Land cover classification

Land cover estimation with ALOS satellite image using a neural-network AVNIR-2 were adopted as a supervisor signal. There are two reasons for using the classifi cation from AVNIR-2. First, the raw AVNIR-2 data contain much information. Therefore, the AVNIR-2 data are divided into eight classes in order to reduce it. Second, in the wide-area PALSAR data, more sample data are included in order to enhance the classifi cation accuracy. In the classifi cation with PALSAR, the number of input layers poses a problem because the PALSAR has only one band sensor. We propose that the input signal employs not only single pixels, but also near pixel values.

Very High Resolution (VHR) satellite images offer a great potential for the extraction of landuse and land-cover related information for urban areas. The available techniques are diverse and need to be further examined before operational use is possible. In this paper we applied two pixel-by-pixel classification techniques and the object-oriented image analysis approach (eCognition) for a land-cover classification of a Quickbird image of a study area in the northern part of the city of Ghent (Belgium). Only small differences in overall Kappa were noted between the best results of the pixel-based approach (neural network classification with Haralick texture measures) and the object-oriented classification (eCognition). A rule-based procedure using ancillary information on elevation derived from a digital surface model was applied on the pixel-based land-cover classification in order to obtain information on the spatial distribution of buildings and artificial surfaces.

Maps of current and potential vegetation spatial patterns can be used to assess land cover changes, and aid in ecosystem management and restoration. The vegetation spatial patterns of subalpine forest species are largely controlled by variation in temperature, water and solar radiation resources. These fundamental resources were quantified across a 1100-km 2 landscape using biophysical models, a digital elevation model (DEM), and weather station data. Field data of species abundances were used to define species-habitat relationships and calibrate maximum likelihood classifications of the biophysical gradients. For comparison, a standard land cover classification of Landsat Thematic Mapper satellite imagery had an overall accuracy 68.3%. Using the biophysical gradients alone gave a similar 67.4% accuracy. The highest accuracy classification (83.2%) used both biophysical and spectral data. The biophysical resources were also used to map the presence or absence of four herb and shrub species that cannot be sensed remotely. These predictions ranged from 60% to 79% accurate. Maps of relative abundance were less accurate, from 61% to 63.2%. This low result may be due to historical and stochastic events, or simply a small data set. The spatial pattern of species and communities that are controlled by resources can be predicted using general biophysical models. The species-habitat relationships can also be used to improve remote sensing products.

In recent years, decision tree classifiers have been successfully used for land cover classification from remote sensing data. Their implementation as a per-pixel based classifier to produce hard or crisp classification has been reported in the literature. Remote sensing images, particularly at coarse spatial resolutions, are contaminated with mixed pixels that contain more than one class on the ground. The per-pixel approach may result in erroneous classification of images dominated by mixed pixels. Therefore, soft classification approaches that decompose the pixel into its class constituents in the form of class proportions have been advocated. In this paper, we employ a decision tree regression approach to determine class proportions within a pixel so as to produce soft classification from remote sensing data. Classification accuracy achieved by decision tree regression is compared with those achieved by the most widely used maximum likelihood classifier, implemented in the soft mode, and a supervised version of the fuzzy cmeans classifier. Root Mean Square Error (RMSE) and fuzzy error matrix based measures have been used for accuracy assessment of soft classification. D

In this paper, the potential use tf spaceborne polarimettic synthetic aperture radar (SAB) data in mapping landcover tripes and monitoring d@restation in tropics is studied, ttere, the emphasis is placed on several clearing practices a~d forest regeneration that can be characteri:.ed by usi~,g the sensitivity of SAB channels to vegetation biomass and canopy str~lcture. A supervised Bayes-Jan classifier designed fi)r SAIl signal statistics is employed to separate five classes: primar*j fi)rest, second-ar~j fi)rest, pasture-crops, quebradao, and disturbed forest. The Land C-band polarimetric SAIl data acquired during the shuttle imaging radar-C (SIII-C)/X-SAB space-shuttle ntission in 1994 are used as input data to the class{tier, The results are verified by field observation and comparison with the LancLsat data acquired in August (f 1994. The SAB data can delineate these five classes with approxinmtely 72% accuracy. The confusion arises wher~ separating old secondary fi)rests from primary .tim'st aad the young ones from pasture-crops. It is shown that Landsat and SAg data camj contplenwntary information about the vegetation structure that, when used i1~ sy~u¢r~,ism, may increase the classification accuracq over secondar~j fi~rest regrowth. When the number (f land-co~)er types was reduced to three classes including primanj fi~rest, pasture-crops, and regrowth-disturbed fi~rest, the acc~tracy of classification increased to 87%. A dinwnsionality anahjsis of the classifier showed that the accurac~j can be fi~rther improved to 92% by reducing the feature space to L-band HH and HV chantwls. Comparison of SIIl-C data acquired in April (wet period) a~ld October (dr~j period) indicates that multi

Random Forests are considered for classification of multisource remote sensing and geographic data. Various ensemble classification methods have been proposed in recent years. These methods have been proven to improve classification accuracy considerably. The most widely used ensemble methods are boosting and bagging. Boosting is based on sample re-weighting but bagging uses bootstrapping. The Random Forest classifier uses bagging, or bootstrap aggregating, to form an ensemble of classification and regression tree (CART)-like classifiers. In addition, it searches only a random subset of the variables for a split at each CART node, in order to minimize the correlation between the classifiers in the ensemble. This method is not sensitive to noise or overtraining, as the resampling is not based on weighting. Furthermore, it is computationally much lighter than methods based on boosting and somewhat lighter than simple bagging. In the paper, the use of the Random Forest classifier for land cover classification is explored. We compare the accuracy of the Random Forest classifier to other better-known ensemble methods on multisource remote sensing and geographic data.

