RVM‐based multi‐class classification of remotely sensed data

The relevance vector machine (RVM), a Bayesian extension of the support vector machine (SVM), has considerable potential for the analysis of remotely sensed data. Here, the RVM is introduced and used to derive a multi‐class classification of land cover with an accuracy of 91.25%, a level comparable to that achieved by a suite of popular image classifiers including the SVM. Critically, however, the output of the RVM includes an estimate of the posterior probability of class membership. This output may be used to illustrate the uncertainty of the class allocations on a per‐case basis and help to identify possible routes to further enhance classification accuracy.     

Classification of multi‐spectral remote sensing data using a local transfer function classifier

This paper introduces a new neural network, called the local transfer function classifier (LTF‐C), for classification of multi‐spectral remote sensing data. The network structure of LTF‐C is similar to that of the radial basis function neural network (RBF), but LTF‐C utilizes an entirely different learning algorithm. In particular, the network structure of LTF‐C is not predetermined, but changes dynamically during the learning. Such a learning algorithm fits well to the classification problem, and guarantees that the size of the network is as large as is needed. The classification results show that LTF‐C evidently has a better classification accuracy than the six other classifiers in the experiment.     

Calibration and validation of and results from a micro-simulation model to explore drivers’ actual use of acceleration lanes

This study investigates the actual use of acceleration lanes by means of traffic micro-simulations in proximity to the convergence between the main and the secondary traffic streams.The data needed to develop the micro-simulation model were collected videotaping two acceleration lanes of an Italian highway, near the village of San Liberato. The relevant microsimulation model has been developed utilizing a commercial software, the TransModeler traffic simulation package, while the calibration and validation phases have been realized based on real-life data.Thanks to the microsimulation model, six different scenarios based on various presentation speeds have been analyzed, in a less expensive, quicker and more secure way compared to field observations. A quantitative and qualitative analysis of the results has been discussed to examine both geometric and kinematic aspects of the merging maneuvers and to compare road users' behaviors with theoretical road design guidelines. The analyses have been focused on vehicles coming from the on-ramp, in terms of entrance points dispersion into the main traffic stream along the acceleration lanes, merging speeds, and reached acceleration rates. The mutual influence between the two flows (main and secondary traffic streams), combined with vehicle performances and characteristics of drivers, led to a first assessment of the main variables from which the required length of the specialized lanes depends.     

A high‐level data fusion and spatial modelling system for change‐detection analysis using high‐resolution airborne digital sensor data

A high‐level data fusion system that uses Bayesian statistics involving weights‐of‐evidence modelling is described to combine disparate information from airborne digital data such as digital surface model (DSM), colour, thermal infrared (TIR) and hyperspectral images at different time periods. To determine the efficacy of the system, an analysis of change detection was performed. The data fusion system is capable of detecting changes in man‐made features automatically in a densely populated area where there is little prior information. Multiclass segmented images were obtained from the data captured by four airborne remote sensing sensors. The system performs data fusion modelling by using binary images of each theme class and a total of 40 binary patterns were obtained. Through Bayesian methods, involving weights‐of‐evidence modelling, all the binary images were analysed and finally four binary patterns (indicator images) were identified automatically as significant for the change‐detection application. A weighted index overlay model available in the system combines these four patterns. Data fusion by weights‐of‐evidence modelling is found to be straightforward and unequivocal for predicting newly transformed locations. The results of the Bayesian method are accurate as the weights are based on statistical analysis. Changes in features such as colour of roofs, parking areas, openland areas, newly built structures, and the presence or absence of vehicles are extracted automatically by using the high‐level data fusion approach. The final predictor image shows the probability of change‐detected areas in a densely populated city in Japan.     

Detection of coal fire location and change based on multi‐temporal thermal remotely sensed data and field measurements

This study focuses on analysis methods for monitoring coal fires, using a combination of multi‐temporal thermal infrared data, high spatial resolution remote sensing data and field measurements. This technical note is prepared as a feasibility study for the detection of coal fire dynamics in the Inner Mongolia Autonomous Region in northern China.     

