Hyperspectral Satellite Imagery Research Articles

Peatland leaf-area index and biomass estimation with ultra-high resolution remote sensing

ABSTRACT There is fine-scale spatial heterogeneity in key vegetation properties including leaf-area index (LAI) and biomass in treeless northern peatlands, and hyperspectral drone data with high spatial and spectral resolution could detect the spatial patterns with high accuracy. However, the advantage of hyperspectral drone data has not been tested in a multi-source remote sensing approach (i.e. inclusion of multiple different remote sensing datatypes); and overall, sub-meter-level leaf-area index (LAI) and biomass maps have largely been absent. We evaluated the detectability of LAI and biomass patterns at a northern boreal fen (Halssiaapa) in northern Finland with multi-temporal and multi-source remote sensing data and assessed the benefit of hyperspectral drone data. We measured vascular plant percentage cover and height as well as moss cover in 140 field plots and connected the structural information to measured aboveground vascular LAI and biomass and moss biomass with linear regressions. We predicted both total and plant functional type (PFT) specific LAI and biomass patterns with random forests regressions with predictors including RGB and hyperspectral drone (28 bands in a spectral range of 500–900 nm), aerial and satellite imagery as well as topography and vegetation height information derived from structure-from-motion drone photogrammetry and aerial lidar data. The modeling performance was between moderate and good for total LAI and biomass (mean explained variance between 49.8 and 66.5%) and variable for PFTs (0.3–61.6%). Hyperspectral data increased model performance in most of the regressions, usually relatively little, but in some of the regressions, the inclusion of hyperspectral data even decreased model performance (change in mean explained variance between −14.5 and 9.1%-points). The most important features in regressions included drone topography, vegetation height, hyperspectral and RGB features. The spatial patterns and landscape estimates of LAI and biomass were quite similar in regressions with or without hyperspectral data, in particular for moss and total biomass. The results suggest that the fine-scale spatial patterns of peatland LAI and biomass can be detected with multi-source remote sensing data, vegetation mapping should include both spectral and topographic predictors at sub-meter-level spatial resolution and that hyperspectral imagery gives only slight benefits.

Dictionary learning for clustering on hyperspectral images

Dictionary learning and sparse coding have been widely studied as mechanisms for unsupervised feature learning. Unsupervised learning could bring enormous benefit to the processing of hyperspectral images and to other remote sensing data analysis because labelled data are often scarce in this field. We propose a method for clustering the pixels of hyperspectral images using sparse coefficients computed from a representative dictionary as features. We show empirically that the proposed method works more effectively than clustering on the original pixels. We also demonstrate that our approach, in certain circumstances, outperforms the clustering results of features extracted using principal component analysis and non-negative matrix factorisation. Furthermore, our method is suitable for applications in repetitively clustering an ever-growing amount of high-dimensional data, which is the case when working with hyperspectral satellite imagery.

Smart feature extraction and classification of hyperspectral images based on convolutional neural networks

Hyperspectral satellite imagery (HSI) is an advanced technology for object detection because it provides a large amount of information. Thus, the classification of HSIs is very complicated, so the methods of reducing spectral or spatial information generally degrade the quality of classification. In order to solve this problem and guarantee faster and more efficient processing, we propose a smart feature extraction (SFE) and classification by convolutional neural network (2D-CNN) method made up of two parts. The first consists in reducing spectral information by a probabilistic method based on the Softmax function. The second is classification by processing batches of data in the proposed CNN network. The method was tested on two public hyperspectral images (Indian Pines and SalinasA) to prove its effectiveness in increasing classification accuracy and reducing computing time.

Geometrical Approximated Principal Component Analysis for Hyperspectral Image Analysis

Principal Component Analysis (PCA) is a method based on statistics and linear algebra techniques, used in hyperspectral satellite imagery for data dimensionality reduction required in order to speed up and increase the performance of subsequent hyperspectral image processing algorithms. This paper introduces the PCA approximation method based on a geometric construction approach (gaPCA) method, an alternative algorithm for computing the principal components based on a geometrical constructed approximation of the standard PCA and presents its application to remote sensing hyperspectral images. gaPCA has the potential of yielding better land classification results by preserving a higher degree of information related to the smaller objects of the scene (or to the rare spectral objects) than the standard PCA, being focused not on maximizing the variance of the data, but the range. The paper validates gaPCA on four distinct datasets and performs comparative evaluations and metrics with the standard PCA method. A comparative land classification benchmark of gaPCA and the standard PCA using statistical-based tools is also described. The results show gaPCA is an effective dimensionality-reduction tool, with performance similar to, and in several cases, even higher than standard PCA on specific image classification tasks. gaPCA was shown to be more suitable for hyperspectral images with small structures or objects that need to be detected or where preponderantly spectral classes or spectrally similar classes are present.