Reliable and current availability of land cover knowledge are essential for many studies regarding planning, management, monitoring and updating activities. The optical satellite sensor data has been utilized for the classification of land use/land cover. In this study, the capability of synthetic aperture radar (SAR) interferometric coherence is practiced for land cover classification in Okara, Pakistan using Sentinel-1A imagery. Two Single Look Complex (SLC) product of months April and May 2016 were used and processed to create backscatter and interferometric coherence layers. From backscatter layers of each month, the mean backscatter and backscatter difference layer were obtained. False color composite (FCC) were developed comprising mean backscatter, backscatter difference and coherence, and performed supervised classification using maximum likelihood method to generate land cover classes i.e. water, barren, vegetation and built-up. Kappa statistics were employed for accuracy assessment of the output map. Results showed the good potential of Sentinel-1C-band for land cover classification having 0.69 Kappa coefficient and 80% overall accuracy. This study investigated the potential of C-band backscatter coefficients and coherence map for land cover discrimination. Coherence proved to be efficient in the examination of vegetative and non-vegetative areas.

Enhanced Vegetation Index (EVI) and Leaf Area Index (LAI) data. Results of the simulations reveal four main groups of time series similarity measures with different sensitivities: (i) D M an , D E , D P CA , D F k quantify the difference in time series values, (ii) D M ah accounts for temporal correlation and non-stationarity of variance, (iii) D CC measures the temporal correlation, and (iv) the Fourier based D F F T and D ξ show their specific sensitivity based on the selected Fourier components. The difference measures show relatively the highest sensitivity to amplitude effects, whereas the correlation based measures are highly sensitive to variations in timing and noise. The Fourier based measures, finally, depend highly on the signal to noise ratio and the balance between amplitude and phase dominance. The heterogeneity in sensitivity of each D stresses the importance of (i) understanding the time series characteristics before applying any classification of change detection approach and (ii) defining the variability one wants to identify/account for. This requires an understanding of the ecosystem dynamics and time series characteristics related to the baseline, amplitude, timing, noise and variability of the ecosystem time series. This is also illustrated in the quantitative comparison, where the different sensitivities of D for the NDVI, EVI, and LAI data relate specifically to the temporal characteristics of each data set. Additionally, the effect of noise and intra-and interclass variability is demonstrated in a case study based on land cover classification.

Based on purely spectral-domain prior knowledge taken from the remote sensing (RS) literature, an original spectral (fuzzy) rule-based per-pixel classifier is proposed. Requiring no training and supervision to run, the proposed spectral rule-based system is suitable for the preliminary classification (primal sketch, in the Marr sense) of Landsat-5 Thematic Mapper and Landsat-7 Enhanced Thematic Mapper Plus images calibrated into planetary reflectance (albedo) and at-satellite temperature. The classification system consists of a modular hierarchical top-down processing structure, which is adaptive to image statistics, computationally efficient, and easy to modify, augment, or scale to other sensors' spectral properties, like those of the Advanced Spaceborne Thermal Emission and Reflection Radiometer and of the Satellite Pour l'Observation de la Terre (SPOT-4 and -5). As output, the proposed system detects a set of meaningful and reliable fuzzy spectral layers (strata) consistent (in terms of one-to-one or manyto-one relationships) with land cover classes found in levels I and II of the U.S. Geological Survey classification scheme. Although kernel spectral categories (e.g., strong vegetation) are detected without requiring any reference sample, their symbolic meaning is intermediate between those (low) of clusters and segments and those (high) of land cover classes (e.g., forest). This means that the application domain of the kernel spectral strata is by no means alternative to RS data clustering, image segmentation, and land cover classification. Rather, prior knowledge-based kernel spectral categories are naturally suitable for driving stratified application-specific classification, clustering, or segmentation of RS imagery that could involve training and supervision. The efficacy and robustness of the proposed rule-based system are tested in two operational RS image classification problems.

We define "operational" here as producing information on a regular basis, and "spatial water resources monitoring systems" (SWRMS) as software that integrates observations into models to produce spatial estimates of current (and past) water resources distribution.

Random Forests are considered for classification of multisource remote sensing and geographic data. Various ensemble classification methods have been proposed in recent years. These methods have been proven to improve classification accuracy considerably. The most widely used ensemble methods are boosting and bagging. Boosting is based on sample re-weighting but bagging uses bootstrapping. The Random Forest classifier uses bagging, or bootstrap aggregating, to form an ensemble of classification and regression tree (CART)-like classifiers. In addition, it searches only a random subset of the variables for a split at each CART node, in order to minimize the correlation between the classifiers in the ensemble. This method is not sensitive to noise or overtraining, as the resampling is not based on weighting. Furthermore, it is computationally much lighter than methods based on boosting and somewhat lighter than simple bagging. In the paper, the use of the Random Forest classifier for land cover classification is explored. We compare the accuracy of the Random Forest classifier to other better-known ensemble methods on multisource remote sensing and geographic data.

This work is devoted to a presentation of the ECOCLIMAP-II database for Western Africa, which is an upgrade for this region of the former initiative, ECOCLIMAP-I, implemented at global scale. ECOCLIMAP-II is a dual database at 1-km resolution that comprises an ecosystem classification and a coherent set of land surface parameters. This new physiographic information (e.g. leaf area index, fractional vegetation cover, albedo and land cover classification), was especially developed in the framework of the African Monsoon Multidisciplinary Analysis (AMMA) programme in order to support the modelling of land-atmosphere interactions, which stresses the importance of the present study. Criteria for coherence between prevalent land cover classifications and the analysis of time series of the satellite leaf area index (LAI) between 2000 and 2007 constitute the analysis tools for setting up ECOCLIMAP-II. The LAI and inferred fraction of vegetation cover are spatially distributed per land cover unit. The fraction of vegetation cover is handled to split the land surface albedo into vegetation and bare soil albedo components, as is required for a large number of applications. The new ECOCLIMAP-II land cover product is improved with regard to the spatial coherence compared to former version. The reliability of the physiographic details is also confirmed through verification with land cover products at higher resolution.

