An assessment of two classification methods for mapping Thames Estuary intertidal habitats using CASI data

The aim of this study was to find the most appropriate method of classification for the Thames intertidal habitat types at Crayford Marsh and Dartford Creek by using Compact Airborne Spectrographic Imager (CASI ) data. Preliminary evaluation of commonly available classification algorithms produced two candidate techniques: the Maximum Likelihood Classifier (MLC) and the Spectral Angle Mapper (SAM). Pre-classification enhancements and the two different classifiers were compared. Ten different dataset combinations were created for two pilot sites: one at Crayford Marsh and one at Dartford Creek. These consisted of the original CASI bandset (15 bands in spatial mode from blue to near-infrared) and nine other combinations resulting from band subsets, Principal Component Analysis (PCA) and Normalized Difference Vegetation Index (NDVI ). Twelve classes were established for each site although only some of these were common to both. Each classified image was accuracy assessed using a combination of field mapping, field photographs and air photograph interpretation as reference data. The most accurate classification (68% for Crayford Marsh and 53% for Dartford Creek) for both sites comprised the use of MLC with a dataset created from PCs 2, 3 and 4 from a PCA carried out on the original 15 band data, combined with an additional NDVI band. CASI data proved useful for the mapping of salt-marsh vegetation and sediments especially in the Crayford Marsh site. In the Dartford Creek site, however, there was significant confusion between some classes. Further work is recommended to test the classification     

Salt-marsh characterization, zonation assessment and mapping through a dual-wavelength LiDAR

Linking intertidal processes to their natural patterns within a framework of coastal erosion requires monitoring techniques providing high-resolution spatio-temporal data from the scale of processes to this of patterns. The Scanning Hydrographic Operational Airborne LiDAR Survey (SHOALS) consists of a ubiquitous topographic and bathymetric LiDAR (Light Detection And Ranging) system that has become an important technology for generating high-resolution Digital Terrain Models (DTM) and Digital Surface Models (DSM) over intertidal landscapes. The objectives of this project are i) to highlight the capacity of SHOALS Topography and intensity data (Red and Near-InfraRed) to detect intertidal vegetation, ii) to assess the salt-marsh zonation, and iii) to map intertidal habitats and its adjacent coastal areas (Gulf of St. Lawrence, Canada). The study area was selected based on the spectrum of land cover types, encompassing beach, salt-marsh, arable farm and urban coastal environments. Surfaces constructed from the LiDAR survey included DSM, DTM, Normalized Surface Model (NSM), Digital Intensity Model for InfraRed (DIMI), Digital Intensity Model for Red (DIMR), and Normalized Difference LiDAR Vegetation Index Model (NDLVIM), derived from the two previous models. The correlation between the so-called NDLVI and the amount of salt-marsh vegetation, measured in situ, was 0.87 (p < 0.01). Then, LiDAR-assessed salt-marsh ecological zonation allowed finding out intermediate and strong relationships between NDLVI and Topography (r² = 0.89, p < 0.038) and Topographic heterogeneity (r² = 0.54, p < 0.1394), respectively. Finally, NDLVI and Topography surfaces were classified using maximum likelihood algorithm into 17 classes, whose overall accuracy and kappa coefficient were 91.89% and 0.9088, respectively. These results support that (1) intertidal vegetation can be discriminated by NDLVI, (2) salt-marsh ecological zonation pattern, and (3) accurate coastal land cover maps can be satisfactorily generated from a single LiDAR survey using the NDLVIM and DTM approach.     

Landscape-scale characterization of cropland in China using Vegetation and Landsat TM images

