Accurate mapping of forest types using dense seasonal Landsat time-series

An accurate map of forest types is important for proper usage and management of forestry resources. Medium resolution satellite images (e.g., Landsat) have been widely used for forest type mapping because they are able to cover large areas more efficiently than the traditional forest inventory. However, the results of a detailed forest type classification based on these images are still not satisfactory. To improve forest mapping accuracy, this study proposed an operational method to get detailed forest types from dense Landsat time-series incorporating with or without topographic information provided by DEM. This method integrated a feature selection and a training-sample-adding procedure into a hierarchical classification framework. The proposed method has been tested in Vinton County of southeastern Ohio. The detailed forest types include pine forest, oak forest, and mixed-mesophytic forest. The proposed method was trained and validated using ground samples from field plots. The three forest types were classified with an overall accuracy of 90.52% using dense Landsat time-series, while topographic information can only slightly improve the accuracy to 92.63%. Moreover, the comparison between results of using Landsat time-series and a single image reveals that time-series data can largely improve the accuracy of forest type mapping, indicating the importance of phenological information contained in multi-seasonal images for discriminating different forest types. Thanks to zero cost of all input remotely sensed datasets and ease of implementation, this approach has the potential to be applied to map forest types at regional or global scales.     

Crop type mapping using LiDAR,Sentinel-2 and aerial imagery with machine learning algorithms

LiDAR data are becoming increasingly available, which has opened up many new applications. One such application is crop type mapping. Accurate crop type maps are critical for monitoring water use, estimating harvests and in precision agriculture. The traditional approach to obtaining maps of cultivated fields is by manually digitizing the fields from satellite or aerial imagery and then assigning crop type labels to each field - often informed by data collected during ground and aerial surveys. However, manual digitizing and labeling is time-consuming, expensive and subject to human error. Automated remote sensing methods is a cost-effective alternative, with machine learning gaining popularity for classifying crop types. This study evaluated the use of LiDAR data, Sentinel-2 imagery, aerial imagery and machine learning for differentiating five crop types in an intensively cultivated area. Different combinations of the three datasets were evaluated along with ten machine learning. The classification results were interpreted by comparing overall accuracies, kappa, standard deviation and f-score. It was found that LiDAR data successfully differentiated between different crop types, with XGBoost providing the highest overall accuracy of 87.8％. Furthermore, the crop type maps produced using the LiDAR data were in general agreement with those obtained by using Sentinel-2 data, with LiDAR obtaining a mean overall accuracy of 84.3％ and Sentinel-2 a mean overall accuracy of 83.6％. However, the combination of all three datasets proved to be the most effective at differentiating between the crop types, with RF providing the highest overall accuracy of 94.4％. These findings provide a foundation for selecting the appropriate combination of remotely sensed data sources and machine learning algorithms for operational crop type mapping.     

In-season crop classification using elements of the Kennaugh matrix derived from polarimetric RADARSAT-2 SAR data

Accurate spatio-temporal classification of crops is of prime importance for in-season crop monitoring. Synthetic Aperture Radar (SAR) data provides diverse physical information about crop morphology. In the present work, we propose a day-wise and a time-series approach for crop classification using full-polarimetric SAR data. In this context, the 4 × 4 real Kennaugh matrix representation of a full-polarimetric SAR data is utilized, which can provide valuable information about various morphological and dielectric attributes of a scatterer. The elements of the Kennaugh matrix are used as the parameters for the classification of crop types using the random forest and the extreme gradient boosting classifiers.The time-series approach uses data patterns throughout the whole growth period, while the day-wise approach analyzes the PolSAR data from each acquisition into a single data stack for training and validation. The main advantage of this approach is the possibility of generating an intermediate crop map, whenever a SAR acquisition is available for any particular day. Besides, the day-wise approach has the least climatic influence as compared to the time series approach. However, as time-series data retains the crop growth signature in the entire growth cycle, the classification accuracy is usually higher than the day-wise data.Within the Joint Experiment for Crop Assessment and Monitoring (JECAM) initiative, in situ measurements collected over the Canadian and Indian test sites and C-band full-polarimetric RADARSAT-2 data are used for the training and validation of the classifiers. Besides, the sensitivity of the Kennaugh matrix elements to crop morphology is apparent in this study. The overall classification accuracies of 87.75% and 80.41% are achieved for the time-series data over the Indian and Canadian test sites, respectively. However, for the day-wise data, a ∼6% decrease in the overall accuracy is observed for both the classifiers.     