Spatial water resource monitoring systems (SWRMS) can provide valuable information in support of water management, but current operational systems are few and provide only a subset of the information required. Necessary innovations include the explicit description of water redistribution and water use from river and groundwater systems, achieving greater spatial detail (particularly in key features such as irrigated areas and wetlands), and improving accuracy as assessed against hydrometric observations, as well as assimilating those observations. The Australian water resources assessment (AWRA) system aims to achieve this by coupling landscape models with models describing surface water and groundwater dynamics and water use. A review of operational and research applications demonstrates that satellite observations can improve accuracy and spatial detail in hydrological model estimation. All operational systems use dynamic forcing, land cover classifications and a priori parameterisation of vegetation dynamics that are partially or wholly derived from remote sensing. Satellite observations are used to varying degrees in model evaluation and data assimilation. The utility of satellite observations through data assimilation can vary as a function of dominant hydrological processes. Opportunities for improvement are identified, including the development of more accurate and higher spatial and temporal resolution precipitation products, and the use of a greater range of remote sensing products in a priori model parameter estimation, model evaluation and data assimilation. Operational challenges include the continuity of research satellite missions and data services, and the need to find computationally-efficient data assimilation techniques. The successful use of observations critically depends on the availability of detailed information on observational error and understanding of the relationship between remotely-sensed and model variables, as affected by conceptual discrepancies and spatial and temporal scaling.

The Boreal Ecosystem Atmosphere Study (BOREAS) was a large, multiyear internationally supported study designed to improve our understanding of the boreal forest biome and its interactions with the atmosphere, biosphere, and the carbon cycle in the face of global climate change. In the initial phase of this study (early 1990s), remote sensing played a key role by providing products needed for planning and modeling. During and after the main BOREAS field campaigns (1994 and 1996), innovative remote sensing approaches and analyses expanded our understanding of the boreal forest in four key areas: (1) definition of vegetation structure, (2) land-cover classification, (3) assessment of the carbon balance, and (4) links between surface properties, weather, and climate. In addition to six BOREAS special issues and over 500 journal papers, a principal legacy of BOREAS is its well-documented and publicly available database, which provides a lasting scientific resource and opportunity to further advance our understanding of this critical northern biome. D

One of key inputs to hydrological modeling is the potential evapotranspiration, either from the interception (PET 0 ) or from the soil water of root zone (PET). The Shuttleworth-Wallace (S-W) model is developed for their estimation. In this parameterization, neither experimental measurement nor calibration is introduced. Based on IGBP land cover classification, the typical thresholds of vegetation parameters are drawn from the literature. The spatial and temporal variation of vegetation LAI is derived from the composite NOAA-AVHRR NDVI using the SiB2 method. The CRU database supplies with the required meteorological data. They are all publicly available. The developed S-W model is applicable at the global scale, particularly to the data-poor or ungauged large basins.

One of the significant environmental consequences of urbanization is the urban heat island (UHI). In this paper, Landsat TM images of 1986 and 2004 were utilized to study the spatial and temporal variations of heat island and their relationships with land cover changes in Suzhou, a Chinese city which experienced rapid urbanization in past decades. Land cover classifications were derived to quantify urban expansions and brightness temperatures were computed from the TM thermal data to express the urban thermal environment. The spatial distributions of surface temperature indicated that heat islands had been largely broadened and showed good agreements with urban expansion. Temperature statistics of main land cover types showed that built-up and bare land had higher surface temperatures than natural land covers, implying the warming effect caused by the urbanization with natural landscape being replaced by urban areas. In addition, the spatial detail distributions of surface temperature were compared with the distribution of land cover by means of GIS buffer analysis. Results show remarkable show good correspondence between heat island variations with urban area expansions.

This study focuses on the comparison between the classical and object-oriented image classifications of remote sensing imagery in the arid area. Due to its special geographic environment and socio-economic contexts, the land cover and its spatio-temporal pattern in aridzone is very different from those in coastal area, thus some conventional methods of remote sensing image classification may not be suitable. In order to investigate an appropriate method for aridzone image classification, pixel-based image classifiers such as the Maximum Likelihood Classifier and an object-oriented image classifier were tested and compared using an Landsat ETM+ image. The accuracy of each method was assessed using reference data sets derived from high-resolution satellite images, aerial photograph and field investigation. The result shows that the object-oriented method has achieved an overall accuracy of 89% with a kappa coefficient of 0.87, compared with 71% (0.66) that was derived from the conventional pixel-based method.

The brightness of a SAR image is affected by topography due to varying projection between ground and image coordinates. For polarimetric SAR (PolSAR) imagery being used for purposes of land cover classification, this radiometric variability is shown to affect the outcome of a Wishart unsupervised classification in areas of moderate topography. The intent of this paper is to investigate the impact of applying a radiometric correction to the PolSAR coherency matrix for a region of boreal forest in interior Alaska. The gamma naught radiometric correction estimates the local illuminated area at each grid point in the radar geometry. Then, each element of the coherency matrix is divided by the local area to produce a polarimetric product that is radiometrically "flat." This paper follows two paths, one with and one without radiometric correction, to investigate the impact upon classification accuracy. Using a Landsat-derived land cover reference, the radiometric correction is shown to bring about significant qualitative and quantitative improvements in the land cover map. Confusion matrix analysis confirms the accuracy for most classes and shows a 15% improvement in the classification of the deciduous forest class.

