Remote Sensing

Data stream partitioning is a fundamental and important mechanism for distributed systems. However, use of an inappropriate partition scheme may generate a data skew problem, which can influence the execution efficiency of many application tasks. Processing of skewed partitions usually takes a longer time, need more computational resources to complete the task and can even become a performance bottleneck. To solve such data skew issues, this paper proposes a novel partition method to divide on demand the image tiles uniformly into partitions. The partitioning problem is then transformed into a uniform and compact clustering problem whereby the image tiles are regarded as image pixels without spectrum and texture information. First, the equal area conversion principle was used to select the seed points of the partitions and then the image tiles were aggregated in an image layout, thus achieving an initial partition scheme. Second, the image tiles of the initial partition were finely adjusted in the vertical and horizontal directions in separate steps to achieve a uniform distribution among the partitions. Two traditional partition methods were adopted to evaluate the efficiency of the proposed method in terms of the image segmentation testing, data shuffle testing, and image clipping testing. The results demonstrated that the proposed partition method solved the data skew problem observed in the hash partition method. In addition, this method is designed specifically for processing of image tiles and makes the related processing operations for large-scale images faster and more efficient. Full article

Information on tree species and changes in forest composition is necessary to understand species-specific responses to change, and to develop conservation strategies. Remote sensing methods have been increasingly used for tree detection and species classification. In mixed species forests, conventional tree detection methods developed with assumptions about uniform tree canopy structure often fail. The main aim of this study is to identify effective methods for tree delineation and species classification in an Australian native forest. Tree canopies were delineated at three different spatial scales of analysis: (i) superpixels representing small elements in the tree canopy, (ii) tree canopy objects generated using a conventional segmentation technique, multiresolution segmentation (MRS), and (iii) individual tree bounding boxes detected using deep learning based on the DeepForest open-source algorithm. Combinations of spectral, texture, and structural measures were tested to assess features relevant for species classification using RandomForest. The highest overall classification accuracies were achieved at the superpixel scale (0.84 with all classes and 0.93 with Eucalyptus classes grouped). The highest accuracies at the individual tree bounding box and object scales were similar (0.77 with Eucalyptus classes grouped), highlighting the potential of tree detection using DeepForest, which uses only RGB, compared to site-specific tuning with MRS using additional layers. This study demonstrates the broad applicability of DeepForest and superpixel approaches for tree delineation and species classification. These methods have the potential to offer transferable solutions that can be applied in other forests. Full article

Soil salinization is prominent environmental issue in arid and semi-arid regions, such as Xinjiang in Northwest China. Salinization severely restricts economic and agricultural development and would lead to ecosystem degradation. Finding a method of rapidly and accurately determining soil salinity (SS) is one of the main challenges in salinity evaluation, saline soil development, and utilization. In situ visible and near infrared (Vis-NIR) spectroscopy has proven to be a promising technique for detecting soil properties since it can realize real-time, rapid detection of SS. However, it still remains challenging whether Vis-NIR in situ spectroscopy can invert SS with high accuracy due to the interference of environmental factors (e.g., light, water vapor, solar altitude angle, etc.) on the spectral in the field. To fill this knowledge gap, we collected Vis-NIR in situ spectral and lab-measured SS data from 135 surface soil samples in the Kongterik Pasture Nature Reserve (KPNR) in the desert oasis ecotone of southern Xinjiang, China. We used genetic algorithm (GA), particle swarm optimization (PSO), and simulated annealing (SA) algorithms to select the feature bands of SS. Subsequently, we combined extreme learning machines (ELM), back-propagation neural networks (BPNN), and convolutional neural networks (CNN) to build inversion models of SS. The results showed that different feature bands selection methods could improve the Vis-NIR in situ spectral prediction model accuracy. Either SS inversion models were built using full-band spectral data or feature-band spectral data. Compared with the full-band (401–2400 nm) spectral modeling, the validation set R² of ELM, BPNN, and CNN models built selected feature bands selected by PSO, GA, and SA, respectively, were improved by more than 0.06. The accuracy of predicting SS varied widely among modeling methods. The accuracy of CNN model was obviously higher than that of BPNN and ELM models. The optimal hybrid model for predicting SS constructed in this study is SA-CNN model (R² = 0.79, RMSE = 9.41 g kg⁻¹, RPD = 1.81, RPIQ = 2.37). This study showed that the spectral feature bands selection methods can reduce the influence of environmental factors on in situ spectroscopy and significantly enhance the inversion accuracy of SS. The present study provided that estimating SS using in situ Vis-NIR spectral is feasible. Full article

