Remote sensing of urban vegetation life form by spectral mixture analysis of high‐resolution IKONOS satellite images

This paper evaluates the techniques of linear spectral unmixing (LSU), comparing high‐ and medium‐resolution images for their ability to obtain separate estimates of tree and grassy surfaces in urban areas. It demonstrates that, unlike on medium‐resolution images, tree and grassy surfaces each constitute distinct endmembers on high‐resolution images. This is because at high resolution, shadows in the urban scene approximate pixel size and therefore can be separately masked, thus avoiding the spectral similarities between shadow and tree canopies on the one hand, and low albedo surfaces on the other. In this study, the ability to mask shadow on IKONOS VHR images removes these spectral overlaps. Spatial autocorrelation, applied to find the characteristic scale lengths of vegetated patches in the study area, demonstrated that at the 4 m spatial resolution of IKONOS almost two thirds of pixels would be mixed, and at the 20 m resolution of SPOT all pixels would be mixed. Accuracies of the tree and grass fractions were found to be very high in the case of IKONOS, with 87% confidence that both the grass and tree fractions within each pixel were within 10% of the actual amount. The somewhat lower accuracy for SPOT supports previous studies based on medium‐resolution sensors, which have noted that trees do not constitute an endmember.     

Land‐cover mixing and spectral vegetation indices

Vegetation indices have been widely used as indicators of seasonal and inter‐annual variations in vegetation caused by either human activities or climate, with the overall goal of observing and documenting changes in the Earth system. While existing satellite remote sensing systems, such as NASA's Multi‐angle Imaging SpectroRadiometer (MISR) and Moderate Resolution Imaging Spectroradiometer (MODIS), are providing improved vegetation index data products through correcting for the distortions in surface reflectance caused by atmospheric particles as well as ground covers below vegetation canopy, the impact of land‐cover mixing on vegetation indices has not been fully addressed. In this study, based on real image spectral samples for two‐component mixtures of forest and common nonforest land‐cover types directly extracted from a 1.1?km MISR image by referencing a 30?m land‐cover classification, the effect of land‐cover mixing on the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) has been quantitatively evaluated. When the areal fraction of forest was lower than 80%, both NDVI and EVI varied greatly with mixed land‐cover types, although EVI varied less than NDVI. Such a phenomenon can cause errors in applications based on use of these vegetation indices. This study suggests that methods that reduce land‐cover mixing effects should be introduced when developing new spectral vegetation indices.     

Quantifying high‐resolution impervious surfaces using spectral mixture analysis

Impervious surface distribution and its temporal changes are considered key urbanization indicators and are utilized for analysing urban growth and influences of urbanization on natural environments. Recently, urban impervious surface information was extracted from medium/coarse resolution remote sensing imagery (e.g. Landsat ETM+ and AVHRR) through spectral analytical methods (e.g. spectral mixture analysis (SMA), regression tree, etc.). Few studies, however, have attempted to generate impervious surface information from high resolution remotely sensed imagery (e.g. IKONOS and Quickbird). High resolution images provide detailed information about urban features and are, therefore, more valuable for urban analysis. The improved spatial resolution, however, also brings new challenges when existing spectral analytical methods are applied. In particular, a higher spatial resolution leads to reduced boundary effects and increased within‐class variability. Taking Grafton, Wisconsin, USA as a study site, this paper analyses the spectral characteristics of IKONOS imagery and explores the applicability of SMA for impervious surface estimation. Results suggest that with improved spatial resolution, IKONOS imagery contains 40–50% of mixed urban pixels for the study area, and the within‐class variability is a severe problem for spectral analysis. To address this problem, this paper proposes two approaches, interior end‐member set selection and spectral normalization, for SMA. Analysis of results indicates that these approaches can reasonably reduce the problems associated with boundary effects and within‐class variability, therefore generating better impervious surface estimates.     

Urban vegetation mapping using sub‐pixel analysis and expert system rules: A critical approach

Since the traditional hard classifier can label each pixel only with one class, urban vegetation (e.g. trees) can only be recorded as either present or absent. The sub‐pixel analysis that can provide the relative abundance of surface materials within a pixel may be a potential solution to effectively identifying urban vegetation distribution. This study examines the effectiveness of a sub‐pixel classifier with the use of expert system rules to estimate varying distributions of different vegetation types in urban areas. The Spearman's rank order correlation between the vegetation output and reference data for wild grass, man‐made grass, riparian vegetation, tree, and agriculture were 0.791, 0.869, 0.628, 0.743, and 0.840 respectively. Results from this study demonstrated that the expert system rule using NDVI threshold procedure is reliable and the sub‐pixel processor picked the signatures relatively well. This study reports a checklist of the sources of limitation in the application of sub‐pixel approaches.     