The Land Cover Map of North and Central America for the year 2000 (GLC 2000-NCA), prepared by NRCan/CCRS and USGS/EROS Data Centre (EDC) as a regional component of the Global Land Cover 2000 project, is the subject of this paper. A new mapping approach for transforming satellite observations acquired by the SPOT4/VGTETATION (VGT) sensor into land cover information is outlined. The procedure includes: (1) conversion of daily data into 10-day composite; (2) post-seasonal correction and refinement of apparent surface reflectance in 10-day composite images; and (3) extraction of land cover information from the composite images. The pre-processing and mosaicking techniques developed and used in this study proved to be very effective in removing cloud contamination, BRDF effects, and noise in Short Wave Infra-Red (SWIR). The GLC 2000-NCA land cover map is provided as a regional product with 28 land cover classes based on modified Federal Geographic Data Committee/Vegetation Classification Standard (FGDC NVCS) classification system, and as part of a global product with 22 land cover classes based on Land Cover Classification System (LCCS) of the Food and Agriculture Organisation. The map was compared on both areal and per-pixel bases over North and Central America to the International Geosphere–Biosphere Programme (IGBP) global land cover classification, the University of Maryland global land cover classification (UMd) and the Moderate Resolution Imaging Spectroradiometer (MODIS) Global land cover classification produced by Boston University (BU). There was good agreement (79%) on the spatial distribution and areal extent of forest between GLC 2000-NCA and the other maps, however, GLC 2000-NCA provides additional information on the spatial distribution of forest types. The GLC 2000-NCA map was produced at the continental level incorporating specific needs of the region.

Land cover data represent environmental information for a variety of scientific and policy applications, so its classification from satellite images is important. Since neural networks (NN) do not require a hypothesis about data distribution, they are valuable tools to classify satellite images. The objectives of this work were to develop NN models to classify land cover data from information from satellite images and to evaluate them when different input variables are used. MODIS-MYD13Q1 satellite images and data of 85 plots in Córdoba, Argentina, were used. Five NN models of multi-layer feed-forward perceptron were designed. Four of these received NDVI (Normalized Difference Vegetation Index), EVI (Enhanced Vegetation Index), red (RED) and near infrared (NIR) reflectance values as input patterns, respectively. The fifth NN had RED and NIR reflectances as input values. By comparing the information taken in the field and the classification made during the validation phase, it can be concluded that all models presented good performance in the classification. The model that shows better performance is the one that jointly considers RED and NIR reflectance as input; this model shows an overall classification accuracy of 93% and an excellent Kappa statistic. The networks constructed with NDVI and EVI values have a similar perfomance (86 and 83% accuracy, respectively). The Kappa statistics correspond to the categories of very good and good, respectively. The networks including only RED or NIR reflectance values get the lowest accuracy results (76 and 81%, respectively) and Kappa values within fair and good ranks, respectively.

Floods are a common feature in rapidly urbanizing Dhaka and its adjoining areas. Though Greater Dhaka experiences flood almost in every year, flood management policies are mostly based on structural options including flood walls, dykes, embankments etc. Many shortcomings of the existing flood management systems are reported in numerous literatures. The objective of this paper is to assess flood hazard in Greater Dhaka for the historical flood event of 1998 using Synthetic Aperture Radar (SAR) data with GIS data. Flood-affected frequency and flood depth calculated from the multi-date SAR imageries were used as hydrologic parameters. Elevation heights, land cover classification, geomorphic division and drainage network data generated from optical remote sensing and analogue maps were used through GIS approach. Using a ranking matrix in three dimensional multiplication mode, flood hazard was assessed. All possible combination of flood hazard maps was prepared using land-cover, geomorphology and elevation heights for flood-affected frequency and floodwater depth. Using two hazard maps which produced the highest congruence for flood frequency and flood depth, a new flood hazard map was developed by considering the interactive effect of flood-affected frequency and floodwater depth, simultaneously. This new hazard map can provide more safety for flood countermeasures because pixels belonging to higher hazard degrees were increased due to the

In this study we evaluated changes in land cover and rainfall in the Upper Gilgel Abbay catchment in the Upper Blue Nile basin and how changes affected stream flow in terms of annual flow, high flows and low flows. Land cover change assessment was through classification analysis of remote sensing based land cover data while assessments on rainfall and stream flow data are by statistical analysis. Results of the supervised land cover classification analysis indicated that 50.9 % and 16.7 % of the catchment area was covered by forest in 1973 and 2001, respectively. This significant decrease in forest cover is mainly due to expansion of agricultural land.

The leaf area index (LAI) product from the Moderate Resolution Imaging Spectroradiometer (MODIS) is important for monitoring and modelling global change and terrestrial dynamics at many scales. The algorithm relies on spectral reflectances and a six biome land cover classification. Evaluation of the specific behaviour and performance of the product for regions of the globe such as Australia are needed to assist with product refinement and validation. We made an assessment of Collection 4 of the MODIS LAI product using four approaches: (a) assessment against a continental scale Structural Classification of Australian Vegetation (SCAV); (b) assessment against a continental scale land use classification (LUC); (c) assessment against historical field-based measurement of LAI collected prior to the Terra Mission; and (d) direct comparison of MODIS LAI with coincident field measurements of LAI, mostly from hemispherical photography. The MODIS LAI product produced a wide variety of geographically and structurally specific temporal response profiles between different classes and even for sub-groups within classes of the SCAV. Historical and concurrent field measurements indicated that MODIS LAI was giving reasonable estimates for LAI for most cover types and land use types, but that major overestimation of LAI occurs in some eastern Australian open forests and woodlands. The six biome structural land cover classification showed some significant deviations in class allocation compared to the SCAV particularly where grasslands are allocated to shrubland, savanna woodlands are allocated to shrubland, savanna and broadleaf forest, and open forests are allocated to savanna and broadleaf forest. The land cover and LAI products could benefit from some additional examination of Australian data addressing the structural representation of Eucalypt canopies in the "space of canopy realisation" for savanna and broadleaf forest classes.