Development of a modified neural network-based land cover classification system using automated data selector and multiresolution remotely sensed data

Integrating multiple images with artificial neural networks (ANN) improves classification accuracy. ANN performance is sensitive to training datasets. Complexity and errors compound when merging multiple data, pointing to needs for new techniques. Kohonen's self-organizing mapping (KSOM) neural network was adapted as an automated data selector (ADS) to replace manual training data processes. The multilayer perceptron (MLP) network was then trained using automatically extracted datasets and used for classification. Two hypotheses were tested: ADS adapted from the KSOM network provides adequate and reliable training datasets, improving MLP classification performance; and fusion of Landsat thematic mapper (TM) and SPOT images using the modified ANN approach increases accuracy. ADS adapted from the KSOM network improved training data quality and increased classification accuracy and efficiency. Fusion of compatible multiple data can improve performance if appropriate training datasets are collected. This proved to be a viable classification scheme particularly where acquiring sufficient and reliable training datasets is difficult.     

Geospatial method for computing supplemental multi-decadal US coastal land use and land cover classification products,using Landsat data and C-CAP products

This paper discusses the development and implementation of a method that can be used with multi-decadal Landsat data for computing general coastal US land use and land cover (LULC) maps consisting of seven classes. With Mobile Bay, Alabama as the study region, the method that was applied to derive LULC products for nine dates across a 34-year time span. Classifications were computed and refined using decision rules in conjunction with unsupervised classification of Landsat data and Coastal Change and Analysis Program value-added products. Each classification’s overall accuracy was assessed by comparing stratified random locations to available high spatial resolution satellite and aerial imagery, field survey data and raw Landsat RGBs. Overall classification accuracies ranged from 83 to 91% with overall κ statistics ranging from 0.78 to 0.89. Accurate classifications were computed for all nine dates, yielding effective results regardless of season and Landsat sensor. This classification method provided useful map inputs for computing LULC change products.     

A kernel functions analysis for support vector machines for land cover classification

Information about the Earth's surface is required in many wide-scale applications. Land cover/use classification using remotely sensed images is one of the most common applications in remote sensing, and many algorithms have been developed and applied for this purpose in the literature. Support vector machines (SVMs) are a group of supervised classification algorithms that have been recently used in the remote sensing field. The classification accuracy produced by SVMs may show variation depending on the choice of the kernel function and its parameters. In this study, SVMs were used for land cover classification of Gebze district of Turkey using Landsat ETM+ and Terra ASTER images. Polynomial and radial basis kernel functions with their estimated optimum parameters were applied for the classification of the data sets and the results were analyzed thoroughly. Results showed that SVMs, especially with the use of radial basis function kernel, outperform the maximum likelihood classifier in terms of overall and individual class accuracies. Some important findings were also obtained concerning the changes in land use/cover in the study area. This study verifies the effectiveness and robustness of SVMs in the classification of remotely sensed images.     

Assessing the suitability of data from Sentinel-1A and 2A for crop classification

Sentinel-1A C-SAR and Sentinel-2A MultiSpectral Instrument (MSI) provide data applicable to the remote identification of crop type. In this study, six crop types (beans, beetroot, grass, maize, potato, and winter wheat) were identified using five C-SAR images and one MSI image acquired during the 2016 growing season. To assess the potential for accurate crop classification with existing supervised learning models, the four different approaches namely kernel-based extreme learning machine (KELM), multilayer feedforward neural networks, random forests, and support vector machine were compared. Algorithm hyperparameters were tuned using Bayesian optimization. Overall, KELM yielded the highest performance, achieving an overall classification accuracy of 96.8%. Evaluation of the sensitivity of classification models and relative importance of data types using data-based sensitivity analysis showed that the set of VV polarization data acquired on 24 July (Sentinel-1A) and band 4 data (Sentinel-2A) had the greatest potential for use in crop classification.     

