODerate Resolution Imaging Spectroradiometer (MODIS) NDVI images were available globally since 2000. This research developed a novel method to produce annual 25 m forest maps during 2007–2016 by fusing the fine spatial resolution, but asynchronous PALSAR/PALSAR-2 with coarse spatial resolution, but synchronous MODIS NDVI data, thus, filling the four-year gap in the ALOS and ALOS-2 time-series, as well as enhancing the existing mapping activity. The method was developed concentrating on two key objectives: 1) producing more accurate 25 m forest maps by integrating PALSAR/PALSAR-2 and MODIS NDVI data during 2007–2010 and 2015–2016; 2) reconstructing annual 25 m forest maps from time-series MODIS NDVI images during 2011–2014. Specifically, a decision tree classification was developed for forest mapping based on both the PALSAR/PALSAR-2 and MODIS NDVI data, and a new spatial-temporal super-resolution mapping was proposed to reconstruct the 25 m forest maps from time-series MODIS NDVI images. Three study sites including Paraguay, the USA and Russia were chosen, as they represent the world's three main forest types: tropical forest, temperate broadleaf and mixed forest, and boreal conifer forest, respectively. Compared with traditional methods, the proposed approach produced the most accurate continuous time-series of fine spatial resolution forest maps both visually and quantitatively. For the forest maps during 2007–2010 and 2015–2016, the results had greater overall accuracy values (>98%) than those of the original JAXA forest product. For the reconstructed 25 m forest maps during 2011–2014, the increases in classifications accuracy relative to three benchmark methods were statistically significant, and the overall accuracy values of the three study sites were almost universally >92%. The proposed approach, therefore, has great potential to support the production of annual 25 m forest maps by fusing PALSAR/PALSAR-2 and MODIS NDVI during 2007–2016. Highlights • More accurate forest maps were produced by integrating PALSAR/PALSAR-2 and MODIS NDVI images. • The missing PALSAR forest maps during 2011–2014 were reconstructed with a new spatial-temporal model. • Forest fractional maps were estimated automatically from time-series MODIS NDVI images. • The reconstructed forest maps have an overall accuracy >92%. [ABSTRACT FROM AUTHOR]

Reliable spatial information on growing stock volume (GSV) and biomass is critical for creating management strategies for plantation forests. This study developed empirical models to map the GSV and biomass of larch plantations (LPs) in Northeast China (1.25 million km2 total area) by integrating L-band synthetic aperture radar (SAR) data with ground-based survey data. The best correlation model was used to map the GSVs and biomasses of LPs. The total GSV and biomass carbon storage were estimated at 224.3 ± 59.0 million m3 and 113.0 ± 29.7 × 1012 g C with average densities of 85.1 m3 ha−1 and 42.9 106 g × C ha−1, respectively, over a total area of 2.64 million ha. The saturation effect of SAR was determined beyond 260 m3 ha−1, which was expected to influence the estimations for a small proportion of the study area. The accuracy of the estimations has limitations mainly due to the uncertainties in the GSV inventories, discrimination of natural larch and the SAR dataset. Based on the mapping results of the GSVs of LPs, a planning strategy for multipurpose management was tentatively proposed. This study can inform policies and management practices to assure broader and sustainable benefits from plantation forests in the future. [ABSTRACT FROM AUTHOR]

The Sentinel-1 satellites provide the formerly unprecedented combination of high spatial and temporal resolution of dual polarization synthetic aperture radar data. The availability of dense time series enables the derivation and analysis of temporally filtered annual backscatter signals. The study concentrates on the use of Sentinel-1 seasonal backscatter signatures for forest area estimation and forest type classification. A classification method based on time series similarity measures is introduced and tested in three test areas covered by various forest types including broadleaf temperate, boreal and montane forests. The results are compared with two European-wide Copernicus high resolution layers, namely forest type and tree cover density (TCD). The correspondence of forest/non-forest maps and TCD is high in all test areas, with overall accuracies for forest/non-forest classification between 86% and 91% and Pearson correlation coefficients for TCD between 0.68 and 0.74. The forest type classification (non-forest, coniferous and broadleaf forest classes) provides best results in temperate forests with an overall accuracy of 85%; in boreal forest, the accuracy decreases to only 65%. Generally, the method provides reliable results for forest area estimation, including regions where methods based on static parameters are often problematic (mountainous areas), and it enables forest type classification in temperate forests. [ABSTRACT FROM AUTHOR]

This paper reports recent advancements in the field of Synthetic Aperture Radar (SAR) for forest mapping by using interferometric short-time-series. In particular, we first present how the interferometric capabilities of the Sentinel-1 satellites constellation can be exploited for the monthly mapping of the Amazon rainforest. Indeed, the evolution in time of the interferometric coherence can be properly modeled as an exponential decay and the retrieved interferometric parameters can be used, together with the backscatter, as input features to the machine learning Random Forests classifier. Furthermore, we present an analysis on the benefits of the use of textural information, derived from Sentinel-1 backscatter, in order to enhance the classification accuracy. These textures are computed through the Sum And Difference Histograms methodology and the final classification accuracy, resulting by adding them to the aforementioned features, is a thematic map that exceeds an overall agreement of 85 % , when validated using the optical external reference Finer Resolution Observation and Monitoring of Global Land Cover (FROM-GLC) map. The experiments presented in the final part of the paper are enriched with a further analysis and discussion on the selected scenes using updated multispectral Sentinel-2 acquisitions.

