Mapping plant communities in a local Arctic landscape applying a scanned infrared aerial photograph in a geographical information system

The vegetation of a 5km2 area in front of the Midtre Lovenbreen glacier, Northwest Spitsbergen, Svalbard, was mapped on the scale 1 10000. The main aim of the study was to develop a new method of vegetation classification based on a probability model, and apply the method on a digitized aerial colour infrared (CIR) photograph with a better ground resolution than provided by the Landsat and SPOT satellites. Large-scale data from different sources such as the CIR-aerial photograph, information layers derived from a digital elevation model (DEM) and vegetation sampling in the field have been integrated in a GIS. Probability models build the links between GIS data layers and plant communities resulting from classification of field data. Eight plant communities were defined by means of vegetation data and mapped automatically by classification of the CIR-photograph. Based on the probability model, maps were produced showing the actual and potential distribution of plant communities. The accuracy of the vegetation map was improved by including additional information from the DEM.     

Classification of Amazonian primary rain forest vegetation using Landsat ETM+ satellite imagery

A lack of spatially and thematically accurate vegetation maps complicates conservation and management planning, as well as ecological research, in tropical rain forests. Remote sensing has considerable potential to provide such maps, but classification accuracy within primary rain forests has generally been inadequate for practical applications. Here we test how accurately floristically defined forest types in lowland tropical rain forests in Peruvian Amazonia can be recognized using remote sensing data (Landsat ETM+ satellite image and STRM elevation model). Floristic data and a vegetation classification with four forest classes were available for eight line transects, each 8 km long, located in an area of ca 800 km². We compared two sampling unit sizes (line transect subunits of 200 and 500 m) and several image feature combinations to analyze their suitability for image classification. Mantel tests were used to quantify how well the patterns in elevation and in the digital numbers of the satellite image correlated with the floristic patterns observed in the field. Most Mantel correlations were positive and highly significant. Linear discriminant analysis was used first to build a function that discriminates between forest classes in the eight field-verified transects on the basis of remotely sensed data, and then to classify those parts of the line transects and the satellite image that had not been visited in the field. Classification accuracy was quantified by 8-fold crossvalidation. Two of the tierra firme (non-inundated) forest types were combined because they were too often misclassified. The remaining three forest types (inundated forest, terrace forest and Pebas formation/intermediate tierra firme forest) could be separated using the 500-m sampling units with an overall classification accuracy of 85% and a Kappa coefficient of 0.62. For the 200-m sampling units, the classification accuracy was clearly lower (71%, Kappa 0.35). The forest classification will be used as habitat data to study wildlife habitat use in the same area. Our results show that remotely sensed data and relatively simple classification methods can be used to produce reasonably accurate forest type classifications, even in structurally homogeneous primary rain forests.     

Sensitivity of vegetation phenology detection to the temporal resolution of satellite data

Vegetation phenology derived from satellite data has increasingly received attention for applications in environmental monitoring and modelling. The accuracy of phenological estimates, however, is unknown at the regional and global level because field validation data are insufficient. To assess the accuracy of satellite‐derived phenology, this study investigates the sensitivity of phenology detection to both the temporal resolution of sampling and the number of consecutive missing values (usually representing cloud cover) in the time series of satellite data. To do this, time series of daily vegetation index data for various ecosystems are modelled and simulated using data from Moderate‐Resolution Imaging Spectroradiometer (MODIS) data. The annual temporal data are then fitted using piecewise logistic functions, which are employed to calculate curvature change rate for detecting phenological transition dates. The results show that vegetation phenology can be estimated with a high precision from time series with temporal resolutions of 6–16 days even if daily data contains some uncertainties. If the temporal resolution is no coarser than 16 days for time series sampled using an average composite, the absolute errors are less than 3 days. On the other hand, the phase shift of temporal sampling is shown to have limited impacts on phenology detection. However, the accuracy of phenology detection may be reduced greatly if missing values in the time series of 16‐day MODIS data occur around the onsets of phenological transition dates. Even so, the probability that the absolute error in phenological estimates is greater than 5 days is less than 4% when only one period is missing in the time series of 16‐day data during vegetation growing seasons; this probability increases to 20% if there are two consecutive missing values.     