单变量特征选择的苏北地区主要农作物遥感识别

遥感识别多源特征综合和特征优选是提高遥感影像分类精度的关键技术。农作物遥感识别中,识别特征的相对单一和数量过多均会导致作物识别精度不理想。随机森林(random forests)采用分类与回归树(CART)算法来生成分类树,结合了bagging和随机选择特征变量的优点,是一种有效的分类方法。单变量特征选择(univariate feature selection)能够对每一个待分类的特征进行测试,衡量该特征和响应变量之间的关系,根据得分舍弃不好的特征,优选得到的特征用于分类。本文基于随机森林和单变量特征选择,利用多时相光谱信息、植被指数信息、纹理信息及波段差值信息,设计多组分类实验方案,对江苏省泗洪县的高分一号(GF-1)和环境一号(HJ-1A)影像进行分类研究,旨在选择最佳的分类方案对实验区主要农作物进行识别和提取。实验结果表明:(1)多源信息综合的农作物分类精度明显高于单一的原始光谱特征分类,说明不同类型特征的引入能改善分类效果;(2)基于单变量特征选择算法的优选特征分类效果最佳,总体精度97.07%,Kappa系数0.96,表明了特征优选在降低维度的同时,也保证了较高的分类精度。随机森林和单变量特征选择结合的方法可以提高遥感影像的分类精度,为农作物的识别和提取研究提供了有效的方法。     

Assessing the potential of integrated Landsat 8 thermal bands,with the traditional reflective bands and derived vegetation indices in classifying urban landscapes

Reliable and up-to-date urban land cover information is valuable in urban planning and policy development. Due to the increasing demand for reliable land cover information there has been a growing need for robust methods and datasets to improve the classification accuracy from remotely sensed imagery. This study sought to assess the potential of the newly launched Landsat 8 sensor’s thermal bands and derived vegetation indices in improving land cover classification in a complex urban landscape using the support vector machine classifier. This study compared the individual and combined performance of Landsat 8’s reflective, thermal bands and vegetation indices in classifying urban land use-land cover. The integration of Landsat 8 reflective bands, derived vegetation indices and thermal bands overall produced significantly higher accuracy classification results than using traditional bands as standalone (i.e. overall, user and producer accuracies). An overall accuracy above 89.33% and a kappa index of 0.86, significantly higher than the one obtained with the use of the traditional reflective bands as a standalone data-set and other analysis stages. On average, the results also indicate high producer and user accuracies (i.e. above 80%) for most of the classes with a McNemar’s Z score of 9.00 at 95% confidence interval showing significant improvement compared with classification using reflective bands as standalone. Overall, the results of this study indicate that the integration of the Landsat 8’s OLI and TIR data presents an invaluable potential for accurate and robust land cover classification in a complex urban landscape, especially in areas where the availability of high resolution datasets remains a challenge.     

Analyzing multi-year rice-fallow dynamics in Odisha using multi-temporal Landsat-8 OLI and Sentinel-1 Data

ABSTRACT

Sustainable intensification of existing cropland is one of the most viable options for meeting the escalating food demands of the ever-increasing population in the world. Accurate geospatial data about the potential single-crop (rice-fallows) areas is vital for policymakers to target the agro-technologies for enhancing crop productivity and intensification. Therefore, the study aimed to evaluate and understand the dynamics of rice-fallows in the Odisha state of India, using SAR (Sentinel-1) and Optical (Landsat OLI) datasets. This study utilized a decision-tree approach and Principal component analysis (PCA) for the segmentation and separation of different vegetation classes. The estimated overall accuracy of extracted rice-fallow maps was in the range of 84 to 85 percent. The study identified about 2.2, 2.0 and 2.1mha of Rice-Fallows in the years 2015–16, 2016–17, and 2017–18, respectively. The combined analysis (all three years) of rice-fallow maps identified about 1.34mha of permanent rice-fallows, whereas the remaining 0.6–0.8mha area was under the current-fallow category. About 50% of the total permanent rice-fallows were detected in the rainfed areas of Mayurbhanj, Bhadrak, Bolangir, Sundargarh, Keonjhar, Baleswar, Nawarangpur and Bargarh districts. The study also illustrated the time-series profiles of SMAP (soil moisture) datasets for the ten agroclimatic zones of the Odisha, which can be utilized (along with rice-fallow maps) for the selection of crop and cultivars (e.g. short or medium duration pulses or oilseeds) to target the rice fallows. The approach utilized in the current study can be scaled up in similar areas of South and South-east Asia and Africa to identify single-crop areas for targeting improved technologies for enhanced crop productivity and intensification.     