A theoretical and empirical investigation of multi‐item on‐line auctions

In this paper we explore and analyze the structure of Internet auctions from an analytical and an empirical perspective. Such web‐based auctions are rapidly emerging as a mercantile process of choice in the electronic marketplace. We observe current Internet auctions for one‐time products, such as rapidly aging hardware, and analyze them within the framework of the existing auction theory. While traditional auction theory focuses on single‐item auctions, we observe that a majority of on‐line auctions are multi‐item auctions. A significant contribution of this work is the theoretical derivation of the structure of the winning bids in multi‐item progressive on‐line auctions. Additionally, for comparative purposes, we explore the structural characteristics of alternative multi‐item auction mechanisms proposed in the auction theory. We derive hypotheses based on our analytical results and compare two different types of auction mechanisms. We test the traditional auction theory assumption regarding the homogeneity of bidders and present the first ever empirically derived classification and performance‐comparison of on‐line bidders. We test our hypotheses using real‐world empirical data obtained by tracking a premier web‐based auction site. Statistical analysis of the data indicates that firms may gain by choosing alternative auction mechanisms. We also provide directions for further exploration of this emerging but important dimension of electronic commerce. This revised version was published online in July 2006 with corrections to the Cover Date.     

Urban land cover multi‐level region‐based classification of VHR data by selecting relevant features

The limited spatial resolution of satellite images used to be a problem for the adequate definition of the urban environment. This problem was expected to be solved with the availability of very high spatial resolution satellite images (IKONOS, QuickBird, OrbView‐3). However, these space‐borne sensors are limited to four multi‐spectral bands and may have specific limitations as far as detailed urban area mapping is concerned. It is therefore essential to combine spectral information with other information, such as the features used in visual interpretation (e.g. the degree and kind of texture and the shape) transposed to digital analysis. In this study, a feature selection method is used to show which features are useful for particular land‐cover classes. These features are used to improve the land‐cover classification of very high spatial resolution satellite images of urban areas. The useful features are compared with a visual feature selection. The features are calculated after segmentation into regions that become analysis units and ease the feature calculation.     

Analysis of urban growth using multi‐temporal satellite data in Istanbul,Turkey

Uncontrolled population growth, especially in developing countries, causes serious problems, such as scarcity of food, informal settlements, environmental pollution, destruction of ecological structure, unemployment, etc. This phenomenon will require advanced methodologies, such as space technologies, to enable city planners, economists, environmentalists, ecologists and resource managers to solve these problems. In Turkey, as a result of the undesired population growth, new settlements are continuously appearing and adverse developments and changes are occurring in the presently populated areas. In Turkey's major cities, such as in Istanbul, Izmir, Ankara, Adana, Bursa and Antalya, which have seen mass migrations of people, considerable urban developments and changes have occurred. Consequently, the mostly negative impacts of uncontrolled population growth on the urban environment must be monitored continuously.

The objective of this study was to investigate the impact of urban growth on land‐use changes, especially the agricultural land in the district of Büyükçekmece in suburban Istanbul. The study is based on the 1984–1997 population database, multi‐temporal satellite data and remote sensing methods. The study revealed significant loss of agricultural and natural land areas to urban developments throughout the period 1984–1998.     

The Jeffries–Matusita distance for the case of complex Wishart distribution as a separability criterion for fully polarimetric SAR data

In multidimensional observations, many classification algorithms (supervised or unsupervised) require the selection of optimum bands in which the classes are most distinct. The Jeffries–Matusita (JM) distance is widely used as a separability criterion for optimal band selection and evaluation of classification results. Its original form is based on the assumption of normal distribution of the data. However, in the case of the covariance/coherency matrix of synthetic aperture radar (SAR) polarimetry, the data follow the complex Wishart distribution. In this article, we calculate the JM separability criterion for the case of the complex Wishart distribution. The updated formulation is used for: (1) the estimation of the separability between classes in fully polarimetric SAR data and to evaluate two standard polarimetric SAR classification algorithms, the Wishart and the expectation maximization algorithms, and (2) the classification of fully polarimetric SAR images based on the derived JM separability for the case of complex Wishart distribution. Fully polarimetric RADARSAT-2 images over sea ice in the Canadian Arctic are used to classify different ice surfaces and open water.     