In this landscape-scale study we explored the potential for multitemporal 10-day composite data from the Vegetation sensor to characterize land cover types, in combination with Landsat TM image and agricultural census data. The study area (175 km by 165 km) is located in eastern Jiangsu Province, China. The Normalized Difference Vegetation Index (NDVI ) and the Normalized Difference Water Index (NDWI ) were calculated for seven 10-day composite (VGT-S10) data from 11 March to 20 May 1999. Multi-temporal NDVI and NDWI were visually examined and used for unsupervised classification. The resultant VGT classification map at 1 km resolution was compared to the TM classification map derived from unsupervised classification of a Landsat 5 TM image acquired on 26 April 1996 at 30 m resolution to quantify percent fraction of cropland within a 1 km VGT pixel; resulting in a mean of 60% for pixels classified as cropland, and 47% for pixels classified as cropland/natural vegetation mosaic. The estimates of cropland area from VGT data and TM image were also aggregated to county-level, using an administrative county map, and then compared to the 1995 county-level agricultural census data. This landscape-scale analysis incorporated image classification (e.g. coarse-resolution VGT data, fineresolution TM data), statistical census data (e.g. county-level agricultural census data) and a geographical information system (e.g. an administrative county map), and demonstrated the potential of multi-temporal VGT data for mapping of croplands across various spatial scales from landscape to region. This analysis also illustrated some of the limitations of per-pixel classification at the 1 km resolution for a heterogeneous landscape.     

Vegetation maps,mapping needs and scope of digital processing of Landsat Thematic Mapper data in tropical region of south-west India

Abstract

This study was aimed at assessing the scope of Landsat Thematic Mapper (TM) data for vegetation classification and mapping needs in a tropical region of south-west India. Outputs generated through common digital enhancement/classification techniques were compared with the vegetation map prepared from visual interpretation of black and white panchromatic aerial photographs with a scale of 1:15000 (approximately), in terms of extractable thematic information and cost/time incurred. It has been shown that digital processing of TM data is capable of satisfying the classification and mapping needs in the country with a reasonable degree of precision (85 per cent), in much less cost and time when compared with the aerial photographs. Supervised classification using raw data was found to be more effective in discriminating vegetation types than enhancements like band ratioing and principal component analysis. It was possible to classify forest vegetation with respect to variability in bioclimatic and structural attributes by classification of digital data. In view of varied vegetation classification and mapping needs in India, it is suggested that as detailed a land cover classification as possible should be attempted initially. Subsequently, the detailed classification output, which is often difficult to read or understand, may be converted to user-specific simplified outputs by appropriate aggregation of classes.     

Caractérisation de formations végétales mediterranéennes à partir de données ‘Thematic Mapper’ Une étude de cas en Andalousie (Espagne)

Abstract

The study is focused on the characterization of vegetation formations in a Mediterranean area (943 km²) located in southern Spain: herbaceous canopies (rangelands), shrubby vegetation (‘matorral’) and complex woody/herbaceous formations (‘dehesa’). Vegetation formations (physiognomical units) have been characterized by their spectral responses in the six reflective TM channels and by vegetation indices. From the ratio index TM4/TM3 there has been derived a map displaying seven classes (water, bare soil and five biomass levels reflecting the hierarchy of vegetation formations). Channels TM3, TM4 and TM5 have been considered for a supervised classification into nine land-cover categories (seven vegetation formations, bare soil and water). The proportion of correct classification of vegetation formations is about 78 per cent when considering test areas. Classification made from three principal components gives similar results.     

A non-parametric,supervised classification of vegetation types on the Kaibab National Forest using decision trees

Traditional land classification techniques for large areas that use Landsat Thematic Mapper (TM) imagery are typically limited to the fixed spatial resolution of the sensors (30 m). However, the study of some ecological processes requires land cover classifications at finer spatial resolutions. We model forest vegetation types on the Kaibab National Forest (KNF) in northern Arizona to a 10-m spatial resolution with field data, using topographical information and Landsat TM imagery as auxiliary variables. Vegetation types were identified by clustering the field variables total basal area and proportion of basal area by species, and then using a decision tree based on auxiliary variables to predict vegetation types. Vegetation types modelled included pinyon-juniper, ponderosa pine, mixed conifer, spruce- and deciduous-dominated mixes, and openings. To independently assess the accuracy of the final vegetation maps using reference data from different sources, we used a post-stratified, multivariate composite estimator. Overall accuracy was 74.5% (Kappa statistic = 49.9%). Sources of error included differentiating between mixed conifer and spruce-dominated types and between openings in the forest and deciduous-dominated mixes. Overall, our non-parametric classification method successfully identified dominant vegetation types on the study area at a finer spatial resolution than can typically be achieved using traditional classification techniques.     