Mapping coffee lands by using intermediate spatial resolution imagery has been a challenge for remote sensing researchers. Several environmental, physiological and crop management factors, increase confusion in both visual interpretation and automatic classification of coffee fields related to other landcover units. It is essential to combine spectral information from satellite imagery with ancillary data, such as environmental factors and contextual interpretation elements (e.g. the degree and kind of texture and shape conditions) transposed to digital analysis. This study shows the outcomes of Object Based Image Analysis (OBIA) of coffee lands supported by ancillary data. Non supervised classification (the ISOSEG algorithm available in the SPRING software) of the same area using Landsat Thematic Mapper (TM) images was carried out in order to compare the results. A reference map was generated by on-screen digitalization of Landsat imagery. Several features were selected on-screen to improve the land cover classification through a semantic net. The object-oriented classification performed better, in terms of accuracy, than the non supervised classification. The highest overall accuracy was 83%. However, some types of coffee crops could not be identified only by expert knowledge based feature selection, so it is advisable to use a formal method, such as data mining, to better select the best thresholds and the suitable features to identify those types of coffee crops.

Global change issues are high on the current international political agenda. A variety of global protocols and conventions have been established aimed at mitigating global environmental risks. A system for monitoring, evaluation and compliance of these international agreements is needed, with each component requiring comprehensive analytical work based on consistent datasets. Consequently, scientists and policymakers have put faith in earth observation data for improved global analysis. Land cover provides in many aspects the foundation for environmental monitoring . Proceedings of the FAO/UNEP Expert Consultation on Strategies for Global Land Cover Mapping and Monitoring. FAO, Rome, Italy, 38 pp.]. Despite the significance of land cover as an environmental variable, our knowledge of land cover and its dynamics is poor [Foody, G.M., 2002. Status of land cover classification accuracy assessment. Rem. Sens. Environ. 80,[185][186][187][188][189][190][191][192][193][194][195][196][197][198][199][200][201]. This study compares four satellite derived 1 km land cover datasets freely available from the internet and in wide use among the scientific community. Our analysis shows that while these datasets have in many cases reasonable agreement at a global level in terms of total area and general spatial pattern, there is limited agreement on the spatial distribution of the individual land classes. If global datasets are used at a continental or regional level, agreement in many cases decreases significantly. Reasons for these differences are many-ranging from the classes and thresholds applied, time of data collection, sensor type, classification techniques, use of in situ data, etc., and make comparison difficult. Results of studies based on global land cover datasets are likely influenced by the dataset chosen. Scientists and policymakers should be made aware of the inherent limitations in using current global land cover datasets, and would be wise to utilise multiple datasets for comparison. #

Marginalisation has been integrated in the debate about rural landscapes already in the eighties, when the first large reform of the Common Agricultural Policy started being discussed and scenarios of large areas of farmland getting out of production became a threat. Since then, the concept of rural marginal areas is often connected both with trends in the sense of socio-economic decay and abandonment of agricultural land, especially in the most peripheral parts of Europe. In South-eastern Portugal these concepts are also connected, both by experts and the public, with the problem of desertification, affecting the areas with most extreme environmental constrains. Since the beginning of the nineties, different type of policy strategies and specific instruments have been created to avoid the effects of marginalisation and to preserve the diversity of European rural landscapes and communities. But these policies do not consider the specific problems of each area affected, and their eff...

Urban and peri-urban environments are composed of a wide variety of materials, making land cover classification challenging. The objective of this research is to determine how effectively multi-season Landsat Enhanced Thematic Mapper Plus (ETM+) and single-season Advanced Land Observing Satellite (ALOS) Phased Array type L-band Synthetic Aperture Radar (PALSAR) data can be combined to map 17 land cover categories in the Greater Boston area of, Massachusetts, USA. The key goal of this work is to test the integration of radar and optical data. The contribution of different dimensions of input data to a random forest classifier was evaluated with map accuracy statistics. PALSAR data produced a 30.99% overall classification accuracy. Higher classification accuracy (72.24%) was achieved by adding texture variables derived from the PALSAR data. A September Landsat image produced a map accuracy of 77.96%. The inclusion of Landsat images from other three seasons increased map accuracy to 86.86% and Landsat derived texture variables further increased the map accuracy to 92.69%. The highest map accuracy (93.82%) was achieved by combining Landsat and PALSAR. Though combining PALSAR and Landsat only increased the overall accuracy by 1.1%, it was a statistically significant increase, whose magnitude was limited by the high accuracy already achieved with Landsat data. Moreover, confusion matrices and land cover maps indicated that most of this increase was from three urban land cover types (low density residential, high density residential, and commercial/industrial). The results demonstrate the value of combining multitemporal Landsat imagery, ALOS PALSAR data, and texture variables for land cover classification in urban and peri-urban environments.

Current studies of land cover change and landscape fragmentation rely predominantly on land cover classifications derived from remotely sensed images. However, limitations of traditional land cover classifications are numerous and well known. This research compares classification-based techniques (discrete data) to the use of vegetation indices (continuous data) for land cover modeling and analyses of landscape fragmentation for a study area in western Honduras. The study area typifies many regions of tropical developing countries, where a complex interaction of social and environmental factors have given rise to a dynamic mosaic of patches of reforestation and deforestation. Understanding the complex human dimensions of land use and land cover change in these parts of the world continues to present a challenge for researchers.