In this paper we present the activities performed at the Microwaves and Radar Institute of the German Aerospace Center (DLR) to derive global forest/non-forest classification mosaics from interferometric synthetic aperture radar (InSAR) data acquired by the TanDEM-X mission. The data have been collected between 2011 and 2016 in bistatic stripmap single polarization (HH) mode, with the main goal of generating a consistent, timely, and highly accurate 3D representation of the global terrain’s surface (digital elevation model, DEM). The global data set of quicklook images, which represent a spatially averaged version of the original full resolution data at a ground independent pixel spacing of 50 m × 50 m, was used as input, in order to limit the computational burden. For classification purposes, several observables, systematically provided by the TanDEM-X system, can be exploited, such as the calibrated amplitude, the digital elevation model (DEM), and the interferometric coherence. Among the several factors contributing to a coherence degradation in InSAR data, the so-called volume correlation factor quantifies the coherence loss due to volume scattering phenomena, which typically occur in presence of vegetation. This quantity is directly derived from the interferometric coherence and used as main indicator for the identification of vegetated areas. For this purpose, a fuzzy multi-clustering classification approach, which takes into account the geometry and acquisition configuration, is applied to each acquired scene separately. A certain variability of the interferometric coherence at X band was observed among different forest types, mainly due to changes in forest structure, density, and tree height, which led to an adjustment of the algorithm settings depending on the considered type of forest. The identification of additional information layers, such as urban settlements or water areas, is also discussed, and the procedure for mosaicking of overlapping acquisitions (two at global scale, up to ten over mountainous terrain, forests, and desert regions) to improve the classification accuracy is detailed. The resulting global forest/non-forest map was validated using external reference information as well as with other existing classification maps and an overall agreement was observed that often exceeds 90 % . Finally, examples for high-resolution (at 12 m × 12 m) forest maps and potentials for deforestation monitoring over selected regions are presented as well, demonstrating the unique capabilities offered by the TanDEM-X bistatic system for a broad range of geoinformation services and scientific applications. The global TanDEM-X forest/non-forest map presented in this paper will be made available to the scientific community for free download and usage. [ABSTRACT FROM AUTHOR]

Efficient methods to monitor forested areas help us to better understand their processes. To date, only a few studies have assessed the usability of multitemporal synthetic aperture radar (SAR) datasets in this context. Here we present an analysis of an unprecedented set of C-band observations of mixed temperate forests. We demonstrate the potential of using multitemporal C-band VV and VH polarisation data for monitoring phenology and classifying forests in northern Switzerland. Each SAR acquisition was first radiometrically terrain corrected using digital elevation model-based image simulations of the local illuminated area. The flattened backscatter values and the local area values were input to a temporal compositing process integrating backscatter values from ascending and descending tracks. The process used local resolution weighting of each input, producing composite backscatter values that strongly mitigated terrain-induced distortions. Several descriptors were calculated to show the seasonal variation of European beech (Fagus sylvatica), oak (Quercus robur, Quercus petraea) and Norway spruce (Picea abies) in C-band data. Using their distinct seasonal signatures, the timing of leaf emergence and leaf fall of the deciduous species were estimated and compared to available ground observations. Furthermore, classifications for the forest types 'deciduous' and 'coniferous' and the investigated species were implemented using random forest classifiers. The deciduous species backscatter was about 1 dB higher than spruce throughout the year in both polarisations. The forest types showed opposing seasonal backscatter behaviours. At VH, deciduous species showed higher backscatter in winter than in summer, whereas spruce showed higher backscatter in summer than in winter. In VV, this pattern was similar for spruce, while no distinct seasonal behaviour was apparent for the deciduous species. The time differences between the estimations and the ground observations of the phenological events were approximately within the error margin (±12 days) of the temporal resolution. The classification performances were promising, with higher accuracies achieved for the forest types (OA of 86% and k = 0.73) than for individual species (OA of 72% and k = 0.58). These results show that multitemporal C-band backscatter data have significant potential to supplement optical remote sensing data for ecological studies and mapping of mixed temperate forests. [ABSTRACT FROM AUTHOR]

This paper report experiences from the processing and mosaicking of 518 TanDEM-X image pairs covering the entirety of Sweden, with two single map products of above-ground biomass (AGB) and forest stem volume (VOL), both with 10 m resolution. The main objective was to explore the possibilities and overcome the challenges related to forest mapping extending a large number of adjacent satellite scenes. Hence, numerous examples are presented to illustrate challenges and possible solutions. To derive the forest maps, the observables backscatter, interferometric phase height and interferometric coherence, obtained from TanDEM-X, were evaluated using empirical robust linear regression models with reference data extracted from 2288 national forest inventory plots with a 10 m radius. The interferometric phase height was the single most important observable, to predict AGB and VOL. The mosaics were evaluated on different datasets with field-inventoried stands across Sweden. The root mean square error (RMSE) was about 21%-25% (27-30 tons/ha and 52-65 m3/ha) at the stand level. It was noted that the most influencing factors on the observables in this study were local temperature and geolocation errors that were challenging to robustly compensate against. Because of this variability at the scene-level, determinations of AGB and VOL for single stands are recommended to be used with care, as an equivalent accuracy is difficult to achieve for all different scenes, with varying acquisition conditions. Still, for the evaluated stands, the mosaics were of sufficient accuracy to be used for forest management at the stand level. [ABSTRACT FROM AUTHOR]