Regional thermal and terrain modelling of the Afar Depression from MODIS multi‐temporal monthly night LST data

Multi‐temporal MODIS thermal imagery isolates geothermal from seasonal land surface temperature (LST) variability in Afar Depression and identifies localized LST thermal anomalies during the September 2005 seismo‐volcanic crisis. The interpretation of the monthly night annual imagery sequences and time series analysis proved that in the northern and the central parts, LST is maximized throughout the year. The two regions lie along the NW‐SE trending Red Sea Propagator, which is the on‐land southern termination of the Red Sea Rift. Linear regression analysis proved that LST variation is terrain elevation dependent. Terrain segmentation outlined 5 terrain classes. The localized geothermal activity during the September 2005 seismic‐volcanic crisis activity in both Erta Ale and Dabbahu vicinity was revealed by combined terrain‐LST modelling. The LST statistics of the segmented terrain classes revealed that in September 2005, a regional minor increase in LST also affected the four classes with positive elevation.     

Integration of multitemporal/polarization C‐band SAR data sets for land‐cover classification

This paper investigates the potential of multitemporal/polarization C‐band SAR data for land‐cover classification. Multitemporal Radarsat‐1 data with HH polarization and ENVISAT ASAR data with VV polarization acquired in the Yedang plain, Korea are used for the classification of typical five land‐cover classes in an agricultural area. The presented methodologies consist of two analytical stages: one for feature extraction and the other for classification based on the combination of features. Both a traditional SAR signal property analysis‐based approach and principal‐component analysis (PCA) are applied in the feature extraction stage. Special concerns are in the interpretation of each principal component by using principal‐component loading. The tau model applied as a decision‐level fusion methodology can provide a formal framework in which the posteriori probabilities derived from different sensor data can be combined. From the case study results, the combination of PCA‐based features showed improved classification accuracy for both Radarsat‐1 and ENVISAT ASAR data, as compared with the traditional SAR signal property analysis‐based approach. The integration of PCA‐based features based on multiple polarization (i.e. HH from Radarsat‐1, and both VV and VH from ENVISAT ASAR) and different incidence angles contributed to a significant improvement of discrimination capability for dry fields which could not be properly classified by using only Radarsat‐1 or ENVISAT ASAR data, and thus showed the best classification accuracy. The results of this case study indicate that the use of multiple polarization SAR data with a proper feature extraction stage would improve classification accuracy in multitemporal SAR data classification, although further consideration should be given to the polarization and incidence angle dependency of complex land‐cover classes through more experiments.     

Estimation and validation of LAI using physical and semi‐empirical BRDF models

This letter deals with estimation of LAI for a wheat crop using physical and semi‐empirical BRDF models and IRS‐1D LISS‐III sensor data. NDVI was computed for both the models with LAI as a free parameter. The model‐computed NDVI was compared with corresponding atmospherically corrected LISS‐III NDVI. The estimation of LAI was carried out on the basis of a look‐up table approach and minimum root mean squared deviation between model computed and observed NDVI. The estimated LAI was validated against field measurements carried out during the months of February and March 2003, at the Central State Farm, Rajasthan, India. It was found that LAI was underestimated in both physical and semi‐empirical models. Results show that inclusion of multiple scattering in physical models may not always lead to a more accurate estimation of LAI and that it may be possible to estimate LAI at early stages of crop growth using semi‐empirical models. The coefficient of determination (R ²) between model estimated and measured LAI was 0.57 (standard error of estimate (SE) 0.156) and 0.63 (SE 0.187) for semi‐empirical and physical models, respectively, in the single scattering approximation, for February data. The corresponding values for March data were 0.57 (SE 0.206) and 0.51 (SE 0.216), respectively.     