Temporal replication of the national land cover database using active machine learning

Regional and national level land cover datasets, such as the National Land Cover Database (NLCD) in the United States, have become an important resource in physical and social science research. Updates to the NLCD have been conducted every 5 years since 2001; however, the procedure for producing a new release is labor-intensive and time-consuming, taking 3 or 4 years to complete. Furthermore, in most countries very few, if any, such releases exist, and thus there is high demand for efficient production of land cover data at different points in time. In this paper, an active machine learning framework for temporal updating (or backcasting) of land cover data is proposed and tested for three study sites covered by the NLCD. The approach employs a maximum entropy classifier to extract information from one Landsat image using the NLCD, and then replicate the classification on a Landsat image for the same geographic extent from a different point in time to create land cover data of similar quality. Results show that this framework can effectively replicate the land cover database in the temporal domain with similar levels of overall and within class agreement when compared against high resolution reference land cover datasets. These results demonstrate that the land cover information encapsulated in the NLCD can effectively be extracted using solely Landsat imagery for replication purposes. The algorithm is fully automated and scalable for applications at landscape and regional scales for multiple points in time.     

Assessing intra-annual vegetation regrowth after fire using the pixel based regeneration index

Several remote sensing studies have discussed the potential of satellite imagery as an alternative for extensive field sampling to quantify fire-vegetation impact over large areas. Most studies depend on Landsat image availability with infrequent image acquisition dates and consequently are limited for assessing intra-annual fire-vegetation dynamics or comparing different fire plots and dates. The control pixel based regeneration index (pRI) derived from SPOT-VEGETATION (VGT) normalized difference vegetation index (NDVI) is used in this study as an alternative to the traditional bi-temporal Landsat approach based on the normalized burn ratio (NBR). The major advantage of the pRI is the use of unburnt control plots which allow the expression of the intra-annual variation due to regeneration processes without external influences. In the comparison of Landsat and VGT data, (i) the inter-annual differences between the bi-temporal and control plot approach were contrasted and (ii) metrics of pRI were derived and compared with the inter-annual dynamics of both VGT and Landsat data. Results of these comparisons, demonstrate the overall similarity between NBR and NDVI data, stress the importance of the elimination of external influences (e.g., phenological variations), and emphasize the failure of including post-fire vegetation responses in bi-temporal Landsat assessments, especially in quickly recovering ecotypes with a strong annual phenological cycle such as savanna. This highlights the importance of using high frequency multi-temporal approaches to estimate fire-vegetation impact in temporally dynamic vegetation types.     

基于随机森林的土地利用分类与景观格局分析

基于景观格局分析方法,分别选取8个反映景观格局类型和6个反映景观水平格局指数用于探索土地利用变化状况.实验结果表明:①基于选择合适的训练样本下,对3期遥感影像采用随机森林的监督分类方法,总体精度均在94％以上,kappa系数均为90％以上;②综合来看,研究期内建设用地为优势景观类型.研究区斑块数量增加,破碎化程度有所加...     

综合多特征的Landsat 8时序遥感图像棉花分类方法

传统的多时相遥感图像分类大多拘泥于单一特征,本文基于多时相的Landsat 8遥感数据,开展了综合多特征的特征提取与特征选择方法研究。综合了NDVI时间序列、最佳时相反射率光谱特征以及纹理特征作为初始分类特征,并采用基于属性重要度的粗糙集特征选择算法对其进行特征约简。分类结果表明:(1)利用初始分类特征,分类的总体精度达到92.81%,棉花提取精度达87.4%,与仅利用NDVI时间序列相比,精度分别提高5.53%和5.05%;(2)利用粗糙集选择后的特征分类,分类总体精度可达93.66%,棉花分类精度达92.73%,与初始分类特征提取结果相比,棉花分类精度提高5.33%。基于属性重要度的粗糙集特征选择不仅提高了分类精度,同时有效降低了分类器的计算复杂度。     

Learning with transductive SVM for semisupervised pixel classification of remote sensing imagery

Land cover classification using remotely sensed data requires robust classification methods for the accurate mapping of complex land cover area of different categories. In this regard, support vector machines (SVMs) have recently received increasing attention. However, small number of training samples remains a bottleneck to design suitable supervised classifiers. On the other hand, adequate number of unlabeled data is available in remote sensing images which can be employed as additional source of information about margins. To fully leverage all of the precious unlabeled data, integration of filtering in a transductive SVM is proposed.Using two labeled image datasets of small size and two large unlabeled image datasets, the effectiveness of the proposed method is explored. Experimental results show that the proposed technique achieves average overall accuracies of around 4.5–7.8%, 0.8–2.6% and 0.9–2.2% more than the standard inductive SVM (ISVM), progressive transductive SVM (PTSVM) and low density separation (LDS) classifiers, respectively on larger domains in case of labeled datasets. Using image datasets, visual interpretation from the classified images as well as the segmentation quality reveal that the proposed method can efficiently filter informative data from the unlabeled samples.     

