Integrated Use of Existing Global Land Cover Datasets for Producing a New Global Land Cover Dataset with a Higher Accuracy: A Case Study in Eurasia

It has been commonly acknowledged that the current global mapping projects have encountered the accuracy challenge. By conducting a comparison among the four existing global land cover datasets (MODIS LC, GLC2000, GLCNMO and GLOBCOVER), it has been identified that certain areas’ accuracy has dragged down the overall accuracy of these global land cover datasets. In this paper, those areas have been defined as the “unreliable area”. This study has recollected the training data from the “unreliable area” within the above four mentioned datasets and reclassified the “unreliable area” by using two supervised classifications. The final result has shown that compared with any existing datasets, a relatively higher accuracy has been able to achieve.

Detection of landuse/landcover changes using remotely-sensed data

We evaluated the use of spatial sampling and satellite images to identify deforested areas in Wonju, South Korea. The changes in land cover were identified using a grid of sample points overlaid onto medium and high-resolution remote sensing （RS） satellite images. Deforestation identified in this way （hereafter, RSD） was compared to administrative data on deforestation. We also compared high-resolution satellite images （HR-RSD） and actual deforestation based on categories which were Intergovernmental Panel on Climate Change data. RSD generated by medium-resolution satellite images overesti- mated the amount of deforested area by 1.5-2.4 times the actual deforested area, whereas RSD generated by HR- RSD underestimated the amount of deforested area by 0.4-0.9 times the actual area. The highest degree of matching （90 %） was found in HR-RSD with a grid interval of 500 m and the accuracy of HR-RSD was the highest, at 67 %. The results also revealed that the largest cause of deforestation was the establishment of settlements followed by conversion to cropland and grassland. We conclude that for the identification of deforestation using satellite images, HR-RSD with a grid interval of 500 m is most suitable.

Land Cover Map Delineation, for Agriculture Development, Case Study in North Sinai, Egypt Using SPOT4 Data and Geographic Information System

Land cover map for a part of North Sinai was produced using the FAO—Land Cover Classification System (LCCS) of 2004. The standard FAO classification scheme provides a standardized system of classification that can be used to analyze spatial and temporal land cover variability in the study area. This approach also has the advantage of facilitating the integration of Sinai land cover mapping products to be included with the regional and global land cover datasets. The total study area is 7450 km2 (1,773,842) feddans. The landscape classification was performed on SPOT4 data acquired in 2011 using combined multi-spectral bands of 20 meter spatial resolution. Geographic Information System (GIS) was used to edit the classification result in order to reach the maximum possible accuracy. GIS was also used to include all necessary information. The identified vegetative land cover classes of the study area are irrigated herbaceous crops, irrigated tree crops and rain fed tree crops. The non-vegetated land covers in the study area include: bare rock, bare soil, bare soil stony, bare soil very stony, bare soil salt crusts, loose and shifting sands and sand dunes. The water bodies were classified as artificial perennial water bodies (fish ponds and irrigated canals) and natural perennial water bodies as lakes (standing) and rivers (flowing). Artificial surfaces in the study area include linear and non-linear. The produced maps and the statistics of the different land covers are included in the following sub-sections.

Land cover mapping of deciduous forest regions using ETM+ data: a case study of Azerbaijan Province, Iran

Up to date information about the existing land cover patterns and changes in land cover over time is one of the prime prerequisites for the preparation of an integrated development plan and economic development program of a region. By using ETM＋ image data from 2002, we provided a land cover map of deciduous forest regions in Azerbaijan Province, Iran. Initial qualitative evaluation of the data showed no significant radiometric errors. Image classification was carried out using a maximum likelihood-based supervised classification method. In the end, we determined five major land cover classes, i.e., grass lands, deciduous broad-leaf forest, cultivated land, river and land without vegetation cover. Accuracy, estimated by the use of criteria such as overall accuracy from a confusion matrix of classification was 86% with a 0.88 Kappa coefficient. Such high accuracy results demonstrate that the combined use of spectral and textural characteristics increased the number of classes in the field classification, also with excellent accuracy. The availability and use of time series of remote sensing data permit the detection and quantification of land cover changes and improve our understanding of the past and present status of forest ecosystems.