Extracting structural characteristics of herbaceous floodplain vegetation under leaf‐off conditions using airborne laser scanner data

Hydrodynamic models of river flow need detailed and accurate friction values as input. Friction values of floodplain vegetation are based on vegetation height and density. To map spatial patterns of floodplain vegetation structure, airborne laser scanning is a promising tool. In a test for the lower Rhine floodplain, vegetation height and density of herbaceous vegetation were measured in the field at 42 georeferenced plots of 200 m² each. Simultaneously, three airborne laser scanning (ALS) surveys were carried out in the same area resulting in three high resolution, first pulse, small‐footprint datasets. The laser data surveys differed in flying height, gain setting and laser diode age. Point density of the laser data varied between 10 and 75 points m^?2. Point heights relative to the DTM derived from the ALS data were used in all analyses. Laser points were labelled as either vegetation or ground using three different methods: (1) a fixed threshold value; (2) a flexible threshold value based on the inflection point in the point height distribution; and (3) using a Gaussian distribution to separate noise in the ground surface points from vegetation. Twenty‐one statistics were computed for each of the resulting vegetation‐point distributions, which were subsequently compared with field observations of vegetation height. Additionally, the percentage index (PI) was computed to relate density of vegetation points to hydrodynamic vegetation density. The vegetation height was best predicted by using the inflection method for labelling and the 95 percentile as a regressor (R ² = 0.74–0.88). Vegetation density was best predicted using the threshold method for labelling and the PI as a predictor (R ² = 0.51). The results of vegetation height prediction were found to depend on the combined effect of flying height, gain setting or laser diode age. The quality of the estimation of vegetation height and density is also affected by point density, for densities lower than 15 points m^?2. We conclude that high resolution ALS data allows to estimate vegetation height and density of herbaceous vegetation in winter condition, but field reference data remains necessary for calibration.     

Multi‐scale thermal pattern monitoring of a large lake (Lake Geneva) using a multi‐sensor approach

The applicability of satellite imagery products from different sensors (AVHRR‐derived multi‐channel sea surface temperature (MCSST), MODIS sea surface temperature (SST) products 5‐Min L2 Swath 1 km and Landsat TM band 6 thermal signature) for the comprehensive monitoring of temperature and its temporal patterns over a large lake is tested in this study. The coverage of cloud‐free satellite data for Lake Geneva is reported throughout a year and, more specifically, during a 13 day period in summer 2003. In a second step, we demonstrate the feasibility of the AVHRR/MODIS imagery to discern day and night temperature patterns, by generating day and night climatologies and various spatial statistics over the 13 day period. The different day and night surface thermal patterns observed by satellite imagery could be linked to the thermal structure existing in deeper water using the concept of the diurnal decoupled layer. The forcing of the persistent patterns, two warm cores divided by a saddle‐shaped cold anomaly, is explained by wind periodicity and insolation conditions. The patterns can be matched to features postulated by findings of different limnologists in the past. Other surface temperature related phenomena such as water upwelling and downwelling and the occurrence of plumes are related to meteorological and hydrological events. The lakewide average lake surface water temperature (LSWT) trends for day and night during the study period are roughly parallel. A sudden loss of stored heat can be explained by episodes of long fetch, synoptic wind (bise) that interrupted the predominant breeze regime.     

Retrieval of sea ice thickness distribution in the seasonal ice zone from airborne L‐band SAR

Although satellite data are known to be useful for obtaining ice thickness distribution for perennial sea ice or in stable thin sea ice areas, their use in the seasonal sea ice zone (SIZ) is still unresolved. In this study, we approached the problem of ice thickness retrieval by using L‐band Synthetic Aperture Radar (SAR). In the SIZ, ice thickness growth is closely related to ridging activity and therefore we expected surface roughness to be correlated to ice thickness. L‐band SAR is suitable for detecting such surface roughness and should be a useful tool for obtaining ice thickness distribution. To verify this correlation, we conducted shipborne electromagnetic (EM) inductive sounding and supersonic profiling observations with an icebreaker, coordinated with airborne L‐band SAR observations in the southern Sea of Okhotsk in February 2005. The surface elevation was estimated by representing the ship's motion with a low‐pass filter. Backscattering coefficients correlated well with ice thickness and surface roughness, defined by the standard deviation of the surface elevation. This result sheds light on the possibility of determining ice thickness distribution in the SIZ.     