Mapping mixed vegetation communities in salt marshes using airborne spectral data

The aim of this study is to evaluate a new neural network classifier using spectrally sampled image data to map mixed halophytic vegetation in tidal environments. The work is based on the concept of vegetation communities, mixtures of several species, characteristic of salt marshes. The study site is the Venice lagoon, and the material available is a spectrally sampled Compact Airborne Spectral Imager (CASI) image, in conjunction with ground truth for precise characterization of vegetation communities. Detailed observations of vegetation species and of their fractional abundance were collected for 36 Regions Of Interest (ROI): such field polygons are used for classification training and accuracy assessment. To select the most significant spectral channels, the Spectral Reconstruction method was applied to the image data: a set of 6 bands was selected as optimal for classification, out of the 15 available. The spatial heterogeneity of salt-marsh vegetation is significant and even at the spatial resolution of the airborne CASI image data, mixed pixels are observed. The Vegetation Community based Neural Network Classifier (VCNNC) is introduced to cope with a situation where no pure pixels exist, and was applied to the set of 6 selected bands. Both quantitative and qualitative comparisons of classification results of VCNNC with those of conventional Neural Network Classifier (NNC), trained and assessed on exactly the same data sets, shows that VCNNC's accuracy is substantially higher (≈ 91%) than that of NNC (≈ 84%), while the Kappa coefficient is 0.87 for VCNNC and 0.75 for the NNC method.     

Discriminating between cool season and warm season grassland cover types in northeastern Kansas

This study assesses the ability of multitemporal Landsat Thematic Mapper (TM) data and the normalized difference vegetation index (NDVI) to spectrally separate grazed cool season and warm season grassland cover types in Douglas County, Kansas. Biophysical data collected during the summer of 1997 suggest that differences in the per cent of total living vegetation cover, per cent of senescent vegetation, and proportion of forb cover between the two grassland cover types could make cool season and warm season grassland cover types spectrally distinct. The results show that the two grassland cover types were spectrally different in several spring (May) and mid-summer (July) bands, but not in any fall (September) bands. Furthermore, the two grassland cover types could be discriminated with a high level of accuracy. Accuracy assessments of the three single dates showed that the mid-summer (July) image and NDVI discriminated between the grassland cover types most accurately (81.8%). The multitemporal TM and NDVI data did not improve the spectral discrimination of the two grassland cover types over the mid-summer image or NDVI and had classification accuracy levels of 63.6% and 68.2%, respectively.     

Vegetation classification by satellite image processing in a dry area of north-eastern Syria

As vegetation classification on the highly diverse rangeland is an inevitable procedure in evaluating total forage resources and assessing human impact in large areas, a supervised classification was conducted by satellite image processing using geocoded bands 2, 3 and 4 of Landsat 5 Thematic Mapper (TM) images, dated 13 April 1994 in the Abdal Aziz Mountain study area in northeastern Syria. The rangeland was categorized into six classes according to the plant contacts of dominant shrubs (Artemisia herba-alba and Noaea mucronata) and herbaceous plants. In addition, cultivated fields were categorized into two classes. An average classification accuracy of 85% in the supervised processing and an average ground verification accuracy of 81% on the Landsat-estimated vegetation classes were achieved for the rangeland. These show that a 30m X 30m resolution of the Landsat TM image had the ability to recognize vegetation at six sub-divided community levels, and the successful classification was conducted on the whole rangeland of the study area. The distinctive feature of this work is that this vegetation classification using Landsat TM images was accomplished at the level of classifying a A. herba-alba and N. mucronata dominant community into six sub-community classes. This detailed vegetation classification was conducted with the final aim of forage resource estimation and human impact assessment in mind.     

Mineral imaging with Landsat Thematic Mapper data for hydrothermal alteration mapping in heavily vegetated terrane

A mineral imaging methodology, which involves processing of Landsat Thematic Mapper (TM) images and integration of ground data, is tested in the Baguio district of the Philippines to map hydrothermally altered zones in heavily vegetated terranes. Based on published reflectance spectra, two band ratio images are created and input into principal components analysis to map each predominant hydrothermal alteration mineral into separate mineral images. Digitized map data of known hydrothermal alteration zones are used for identifying training pixels for the known alteration zones. The mineral images and the training pixels are used in a supervised classification to map hydrothermally altered zones; classification accuracy reaches 69%. Inclusion of an image of a digital elevation model improves the classification accuracy to 82%. The mineral imaging methodology proved more successful in remote mapping of known hydrothermally altered zones in the Baguio district than remote mapping of limonitic and clay alteration using previously developed techniques.     