Spectral enhancement of Landsat OLI images by using Hyperion data: a comparison between multilayer perceptron and radial basis function networks

The deactivation of Earth Observing-1 satellite has resulted in the termination of capturing Hyperion data as a unique source of hyperspectral satellite imagery. These images also were collected through an on-demand service and thereby are not available for the entire Earth’s surface. The Operational Land Imager (OLI) sensor, on the other hand, provides a free source of multi-spectral images with global coverage. Recognized these facts, the aim of this paper is to enhance the spectral resolution of OLI images by using existing Hyperion imageries to generate a high spectral resolution image for a desired date and site. This was conducted through the artificial neural network (ANN). To find the suitable ANN, we compared the performance of multilayer perceptron (MLP) and radial basis function (RBF) networks for spectral enhancement. The research obtained two Hyperion and OLI images covering West Region of Tehran, Iran, on 4 January 2016. From 242 original Hyperion spectral bands, we selected 31 bands to reproduce from OLI spectral bands. These were determined through visual inspection, principal component analysis and Pearson’s correlation test. The MLP and RBF networks were generated based on the OLI bands 1–7 and per 31 Hyperion bands as input and output layers respectively. The comparison between the spectral bands of spectra-enhanced image and original Hyperion data indicated a good agreement (0.884 > R2 > 0.692). This study also found MLP network delivered higher accuracy against RBF network for spectral enhancement. The spectra-enhanced image can be used in studies with the need of images with continuous spectral bands.

Геоинформационное обеспечение геологоразведки месторождений руд цветных металлов

Geoinformation technologies are applied, without exaggeration, in all spheres of human activity, and the geological industry is no exception. Geoinformation systems in geology help to solve a number of practical problems. Geologists specializing in mineral exploration use a variety of data sets to search for new, cost-effective deposits, from geological maps, hyperspectral aerial photographs and multispectral satellite imagery to databases of various formats. The geographic information system is an ideal platform for combining such heterogeneous information and its subsequent analysis. The article describes the experience of the national mining company of the Republic of Azerbaijan CJSC AzerGold in the use of geographic information systems in the exploration of non-ferrous metal deposits using the example of the development of the Filizchay and Aghyokhush deposits. The authors considered production problems, described the problems that arise at the stage of exploration and proposed methods for solving them, the need for the integrated use of information systems in exploration when searching for deposits of non-ferrous metals is noted. A database structure has been developed for the purpose of geological exploration of non-ferrous metal ores, including sections such as topography, geochemistry, geophysics, structural geology. The structure of geoinformation support for geological exploration of non-ferrous metal ores is proposed, which combines the functionality of geographic information systems (GIS) and mining geological information systems (MGIS). The stages of geoinformation processing of initial data are illustrated, on the basis of which a geological exploration strategy was formed at the Filizchay pyrite-polymetallic deposit, and then the optimal places and number of wells, coordinates, directions and angles of incidence of wells were determined quickly and with high accuracy, as well as the resource potential was specified. The role of interpolation methods in determining the primary exploration wells in the exploration of the Agyokhush gold deposit is disclosed.