Traditional spectral classification of remote sensing data applied on per pixel basis ignores the potentially useful spatial information between the values of proximate pixels. Although spatial information extraction has been greatly explored, there have been limited attempts to enhance classification by combining spectral and spatial information. This improvement would arise from the hypothesis that a pixel is not independent of its neighbors and, furthermore, that its dependence can be quantified and incorporated into the classifier. This study aims to explore the potential of utilizing texture spatial variability using geostatistics and Grey Level Co-occurrence Matrix (GLCM) texture measures. Different texture layers derived from geostatistics method, namely fractal dimension, semivariogram, madogram, rodogram, pseudo-cross variogram and pseudo-cross madogram, were incorporated for the land cover classification of tropical rainforests in East Kalimantan, Indonesia. Texture layers of grey level co-occurrence matrix (GLCM) channels, i.e. variance, contrast, dissimilarity, and homogeinity, were also used for the classification. Two classification methods, using Support Vector Machine and Minimum distance were applied for image classification. Landsat 7 ETM images combined with textural information is used for land cover classification of tropical rainforest area. Band 5 of Landsat data was used to compute texture layers using the GLCM and geostatistics methods. This band was chosen because it has the highest variance of training data compared to other spectral bands. The results were compared to find out how the extra information given by the texture enhances the classification. According to the accuracy assessment using error matrix, combinations of image and texture data performed better with 81% of accuracy compared to those of image data only with 76% of accuracy.

Remotely sensed images and processing techniques are a primary tool for mapping changes in tropical forest types important to biodiversity and environmental assessment. Detailed land cover data are lacking for most wet tropical areas that present special challenges for data collection. For this study, we utilize decision tree (DT) classifiers to map 32 land cover types of varying ecological and economic importance over an 8000 km 2 study area and biological corridor in Costa Rica. We assess multivariate QUEST DTs with unbiased classification rules and linear discriminant node models for integrated vegetation mapping and change detection. Predictor variables essential to accurate land cover classification were selected using importance indices statistically derived with classification trees. A set of 35 variables from SRTM-DEM terrain variables, WorldClim grids, and Landsat TM bands were assessed.

In this paper the potential of neural networks has been applied to hyperspectral data and exploited either for classification purposes or for data feature extraction and dimensionality reduction. For this latter task, a topology named autoassociative neural network has been used. In its complete form, the processing scheme uses a neural network architecture consisting of two stages: the first stage

The MODIS land science team produces a number of standard products, including land cover and leaf area index (LAI). Critical to the success of MODIS and other sensor products is an independent evaluation of product quality. In that context, we describe a study using field data and Landsat ETM+ to map land cover and LAI at four 49-km 2 sites in North America containing agricultural cropland (AGRO), prairie grassland (KONZ), boreal needleleaf forest, and temperate mixed forest. The purpose was to: (1) develop accurate maps of land cover, based on the MODIS IGBP (International Geosphere -Biosphere Programme) land cover classification scheme; (2) derive continuous surfaces of LAI that capture the mean and variability of the LAI field measurements; and (3) conduct initial MODIS validation exercises to assess the quality of early (i.e., provisional) MODIS products. ETM+ land cover maps varied in overall accuracy from 81% to 95%. The boreal forest was the most spatially complex, had the greatest number of classes, and the lowest accuracy. The intensive agricultural cropland had the simplest spatial structure, the least number of classes, and the highest overall accuracy. At each site, mapped LAI patterns generally followed patterns of land cover across the site. Predicted versus observed LAI indicated a high degree of correspondence between field-based measures and ETM+ predictions of LAI. Direct comparisons of ETM+ land cover maps with Collection 3 MODIS cover maps revealed several important distinctions and similarities. One obvious difference was associated with image/map resolution. ETM+ captured much of the spatial complexity of land cover at the sites. In contrast, the relatively coarse resolution of MODIS did not allow for that level of spatial detail. Over the extent of all sites, the greatest difference was an overprediction by MODIS of evergreen needleleaf forest cover at the boreal forest site, which consisted largely of open shrubland, woody savanna, and savanna. At the agricultural, temperate mixed forest, and prairie grassland sites, ETM+ and MODIS cover estimates were similar. Collection 3 MODIS-based LAI estimates were considerably higher (up to 4 m 2 m À 2 ) than those based on ETM+ LAI at each site. There are numerous probable reasons for this, the most important being the algorithms' sensitivity to MODIS reflectance calibration, its use of a prelaunch AVHRR-based land cover map, and its apparent reliance on mainly red and near-IR reflectance. Samples of Collection 4 LAI products were examined and found to consist of significantly improved LAI predictions for KONZ, and to some extent for AGRO, but not for the other two sites. In this study, we demonstrate that MODIS reflectance data are highly correlated with LAI across three study sites, with relationships increasing in strength from 500 to 1000 m spatial resolution, when shortwave-infrared bands are included. D

This paper presents the land cover classification capabilities of fully versus partially polarimetric SAR data for C-and L-band frequencies. Maximum Likelihood classifier with complex Wishart distribution and artificial neural network classifier (ANN) have been used for classification. The change in accuracy due to the phase information of SAR data is also assessed by comparing the classified results of intensity and complex images for all the possible polarization combinations at L-and C-band. In all the combinations, fully polarimetric data provides highest accuracy and it is not much different from that of complex partial polarimetric (HH, VV) combination. The accuracies obtained with various partial polarimetric combinations are dependent on the land cover types. Among L-, C-and X-bands, L-band offers better accuracy. By combining all bands data, accuracy improved by 7%.The accuracy has been improved slightly by combining the three components of van Zyl decomposition with the combination of X-, C-and L-band. IRS-P6 optical data over the same area has been used to compare the classification accuracy between optical and SAR data.

Land cover classification over large geographic areas using remotely sensed data is increasingly common as a result of the requirements of national inventory and monitoring programmes, scientific modelling and international environmental treaties. Although large-area land cover products are more prevalent, standard operational protocols for their validation do not exist. This paper provides a framework for the accuracy assessment of large-area land cover products and synthesizes some of the key decision points in the design and implementation of an accuracy assessment from the literature. The fundamental components of a validation plan are addressed and the framework is then applied to the land cover map of the forested area of Canada that is currently being produced by the Earth Observation for Sustainable Development programme. This example demonstrates the compromise between the theoretical aspects of accuracy assessment and the practical realities of implementation, over a specific jurisdiction. The framework presented in this paper provides an example for others embarking on the assessment of large-area land cover products and can serve as the foundation for planning a statistically robust validation.