Predicting aboveground biomass in Arctic landscapes using very high spatial resolution satellite imagery and field sampling

Remote sensing based biomass estimates in Arctic areas are usually produced using coarse spatial resolution satellite imagery, which is incapable of capturing the fragmented nature of tundra vegetation communities. We mapped aboveground biomass using field sampling and very high spatial resolution (VHSR) satellite images (QuickBird, WorldView-2 and WorldView-3) in four different Arctic tundra or peatland sites with low vegetation located in Russia, Canada, and Finland. We compared site-specific and cross-site empirical regressions. First, we classified species into plant functional types and estimated biomass using easy, non-destructive field measurements (cover, height). Second, we used the cover/height-based biomass as the response variable and used combinations of single bands and vegetation indices in predicting total biomass. We found that plant functional type biomass could be predicted reasonably well in most cases using cover and height as the explanatory variables (adjusted R² 0.21–0.92), and there was considerable variation in the model fit when the total biomass was predicted with satellite spectra (adjusted R² 0.33–0.75). There were dissimilarities between cross-site and site-specific regression estimates in satellite spectra based regressions suggesting that the same regression should be used only in areas with similar kinds of vegetation. We discuss the considerable variation in biomass and plant functional type composition within and between different Arctic landscapes and how well this variation can be reproduced using VHSR satellite images. Overall, the usage of VHSR images creates new possibilities but to utilize them to full potential requires similarly more detailed in-situ data related to biomass inventories and other ecosystem change studies and modelling.     

A double-sampling approach to deriving training and validation data for remotely-sensed vegetation products

The need for large sample sizes to train, calibrate, and validate remote-sensing products has driven an emphasis towards rapid, and in many cases qualitative, field methods. Double-sampling is an option for calibrating less precise field measurements with data from a more precise method collected at a subset of sampling locations. While applicable to the creation of training and validation datasets for remote-sensing products, double-sampling has rarely been used in this context. Our objective was to compare vegetation indicators developed from a rapid qualitative (i.e. ocular estimation) field protocol with the quantitative field protocol used by the Bureau of Land Management’s Assessment, Inventory and Monitoring (AIM) programme to determine whether double-sampling could be used to adjust the qualitative estimates to improve the relationship between rapidly collected field data and high-resolution satellite imagery. We used beta regression to establish the relationship between the quantitative and qualitative estimates of vegetation cover from 50 field sites in the Piceance Basin of northwestern Colorado, USA. Using the defined regression models for eight vegetation indicators we adjusted the qualitative estimates and compared the results, along with the original measurements, to 5 m-resolution RapidEye satellite imagery. We found good correlation between quantitative and ocular estimates for dominant site components such as shrub cover and bare ground, but low correlations for minor site components (e.g. annual grass cover) or indicators where observers were required to estimate over multiple life forms (e.g. total canopy cover). Using the beta-regression models to adjust the qualitative estimates with the quantitative data significantly improved correlation with the RapidEye imagery for most indicators. As a means of improving training data for remote-sensing projects, double-sampling should be used where a strong relationship exists between quantitative and qualitative field techniques. Accordingly, ocular techniques should be used only when they can generate reliable estimates of vegetation cover.     

Mapping vegetation in a heterogeneous mountain rangeland using landsat data: an alternative method to define and classify land-cover units

Three major problems are faced when mapping natural vegetation with mid-resolution satellite images using conventional supervised classification techniques: defining the adequate hierarchical level for mapping; defining discrete land cover units discernible by the satellite; and selecting representative training sites. In order to solve these problems, we developed an approach based on the: (1) definition of ecologically meaningful units as mosaics or repetitive combinations of structural types, (2) utilization of spectral information (indirectly) to define the units, (3) exploration of two alternative methods to classify the units once they are defined: the traditional, Maximum Likelihood method, which was enhanced by analyzing objective ways of selecting the best training sites, and an alternative method using Discriminant Functions directly obtained from the statistical analysis of signatures. The study was carried out in a heterogeneous mountain rangeland in central Argentina using Landsat data and 251 field sampling sites. On the basis of our analysis combining terrain information (a matrix of 251 stands×14 land cover attributes) and satellite data (a matrix of 251 stands×8 bands), we defined 8 land cover units (mosaics of structural types) for mapping, emphasizing the structural types which had stronger effects on reflectance. The comparison through field validation of both methods for mapping units showed that classification based on Discriminant Functions produced better results than the traditional Maximum Likelihood method (accuracy of 86% vs. 78%).     

