Showing papers by "Dengsheng Lu published in 2019"
TL;DR: Using multiple source data—spectral bands, vegetation indices, textures, and topographic factors—considerably improved land-cover and forest classification accuracies compared to spectral bands alone, and no one classification algorithm provided the best accuracy for all tree species classes.
Abstract: The global availability of high spatial resolution images makes mapping tree species distribution possible for better management of forest resources. Previous research mainly focused on mapping single tree species, but information about the spatial distribution of all kinds of trees, especially plantations, is often required. This research aims to identify suitable variables and algorithms for classifying land cover, forest, and tree species. Bi-temporal ZiYuan-3 multispectral and stereo images were used. Spectral responses and textures from multispectral imagery, canopy height features from bi-temporal stereo imagery, and slope and elevation from the stereo-derived digital surface model data were examined through comparative analysis of six classification algorithms including maximum likelihood classifier (MLC), k-nearest neighbor (kNN), decision tree (DT), random forest (RF), artificial neural network (ANN), and support vector machine (SVM). The results showed that use of multiple source data—spectral bands, vegetation indices, textures, and topographic factors—considerably improved land-cover and forest classification accuracies compared to spectral bands alone, which the highest overall accuracy of 84.5% for land cover classes was from the SVM, and, of 89.2% for forest classes, was from the MLC. The combination of leaf-on and leaf-off seasonal images further improved classification accuracies by 7.8% to 15.0% for land cover classes and by 6.0% to 11.8% for forest classes compared to single season spectral image. The combination of multiple source data also improved land cover classification by 3.7% to 15.5% and forest classification by 1.0% to 12.7% compared to the spectral image alone. MLC provided better land-cover and forest classification accuracies than machine learning algorithms when spectral data alone were used. However, some machine learning approaches such as RF and SVM provided better performance than MLC when multiple data sources were used. Further addition of canopy height features into multiple source data had no or limited effects in improving land-cover or forest classification, but improved classification accuracies of some tree species such as birch and Mongolia scotch pine. Considering tree species classification, Chinese pine, Mongolia scotch pine, red pine, aspen and elm, and other broadleaf trees as having classification accuracies of over 92%, and larch and birch have relatively low accuracies of 87.3% and 84.5%. However, these high classification accuracies are from different data sources and classification algorithms, and no one classification algorithm provided the best accuracy for all tree species classes. This research implies the same data source and the classification algorithm cannot provide the best classification results for different land cover classes. It is necessary to develop a comprehensive classification procedure using an expert-based approach or hierarchical-based classification approach that can employ specific data variables and algorithm for each tree species class.
97 citations
TL;DR: The method demonstrated the effectiveness of using a fusion of optical and SAR data for large area urban land extraction especially in areas where optical data fail to distinguish urban land from spectrally similar objects.
Abstract: Accurate and timely urban land mapping is fundamental to supporting large area environmental and socio-economic research. Most of the available large-area urban land products are limited to a spatial resolution of 30 m. The fusion of optical and synthetic aperture radar (SAR) data for large-area high-resolution urban land mapping has not yet been widely explored. In this study, we propose a fast and effective urban land extraction method using ascending/descending orbits of Sentinel-1A SAR data and Sentinel-2 MSI (MultiSpectral Instrument, Level 1C) optical data acquired from 1 January 2015 to 30 June 2016. Potential urban land (PUL) was identified first through logical operations on yearly mean and standard deviation composites from a time series of ascending/descending orbits of SAR data. A Yearly Normalized Difference Vegetation Index (NDVI) maximum and modified Normalized Difference Water Index (MNDWI) mean composite were generated from Sentinel-2 imagery. The slope image derived from SRTM DEM data was used to mask mountain pixels and reduce the false positives in SAR data over these regions. We applied a region-specific threshold on PUL to extract the target urban land (TUL) and a global threshold on the MNDWI mean, and slope image to extract water bodies and high-slope regions. A majority filter with a three by three window was applied on previously extracted results and the main processing was carried out on the Google Earth Engine (GEE) platform. China was chosen as the testing region to validate the accuracy and robustness of our proposed method through 224,000 validation points randomly selected from high-resolution Google Earth imagery. Additionally, a total of 735 blocks with a size of 900 × 900 m were randomly selected and used to compare our product’s accuracy with the global human settlement layer (GHSL, 2014), GlobeLand30 (2010), and Liu (2015) products. Our method demonstrated the effectiveness of using a fusion of optical and SAR data for large area urban land extraction especially in areas where optical data fail to distinguish urban land from spectrally similar objects. Results show that the average overall, producer’s and user’s accuracies are 88.03%, 94.50% and 82.22%, respectively.