HYPERSPECTRAL CHANGE DETECTION IN WETLAND AND WATER-BODY AREAS BASED ON MACHINE LEARNING

Abstract. Wetlands and water-bodies are transitional lands between terrestrial and aquatic ecosystems that give many advantages. The presence of phenomena activities and artificial human activities cause changes on the Earth’s surface. This process causes the changes in land cover type especially in wetlands and water-bodies’ area. For monitoring and assessing resources like wetlands and water-bodies, it is necessary to be aware of these changes. Change detection and attribution of wetlands and water-bodies change over time present more challenges for correctly analysing remote sensing imagery. Hyperspectral images now have potential applications in many scientific areas due to their high spectral resolution and so their good information contents. The aim of this study is to propose a procedure for determining land surface changes within the semi-arid wetland and surrounding upland areas using the new method by combining machine Learning method for detecting change using EO-1 Hyperion satellite hyperspectral imagery. The study area is Shadegan wetlands in the south-west of Iran in Khuzestan province. The most critical water resources of the province are depleted and contain unprecedented levels of toxic waste. In addition, the results show the superiority of the implemented method to extract change map with overall accuracy by a margin of nearly 94% using multi-temporal hyperspectral.

A Robust Rule-Based Ensemble Framework Using Mean-Shift Segmentation for Hyperspectral Image Classification

This paper assesses the performance of DoTRules—a dictionary of trusted rules—as a supervised rule-based ensemble framework based on the mean-shift segmentation for hyperspectral image classification. The proposed ensemble framework consists of multiple rule sets with rules constructed based on different class frequencies and sequences of occurrences. Shannon entropy was derived for assessing the uncertainty of every rule and the subsequent filtering of unreliable rules. DoTRules is not only a transparent approach for image classification but also a tool to map rule uncertainty, where rule uncertainty assessment can be applied as an estimate of classification accuracy prior to image classification. In this research, the proposed image classification framework is implemented using three world reference hyperspectral image datasets. We found that the overall accuracy of classification using the proposed ensemble framework was superior to state-of-the-art ensemble algorithms, as well as two non-ensemble algorithms, at multiple training sample sizes. We believe DoTRules can be applied more generally to the classification of discrete data such as hyperspectral satellite imagery products.

Hyperspectral Remote Sensing of Phytoplankton Species Composition Based on Transfer Learning

Phytoplankton species composition research is key to understanding phytoplankton ecological and biogeochemical functions. Hyperspectral optical sensor technology allows us to obtain detailed information about phytoplankton species composition. In the present study, a transfer learning method to inverse phytoplankton species composition using in situ hyperspectral remote sensing reflectance and hyperspectral satellite imagery was presented. By transferring the general knowledge learned from the first few layers of a deep neural network (DNN) trained by a general simulation dataset, and updating the last few layers with an in situ dataset, the requirement for large numbers of in situ samples for training the DNN to predict phytoplankton species composition in natural waters was lowered. This method was established from in situ datasets and validated with datasets collected in different ocean regions in China with considerable accuracy (R2 = 0.88, mean absolute percentage error (MAPE) = 26.08%). Application of the method to Hyperspectral Imager for the Coastal Ocean (HICO) imagery showed that spatial distributions of dominant phytoplankton species and associated compositions could be derived. These results indicated the feasibility of species composition inversion from hyperspectral remote sensing, highlighting the advantages of transfer learning algorithms, which can bring broader application prospects for phytoplankton species composition and phytoplankton functional type research.

Forest health assessment for geo-environmental planning and management in hilltop mining areas using Hyperion and Landsat data

Abstract This work focuses on assessing the health condition of the forest using Hyperspectral and Multispectral satellite imagery and validating it with field spectra data. The field-based spectroradiometer and PCE instrument were used for collection of forest health spectra and measurement of dust accumulation on leaves respectively. In the study area, mining activities have very high potential to induce forestry health and environmental problem which necessitated a proper forest health assessment and its monitoring. The result of the Discrimination analysis (21 spectral wavebands) were used for forest health classification. The result shows that healthy forest parts are found in the upper as well as the lower hilly side of Kiriburu and Meghahatuburu mines. Some healthy pixels are located within 1.5 km from mines because it was situated off the hillside. Furthermore, it also exhibits positive relation amongst different forest health class, distance from mines and foliar dust concentration. Hyperspectral (narrow-bands) Hyperion data used with Vegetation Indices (VIs) model shows better accuracy for forest health assessment (overall accuracy 81.52%, kappa statistic 0.79) than Spectral Angle Mapper (overall accuracy 79.99%, kappa statistic 0.75) as well as Support Vector Machine (overall accuracy 76.53%, kappa statistic 0.71). It was observed that the health assessment accuracy (using SVM algorithm) achieved with Hyperion bands was significantly higher than multispectral (broad-bands) Landsat-OLI data (overall accuracy 67.27%, kappa statistic 0.62). Finally, the forest health assessment results were validated by 60 field sampled spectra data obtained from spectroradiometer. The forest health map obtained provides a guideline for geo-environmental planning and management in mining proximity forest area.