Using remote sensing to estimate the change of carbon storage: A case study in the estuary of Yellow River delta

Land cover change can exert a crucial effect on the terrestrial carbon cycle. To estimate changes in the carbon pools and carbon fluxes to the atmosphere, Landsat Thematic Mapper (TM) data of 1992 and 1996 were used to calculate the extent of different land cover types and their changes in the estuary area of the Yellow River delta. Image processing and the unsupervised classification allowed accurate land cover maps for 1992 and 1996 to be generated, by which the changes in the carbon pools were detected. Estimation of the carbon pools and the carbon fluxes to the atmosphere was carried out employing the results of Landsat image analysis and the published data on carbon stocks in vegetation and soil. By calculating the area changed between different types of vegetation and their different carbon stocks, the quantity of the terrestrial carbon cycle in the estuary area of the Yellow River delta was acquired. The results shows that the vegetation carbon storage was 11.43 2 10 11 g and soil carbon pool 7.24 2 10 12 g in 1992, and the vegetation carbon pool increased by 3.77 2 10 11 g during the 4 years from 1992 to 1996.     

The use of SPOT VEGETATION data in a classification tree approach for burnt area mapping in Australian savanna

An algorithm to map burnt areas has been developed for SPOT VEGETATION (VGT) data in Australian woodland savannas. A time series of daily VGT images (15 May to 15 July 1999) was composited into 10-day periods by applying a minimum value criterion to the near-infrared band (0.78-0.89 @m). The algorithm was developed using a classification tree methodology that was confirmed as a powerful means of image classification. This methodology allowed the identification of three classes of burnt surfaces that appear to be differentiated by the proportion of the pixel that is burnt, the intensity of the fire and the density of the tree layer. The performance of the algorithm was assessed by classification of one VGT composite image (31 May-9 June) using, as representative of the ground truth, burnt areas extracted from two Landsat TM scenes (9 June). We randomly extracted 30 windows (each of ～14 km by 14 km) for which we compared the percentage of area burnt as derived from TM and VGT. The estimated mean absolute deviation in the percentage of the area burnt in each window is - 6.3%. In the area common to the two datasets a total amount of 6473 km 2 was estimated to be burnt in the VGT classification against 7536 km 2 that was burnt according to TM images. The accuracy of the classification was found to vary with the vegetation type being the most accurate estimate in low woodland with an underestimation error of 8.6%. These results show that VGT could be a very useful sensor for burnt area mapping over large woodland areas, although the low spatial resolution and the lack of a thermal band can be a limitation in certain conditions (e.g. understorey burns). The same methodology will be applied to map burnt areas for the entire Australian continent.     

Mapping of boreal vegetation of a temperate mountain in China by multitemporal Landsat TM imagery

The Changbai Mountain Natural Reserve (2000 km 2 ), north-east China, is a very important ecosystem representing the temperate biosphere. The cover types were derived by using multitemporal Landsat TM imagery, which was modified with DEM data on the relationship between vegetation distribution and elevation. It was classified into 20 groups by supervised classification. By comparing the results of the classification of different band combinations, bands 4 and 5 of an image from 18 July 1997 and band 3 of an image from 22 October 1997 were used to make a false colour image for the final output, a vegetation map, which showed the best in terms of classification accuracy. The overall accuracy by individual images was less than 70%, while that of the multitemporal classification was higher than 80%. Further, on the basis of the relationship of vegetation distribution and elevation, the accuracy of multitemporal classification was raised from 85.8 to 89.5% by using DEM. Bands 4 and 5 showed a high ability for discriminating cover types. Images acquired in late spring and mid-summer were recognized better than other seasons for cover type identification. NDVI and band ratio of B4/B3 proved useful for cover type discrimination, but were not superior to the original spectral bands. Other band ratios like B5/B4 and B7/B5 were less important for improving classification accuracy. The changes of spectral reflectance and NDVI with season were also analysed with 10 images ranging from 1984 to 1997. Seperability of images in terms of classification accuracy was high in late spring and summer, and decreased towards winter. There were five vegetation zones on the mountain, from the base to the peak: deciduous forest zone, mixed forest zone, conifer forest zone, birch forest zone and tundra zone. Spruce-fir conifer dominated forest was the most dominant vegetation (33%), followed by mixed forest (26%), Korean pine forest (8%) and mountain birch forest (5%).     