Estimates of forest area were obtained for the states of Indiana, Iowa, Minnesota, and Missouri in the United States using stratified analyses and observations from forest inventory plots measured in federal fiscal year 1999. Strata were created by aggregating the land cover classes of the National Land Cover Data (NLCD), and strata weights were calculated as proportions of strata pixel counts. The analyses focused on improving the precision of unbiased forest area estimates and included evaluation of the correspondence between forest/nonforest aggregations of the NLCD classes and observed attributes of forest inventory plots, evaluation of the utility of the NLCD as a stratification tool, and estimation of the effects on precision of image registration and plot location errors. The results indicate that the combination of NLCD-based stratification of inventory plots and stratified analyses increases the precision of forest area estimates and that the estimates are only slightly adversely affected by image registration and plot location errors. D

In Queensland, Australia, forest areas are discriminated from non-forest by applying a threshold (¨12%) to Landsat-derived Foliage Projected Cover (FPC) layers (equating to¨20% canopy cover), which are produced routinely for the State. However, separation of woody regrowth following agricultural clearing cannot be undertaken with confidence, and is therefore not mapped routinely by State Agencies. Using fully polarimetric C-, L-and P-band NASA AIRSAR and Landsat FPC data for forests and agricultural land near Injune, central Queensland, we corroborate that woody regrowth dominated by Brigalow (Acacia harpophylla) cannot be discriminated using either FPC or indeed C-band data alone, because the rapid attainment of a canopy cover leads to similarities in both reflectance and backscatter with remnant forest. We also show that regrowth cannot be discriminated from non-forest areas using either L-band or P-band data alone. However, mapping can be achieved by thresholding and intersecting these layers, as regrowth is unique in supporting both a high FPC (>¨12%) and Cband SAR backscatter (>~À 18 dB at HV polarisation) and low L-band and P-band SAR backscatter (e.g. <=¨14 dB at L-band HH polarisation). To provide a theoretical explanation, a wave scattering model based on that of Durden et al. [Durden, S.L., Van Zyl, J.J. & Zebker, H.A. (1989). Modelling and observation of radar polarization signature of forested areas. IEEE Trans. Geoscience and Remote Sensing, 27, 290 -301.] was used to demonstrate that volume scattering from leaves and small branches in the upper canopy leads to increases in C-band backscattering (particularly HV polarisations) from regrowth, which increases proportionally with FPC. By contrast, low L-band and P-band backscatter occurs because of the lack of double bounce interactions at co-polarisations (particularly HH) and volume scattering at HV polarisation from the stems and branches, respectively, when their dimensions are smaller than the wavelength. Regrowth maps generated by applying simple thresholds to both FPC and AIRSAR L-band data showed a very close correspondence with those mapped using samedate 2.5 m Hymap data and an average 73.7% overlap with those mapped through time-series comparison of Landsat-derived land cover classifications. Regrowth mapped using Landsat-derived FPC from 1995 and JER-1 SAR data from 1994 -1995 also corresponded with areas identified within the time-series classification and true colour stereo photographs for the same period. The integration of Landsat FPC and Lband SAR data is therefore expected to facilitate regrowth mapping across Queensland and other regions of Australia, particularly as Japan's Advanced Land Observing System (ALOS) Phase Arrayed L-band SAR (PALSAR), to be launched in 2006, will observe at both L-band HH and HV polarisations. D

This study is a first assessment of fully polarimetric SAR data from the Advanced Land Observation Satellite (ALOS) -Phased Array type L-band Synthetic Aperture Radar (PALSAR) over the test-site of Oberpfaffenhofen in Germany. The polarimetric calibration is checked by using corner reflectors on the Oberpfaffenhofen test-site. Further, the fully polarimetric images are classified by using the unsupervised H / A / α -Wishart Classification, based on entropy, anisotropy and alpha decompositions and supervised Wishart Classification. The potential in discerning land-cover classes is assessed by comparing them to optical images.

Geosciences Laser Altimeter System (GLAS) space LiDAR data are used to attribute a MODerate resolution Imaging Spectrometer (MODIS) 500 m land cover classification of a 10°latitude by 12°longitude study area in south-central Siberia. Timber volume estimates are generated for 16 forest classes, i.e., four forest cover types × four canopy density classes, across this 811,414 km 2 area and compared with a ground-based regional volume estimate. Two regional GLAS/MODIS timber volume products, one considering only those pulses falling on slopes ≤10°and one utilizing all GLAS pulses regardless of slope, are generated. Using a two-phase (GLAS-ground plot) sampling design, GLAS/MODIS volumes average 163.4 ± 11.8 m 3 /ha across all 16 forest classes based on GLAS pulses on slopes ≤ 10°and 171.9 ± 12.4 m 3 /ha considering GLAS shots on all slopes. The increase in regional GLAS volume per-hectare estimates as a function of increasing slope most likely illustrate the effects of vertical waveform expansion due to the convolution of topography with the forest canopy response. A comparable, independent, ground-based estimate is 146 m 3 /ha . Phytomass (live biomass) and carbon of Siberian forests. Biomass and Bioenergy, 14, 21-31], a difference of 11.9% and 17.7% for GLAS shots on slopes ≤10°and all GLAS shots regardless of slope, respectively. A ground-based estimate of total volume for the entire study area, 7.46 × 10 9 m 3 , is derived using Shepashenko et al.'s per-hectare volume estimate in conjunction with forest area derived from a 1990 forest map [Grasia, M.G. (ed.). (1990). Forest Map of USSR. Soyuzgiproleskhoz, Moscow, RU. Scale: 1:2,500,000]. The comparable GLAS/MODIS estimate is 7.38 × 10 9 m 3 , a difference of less than 1.1%. Results indicate that GLAS data can be used to attribute digital land cover maps to estimate forest resources over subcontinental areas encompassing hundreds of thousands of square kilometers.