Relationship between solar radiation and surface distribution of vegetation in Fildes Peninsula and Ardley Island,Maritime Antarctica

The aim of this work is to produce a simplified vegetation map of ice-free areas of the Fildes Peninsula (FP) and Ardley Island (AI) thought object-oriented classification using a QuickBird satellite image and to evaluate the influence of the global solar radiation (GSR) over the vegetation distribution. The vegetation data were generated from multiresolution segmentation using the panchromatic and infrared layers, and for the classification we calculated the normalized vegetative difference index (NVDI) and the green NVDI. Two classes were created – Lichen and Moss Cushion SubFormation and Moss Subformation – with 48 vegetation samples collected on surveys during the austral summers of 2008 and 2009. We used a kappa index to evaluate the classification efficiency using 100 sampled points obtained in austral summer of 2013. The GSR was estimated, and in order to evaluate the effect of meteorological phenomena and cloudless, we measured the GSR using a net radiometer model CNR4 installed in FP between 2014 and 2016. The estimate of GSR was done for seasons of 2015, in order to estimate the light compensation point and the saturation point for the plant communities in FP and AI. The kappa index was 0.73 and the global accuracy was 0.78, showing consistency between the classification and ground truth. The area was covered by vegetation in FP was 16.7% and in AI is 59.1%. The vegetation cover is distributed differently at FP and AI, and our results suggest GSR plays an important role in vegetation distribution and these tendencies could be related to greater GSR demand by mosses when compared to lichens.     

Relaxation vegetation index in non-linear modelling of ground plant cover by satellite remote-sensing data

Abstract

In order to obtain a model equation for the calculation of percentage plant cover by multi-spectral radiances remotely-sensed by satellites, a regression procedure is used to connect space remote-sensing data to ground plant cover measurement. A traditional linear regression model using the normalized difference vegetation index (NDVI) is examined by remote-sensing data of the SPOT satellite and ground measurement of LCTA project for a test site at Hohenfels. Germany. A relaxation vegetation index (RVI) is proposed in a non-linear regression modelling to replace the NDVI in linear regression modelling to get a better calculation of percentage plant cover. The definition of the RVI is

where X _i is raw remote-sensing data in channel i. Using the RVI, the correlation coefficient between calculated and observed percentage plant cover for a test scene in 1989 reaches 0·9 while for the NDVI it is only 0·7; the coefficient of multiple determination R ² reaches 0·8 for the RVI while it is only 0·5 for the NDVI. Numerical testing shows that the ability of using the RVI to predict percentage plant cover by space remote-sensing data for the same scene or the scene in other years is much stronger than the NDVI.     

Multiscale assessment of green leaf cover in a semi-arid rangeland with a small unmanned aerial vehicle

Spatial variability in green leaf cover of a semi-arid rangeland was studied by comparing field measurements on 50 m crossed transects to aerial and satellite imagery. The normalized difference vegetation index was calculated for 2 cm resolution images collected with a multispectral digital camera mounted on a radio-controlled helicopter, as well as a 30 m resolution Landsat Thematic Mapper image. Variograms of green cover from these two sources show that the range of influence for spatial autocorrelation extended to a distance of approximately 200 m. Field transects that are much smaller than the extent of this spatial autocorrelation are more likely to fall within local deviations from the mean landscape condition. A sampling scheme that exceeds the spatial scale of these localized deviations is shown to reduce sample variance and require fewer sampling locations to reach a given level of measurement uncertainty. The time and cost of more spatially extensive sampling at each location may be less than deploying to a larger number of locations with smaller transects, and unmanned aerial vehicles may be a valuable tool in extending current field sampling strategies for quantifying the health of shrub-dominated rangelands.     