Land cover mapping in support of LAI and FPAR retrievals from EOS-MODIS and MISR: Classification methods and sensitivities to errors

Land cover maps are used widely to parameterize the biophysical properties of plant canopies in models that describe terrestrial biogeochemical processes. In this paper, we describe the use of supervised classification algorithms to generate land cover maps that characterize the vegetation types required for Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation (FPAR) retrievals from MODIS and MISR. As part of this analysis, we examine the sensitivity of remote sensing-based retrievals of LAI and FPAR to land cover information used to parameterize vegetation canopy radiative transfer models. Specifically, a decision tree classification algorithm is used to generate a land cover map of North America from Advanced Very High Resolution Radiometer (AVHRR) data with 1 km spatial resolution using a six-biome classification scheme. To do this, a time series of normalized difference vegetation index data from the AVHRR is used in association with extensive site-based training data compiled using Landsat Thematic Mapper (TM) and ancillary map sources. Accuracy assessment of the map produced via decision tree classification yields a cross-validated map accuracy of 73%. Results comparing LAI and FPAR retrievals using maps from different sources show that disagreement in land cover labels generally do not translate into strong disagreement in LAI and FPAR maps. Further, the main source of disagreement in LAI and FPAR maps can be attributed to specific biome classes that are characterized by a continuum of fractional cover and canopy structure.     

Using MERIS fused images for land-cover mapping and vegetation status assessment in heterogeneous landscapes

In this paper we evaluate the potential of ENVISAT–Medium Resolution Imaging Spectrometer (MERIS) fused images for land-cover mapping and vegetation status assessment in heterogeneous landscapes. A series of MERIS fused images (15 spectral bands; 25 m pixel size) is created using the linear mixing model and a Landsat Thematic Mapper (TM) image acquired over the Netherlands. First, the fused images are classified to produce a map of the eight main land-cover types of the Netherlands. Subsequently, the maps are validated using the Dutch land-cover/land-use database as a reference. Then, the fused image with the highest overall classification accuracy is selected as the best fused image. Finally, the best fused image is used to compute three vegetation indices: the normalized difference vegetation index (NDVI) and two indices specifically designed to monitor vegetation status using MERIS data: the MERIS terrestrial chlorophyll index (MTCI) and the MERIS global vegetation index (MGVI).

Results indicate that the selected data fusion approach is able to downscale MERIS data to a Landsat-like spatial resolution. The spectral information in the fused images originates fully from MERIS and is not influenced by the TM data. Classification results for the TM and for the best fused image are similar and, when comparing spectrally similar images (i.e. TM with no short-wave infrared bands), the results of the fused image outperform those of TM. With respect to the vegetation indices, a good correlation was found between the NDVI computed from TM and from the best fused image (in spite of the spectral differences between these two sensors). In addition, results show the potential of using MERIS vegetation indices computed from fused images to monitor individual fields. This is not possible using the original MERIS full resolution image. Therefore, we conclude that MERIS–TM fused images are very useful to map heterogeneous landscapes.     

加权KNN的图文数据融合分类

目的 图文数据在不同应用场景下的最佳分类方法各不相同,而现有语义级融合算法大多适用于图文数据分类方法相同的情况,若将其应用于不同分类方法时由于分类决策基准不统一导致分类结果不理想,大幅降低了融合分类性能。针对这一问题,提出基于加权KNN的融合分类方法。方法 首先,分别利用softmax多分类器和多分类支持向量机(SVM)实现图像和文本分类,同时利用训练数据集各类别分类精确度加权后的图像和文本正确判别实例的分类决策值分别构建图像和文本KNN模型;再分别利用其对测试实例的图像和文本分类决策值进行预测,通过最邻近k个实例属于各类别的数目确定测试实例的分类概率,统一图像和文本的分类决策基准;最后利用训练数据集中图像和文本分类正确的数目确定测试实例中图像和文本分类概率的融合系数,实现统一分类决策基准下的图文数据融合。结果 在Attribute Discovery数据集的图像文本对上进行实验,并与基准方法进行比较,实验结果表明,本文融合算法的分类精确度高于图像和文本各自的分类精确度,且平均分类精确度相比基准方法提高了4.45%;此外,本文算法对图文信息的平均整合能力相比基准方法提高了4.19%。结论 本文算法将图像和文本不同分类方法的分类决策基准统一化,实现了图文数据的有效融合,具有较强的信息整合能力和较好的融合分类性能。     