77 citations
TL;DR: Wang et al. as discussed by the authors used spectral differences of the bamboo forest to distinguish between on-year and off-year bamboo forests and found that the spectral differences between the two types are mainly reflected in the red-edge 3 near-infrared (NIR) and NIR narrow bands (740-865 nm).
38 citations
TL;DR: This research indicated that the combined use of RCH-variables and spectral data provided more accurate AGB estimation for larch plantations than the use of spectral data alone.
Abstract: Data saturation in optical sensor data has long been recognized as a major factor that causes underestimation of aboveground biomass (AGB) for forest sites having high AGB, but there is a lack of suitable approaches to solve this problem. The objective of this research was to understand how incorporation of forest canopy features into high spatial resolution optical sensor data improves forest AGB estimation. Therefore, we explored the use of ZiYuan-3 (ZY-3) satellite imagery, including multispectral and stereo data, for AGB estimation of larch plantations in North China. The relative canopy height (RCH) image was calculated from the difference of digital surface model (DSM) data at leaf-on and leaf-off seasons, which were extracted from the ZY-3 stereo images. Image segmentation was conducted using eCognition on the basis of the fused ZY-3 multispectral and panchromatic data. Spectral bands, vegetation indices, textural images, and RCH-based variables based on this segment image were extracted. Linear regression was used to develop forest AGB estimation models, where the dependent variable was AGB from sample plots, and explanatory variables were from the aforementioned remote-sensing variables. The results indicated that incorporation of RCH-based variables and spectral data considerably improved AGB estimation performance when compared with the use of spectral data alone. The RCH-variable successfully reduced the data saturation problem. This research indicated that the combined use of RCH-variables and spectral data provided more accurate AGB estimation for larch plantations than the use of spectral data alone. Specifically, the root mean squared error (RMSE), relative RMSE, and mean absolute error values were 33.89 Mg/ha, 29.57%, and 30.68 Mg/ha, respectively, when using the spectral-only model, but they become 24.49 Mg/ha, 21.37%, and 20.37 Mg/ha, respectively, when using the combined model with RCH variables and spectral band. This proposed approach provides a new insight in reducing the data saturation problem.
25 citations
TL;DR: Wang et al. as discussed by the authors developed a hybrid approach combining expert knowledge, decision tree, threshold technique, unsupervised classification and postprocessing for wetland classification and three typical sites were selected to analyze different change patterns.
Abstract: Wetland monitoring is crucial for understanding wetland changes and responses to natural and anthropogenic actions. In this research, Hangzhou Bay coastal wetland was selected as the study area to analyze spatial patterns and dynamic changes based on Landsat multitemporal imagery. A hybrid approach combining expert knowledge, decision tree, threshold technique, unsupervised classification and postprocessing was developed for wetland classification. Three typical sites were selected to analyze different change patterns. Over the last 10 years, the wetlands have undergone dramatic changes, continued to expand outwards due to natural accumulation and increased in area. At the same time, these wetlands have also partially decreased with artificial reclamation and urban construction. An analysis of typical sites found that the change in the western part of this region was dominated by natural accumulation and that in the central and eastern parts was dominated by reclamation due to geographical location and hydrological power. In general, wetlands are continually changing under the joint action of natural accumulation, artificial reclamation and urban construction. The local terrain, hydrology, soil, population, economy and policy also influence wetland changes. The results of wetland monitoring are essential for the protection and management of local wetlands.
19 citations
TL;DR: In this article, the difference of major land cover change results caused in various time periods and the impacts of human-induced factors on land-cover changes along the land cover along the route of land cover changes.
Abstract: This research aims to understand the difference of major land-cover change results caused in various time periods and to examine the impacts of human-induced factors on land-cover changes along the
16 citations
TL;DR: Wang et al. as discussed by the authors proposed that surface shortwave, longwave, and net radiations are key components of surface short-wave and long-wave radiations in metropolitan Beijing.
Abstract: Metropolitan Beijing is facing many environmental problems such as haze and urban heat island due to the rapid urbanization. Surface shortwave, longwave, and net radiations are key components of th...
16 citations
TL;DR: A comparative study on the performance of eight linear regression parameter estimation methods in the subtropical forest biomass remote sensing model development shows that BIC performs best in comprehensive evaluation, while NNG, Cp and AIC perform poorly as a whole.