Spectral vegetation indices for estimating shrub cover,green phytomass and leaf turnover in a sedge‐shrub tundra

Using field observations, we determined the relationships between spectral indices and the shrub ratio, green phytomass and leaf turnover of a sedge-shrub tundra community in the Arctic National Wildlife Refuge, Alaska, USA. We established a 50‐m × 50‐m plot (69.73°N 143.62°W) located on a floodplain of the refuge. The willow shrub (Salix lanata) and sedge (Carex bigelowii) dominated the plot vegetation. In July to August 2007, we established ten 0.5‐m × 0.5‐m quadrats on both shrub‐covered ground (shrub quadrats) and on ground with no shrubs (sedge quadrats). The shrub ratio was more strongly correlated with the normalized difference vegetation index (NDVI, R2 of 0.57) than the normalized difference infrared index (NDII), the soil-adjusted vegetation index (SAVI) or the enhanced vegetation index (EVI). On the other hand, for both green phytomass and leaf turnover, the strongest correlation was with NDII (R ² of 0.63 and 0.79, respectively).     

Modelling land‐cover types using multiple endmember spectral mixture analysis in a desert city

Spectral mixture analysis is probably the most commonly used approach among sub‐pixel analysis techniques. This method models pixel spectra as a linear combination of spectral signatures from two or more ground components. However, spectral mixture analysis does not account for the absence of one of the surface features or spectral variation within pure materials since it utilizes an invariable set of surface features. Multiple endmember spectral mixture analysis (MESMA), which addresses these issues by allowing endmembers to vary on a per pixel basis, was employed in this study to model Landsat ETM+ reflectance in the Phoenix metropolitan area. Image endmember spectra of vegetation, soils, and impervious surfaces were collected with the use of a fine resolution Quickbird image and the pixel purity index. This study employed 204 (3×17×4) total four‐endmember models for the urban subset and 96 (6×6×2×4) total five‐endmember models for the non‐urban subset to identify fractions of soil, impervious surface, vegetation and shade. The Pearson correlation between the fraction outputs from MESMA and reference data from Quickbird 60 cm resolution data for soil, impervious, and vegetation were 0.8030, 0.8632, and 0.8496 respectively. Results from this study suggest that the MESMA approach is effective in mapping urban land covers in desert cities at sub‐pixel level.     

Phase correlation pixel‐to‐pixel image co‐registration based on optical flow and median shift propagation

With singular value decomposition (SVD) and robust 2‐dimensional fitting phase correlation algorithms, it is possible to achieve pixel‐to‐pixel image co‐registration at sub‐pixel accuracy via local feature matching. However, the method often fails in featureless and low correlation areas making it not robust for co‐registration of images with considerable spectral differences and large featureless ground objects. A median shift propagation (MSP) technique is proposed to eliminate the problem, in a phase correlation and Normalized Cross‐Correlation (NCC) combined approach. The experiment results using images from different sensor platforms and spectral bands indicate that the new method is very robust to featureless and low correlation areas and can achieve very accurate pixel‐to‐pixel image co‐registration with good tolerance of spectral and spatial differences between images. The method will significantly improve change detection in various remote sensing applications.     

Mapping burned areas in Mediterranean countries using spectral mixture analysis from a uni‐temporal perspective

The main aim of this study was to evaluate the usefulness of spectral mixture analysis (SMA) for mapping forest areas burned by fires in the Mediterranean area using low and medium spatial resolution satellite sensor data. A methodology requiring only one single post‐fire image was used to carry out the study (uni‐temporal techniques). This methodology is based on the contextual classification of the fraction images obtained after applying SMA to the original post‐fire image. The results showed that the proposed method, using only one image acquired post‐fire, could accurately identify the burned surface area (Kappa coefficient>0.8). The spatial resolution of the satellite images had practically no influence on the accuracy of the burned area estimate but did affect the possibility of detecting areas inside the perimeter of the burned area which were only slightly damaged.     