Operational maize yield model development and validation based on remote sensing and agro‐meteorological data in Kenya

Remote‐sensing data acquired by satellite have a wide scope for agricultural applications owing to their synoptic and repetitive coverage. On the one hand, spectral indices deduced from visible and near‐infrared remote‐sensing data have been extensively used for crop characterization, biomass estimation, and crop yield monitoring and forecasting. On the other hand, extensive research has been conducted using agrometerological models to estimate soil moisture to produce indicators of plant‐water stress. This paper reports the development of an operational spectro‐agrometeorological yield model for maize using a spectral index, the Normalized Difference Vegetation Index (NDVI) derived from SPOT‐VEGETATION, meteorological data obtained from the European Centre for Medium‐Range Weather Forecast (ECMWF) model, and crop‐water status indicators estimated by the Crop‐Specific Water Balance model (CSWB). Official figures produced by the Government of Kenya (GoK) on crop yield, area planted, and production were used in the model. The statistical multiple regression linear model has been developed for six large maize‐growing provinces in Kenya. The spectro‐agrometerological yield model was validated by comparing the predicted province‐level yields with those estimated by GoK. The performance of the NDVI and land cover weighted NDVI (CNDVI) on the yield model was tested. Using CNDVI instead of NDVI in the model reduces 26% of the unknown variance. Of the output indicators of the CSWB model, the actual evapotranspiration (ETA) performs best. CNDVI and ETA in the model explain 83% of the maize crop yield variance with a root square mean error (RMSE) of 0.3298 t ha^?1. Very encouraging results were obtained when the Jack‐knife re‐sampling technique was applied, thus proving the validity of the forecast capability of the model (r ² = 0.81 and RMSE = 0.359 t ha^?1). The optimal prediction capability of the independent variables is 20 days and 30 days for the short and long maize crop cycles, respectively. The national maize production during the first crop season for the years 1998–2003 was estimated with an RMSE of 185 060 t and coefficient of variation of 9%.     

Multi‐sensor data fusion and comparison of total column ozone

With many remote‐sensing instruments onboard satellites exploring the Earth's atmosphere, most data are processed to gridded daily maps. However, differences in the original spatial, temporal, and spectral resolution—as well as format, structure, and temporal and spatial coverage—make the data merging, or fusion, difficult. NASA Goddard Earth Sciences Data and Information Services Center (GES‐DISC) has archived several data products for various sensors in different formats, structures, and multi‐temporal and spatial scales for ocean, land, and atmosphere. In this investigation using Earth science data sets from multiple sources, an attempt was made to develop an optimal technique to merge the atmospheric products and provide interactive, online analysis tools for the user community. The merged/fused measurements provide a more comprehensive view of the atmosphere and improve coverage and accuracy, compared with a single instrument dataset. This paper describes ways of merging/fusing several NASA Earth Observing Systems (EOS) remote‐sensing datasets available at GES‐DISC. The applicability of various methods was investigated for merging total column ozone to implement these methods into Giovanni, the online interactive analysis tool developed by GES‐DISC. Ozone data fusion of MODerate resolution Imaging Spectrometer (MODIS) Terra and Aqua Level‐3 daily data sets was conducted, and the results were found to provide better coverage. Weighted averaging of Terra and Aqua data sets, with the consequent interpolation through the remaining gaps using Optimal Interpolation (OI), also was conducted and found to produce better results. Ozone Monitoring Instrument (OMI) total column ozone is reliable and provides better results than Atmospheric Infrared Sounder (AIRS) and MODIS. However, the agreement among these instruments is reasonable. The correlation is high (0.88) between OMI and AIRS total column ozone, while the correlation between OMI and MODIS Terra/Aqua fused total column ozone is 0.79.     