This paper investigates the contribution of multi-temporal enhanced vegetation index (EVI) data to the improvement of object-based classification accuracy using multi-spectral moderate resolution imaging spectral-radiometer (MODIS) imagery. In object-oriented classification, similar pixels are firstly grouped together and then classified; the produced result does not suffer the speckled appearance and closer to human vision. EVI data are from the MODIS sensor aboard Terra spacecraft. 69 EVI data (scenes) were collected during the period of three years (2001Á2003) in a mountainous vegetated area. These data sets were used to study the phenology of the land cover types. Different land cover types show distinct fluctuations over time in EVI values and this information might be used to improve object-oriented land cover classification. Two experiments were carried out: one was only with single date MODIS multispectral data, and the other one including also the 69 EVI images. Eight classes were distinguished: temperate forest, tropical dry forest, grassland, irrigated agriculture, rain-fed agriculture, orchards, lava flows and human settlement. The two classifications were evaluated with independent verification data, and the results showed that with multitemporal EVI data, the classification accuracy was improved 5.2%. Evaluated by McNemar's test, this improved was significant, with significance level p 00.01.

A set of ERS SAR and optical MODIS-images were classified to land cover and tree species classes. Different methods for pixel and decision based data fusion were tested. Classifications of featuresets were carried out using Bayes rule for minimum error. The results were not very successful, the classification accuracies of land cover classes varied from 43% to 75%, depending on the used features and classes. The decision based data fusion method, where the a'posteriori probabilities representing the proportions of different land cover classes of low resolution classification are used as a'prior probabilities in high resolution classification looks promising. Using this method, the increase of overall and classwise accuracies can be more than 10 and 25 %-units, respectively.

Despite its very large territory and the best Landsat archive in the world, Canada has made very limited use of Landsat data for land cover mapping. The primary difficulty has been the prohibitive cost of information extraction and the earlier (and now overcome for Landsat-7 enhanced thematic mapper plus data) high cost of data purchase. The solution to this remaining obstacle lies in decreasing the cost of Landsat data processing and analysis while ensuring the high quality of the extracted information. In this paper, we present an efficient and effective approach to mapping land cover in Canada from Landsat thematic mapper data (single or multiple satellites). The key features of this approach are an increase in the ratio of computer to human analysis and automation for high data volume or large area processing. However, it is essential that the final product quality not suffer because of the greater reliance on computer processing, thus the algorithm performance becomes critical. We describe the overall approach, discuss key challenges, explain the principles behind key algorithms developed to respond to the challenges, present evidence demonstrating the effectiveness of these algorithms in a boreal landscape setting, and consider implementation issues. With a processing system developed to handle large numbers (tens to hundreds) of Landsat scenes, which incorporates most of the algorithms discussed here, the stage is nearly set for large-scale processing leading to a Landsat-based land cover classification product(s) for Canada. Résumé. En dépit de l'étendue de son territoire et de la disponibilité de la meilleure archive Landsat au monde, le Canada a jusqu'à maintenant très peu fait usage des données Landsat pour les besoins de la cartographie du couvert. La difficulté première provient du coût prohibitif de l'extraction de l'information et des coûts élevés au départ pour l'acquisition des données (problème réglé depuis avec les données Landsat-7 ETM+). La solution à ce problème qui perdure réside dans la diminution des coûts du traitement et de l'analyse des données Landsat tout en assurant la haute qualité de l'information extraite. Dans cet article, nous présentons une approche efficace et rentable pour la cartographie du couvert au Canada à partir des données Landsat TM (uni-ou multi-satellitaires). Sa caractéristique principale réside dans l'accroissement du rapport ordinateur versus analyse humaine et sur l'automatisation du traitement des données à grand volume-grande superficie. Toutefois, il est essentiel que la qualité du produit final ne souffre pas en raison de la dépendance plus grande sur le traitement par ordinateur; dans ce contexte, la performance de l'algorithme devient critique. Nous décrivons l'approche globale, nous discutons des enjeux fondamentaux, nous expliquons les principes sous-jacents des algorithmes développés pour répondre aux défis posés, nous présentons l'évidence démontrant l'efficacité de ces algorithmes dans le contexte des paysages boréaux et nous considérons les concepts de mise en place. Avec un système de traitement conçu pour traiter des grandes quantités de scènes Landsat (des dizaines à des centaines) incorporant la plupart des algorithmes décrits dans cet article, le scénario est presque en place pour le traitement des données sur une grande échelle menant à un ou des produits de classification du couvert basés sur Landsat pour le Canada.

To better understand how terrestrial vegetative ecosystems respond to climate and/or anthropogenic effects, the scientific community is increasingly interested in developing methods to employ satellite data to track changes in land surface phenology (e.g., timing and rate of green-up, amplitude and duration of growing season, timing and rate of senescence of plant classes). By increasing the inherent resolution of signal extraction procedures, while minimizing the effects of cloud cover and prolonged data gaps, such tools can significantly improve land cover classification and land cover change monitoring on multiple scales. This report describes an intuitive approach for tracking the intra-annual details and inter-annual variability of multiyear time series employing a sequence of annual high order polynomial splines (up to 14th order), stabilized by minimizing model roughness, and weighted to fit the upper data envelope to minimize cloud cover bias. The algorithm is tested using a multi-year period-of-record (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002) for three very different classes of vegetation -stable agriculture, high elevation montane shrubland and semi-arid grassland with high inter-annual variability. The results accurately track both short and long-term land surface phenology, and illustrate a robust potential for extracting temporal and spatial detail from a variety of satellite-based, multi-year vegetation signals.