Vegetation Cover Density and Land Surface Temperature Interrelationship Using Satellite Data, Case Study of Wadi Bisha, South KSA

The aim of this study is to identify the relationship between Vegetation Cover (VC) and the land Surface Temperature (LST), using satellite data of Wadi Bisha, south the Kingdome of Saudi Arabia (KSA). The Landsat 7 Thematic Mapper (ETM) thermal band (band 6) was used for calculating the (LST) values. The near-infrared (NIR) and red band (bands 3 and 4 respectively) were used for estimating the vegetation cover. ERDAS Imagine 9.3 and ArcGIS 10.2 were used in the current study. The results of the study show that the increase of vegetation cover (VC) coincides with decrease of (LST), while the decrease in vegetation cover is linked with increase of (LST). It was found that there was no vegetation observed in areas practiced the highest temperature of 49℃, while areas of lowest temperature of 28℃ were characterized by dense vegetation cover. Thus, a quite significant correlation is approved between the (VC) and the (LST), based on the validation of (50) locations. It was concluded that availability and continuity of Satellite remote sensing data was required for elaborating a continuous monitoring of vegetation cover conditions and mapping was recommended in Wadi Bisha. Operational monitoring is recommended to ensure the adoption of flexible land cover validation protocols.

Inventory of Atmospheric Pollutant Emissions from Burning of Crop Residues in China Based on Satellite-retrieved Farmland Data

The burning of crop residues emits large quantities of atmospheric aerosols.Published studies have developed inventories of emissions from crop residue burning based on statistical data.In contrast,this study used satellite-retrieved land-cover data(1 km×1 km)as activity data to compile an inventory of atmospheric pollutants emitted from the burning of crop residues in China in 2015.The emissions of PM10,PM2.5,VOCs,NOx,SO2,CO,and NH3 from burning crop straw on nonirrigated farmland in China in 2015 were 610.5,598.4,584.4,230.6,35.4,3329.3,and 36.1 Gg(1 Gg=109 g),respectively;the corresponding emissions from burning paddy rice residues were 234.1,229.7,342.3,57.5,57.5,1122.1,and 21.5 Gg,respectively.The emissions from crop residue burning showed large spatial and temporal variations.The emissions of particulate matter and gaseous pollutants from crop residue burning in nonirrigated farmland were highest in east China,particularly in Shandong,Henan,Anhui,and Sichuan provinces.Emissions from burning paddy rice residue were highest in east and central China,with particularly high levels in Shandong,Jiangsu,Zhejiang,and Hunan provinces.The monthly variations in atmospheric pollutant emissions were similar among different regions,with the highest levels observed in October in north,northeast,northwest,east,and southwest China and in June and July in central and south China.The developed inventory of emissions from crop residue burning is expected to help improve air quality models by providing high-resolution spatial and temporal data.

Forest Change and Its Effect on Biomass in Yok Don National Park in Central Highlands of Vietnam Using Ground Data and Geospatial Techniques

This paper assesses the changes in forest cover in Yok Don National Park of Vietnam between 2004 and 2010, and the implications of such changes on the biomass stocks of this national park. Remote sensing and GIS tools along with the ground truth data collected from the field were employed for classifying the forest types of the study area from SPOT HRV satellite imagery for years 2004 and 2010. The total area considered in this study is 115.5 thousand ha. Five different categories of forests were identified. The results demonstrated that between 2004 and 2010, the Evergreen broad leaved rich quality forest decreased by 11.2 thousand ha (3.5 Mega tons of biomass) and the Dry open dipterocarps medium quality forest decreased by 15.3 thousand ha (2.5 Mega tons of biomass). In that time period, the Evergreen broad leaved medium quality forest increased by 3.2 thousand ha (0.8 Mega tons of biomass), the Evergreen broad leaved poor quality forest increased by 2.5 thousand ha (0.24 Mega tons of biomass), and the Dry open dipterocarps poor quality forest increased by 3.2 thousand ha (0.69 Mega tons of biomass). Total biomass of the study area decreased by 4.3 Mega tons.

Assessment of Land Use Land Cover Change Detection Using Geospatial Techniques in Southeast Rajasthan

Change analysis acquires effective information in the form of maps and statistical data which becomes the central component in spatial planning, monitoring environmental changes, management and utilization of land. The present study makes an attempt to assess the changes in land use land cover using multi-temporal satellite data in south-east Rajasthan. These maps were derived from geocoded dia-positive False Color Composites (FCC’s) of IRS 1991, 2001, 2010 & 2018 using Arc GIS platform. The present study demonstrates the extension, approach and result of change analysis which might be helpful for decision making and sustainable growth. The landscape has been divided into 12 categories. Mining and its associated features were increased whereas forest and open scrub cover shows decreasing trend during the study period. The former increased by 23.82 km2 while the later shrunk by 26.08 km2. Most significant changes are also witnessed in settlement and industrial areas which shows increment by 8.8 km2 and 1.33 km2. Stone quarrying has destroyed arable land, natural vegetation cover, topsoil, subsoil and consequently the soil profile of the area. On the other hand cultivated land is increasing due to the conversion of uncultivated land and scrub cover with facilitation of irrigation and modern agricultural activities under different government schemes. The study shows that the area of 184.88 km2 has undergone significant spatial and temporal changes during the study period.