Depth‐variant spectral characteristics of submersed aquatic vegetation detected by Landsat 7 ETM+

Spectral variations along depth profiles were compared using two subsets of a Landsat 7 Enhanced Thematic Mapper (ETM+) scene to test the difference between submersed aquatic vegetation (SAV) and non‐vegetated bare substrate in their depth‐induced spectral variation. Field‐surveyed water depth and SAV cover along transects were overlaid with the satellite image of Lake Pontchartrain, LA, USA. Digital numbers on the survey transects for each band and for band ratios were correlated with depth and vegetation cover. Band 1/band 3 correlated well with depth in both SAV and bare substrates, indicating that this ratio least reflects the effect of SAV. The ratio of bands 2 and 1 correlated best with vegetation cover within the shallow estuarine waters. Correlations between depth and the ratio of band 2/band3 showed contrasting results between the two substrate types (SAV and bare), suggesting that the depth‐induced variations in the band ratio can be used as indicators of SAV.     

Biophysical modelling and NDVI time series to project near‐term forage supply: spectral analysis aided by wavelet denoising and ARIMA modelling

Point‐based biophysical simulation of forage production coupled with 1‐km AVHRR NDVI data was used to determine the feasibility of projecting forage conditions 84 days into the future to support stocking decision making for livestock production using autoregressive integrated moving average (ARIMA) with Box and Jenkins methodology. The study was conducted at three highly contrasting ecosystems in South Texas over the period 1989–2000. Wavelet transform was introduced as a mathematical tool to denoise the NDVI time series. The simulated forage production, NDVI and denoised NDVI (DeNDVI) were subject to spectral decomposition for the detection of periodicities. Spectral analysis revealed bimodal vegetation growth patterns in Southwestern Texas. A yearly cycle (364 days) of peak vegetation production was detected for the three study sites, another peak forage production was revealed by spectral analysis at 182 days following the first peak in vegetation production. A similar trend was found for the NDVI imageries sensing the study sites. Wavelet denoising of NDVI signal was effective in revealing clear periodicities in one study site where maximum variability of NDVI was noted.

The Box and Jenkins ARIMA modelling approach was used as a forecasting method for near‐term forage production to assist range managers in proactive operational stocking decisions to mitigate drought risk. Using denoised NDVI provided forage projections with the lowest standard error prediction (SEP) throughout the forecast 84‐day periods. However, acceptable SEP was only achieved up to 6 weeks into a projection for the forage‐only based forecasts. The ARIMA forecasting methodology appears to offer a new approach to help managers of livestock production through the creation of near real‐time early warning systems. Using satellite‐derived NDVI data as a covariate improved the forecast quality and reduced the standard error of forecast in three highly contrasting sites. Denoising the NDVI data using wavelet methods further improved the forecast quality in all study sites.

The integration of AVHRR NDVI data and biophysical simulation of forage production appears a promising approach for assisting decision makers in a positive manner by assessing forage conditions in response to emerging weather conditions and near real‐time projection of available forage for grazing animals.     

Comparative analysis of SRTM–NED vegetation canopy height to LIDAR‐derived vegetation canopy metrics

Vegetation canopy heights derived from the SRTM 30 m grid DEM minus USGS National Elevation Data (NED) DTM were compared to three vegetation metrics derived from a medium footprint LIDAR data (LVIS) for the US Sierra Nevada forest in California. Generally the SRTM minus NED was found to underestimate the vegetation canopy height. Comparing the SRTM–NED‐derived heights as a function of the canopy percentile height (shape/vertical structure) derived from LVIS, the SRTM SAR signal was found to penetrate, on average, into about 44% of the canopy and 85% after adjustment of the data. On the canopy type analysis, it was found that the SRTM phase scattering centres occurred at 60% for red fir, 53% for Sierra mixed conifer, 50% for ponderosa pine and 50% for montane hardwood‐conifer. Whereas analysing the residual errors of the SRTM–NED minus the LVIS‐derived canopy height as a function of LVIS canopy height and cover it was observed that the residuals generally increase with increasing canopy height and cover. Likewise, the behaviour of the RMSE as a function of canopy height and cover was observed to initially increase with canopy height and cover but saturates at 50 m canopy height and 60% canopy cover. On the other hand, the behaviour of the correlation coefficient as a function of canopy height and cover was found to be high at lower canopy height (<15 m) and cover (<20%) and decrease rapidly making a depression at medium canopy heights (>15 m and <50 m) and cover (>20% and <50%). It then increases with increasing canopy height and cover yielding a plateau at canopies higher than 50 m and cover above 70%.     