Accurate and timely access to building rooftop information is very important for urban management. The era of big data brings new opportunities for rooftop extraction based on deep learning and high-resolution satellite imagery. However, collecting representative datasets from such big data to train deep learning models efficiently is an essential problem that still needs to be explored. In this study, geospatial stratified and optimized sampling (GSOS) based on geographical priori information and optimization of sample spatial location distribution is proposed to acquire representative samples. Specifically, the study area is stratified based on land cover to divide the rooftop-dense stratum and the rooftop-sparse stratum. Within each stratum, an equal amount of samples is collected and their spatial locations are optimized. To evaluate the effectiveness of the proposed strategy, several qualitive and quantitative experiments are conducted. As a result, compared with other common sampling approaches (e.g., random sampling, stratified random sampling, and optimized sampling), GSOS is superior in terms of the abundance and types of collected samples. Furthermore, two quantitative metrics, the F₁-score and Intersection over Union (IoU), are reported for rooftop extraction based on deep learning methods and different sampling methods, in which the results based on GSOS are on average 9.88% and 13.20% higher than those based on the other sampling methods, respectively. Moreover, the proposed sampling strategy is able to obtain representative training samples for the task of building rooftop extractions and may serve as a viable method to alleviate the labour-intensive problem in the construction of rooftop benchmark datasets. Full article

Atmospheric lidar is susceptible to the influence of light attenuation, sky background light, and detector dark currents during the detection process. This results in a large amount of noise in the lidar return signal. To reduce noise and extract a useful signal, a novel denoising method combined with variational modal decomposition (VMD), the sparrow search algorithm (SSA) and singular value decomposition (SVD) is proposed. The SSA is used to optimize the number of decomposition layers K and the quadratic penalty factor α values of the VMD algorithm. Some intrinsic mode function (IMF) components obtained from the VMD-SSA decomposition are grouped and reconstructed according to the interrelationship number selection criterion. Then, the reconstructed signal is further denoised by combining the strong noise-reduction ability of SVD to obtain a clean lidar return signal. To verify the effectiveness of the VMD-SSA-SVD method, the method is compared and analysed with wavelet packet decomposition, empirical modal decomposition (EMD), ensemble empirical modal decomposition (EEMD), and adaptive noise-complete ensemble empirical modal decomposition (CEEMD), and its noise-reduction effect is considerably improved over that of the other four methods. The method can eliminate the complex noise in the lidar return signal while retaining all the details of the signal. The signal is not distorted, the waveform is smoother, and far-field noise interference can be suppressed. The denoised signal is closer to the real signal with higher accuracy, which shows the feasibility and the practicality of the proposed method. Full article

Controlled tile drainage (CTD) practices are a promising tool for improving water balance, water quality and increasing crop yield by raising shallow groundwater level and capillary rise due to drainage flow retardation. We tested the effect of CTD on growth and grain yield of spring barley, at a study site in central Bohemia using vegetation indices from unmanned aerial vehicle (UAV) imagery and Sentinel-2 satellite imagery. Tile drainage flow was slowed by fixed water level control structures that increased soil moisture in the surrounding area according to the terrain slope. Vegetation indices based on red-edge spectral bands in combination with near-infrared and red bands were selected, of which the Normalized Red Edge-Red Index (NRERI) showed the closest relationships with shoot biomass parameters (dry biomass, nitrogen concentration and uptake, nitrogen nutrition index) from point sampling at the tillering stage. The CTD sites showed significantly more biomass using NRERI compared to free tile drainage (FTD) sites. In contrast, in the period prior to the implementation of CTD practices, Sentinel-2 satellite imagery did not demonstrate higher biomass based on NRERI at CTD sites compared to FTD sites. The grain yields of spring barley as determined from the yield map also increased due to CTD (by 0.3 t/ha, i.e., by 4%). The positive impact of CTD on biomass development and grain yield of spring barley was confirmed by the increase in soil moisture at depths of 20, 40 and 60 cm compared to FTD. The largest increase in soil water content of 3.5 vol% due to CTD occurred at the depth of 40 cm, which also had a higher degree of saturation of available water capacity and the occurrence of crop water stress was delayed by 14 days compared to FTD. Full article