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

Abstract

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

Fusion of multisensor multitemporal satellite data for land cover mapping

Efficient integration of remote sensing information with different temporal, spectral and spatial resolutions is important for accurate land cover mapping. A new temporal fusion classification (TFC) model is presented for land cover classification, based on statistical fusion of multitemporal satellite images. In the proposed model, the temporal dependence of multitemporal images is taken into account by estimating transition probabilities from the change pattern of a vegetation dynamics indicator (VDI). Extension of this model is applicable to Synthetic Aperture Radar (SAR) images and integration of multisensor multitemporal satellite images, concerning both temporal attributes and reliability of multiple data sources. The feasibility of the new method is verified using multitemporal Landsat Thematic Mapper (TM) and ERS SAR satellite images, and experimental results show improved performance over conventional methods.     

Automated interpretation of sub-pixel vegetation from IRS LISS-II images

Satellite sensor data are important for monitoring and assessment of natural resources. As vegetation is one of the most valuable natural resources, automated interpretation of vegetative cover from satellite images is prerequisite for various applications and decision processes. This paper defines a system that classifies as well as interprets vegetation from satellite images automatically. The system applies a knowledge-based approach wherein features are represented by linguistic variables in terms of their fuzzy labels. The accuracy of the system has been found to be more than 95% for hard class and more than 85% in the case of sub-pixel classification. Thus, it can be concluded that the approach adopted can be utilized in developing any automated image understanding system.     

Coastal wetland vegetation classification with a Landsat Thematic Mapper image

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

Heterogeneous forest classification by creating mixed vegetation classes using EO-1 Hyperion

Hyperspectral satellite data is an efficient tool in vegetation mapping; however, previous studies indicate that classifying heterogeneous forests might be difficult. In this study, we propose a mapping method for a heterogeneous forest using the data of the Earth Observing-1 (EO-1) Hyperion supplemented by field survey. We introduced a band reduction method to raise classification accuracy of the Support Vector Machine classification algorithm and compared the results to the one reduced by principal component analysis (PCA), stepwise discriminant analysis (SDA), and the original data set. We also used a modified version of the Vegetation–Impervious–Soil model to create mixed vegetation classes consisting of the commonly mixing species in the area and classified them using Decision Tree classification method. We managed to achieve 84.28% approximately using our band reduction method which is 2.36% increase compared to PCA (81.92%), 1.43% compared to the SDA (82.85%), and 7.61% compared to the original data set (76.67%). Introducing the mixed vegetation classes raised the overall accuracy even higher (85.79%).     

Mapping percentage tree cover from Envisat MERIS data using linear and nonlinear techniques

The aim of this study was to predict percentage tree cover from Envisat Medium Resolution Imaging Spectrometer (MERIS) imagery with a spatial resolution of 300 m by comparing four common models: a multiple linear regression (MLR) model, a linear mixture model (LMM), an artificial neural network (ANN) model and a regression tree (RT) model. The training data set was derived from a fine spatial resolution land cover classification of IKONOS imagery. Specifically, this classification was aggregated to predict percentage tree cover at the MERIS spatial resolution. The predictor variables included the MERIS wavebands plus biophysical variables (the normalized difference vegetation index (NDVI), leaf area index (LAI), fraction of photosynthetically active radiation (fPAR), fraction of green vegetation covering a unit area of horizontal soil (fCover) and MERIS terrestrial chlorophyll index (MTCI)) estimated from the MERIS data. An RT algorithm was the most accurate model to predict percentage tree cover based on the Envisat MERIS bands and vegetation biophysical variables. This study showed that Envisat MERIS data can be used to predict percentage tree cover with considerable spatial detail. Inclusion of the biophysical variables led to greater accuracy in predicting percentage tree cover. This finer-scale depiction should be useful for environmental monitoring purposes at the regional scale.     

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.     