Fast fractal image compression by pixel peaks and valleys classification

To overcome the long-time consuming problem of the existing fractal image compression methods, combining with the characteristics of fractal image encoding, this paper selects the number and the positions of both pixel peaks and valleys in row direction within an image block as the classification features. And a three-layer tree classifier which provides a stepwise precise classification is utilized. The presented classification method has the advantages such as simple principle, convenient implementation and more accurate classification. Experiments show the validity of the presented approach in improving fractal encoding speed and holding the quality of the reconstructed image.     

High dynamic adaptive mobility network model and performance analysis

Since mobile networks are not currently deployed on a large scale, research in this area is mostly by simulation. Among other simulation parameters, the mobility model plays a very important role in determining the protocol performance in MANET. Based on random direction mobility model, a high dynamic adaptive mobility network model is proposed in the paper. The algorithms and modeling are mainly studied and explained in detail. The technique keystone is that normal distribution is combined with uniform distribution and inertial feedback control is combined with kinematics, through the adaptive control on nodes speed and prediction tracking on nodes routes, an adaptive model is designed, which can be used in simulations to produce realistic and dynamic network scenarios. It is the adaptability that nodes mobile parameters can be adjusted randomly in threedimensional space. As a whole, colony mobility can show some rules. Such random movement processes as varied speed and dwells are simulated realistically. Such problems as sharp turns and urgent stops are smoothed well. The model can be adapted to not only common dynamic scenarios, but also high dynamic scenarios. Finally, the mobility model performance is analyzed and validated based on random dynamic scenarios simulations.     

How to account for endmember variability in spectral mixture analysis of night-time light imagery?

ABSTRACT

Night-Time light imagery has become a very popular data source for monitoring the intensity of human activity in urban environments. Subpixel information is required in many applications, however, the widely used low-spatial-resolution night-time light imagery suffers from the mixed-pixel problem. In this paper, using the Visible Infrared Imaging Radiometer Suite (VIIRS) data, we presented the first spectral mixture analysis (SMA) on night-time light imagery. Specifically, we proposed to define two endmembers (light and dark) for endmember selection. In order to address the severe endmember variability problem caused by the various light sources and intensities, we adopted the Bayesian SMA (BSMA) method which is based on Bayes’ theorem. The results indicate that the approach can obtain subpixel light fraction accurately with an overall root mean square error (RMSE) of 0.17 and a coefficient of determination (R²) of 0.95. Although BSMA achieved similar results with the traditional linear SMA, BSMA allows one to determine the uncertainty of the estimated fraction as a distribution.     

Relationships between tree growth and NDVI of grassland in the semi‐arid grassland of north China

This study explores the relationships between the Normalized Difference Vegetation Index (NDVI), recorded above‐ground grass biomass and tree‐ring width index of relict Meyer spruce (Picea meyeri Rehd. et Wils.) forest in the typical steppe, north China. The average NDVI in May, June and August derived from an area of 0.5°×0.5° shows a large correlation with measured above‐ground production, indicating that NDVI can reflect the approximate variability of above‐ground biomass in the typical steppe. The integrated NDVI from 20 May to 10 July also exhibits high agreement with tree‐ring width series of Meyer spruce from 1982 to 1994, which is attributed to their common response to precipitation in the previous August–October and current May. This study provides a basis for linking remotely sensed NDVI of grassland to tree growth in semi‐arid grassland.     

Reply to ‘Feasibility of neural network approach in spectral mixture analysis of reflectance spectra’

Neural Networks (NN) have proliferated during recent years, and are widely used in the scientific environment, particularly providing interpretation of results acquired by spectroscopic techniques. Separately and independently, these results were historically analysed and interpreted with ‘classical techniques’, derived from statistical formulations. The purpose of this reply is to analyse under what conditions NN methods have a better performance than the statistical methods, when it is necessary to process a spectrum obtained by a linear spectroscopic technique. The use of Neural Networks methods instead of purely statistical methods for linear spectra analysis and interpretation is discussed.     