Global Navigation Satellite System (GNSS) radio occultation (RO) technology has been widely used in Earth atmospheric detection and has a significant impact on numerical weather prediction (NWP), climate detection, and other fields. Cosmic-2 LEO-1 (C2E1) is a well-known RO data provider; however, its observations are confined to 45°S and 45°N. Recently, the Binhu meteorological observation test satellite (BH) has provided global coverage of RO data, including refractivity, specific humidity, and temperature data. In this study, RO data from BH and C2E1 are analyzed and compared from 8 February 2022 to 17 February 2022. Employing the European Centre for Medium-Range Weather Forecasts (ECMWF) reanalysis (ERA-5) data as a reference, both BH and C2E1 RO data agree with the ERA data, with the refractivity, temperature, and specific humidity profiles reflecting the real conditions of the natural atmosphere. In addition, BH data are comparable to C2E1 data at low and middle latitudes (0–45°), and BH data at middle and high latitudes (45–90°) are of better quality than those at low and middle latitudes (0–45°). For example, without considering the errors introduced by the interpolation of the ERA-5 data for comparative analysis, the BH refractivity profiles show a mean absolute bias of 0.73 N at low and middle latitudes and only 0.23 N at middle and high latitudes, and that for BH specific humidity profiles at middle and high latitudes is 0.015 g/kg, which is only half of that at low and middle latitudes. The BH temperature and specific humidity data show promising data accuracy. Therefore, BH RO data may provide important supplementary data at higher latitudes and may improve future NWPs through assimilation. Full article

Land cover change (LCC) studies are increasingly using deep learning (DL) modeling techniques. Past studies have leveraged temporal or spatiotemporal sequences of historical LC data to forecast changes with DL models. However, these studies do not adequately assess the association between neighborhood size and DL model capability to forecast LCCs, where neighborhood size refers to the spatial extent captured by each data sample. The objectives of this research study were to: (1) evaluate the effect of neighborhood size on the capacity of DL models to forecast LCCs, specifically Temporal Convolutional Networks (TCN) and Convolutional Neural Networks (CNN-TCN), and (2) assess the effect of auxiliary spatial variables on model capacity to forecast LCCs. First, each model type and neighborhood setting configuration was assessed using data derived from multitemporal MODIS LC for the Regional District of Bulkley-Nechako, Canada, comparing subareas exhibiting different amounts of LCCs with trends obtained for the full region. Next, outcomes were compared with three other study regions. The modeling results were evaluated with three-map comparison measures, where the real-world LC for the next timestep, the real-world LC for the previous timestep, and the forecasted LC for the next year were used to calculate correctly transitioned areas. Across all regions explored, it was observed that increasing neighborhood sizes improved the DL model’s capabilities to forecast short-term LCCs. CNN–TCN models forecasted the most correct LCCs for several regions while reducing error due to quantity when provided additional spatial variables. This study contributes to the systematic exploration of neighborhood sizes on selected spatiotemporal DL techniques for geographic applications. Full article

Interferometric Synthetic Aperture Radar (InSAR) data processing applications, such as deformation monitoring and topographic mapping, require an interferometric phase filtering step. Indeed, the filtering quality significantly impacts the deformation and terrain height estimation accuracy. However, the existing classical and deep learning-based phase filtering methods provide artefacts in the filtered areas where a large amount of noise prevents retrieving the original signal. In this way, we can no longer distinguish the underlying informative signal for the next processing step. This paper proposes a deep convolutional neural network filtering method, developing a novel learning strategy to preserve the initial phase noise input into these crucial areas. Thanks to the encoder–decoder powerful phase feature extraction ability, the network can predict an accurate coherence and filtered interferometric phase, ensuring reliable final results. Furthermore, we also address a novel Synthetic Aperture Radar (SAR) interferograms simulation strategy that, using initial parameters estimated from real SAR images, considers physical behaviors typical of a real acquisition. According to the results achieved on simulated and real InSAR data, the proposed filtering method significantly outperforms the classical and deep learning-based ones. Full article