Abstract: In the field of quantitative remote sensing of forest biomass, a prominent phenomenon is the increasing number of explanatory variables. Then how to effectively select explanatory variables has become an important issue. Linear regression model is one of the commonly used remote sensing models. In the process of establishing the linear regression model, a vital step is to select explanatory variables. Focusing on variable selection and model stability, this paper conducts a comparative study on the performance of eight linear regression parameter estimation methods (Stepwise Regression Method (SR), Criterions Based on The Bayes Method (BIC), Criterions Based on The Bayes Method (AIC), Criterions Based on Prediction Error (Cp), Least Absolute Shrinkage and Selection Operator (Lasso), Adaptive Lasso, Smoothly Clipped Absolute Deviation (SCAD), Non-negative garrote (NNG)) in the subtropical forest biomass remote sensing model development. For the purpose of comparison, OLS and RR, are commonly used as methods with no variable selection ability, and are also compared and discussed. The performance of five aspects are evaluated in this paper: (i) Determination coefficient, prediction error, model error, etc., (ii) significance test about the difference between determination coefficients, (iii) parameter stability, (iv) variable selection stability and (v) variable selection ability of the methods. All the results are obtained through a five ten-fold CV. Some evaluation indexes are calculated with or without degrees of freedom. The results show that BIC performs best in comprehensive evaluation, while NNG, Cp and AIC perform poorly as a whole. Other methods show a great difference in the performance on each index. SR has a strong capability in variable selection, although it is poor in commonly used indexes. The short-wave infrared band and the texture features derived from it are selected most frequently by various methods, indicating that these variables play an important role in forest biomass estimation. Some of the conclusions in this paper are likely to change as the study object changes. The ultimate goal of this paper is to introduce various model establishment methods with variable selection capability, so that we can have more choices when establishing similar models, and we can know how to select the most appropriate and effective method for specific problems.
14 citations
TL;DR: Land cover change may affect water and carbon cycles when transitioning from one land cover category to another (land cover conversion, LCC) or when the characteristics of the land cover type are a...
Abstract: Land-cover change may affect water and carbon cycles when transitioning from one land-cover category to another (land-cover conversion, LCC) or when the characteristics of the land-cover type are a...
10 citations
TL;DR: Li et al. as mentioned in this paper used the Gaofen-1 images in combination with digital elevation model (DEM) data to develop a land-cover map with emphasis on young and mature hickory plantations using the random forest (RF) approach.
8 citations
TL;DR: Li et al. as mentioned in this paper proposed a hierarchical principle on remotely sensed urban land-use/cover classification for mapping intra-urban structure/component dynamics, which can be used to enhance our understanding of urbanization impacts on ecological and regional climatic conditions and urban dwellers' environments.
Abstract: . Accurate urban land-cover datasets are essential for mapping urban environments. However, a series of national urban land-cover data covering more than 15 years that characterizes urban environments is relatively rare. Here we propose a hierarchical principle on remotely sensed urban land-use/cover classification for mapping intra-urban structure/component dynamics. China's Land Use/cover Dataset (CLUD) is updated, delineating the imperviousness, green surface, waterbody and bare land conditions in cities. A new data subset called CLUD-Urban is created from 2000 to 2015 at five-year intervals with a medium spatial resolution (30 m). The first step is a prerequisite to extract the vector boundaries covered with urban areas from CLUD. A new method is then proposed using logistic regression between urban impervious surface area (ISA) and the annual maximum Normalized Difference Vegetation Index (NDVI) value retrieved from Landsat images based on a big-data platform with Google Earth Engine. National ISA and urban green space (UGS) fraction datasets for China are generated at 30-meter resolution with five-year intervals from 2000 to 2015. The overall classification accuracy of national urban areas is 92 %. The root mean square error values of ISA and UGS fractions are 0.10 and 0.14, respectively. The datasets indicate that the total urban area of China was 6.28 × 104 km2 in 2015, with average fractions of 70.70 % and 26.54 % for ISA and UGS, respectively. The ISA and UGS increased between 2000 and 2015 with unprecedented annual rates of 1,311.13 km2/yr and 405.30 km2/yr, respectively. CLUD-Urban can be used to enhance our understanding of urbanization impacts on ecological and regional climatic conditions and urban dwellers' environments. CLUD-Urban can be applied in future researches on urban environmental research and practices in the future. The datasets can be downloaded from https://doi.org/10.5281/zenodo.2644932 .