A Spatial Database of Cropping System and its Characteristics to Aid Climate Change Impact Assessment Studies

Cropping system study is not only useful to understand the overall sustainability of agricultural system, but also it helps in generating many important parameters which are useful in climate change impact assessment. Considering its importance, Space Applications Centre, took up a project for mapping and characterizing major cropping systems of Indo-Gangetic Plains of India. The study area included the five states of Indo-Gangetic Plains (IGP) of India, i.e. Punjab, Haryana, Uttar Pradesh, Bihar and West Bengal. There were two aspects of the study. The first aspect included state and district level cropping system mapping using multi-date remote sensing (IRS-AWiFS and Radarsat ScanSAR) data. The second part was to characterize the cropping system using moderate spatial resolution multi-date remote sensing data (SPOT VGT NDVI) and ground survey. The remote sensing data was used to compute three cropping system performance indices (Multiple Cropping Index, Area Diversity Index and Cultivated Land Utilization Index). Ground survey was conducted using questionnaires filled up by 1,000 farmers selected from 103 villages based on the cropping systems map. Apart from ground survey, soil and water sampling and quality analysis were carried out to understand the effect of different cropping systems and their management practices. The results showed that, rice-wheat was the major cropping system of the IGP, followed by Rice-Fallow-Fallow and Maize-Wheat. Other major cropping systems of IGP included Sugarcane based, Pearl millet-Wheat, Rice-Fallow-Rice, Cotton-Wheat. The ground survey could identify 77 cropping systems, out of which 38 are rice-based systems. Out of these 77 cropping systems, there were 5 single crop systems, occupying 6.5% coverage (of all cropping system area), 56 double crop systems with 72.7% coverage, and 16 triple crop systems with 20.8% coverage. The cropping system performance analysis showed that the crop diversity was found to be highest in Haryana, while the cropping intensity was highest in Punjab state.     

Performance Analysis of Maximum Likelihood and Artificial Neural Network Classifiers for Training Sets with Mixed Pixels

This study evaluates the performance of an artificial neural network, specifically a multilayer perceptron, and a maximum likelihood algorithm to classify multitemporal Landsat ETM+ remote sensor data. The study area in Turkey is a mountainous region that contains many small scattered fields, usually 5-10 pixels in size. The classifiers were employed to identify eight land cover/use features covering the bulk of the study area using the same training and test datasets in order to avoid any difference resulting from sampling variations. Results show that the neural network approach performed better in extracting land cover information from multispectral and multitemporal images with training data sets including a large amount of mixed and atypical pixels. The maximum likelihood classifier was found to be ineffective, particularly in classifying spectrally similar categories and classes having subclasses.     

Investigating Rice Cropping Practices and Growing Areas from MODIS Data Using Empirical Mode Decomposition and Support Vector Machines

This study investigated rice cropping practices and rice growing areas in the Vietnamese Mekong Delta using MODIS 250 × 250 m normalized difference vegetation index (NDVI) data acquired during the 2002 and 2007 rice cropping seasons. Data processing was conducted in five main steps: (1) constructing time-series MODIS NDVI data; (2) noise filtering of the time-series MODIS NDVI data using empirical mode decomposition (EMD); (3) extracting and evaluating phenological rice training patterns from the smooth time profiles of NDVI; (4) classifying rice cropping systems using support vector machines (SVMs); and (5) conducting an error analysis using ground reference data and government rice statistics. The results indicated that EMD was an efficient filter for noise removal in the time-series MODIS NDVI data. The filtered temporal NDVI profile characterized the distinct behaviors of the rice cropping systems. The estimated sowing and harvesting dates were compared with the field-survey data and indicated root mean square error (RMSE) values of 7.5 and 8.2 days, respectively. The comparison results between the 2002 classification map and the ground reference data indicated that the overall accuracy for the 2002 data was 92.9% with a Kappa coefficient of 0.89, while in 2007 these values were 93.8% and 0.90, respectively. At the district level, there was good agreement between the MODIS-based estimated areas and government rice statistics for 2002 and 2007 (R ² ≥ 0.85). An investigation of changes in cropping practices from 2002 to 2007 showed that 12.9% of the area used for double-cropped irrigated rice in 2002 had been converted to triple-cropped irrigated rice by 2007, whereas 27.4% of the area used for triple-cropped irrigated rice in 2002 had been converted to double-cropped irrigated rice by 2007.     