Coastal wetland vegetation classification with a Landsat Thematic Mapper image

Coastal wetland vegetation classification with remotely sensed data has attracted increased attention but remains a challenge. This paper explored a hybrid approach on a Landsat Thematic Mapper (TM) image for classifying coastal wetland vegetation classes. Linear spectral mixture analysis was used to unmix the TM image into four fraction images, which were used for classifying major land covers with a thresholding technique. The spectral signatures of each land cover were extracted separately and then classified into clusters with the unsupervised classification method. Expert rules were finally used to modify the classified image. This research indicates that the hybrid approach employing sub-pixel information, an analyst's knowledge and characteristics of coastal wetland vegetation distribution shows promise in successfully distinguishing coastal vegetation classes, which are difficult to separate with a maximum likelihood classifier (MLC). The hybrid method provides significantly better classification results than MLC.     

TM digital processing of a tropical forest region in southeastern Mexico

A humid forest in the neotropical area of Los Tuxtlas, in southeastern Mexico has been used as a test area (900km²) for classification of landscape and vegetation by means of Landsat Thematic Mapper (TM) data, aerial photography and 103 ground samples. The area presents altitudinal variations from sea level to 1640m, providing a wide variety of vegetation types. A hybrid (supervised/unsupervised) classification approach was used, defining spectral signatures for 14 clustering areas with data from the reflective bands of the TM. The selected clustering areas ranged from vegetation of the highlands and the rain forest to grassland, barren soil, crops and secondary vegetation. The digital classification compared favourably with results from aerial photography and with those from a multivariate analysis of the 103 ground data. The statistical evaluation (error matrix) of the classified image indicated an overall 84·4 per cent accuracy with a kappa coefficient of agreement of 0·83. A geographical information system (GIS) was used to compile a land unit and a vegetation map. The TM data allowed for delineation of boundaries in the land unit map, and for a finer differentiation of vegetation types than those identified during field work. Digital value patterns of several information classes are shown and discussed as an indirect guide of the spectral behaviour of vegetation of highlands, rain forest, secondary vegetation and crops. The method is considered applicable to the inventory of other forested areas, especially those with significant variations in vegetation.     

Quantitative accuracy assessment of contextually filtered classified images

Relative radiance recorded by the Landsat Thematic Mapper (TM) for l–1O-year-old plantations of oil palm (Elaeis guineensis Jacq.) in Sabah, Malaysia, was negatively correlated with stand age. Remotely sensed response is determined by biophysical variables related to age since field planting, notably leaf area, canopy architecture and progressive masking of ground cover vegetation. The relation was asymptotic. Age and age-related variables will be most accurately inferred from TM data for stands under 5-years old, especially using short wave infrared and thermal bands. Narrower age classes will be required to represent younger stands in image classification.     

Mapping Arctic tundra vegetation types using digital SPOT/HRV-XS data A preliminary assessment

Abstract

Multispectral (XS) image data recorded by the High Resolution Visible (HRV) sensor aboard the SPOT-1 satellite are being evaluated for the mapping of Arctic tundra vegetation in the Arctic Foothill Province of Alaska. This research is part of a current ecosystems study that requires an efficient means for mapping vegetation types over large areas. Conventional spectral-based image classification techniques were applied to SPOT/HRV-XS data from a single date. The unique characteristics of the vegetation cover (mainly tussock tundra) and illumination conditions of the location necessitated a detailed examination of classification approaches that have generally been applied in mid-latitude studies. Preliminary results suggest that areal estimates of Arctic tundra vegetation types can be made accurately (±2·5 per cent per category), but maps generated by classifying spectral features of SPOT/HRV-XS data alone arc unsuitably accurate (56 per cent). This is partly due to the high occurrence of relatively small vegetation parcels, determined by measuring the characteristic lengths of vegetation parcels from a ‘ground reference’ map covering the same area as the SPOT/HRV-XS subscene.