Global Navigation Satellite System interferometric reflectometry (GNSS-IR), a new and inexpensive technique, has become available to the broader scientific community for detecting surface environmental information, such as soil moisture, snow depth and vegetation growth. However, there have been limited experiments focusing on the potential of crop height retrieval, especially the performance evaluation of BeiDou Navigation Satellite System (BDS) with other GNSS. Accuracy and reliability are challenging to achieve with traditional methods utilizing a single GNSS, and few measured verification data. In this study, an improved method that includes segmentation processing and multi-GNSS fusion is proposed based on GPS/GLONASS/Galileo/BDS multi-frequency data. Furthermore, experiments were carried out on a farmland in Fengqiu County, Henan Province, China. The results show that the height retrievals from four GNSS were in good agreement with the in situ observations during the whole growth cycle of the wheat after overwintering. Meanwhile, the retrievals based on the proposed method exhibited greater correspondence than the single frequency results, the correlation coefficient was increased and the root-mean-square error (RMSE) was reduced, respectively. Therefore, this study illustrates the feasibility of the proposed method to precisely estimate wheat height and its potential for use in the early warning of wheat lodging based on GNSS-IR. Full article

Precision Agriculture (PA) is an approach to maximizing crop productivity in a sustainable manner. PA requires up-to-date, accurate and georeferenced information on crops, which can be collected from different sensors from ground, aerial or satellite platforms. The use of optical and thermal sensors from Unmanned Aerial Vehicle (UAV) platform is an emerging solution for mapping and monitoring in PA, yet many technological challenges are still open. This technical note discusses the choice of UAV type and its scientific payload for surveying a sample area of 5 hectares, as well as the procedures for replicating the study on a larger scale. This case study is an ideal opportunity to test the best practices to combine the requirements of PA surveys with the limitations imposed by local UAV regulations. In the field area, to follow crop development at various stages, nine flights over a period of four months were planned and executed. The usage of ground control points for optimal georeferencing and accurate alignment of maps created by multi-temporal processing is analyzed. Output maps are produced in both visible and thermal bands, after appropriate strip alignment, mosaicking, sensor calibration, and processing with Structure from Motion techniques. The discussion of strategies, checklists, workflow, and processing is backed by data from more than 5000 optical and radiometric thermal images taken during five hours of flight time in nine flights throughout the crop season. The geomatics challenges of a georeferenced survey for PA using UAVs are the key focus of this technical note. Accurate maps derived from these multi-temporal and multi-sensor surveys feed Geographic Information Systems (GIS) and Decision Support Systems (DSS) to benefit PA in a multidisciplinary approach. Full article

Suitability of the vegetation indices of normalized difference vegetation index (NDVI), blue normalized difference vegetation index (BNDVI), and normalized difference yellowness index (NDYI) obtained by means of UAV at the flowering stage of oil seed rape for the prediction of seed yield and usability of these vegetation indices in the identification of anomalies in the condition of the flowering growth were verified based on the regression analysis. Correlation analysis was performed to find the degree of yield dependence on the values of NDVI, BNDVI, and NDYI indices, which revealed a strong, significant linear positive dependence of seed yield on BNDVI (R = 0.98) and NDYI (R = 0.95). The level of correlation between the NDVI index and the seed yield was weaker (R = 0.70) than the others. Regression analysis was performed for a closer determination of the functional dependence of NDVI, BNDVI, and NDYI indices and the yield of seeds. Coefficients of determination in the linear regression model of NDVI, BNDVI, and NDYI indices reached the following values: R² = 0.48 (NDVI), R² = 0.95 (BNDVI), and R² = 0.90 (NDYI). Thus, it was shown that increased density of yellow flowers decreased the relationship between NDVI and crop yield. The NDVI index is not appropriate for assessing growth conditions and prediction of yields at the flowering stage of oil seed rape. High accuracy of yield prediction was achieved with the use of BNDVI and NDYI. The performed analysis of NDVI, BNDVI, and NDYI demonstrated that particularly the BNDVI and NDYI indices can be used to identify problems in the development of oil seed rape growth at the stage of flowering, for their precise localization, and hence to targeted and effective remedial measures in line with the principles of precision agriculture. Full article