Correlation between potato yield and MODIS‐derived vegetation indices

Monitoring of crop growth and forecasting its yield well before harvest is very important for crop and food management. Remote sensing images are capable of identifying crop health, as well as predicting its yield. Vegetation indices (VIs), such as the normalized difference vegetation index (NDVI), leaf area index (LAI) and fraction of photosynthetically active radiation (fPAR) calculated from remotely sensed data have been widely used to monitor crop growth and to predict crop yield. This study used 8 day TERRA MODIS reflectance data of 500 m resolution for the years 2005 to 2006 to estimate the yield of potato in the Munshiganj area of Bangladesh. The satellite data has been validated using ground truth data from fields of 50 farmers. Regression models are developed between VIs and field level potato yield for six administrative units of Munshiganj District. The yield prediction equations have high coefficients of correlation (R ²) and are 0.84, 0.72 and 0.80 for the NDVI, LAI and fPAR, respectively. These equations were validated by using data from 2006 to 2007 seasons and found that an average error of estimation is about 15% for the study region. It can be concluded that VIs derived from remote sensing can be an effective tool for early estimation of potato yield.     

A pixel‐based method to estimate urban compactness and its preliminary application

A pixel‐based method was designed to estimate urban compactness. Remote sensing technology was used to extract urban land use, while urban pixels in the classification image were directly used to estimate urban compactness. The method was based on the assumption that an urban area in circular form was the most compact. Forty virtual circular cities with different radii were fabricated in order to derive a formula to quantify urban compactness and a curve function in the form of y = 6.9351x ^2.0039 can thus be derived by using two data: average radius and area. Obviously, this curve can be applied to countless virtual cities in circular form of the same compactness but with different radii. Actually, for each real city in the classification image, specific urban area and average distance can be calculated directly. For every specific data pairs being thus derived, an exclusive curve function in the general form of y = 6.9351x^D can be found. Our studies show that the value of D in the curve function could be used to quantify urban compactness.     

An adaptive speech source separation algorithm under overcomplete-cases using Laplacian mixture modeling for mixture matrix estimation by adaptive EM-type algorithm in wavelet packet domain

Speech process has benefited a great deal from the wavelet transforms. Wavelet packets decompose signals in to broader components using linear spectral bisecting. In this paper, mixtures of speech signals are decomposed using wavelet packets, the phase difference between the two mixtures are investigated in wavelet domain. In our method Laplacian Mixture Model (LMM) is defined. An Expectation Maximization (EM) algorithm is used for training of the model and calculation of model parameters which is the mixture matrix. And then we compare estimation of mixing matrix by LMM-EM with different wavelets. And then we use adaptive algorithm in each wavelet packet for speech separation and we see better results are obtained. Therefore individual speech components of speech mixtures are separated.     

Sub‐pixel reflectance unmixing in estimating vegetation water content and dry biomass of corn and soybeans cropland using normalized difference water index (NDWI) from satellites

Estimating vegetation cover, water content, and dry biomass from space plays a significant role in a variety of scientific fields including drought monitoring, climate modelling, and agricultural prediction. However, getting accurate and consistent measurements of vegetation is complicated very often by the contamination of the remote sensing signal by the atmosphere and soil reflectance variations at the surface. This study used Landsat TM/ETM+ and MODIS data to investigate how sub‐pixel atmospheric and soil reflectance contamination can be removed from the remotely sensed vegetation growth signals. The sensitivity of spectral bands and vegetation indices to such contamination was evaluated. Combining the strengths of atmospheric models and empirical approaches, a hybrid atmospheric correction scheme was proposed. With simplicity, it can achieve reasonable accuracy in comparison with the 6S model. Insufficient vegetation coverage information and poor evaluation of fractional sub‐pixel bare soil reflectance are major difficulties in sub‐pixel soil reflectance unmixing. Vegetation coverage was estimated by the Normalized Difference Water Index (NDWI). Sub‐pixel soil reflectance was approximated from the nearest bare soil pixel. A linear reflectance mixture model was employed to unmix sub‐pixel soil reflectance from vegetation reflectance. Without sub‐pixel reflectance contamination, results demonstrate the true linkage between the growth of sub‐pixel vegetation and the corresponding change in satellite spectral signals. Results suggest that the sub‐pixel soil reflectance contamination is particularly high when vegetation coverage is low. After unmixing, the visible and shortwave infrared reflectances decrease and the near‐infrared reflectances increase. Vegetation water content and dry biomass were estimated using the unmixed vegetation indices. Superior to the NDVI and the other NDWIs, the SWIR (1650 nm) band‐based NDWI showed the best overall performance. The use of the NIR (1240 nm), which is a unique band of MODIS, was also discussed.