全球30m地表覆盖遥感数据产品-Globe Land30

在国家863重点项目支持下,我国成功研制出全球30 m地表覆盖数据产品Globe Land30。该成果包括2000基准年和2010基准年两期,有耕地、森林、草地、灌木地、湿地、水体、苔原、人造地表、裸地和冰雪十大类型,第三方评价总体精度为83.50%。2014年9月22日,中国政府将其赠送给联合国使用,是中国向联合国提供的首个全球地理信息产品。该成果成为全球变化和可持续发展研究的重要科学数据,目前已有近120个国家的用户下载使用,推动了国际对地观测与地理信息的开放共享,彰显了中国负责任大国的形象。本文主要介绍了Globe Land30产品的技术创新、精度评价与成果应用。

基于Landsat TM/ETM数据的锡林河流域土地覆盖变化

根据1987年、1991年、1997年和2000年4期Landsat TM/ETM影像的土地利用/土地覆盖分类结果,运用地理信息系统空间分析方法,分析了内蒙古锡林河流域1987～2000年间各土地利用类型及草甸草原、典型草原、荒漠草原的数量变化和空间变化特征.分析结果显示,锡林河流域土地利用/土地覆盖变化的主要特征为草甸草原、典型草原面积的大幅减少和荒漠草原、农田和沙漠化土地面积的大幅增加及城镇的扩张.其中面积增加最大的是荒漠草原,增加了2328 km2;相当于1987年荒漠草原面积的56%.农出和城镇面积逐年增大,分别从1987年的114.3 km2和25.2 km2增加到2000年的332.1 km2和43.6 km2.面积减少最多的是羊草+丛生禾草、羊草+杂类草等优良高产温带典型草原类型,共减少2040km2.草甸草原面积亦呈逐年减少的趋势,从1987年的1103 km2减少到2000年375 km2,面积减少了65.9%.农出、沙化地及城镇等非草原土地利用类型面积增加了62.5%.

高分一号与Landsat-8影像在荒漠绿洲过渡带应用对比

针对国产高分一号卫星(GF-1)成像质量是否可以满足区域生态环境监测需求的问题,开展了宽幅多光谱相机(wide field view,WFV)在荒漠绿洲过渡带的成像质量评估研究。从辐射质量、纹理、地类识别精度和归一化植被指数等方面构建评估指标,定量分析了GF-1 WFV和Landsat-8OLI在荒漠绿洲过渡带的成像质量差异。结果表明:GF-1 WFV影像虽然具有较高的空间分辨率,但在辐射质量、地类识别效果、纹理信息及植被指数等方面与Landsat-8OLI相比有一定差距;GF-1 WFV影像的信噪比优势明显,对噪声的抑制效果较好;通过与纹理信息的波段组合,可以有效提高GF-1WFV影像的地物识别效果,缩小与Landsat-8OLI在分类精度上的差距;鉴于明显的光谱范围差异,二者归一化植被指数数据在协同应用的过程中宜分地物类型转换,在西北荒漠绿洲过渡带的国土资源调查、城市规划、农情监测等方面可发挥积极作用。

鄱阳湖滨湖区多季相Landsat 8 OLI数据遴选及其土地利用/覆被遥感分类提取

鄱阳湖滨湖地区是中国南方内陆浅水型湖泊滨岸带与红壤丘陵交错的典型区域。研究以江西省上饶市鄱阳县为例,在ENVI遥感平台的支持下,利用2014年多季相Landsat 8 OLI遥感影像数据资料,通过实地土地利用/覆被调查与地物光谱分离度分析进行遥感数据遴选,对优选数据采用马氏距离(MD)法、最大似然(ML)分类器、人工神经网络(ANN)分类器、支持向量机(SVM)分类器、光谱信息散度(SID)分类器、波谱角分类器(SAM)等8种监督分类方法进行监督分类。在先验知识和野外调查数据的基础上,对分类结果进行分析;然后通过总体分类精度、混淆矩阵和Kappa系数对几种分类器的精度进行评价和比较。并针对鄱阳湖滨湖地区土地利用/覆盖分类问题探究几种遥感分类方法的优劣,挑选最适用遥感分类方法。结果表明:在鄱阳湖滨湖地区,各季节影像中地表光谱可分性在8月达到最佳值,期间所获取是优选数据,有利于遥感分类识别;人工神经网络法、支持向量机法(SVM)具有较好的的分类效果和分类精度,总体分类精度>96%,Kappa系数>0.94。总体上讲,SVM分类器的分类精度和Kappa系数都要优于其他分类器,相比之下7-10月获取的影像更适合于鄱阳湖滨湖地区土地利用/覆被的遥感识别和提取。通过遥感数据最优遴选,并结合优选遥感分类器,可以显著提高鄱阳湖滨湖区土地利用/覆盖信息提取效率和遥感高精度分类的总体精度,有助于提升该地区大范围土地利用/覆盖遥感宏观监测的可靠性和精准度。