Potential of polarimetric Radarsat-2 data in geological mapping – a case study in parts of Dharwar craton,India

Polarimetry is a technique for quantitatively estimating the change in polarization in a backscattered electromagnetic signal (in this case a radar signal) with reference to the polarization of the incident signal for estimating the geophysical properties (roughness and dielectric constant) of surface elements. In this study, quadrapole Radarsat-2 data have been utilized for mapping geological elements in Archaean terrain based on their geophysical parameters preserved in the four polarization channels of Radarsat-2 data. The study illustrates how single-look complex polarimetric data can be processed to extract information on lithology, geological structure, and land covers (i.e. soil type, etc.) preferentially developed over each rock type. The study revealed that C-band polarization power images can be used indirectly to delineate lithounits provided each unit has characteristic surface roughness and moisture resulting from the surface cover or exposure pattern developed over each lithounit. Moreover, polarimetric data (inclusive of phase and amplitude) provide complementary information when we compare this information with the information derived from simple polarization composites, derived from the backscattered image of each polarization channel. It is also observed that structural imprints are enhanced in co-polarized (HH, VV) data, whereas the cross-polarized channel (HV, VH) provides information on land cover.     

Temporal and spatial soil moisture change pattern detection in an agricultural area using multi‐temporal Radarsat ScanSAR data

Monitoring the characteristics of spatially and temporally distributed soil moisture is important to the study of hydrology and climatology for understanding and calculating the surface water balance. The major difficulties in retrieving soil moisture with Synthetic Aperture Radar (SAR) measurements are due to the effects of surface roughness and vegetation cover. In this study we demonstrate a technique to estimate the relative soil moisture change by using multi‐temporal C band HH polarized Radarsat ScanSAR data. This technique includes two components. The first is to minimize the effects of surface roughness by using two microwave radar measurements with different incidence angles for estimation of the relative soil moisture change defined as the ratio between two soil volumetric moistures. This was done by the development of a semi‐empirical backscattering model using a database that simulated the Advanced Integral Equation Model for a wide range of soil moisture and surface roughness conditions to characterize the surface roughness effects at different incidence angles. The second is to reduce the effects of vegetation cover on radar measurements by using a semi‐empirical vegetation model and the measurements obtained from the optical sensors (Landsat TM and AVHRR). The vegetation correction was performed based on a first‐order semi‐empirical backscattering vegetation model with the vegetation water content information obtained from the optical sensors as the input. For the validation of this newly developed technique, we compared experimental data obtained from the Southern Great Plain Soil Moisture Experiment in 1997 (SGP97) with our estimations. Comparison with the ground soil moisture measurements showed a good agreement for predication of the relative soil moisture change, in terms of ratio, with a Root Mean Square Error (RMSE) of 1.14. The spatially distributed maps of the relative soil moisture change derived from Radarsat data were also compared with those derived from the airborne passive microwave radiometer ESTAR. The maps of the spatial characteristics of the relative soil moisture change showed comparable results.     

Performance variability as an indicator of fatigue and boredom effects in a VDT data‐entry task

This study evaluated the utility of performance variability as an indicator of fatigue and boredom effects in a computer‐based data‐entry task. Twenty‐four data transcribers entered numeric data from tax forms at a video‐display terminal (VDT) for 3 consecutive workdays under one of the following work management conditions: Electronic Performance Monitoring (EPM) or No Electronic Performance (No EPM). Self‐ratings of fatigue and boredom were assessed at periodic intervals. The speed (keystrokes) and the accuracy (errors) of task performance were recorded on a continuous basis. Irrespective of the type of work management condition, there was an increase in fatigue and boredom from the morning to afternoon work sessions. Correspondingly, there was an increase in the variability of speed performance from the morning to afternoon work sessions for both the EPM and No EPM conditions. Trend analyses revealed a correspondence of the linear increase in fatigue and boredom with the linear increase in speed variability across the afternoon work periods under the No EPM condition. The findings of this study suggest that performance variability may be an objective indicator of fatigue and boredom effects in repetitive computer tasks.