With natural disasters continuing to become more prevalent in recent years, the need for effective disaster management efforts becomes even more critical. Specifically, flooding is an extremely common natural disaster which can cause significant damage to homes and other property. In this article, we look at an area in Hurley, Virginia which suffered a significant flood event in August 2021. A drone is used to capture aerial imagery of the area and reconstructed to produce 3-dimensional models, Digital Elevation Models, and stitched orthophotos for flood modeling and damage assessment. Pre-flood Digital Elevation Models and available weather data are used to perform simulations of the flood event using HEC-RAS software. These were validated with measured water height values and found to be very accurate. After this validation, simulations are performed using the Digital Elevation Models collected after the flood and we found that a similar rainfall event on the new terrain would cause even worse flooding, with water depths between 29% and 105% higher. These simulations could be used to guide recovery efforts as well as aid response efforts for any future events. Finally, we look at performing semantic segmentation on the collected aerial imagery to assess damage to property from the flood event. While our segmentation of debris needs more work, it has potential to help determine the extent of damage and aid disaster response. Based on our investigation, the combination of techniques presented in this article has significant potential to aid in preparation, response, and recovery efforts for natural disasters. Full article

Multiclass geospatial object detection in high-spatial-resolution remote-sensing images (HSRIs) has recently attracted considerable attention in many remote-sensing applications as a fundamental task. However, the complexity and uncertainty of spatial distribution among multiclass geospatial objects are still huge challenges for object detection in HSRIs. Most current remote-sensing object-detection approaches fall back on deep convolutional neural networks (CNNs). Nevertheless, most existing methods only focus on mining visual characteristics and lose sight of spatial or semantic relation discriminations, eventually degrading object-detection performance in HSRIs. To tackle these challenges, we propose a novel hybrid attention-driven multistream hierarchical graph embedding network (HA-MHGEN) to explore complementary spatial and semantic patterns for improving remote-sensing object-detection performance. Specifically, we first constructed hierarchical spatial graphs for multiscale spatial relation representation. Then, semantic graphs were also constructed by integrating them with the word embedding of object category labels on graph nodes. Afterwards, we developed a self-attention-aware multiscale graph convolutional network (GCN) to derive stronger for intra- and interobject hierarchical spatial relations and contextual semantic relations, respectively. These two relation networks were followed by a novel cross-attention-driven spatial- and semantic-feature fusion module that utilizes a multihead attention mechanism to learn associations between diverse spatial and semantic correlations, and guide them to endowing a more powerful discrimination ability. With the collaborative learning of the three relation networks, the proposed HA-MHGEN enables grasping explicit and implicit relations from spatial and semantic patterns, and boosts multiclass object-detection performance in HRSIs. Comprehensive and extensive experimental evaluation results on three benchmarks, namely, DOTA, DIOR, and NWPU VHR-10, demonstrate the effectiveness and superiority of our proposed method compared with that of other advanced remote-sensing object-detection methods. Full article

The quantitative retrieval of the chlorophyll-a concentration is an important remote sensing method that is used to monitor the nutritional status of water bodies. The high spatial resolution of the Sentinel-2 MSI and its subdivision in the red-edge band highlight the characteristics of water chlorophyll-a, which is an important detection tool for assessing water quality parameters in plateau lakes. In this study, the Nine Plateau Lakes in the Yunnan-Kweichow Plateau of China were selected as the study area. Using Sentinel-2 MSI transit images and in situ measured chlorophyll-a concentration as the data source, the chlorophyll-a concentrations of plateau lakes (CCAPLs) were investigated, and the surface temperatures of plateau lakes (STPLs) were retrieved to verify the hypothesis that the lake surface temperature could increase the chlorophyll-a concentration. By comparing feature importance using a random forest (RF), the Sentinel-2 MSI surface reflectance and in situ data were linearly fitted using four retrieval spectral indices with high feature importance, and the accuracy of the estimated concentration of chlorophyll-a was evaluated by monitoring station data in the same period. Then, Landsat-8 TIRS Band 10 data were used to retrieve the STPL with a single-channel temperature retrieval algorithm and to verify the correlation between the STPL and the CCAPL. The results showed that the retrievals of the CCAPL and the STPL were consistent with the actual situation. The root-mean-square error (RMSE) of the fifteenth normalized difference chlorophyll-a index (NDCI15) was 0.0249. When the CCAPL was greater than 0.05 mg/L and the STPL was within 28–34 °C, there was a positive linear correlation between the CCAPL and the STPL. These results will provide support for the remote sensing monitoring of eutrophication in plateau lakes and will contribute to the scientific and effective management of plateau lakes. Full article