基于Landsat TM与CBER-02B的土地覆盖变化分析

以Landsat TM和CBERS-02B图像为遥感数据源,运用RS与GIS技术,对遥感数据的处理方法和技术流程进行系统的研究,并对汉寿县三期遥感影像进行处理,增强影像的可识别性。在野外调查和室内判别的基础上,建立研究区解译标志,采用机助目视解译方法,建立了1994年、2003年及2008年汉寿县土地覆盖数据库,提取并分析汉寿县土地覆盖变化信息。

GlobeLand30湿地细化分类研究

基于30 m地表覆盖数据产品完成湿地精细化分类,能够更好地满足当前较高分辨率及较详尽全球湿地数据的应用需求。本文在深入分析湿地分类体系与细化方法的基础上,提出以湿地细化类别的定义、多元知识的分层分类、亚类数据精细化提取为主线的总体研究思路,制定了基于先验知识的对象系统筛选、基于森林数据的同位像元提取、基于最佳阈值的极大似然掩膜的主体分类方法,并应用于数据生产实践获得8个亚类信息。该方法克服了常规手段普遍存在的周期长、效率低等弊端,实现了全球较高分辨率湿地亚类数据的快速精确制图,总体分类精度达82.6%,对地理世情及其他地表覆盖研究具有借鉴意义。

基于众源地理数据的GlobeLand30人造地表数据精度验证

GlobeLand30数据已被广泛应用于生态、环境、气候等研究领域,对其进行精度验证是保证GlobeLand30产品推广应用的首要前提。本文提出了一种基于众源地理数据的GlobeLand30人造地表精度验证方法:首先利用随机抽样的方法布设样本点,其次应用LDA(latent Dirichlet allocation)主题模型计算众源数据文档的地表覆盖主题,最后通过构建样本点与GlobeLand30数据产品之间的混淆矩阵,以实现人造地表数据的精度验证。研究区域为我国27个省级行政区,众源地理数据选用POI数据,对本文提出的方法进行了实例验证。结果表明:2000年和2010年的人造地表总体精度分别为87.63%和88.81%,这与GlobeLand30人造地表第三方评价精度86.97%相近,验证了本文方法的有效性。

一种利用Landsat年度时序数据的土地覆盖分类方法

针对大面积土地覆盖遥感分类中数据获取难度大、复杂度高、分类结果不够精确且易受季候变化影响等问题,提出了一种利用Landsat时间序列数据,生成年度时序特征,并结合特定算法(UniBagging)进行土地覆盖分类的方法(LandUTime)。该方法定义了一种基于时间序列数据的特征生成方式,根据时序数据特点,设计了一种基于特征子空间的集成分类算法。实现过程分为2个阶段,首先基于特定模型,在像元级别上对Landsat时间序列图像进行回归分析,生成模式特征,然后将所有特征整合成"特征块",根据特征子空间将基分类器集成到相互分离的集合中,最后通过加权投票的方法进行分类结果输出。实验结果与定量分析表明,与传统的特征提取及分类方法相比,该方法提高了分类精度,而且对高维数据具有鲁棒性;可以有效克服大面积土地覆盖分类中云遮掩、数据条带和物候变化等问题的影响,具有较高的准确性和实用性。

基于LandsatTM/OLI数据的吐鲁番绿洲土地利用/覆被变化研究

[目的]运用3S技术研究1990-2015年吐鲁番地区土地利用/覆被变化情况。[方法]以吐鲁番地区1990年landsat5 TM影像及2015年landsat8 OLI影像数据为基础，运用3S技术，基于光谱特征的最大似然监督分类法和人机交互式目视解译方法，获取1990、2015年吐鲁番绿洲土地利用类型数据。最后运用土地利用格局变化指标分析方法及空间转移矩阵方法，探讨了1990-2015年吐鲁番绿洲土地资源的结构特点和变化规律。[结果]吐鲁番绿洲呈扩大趋势，绿洲面积增加，未利用地减少。其中，耕地、草地、园林地面积增加显著，裸地和盐碱地面积减少，沙地面积小幅增加。绿洲区域内建设用地和水域面积增加，但绿洲外围的尾闾湖—艾丁湖却呈萎缩趋势。[结论]该研究为吐鲁番地区生态环境保护提供了基础数据与科学依据。