Improving the mapping of crop types in the Midwestern U.S. by fusing Landsat and MODIS satellite data

Mapping crop types is of great importance for assessing agricultural production, land-use patterns, and the environmental effects of agriculture. Indeed, both radiometric and spatial resolution of Landsat’s sensors images are optimized for cropland monitoring. However, accurate mapping of crop types requires frequent cloud-free images during the growing season, which are often not available, and this raises the question of whether Landsat data can be combined with data from other satellites. Here, our goal is to evaluate to what degree fusing Landsat with MODIS Nadir Bidirectional Reflectance Distribution Function (BRDF)-Adjusted Reflectance (NBAR) data can improve crop-type classification. Choosing either one or two images from all cloud-free Landsat observations available for the Arlington Agricultural Research Station area in Wisconsin from 2010 to 2014, we generated 87 combinations of images, and used each combination as input into the Spatial and Temporal Adaptive Reflectance Fusion Model (STARFM) algorithm to predict Landsat-like images at the nominal dates of each 8-day MODIS NBAR product. Both the original Landsat and STARFM-predicted images were then classified with a support vector machine (SVM), and we compared the classification errors of three scenarios: 1) classifying the one or two original Landsat images of each combination only, 2) classifying the one or two original Landsat images plus all STARFM-predicted images, and 3) classifying the one or two original Landsat images together with STARFM-predicted images for key dates. Our results indicated that using two Landsat images as the input of STARFM did not significantly improve the STARFM predictions compared to using only one, and predictions using Landsat images between July and August as input were most accurate. Including all STARFM-predicted images together with the Landsat images significantly increased average classification error by 4% points (from 21% to 25%) compared to using only Landsat images. However, incorporating only STARFM-predicted images for key dates decreased average classification error by 2% points (from 21% to 19%) compared to using only Landsat images. In particular, if only a single Landsat image was available, adding STARFM predictions for key dates significantly decreased the average classification error by 4 percentage points from 30% to 26% (p < 0.05). We conclude that adding STARFM-predicted images can be effective for improving crop-type classification when only limited Landsat observations are available, but carefully selecting images from a full set of STARFM predictions is crucial. We developed an approach to identify the optimal subsets of all STARFM predictions, which gives an alternative method of feature selection for future research.     

Forest aboveground biomass estimation in Zhejiang Province using the integration of Landsat TM and ALOS PALSAR data

In remote sensing–based forest aboveground biomass (AGB) estimation research, data saturation in Landsat and radar data is well known, but how to reduce this problem for improving AGB estimation has not been fully examined. Different vegetation types have their own species composition and stand structure, thus they have different data saturation values in Landsat or radar data. Optical and radar data also have different characteristics in representing forest stand structures, thus effective use of their features may improve AGB estimation. This research examines the effects of Landsat Thematic Mapper (TM) and ALOS PALSAR L-band data and their integrations in forest AGB estimation of Zhejiang Province, China, and the roles of textural images from both datasets. The linear regression models of AGB were conducted by using (1) Landsat TM alone, (2) ALOS PALSAR data alone, (3) their combination as extra bands, and (4) their data fusion, based on non-stratification and stratification of vegetation types, respectively. The results show that (1) overall, Landsat TM data perform better than PALSAR data, but the latter can produce more accurate estimates for bamboo and shrub, and for forests with AGB values less than 60 Mg/ha; (2) the combination of TM and PALSAR data as extra bands can greatly improve AGB estimation performance, but their fusion using the modified high-pass filter resolution-merging technique cannot; (3) textures are indeed valuable in AGB estimation, especially for forests with complex stand structures such as mixed forests and pine forests with understories of broadleaf species; (4) stratification of vegetation types can improve AGB estimation performance; and (5) the results from the linear regression models are characterized by overestimation and underestimation for the smaller and larger AGB values, respectively, and thus, selecting non-linear models or non-parametric algorithms may be needed in future research.