Combining vegetation indices, constrained ordination and fuzzy classification for mapping semi-natural vegetation units from hyperspectral imagery

Vegetation mapping of plant communities at fine spatial scales is increasingly supported by remote sensing technology. However, combining ecological ground truth information and remote sensing datasets for mapping approaches is complicated by the complexity of ecological datasets. In this study, we present a new approach that uses high spatial resolution hyperspectral datasets to map vegetation units of a semiarid rangeland in Central Namibia. Field vegetation surveys provide the input to the workflow presented in this study. The collected data were classified by hierarchical cluster analysis into seven vegetation units that reflect different ecological states occurring in the study area. Spectral indices covering vegetation and soil characteristics were calculated from hyperspectral remote sensing imagery and used as environmental variables in a constrained ordination by applying redundancy analysis (RDA). The resulting statistical relationships between vegetation data and spectral indices were transferred into images of ordination axes, which were subsequently used in a supervised fuzzy c-means classification approach relying on a k-NN distance metric. Membership images for each vegetation unit as well as a confusion image of the classification result allowed a sound ecological interpretation of the resulting hard classification map. Classification results were validated with two independent reference datasets. For an internal and external validation dataset, overall accuracy reached 98% and 64% with kappa values of 0.98 and 0.53, respectively. Critical steps during the mapping workflow were highlighted and compared with similar mapping approaches.     

Estimating forest canopy density of tropical mixed deciduous vegetation using Landsat data: a comparison of three classification approaches

Although a number of image classification approaches are available to estimate forest canopy density (FCD) using satellite data, assessment of their relative performances with tropical mixed deciduous vegetation is lacking. This study compared three image classification approaches – maximum likelihood classification (MLC), multiple linear regression (MLR) and FCD Mapper – in estimating the FCD of mixed deciduous forest in Myanmar. The application of MLC and MLR was based on spectral reflectance of vegetation, whereas FCD Mapper was operated on integrating the biophysical indices derived from the reflectance of the vegetation. The FCD was classified into four categories: closed canopy forest (CCF; FCD ≥ 70%), medium canopy forest (MCF; 40% ≥ FCD < 70%), open canopy forest (OCF; 10% ≥ FCD < 40%) and non-forest (NF; FCD < 10%). In the three classification approaches, producer's and user's accuracies were higher for more homogeneous vegetation such as NF and CCF than for heterogeneous vegetation density (VD) such as OCF and MCF. FCD Mapper produced the best overall accuracy and kappa coefficient. This study revealed that only spectral reflectance is not enough to get good results in estimating FCD in tropical mixed deciduous vegetation. This study indicates that FCD Mapper, an inexpensive approach because it requires only validation data and thus saves time, can be applied to monitor tropical mixed deciduous vegetation over time at lower cost than alternative methods.     

A geometric approach to a mixed pixel problem: Detecting subpixel woody vegetation

In digital satellite imagery, small fragments of woody vegetation are difficult to detect because they frequently are smaller than the pixel size and are mixed with other land cover classes. A method for detecting subpixel woody vegetation, which analyzes mixture phenomena at the individual pixel level, is presented. This method relies on a moving window to collect training sets for adjacent land cover. In order to locate pixels of interest and to decrease noise, image-derived masks are integrated with the original digital imagery in a geocoded information system. A rule-based scheme is employed to organize relative spatial and spectral information into classification decision procedures. Tests using simulated multispectral and panchromatic SPOT HRV imagery of lowland Britain have shown that the developed method discriminates significantly more woody vegetation than standard multispectral classification.     

Mapping vegetation community types in a highly disturbed landscape: integrating hierarchical object-based image analysis with lidar-derived canopy height data

Focusing on the semi-arid and highly disturbed landscape of San Clemente Island (SCI), California, we test the effectiveness of incorporating a hierarchical object-based image analysis (OBIA) approach with high-spatial resolution imagery and canopy height surfaces derived from light detection and ranging (lidar) data for mapping vegetation communities. The hierarchical approach entailed segmentation and classification of fine-scale patches of vegetation growth forms and bare ground, with shrub species identified, and a coarser-scale segmentation and classification to generate vegetation community maps. Such maps were generated for two areas of interest on SCI, with and without vegetation canopy height data as input, primarily to determine the effectiveness of such structural data on mapping accuracy. Overall accuracy is highest for the vegetation community map derived by integrating airborne visible and near-infrared imagery having very high spatial resolution with the lidar-derived canopy height data. The results demonstrate the utility of the hierarchical OBIA approach for mapping vegetation with very high spatial resolution imagery, and emphasizes the advantage of both multi-scale analysis and digital surface data for accurately mapping vegetation communities within highly disturbed landscapes.