Rangeland ecosystems comprise more than a third of the global land surface, sustaining essential ecosystem services and livelihoods. In Spain, Southeast Spain includes some of the driest regions; accordingly, rangelands from Murcia and Almeria provinces were selected for this study. We used time series metrics and the Hurst Exponent from rescale range and detrended fluctuation analysis to cluster different rangeland dynamics to classify temporally and spatially diverse rangelands. The metrics were only calculated for three time periods that showed significant NDVI changes: March to April, April to July, and September to December. Detrended fluctuation analysis was not previously employed to cluster vegetation. This study used it to improve rangeland classification. K-means and unsupervised random forest were used to cluster the pixels using time series metrics and Hurst exponents. The best clustering results were obtained when unsupervised random forest was used with the Hurst exponent calculated with detrended fluctuation analysis. We used the Silhouette Index to evaluate the clustering results and a spatial comparison with topographical data. Our results show that adding the Hurst exponent, calculated with detrended fluctuation analysis, provided a better classification when clustering NDVI time series, while classifications without the Hurst exponent or with the Hurst exponent calculated with the rescale range method showed lower silhouette values. Overall, this shows the importance of using detrending when calculating the Hurst exponent on vegetation time series, and its usefulness in studying rangeland dynamics for management and research. Full article

A detailed picture of internal solitary waves (ISWs) in the White Sea is presented based on an analysis of historical spaceborne synthetic aperture radar (SAR) data and field measurements. The major hot-spot of ISW generation locates in the southwestern (SW) Gorlo Strait (GS), characterized by the presence of strong tides, complex topography, and two distinct fronts. Here, pronounced high-frequency isopycnal depressions of 5–8 m were regularly observed during flood and flood/ebb slackening. Other regions of pronounced ISW activity are found near Solovetsky Islands and in the northwestern Onega Bay. The spatial and kinematic properties of the observed ISWs are linked to water depth, with larger wave trains and higher propagation speeds being observed over the deep regions. Direct estimates of ISW propagation speeds from sequential and single SAR images agree well, while theoretical ones obtained using a two-layer model overestimate the observed values by 2–3 times. This is explained by the effective modulation of ISW propagation speed during the tidal cycle by background currents that are not accounted for in the model. Enhanced values of diapycnic diffusion coefficient in the pycnocline layer were registered near the frontal zones, where intense 14–17 m high ISWs were regularly observed. Full article

The intensification of extreme precipitation has aggravated urban flood disasters, which makes timely and reliable precipitation information urgently needed. As the high-quality and widely used satellite precipitation products, Integrated Multi-satellitE Retrievals for GPM (IMERG), have not been well investigated in coastal urban agglomerations where damages from precipitation-related disasters are more severe. With precipitation measurements from local high-density gauge stations, this study evaluates three IMERG runs (IMERG ER, IMERG LR, and IMERG FR) in the southeast coastal urban region of China. The evaluation shows that the three IMERG products severely overestimate weak precipitation and underestimate heavy precipitation. Among the three runs, the post-corrected IMERG FR does not show a substantial improvement compared to the near-real-time IMERG ER and IMERG LR. The performance of IMERG varies depending on the precipitation pattern and intensity, with the best estimation ability occurring in the coastal urban region in summer and in the northern forests in winter. Due to the year-round urban effect on precipitation variability, IMERG cannot detect precipitation events well in the central high-density urban areas, and has its best detection ability on cultivated lands in summer and forests in winter. Within the urban agglomeration, IMERG shows a poorer performance in areas with higher urbanization levels. Thus, the IMERG products for coastal urban areas need considerable improvements, such as regionalized segmental corrections based on precipitation intensity and the adjustment of short-duration estimates by daily or sub-daily precipitation measurements. Full article

The expansion of plantation poses new challenges for mapping forest, especially in mountainous regions. Using multi-source data, this study explored the capability of the random forest (RF) algorithm for the extraction and mapping of five forest types located in Yanqing, north China. The Google Earth imagery, forest inventory data, GaoFen-1 wide-field-of-view (GF-1 WFV) images and DEM were applied for obtaining 125 features in total. The recursive feature elimination (RFE) method selected 32 features for mapping five forest types. The results attained overall accuracy of 87.06%, with a Kappa coefficient of 0.833. The mean decrease accuracy (MDA) reveals that the DEM, LAI and EVI in winter and three texture features (entropy, variance and mean) make great contributions to forest classification. The texture features from the NIR band are important, while the other texture features have little contribution. This study has demonstrated the potential of applying multi-source data based on RF algorithm for extracting and mapping plantation forest in north China. Full article