Applying the Distributed Hydrological Model for Tropical Monsoon Basins by Using Earth Observation Data (Case Studies: Kone and Ba River Basins)

Due to the limitation of data sources, the application of Distributed Hydrological Models (DHMs) using earth observation data to research water resources is necessary. In this study, the BTOPMC (Block-wise use of TOPMODEL) model was applied for 2 basins in the tropical monsoon region. This is the first time that the land cover map of the CCI (Climate Change Initiative Land Cover Team) was prepared for input data instead of IGBP (International Geosphere-Biosphere Programme) land cover map as proposed in the demo version of the BTOPMC model. The calibration and validation results showed that the Nash-Sutcliffe coefficients for daily stream discharge were 77.5% and 68.7% at Cung Son station (Ba basin). The Nash-Sutcliffe coefficients for daily stream discharge were 79.4% and 69.0% at Binh Tuong station (Kone basin), respectively. Because of a stop in measuring the discharge at Binh Tuong station in 2007, this model was applied to simulate discharge during the period of 2008-2015. Furthermore, the effect of land cover on discharge at Cung Son station was considered. The annual discharge in 2010 at Cung Son decreased 8 m3/s in the comparison between two scenarios (land cover of 2000 and 2010). According to this result, it is possible to propose a wide application range of the DHMs model to the tropical monsoon river basins using earth observation data.

面向多源异质遥感影像地物分类的自监督预训练方法

近年来,深度学习改变了遥感图像处理的方法。由于标注高质量样本费时费力,标签样本数量不足的现实问题会严重影响深层神经网络模型的性能。为解决这一突出矛盾,本文提出了用于多源异质遥感影像地物分类的自监督预训练和微调分类方案,旨在缓解模型对于标签样本的严重依赖。具体来讲,生成式自监督学习模型由非对称的编码器-解码器结构组成,其中深度编码器从多源遥感数据中学习高阶关键特征,任务特定的解码器用于重建原始遥感影像。为提升特性表示能力,交叉注意力机制模型用于融合异源特征中的信息,进而从多源异质遥感影像中学习更多的互补信息。在微调分类阶段,预训练好的编码器作为无监督特征提取器,基于Transformer结构的轻量级分类器将学习到的特征与光谱信息结合并用于地物分类。这种自监督预训练方案能够从多源异质遥感影像中学习到刻画原始数据的高级关键特征,并且此过程不需要任何人工标注信息,从而缓解了对标签样本的依赖。与现有的分类范式相比,本文提出的自监督预训练和微调方案在多源遥感影像地物分类中能够取得更优的分类结果。

新疆天山北坡城市群土地覆被数据产品精度评价

土地覆被产品为全球各种地球系统科学应用提供了重要的地表覆被信息,如CLCD30(30 m)、GlobeLand30(30 m,简称Globe30)、GLC_FCS30(30 m)、FROM-GLC10(10 m)、Esri_Land_Cover_2020(10 m,简称Esri10),以及ESAWorldCover2020(10 m,简称ESA10),然而其在局地的精度和适用性如何尚不明确。基于Sentinel-2影像,通过样本精度评价和类别混淆评价对上述6套土地覆被数据产品在天山北坡城市群的精度及误差情况进行了研究和分析,并探讨了误差成因和不同数据产品的适用性。结果表明:(1)6套数据产品中,除Esri10外,其余5套数据产品的类型构成、面积占比相对一致。(2)GLC-FCS30、Globe30、CLCD30、FROM-GLC10、ESA10、Esri10的总体精度分别为0.8080、0.8147、0.7880、0.8531、0.8047、0.4725。(3)从产品适用性来看,GLC_FCS30适用于对耕地、裸地的分析,CLCD30适用于对林地、裸地的分析,FROM-GLC10适用于对草地、水体、冰/雪以及建筑的分析,ESA10适用于对耕地和草地的分析,Esri10适用于对林地、冰/雪以及建筑的分析,Globe30在各类别精度评价结果上更均衡。(4)类别混淆主要是耕地、林地以及草地之间及与其他类别的相互混淆,尤其是在土地覆被复杂的地区,如城市边缘区域。