Reconstructing Cloud Contaminated Pixels Using Spatiotemporal Covariance Functions and Multitemporal Hyperspectral Imagery

One of the major challenges in optical-based remote sensing is the presence of clouds, which imposes a hard constraint on the use of multispectral or hyperspectral satellite imagery for earth observation. While some studies have used interpolation models to remove cloud affected data, relatively few aim at restoration via the use of multi-temporal reference images. This paper proposes not only the use of image time-series, but also the implementation of a geostatistical model that considers the spatiotemporal correlation between them to fill the cloud-related gaps. Using Hyperion hyperspectral images, we demonstrate a capacity to reconstruct cloud-affected pixels and predict their underlying surface reflectance values. To do this, cloudy pixels were masked and a parametric family of non-separable covariance functions was automated fitted, using a composite likelihood estimator. A subset of cloud-free pixels per scene was used to perform a kriging interpolation and to predict the spectral reflectance per each cloud-affected pixel. The approach was evaluated using a benchmark dataset of cloud-free pixels, with a synthetic cloud superimposed upon these data. An overall root mean square error (RMSE) of between 0.5% and 16% of the reflectance was achieved, representing a relative root mean square error (rRMSE) of between 0.2% and 7.5%. The spectral similarity between the predicted and reference reflectance signatures was described by a mean spectral angle (MSA) of between 1° and 11°, demonstrating the spatial and spectral coherence of predictions. The approach provides an efficient spatiotemporal interpolation framework for cloud removal, gap-filling, and denoising in remotely sensed datasets.

A Deep Convolution Neural Network Method for Land Cover Mapping: A Case Study of Qinhuangdao, China

Land cover and its dynamic information is the basis for characterizing surface conditions, supporting land resource management and optimization, and assessing the impacts of climate change and human activities. In land cover information extraction, the traditional convolutional neural network (CNN) method has several problems, such as the inability to be applied to multispectral and hyperspectral satellite imagery, the weak generalization ability of the model and the difficulty of automating the construction of a training database. To solve these problems, this study proposes a new type of deep convolutional neural network based on Landsat-8 Operational Land Imager (OLI) imagery. The network integrates cascaded cross-channel parametric pooling and average pooling layer, applies a hierarchical sampling strategy to realize the automatic construction of the training dataset, determines the technical scheme of model-related parameters, and finally performs the automatic classification of remote sensing images. This study used the new type of deep convolutional neural network to extract land cover information from Qinhuangdao City, Hebei Province, and compared the experimental results with those obtained by traditional methods. The results show that: (1) The proposed deep convolutional neural network (DCNN) model can automatically construct the training dataset and classify images. This model performs the classification of multispectral and hyperspectral satellite images using deep neural networks, which improves the generalization ability of the model and simplifies the application of the model. (2) The proposed DCNN model provides the best classification results in the Qinhuangdao area. The overall accuracy of the land cover data obtained is 82.0%, and the kappa coefficient is 0.76. The overall accuracy is improved by 5% and 14% compared to the support vector machine method and the maximum likelihood classification method, respectively.

Hyperspectral satellite imagery detection of ancient raw material sources: Soft‐stone vessel production at Aqir al‐Shamoos (Oman)

AbstractSoft‐stone vessels (made of the rock chloritite, which is comprised of the mineral chlorite) are an important component of the material culture record of ancient southeast Arabia. Given the prevalence of soft‐stone vessels in domestic and funerary contexts, and the natural geological occurrence of chloritite in the region, archaeologists have long suspected that soft‐stone vessels could have been produced in United Arab Emirates (UAE) or Oman. However, over many decades of research very few pieces of worked raw material or unfinished vessel fragments have been recovered in Arabia. This changed dramatically in 2015 with the discovery of the Iron Age village of Aqir al‐Shamoos where thousands of raw, worked, and unfinished pieces attest to soft‐stone vessel manufacture. Aqir al‐Shamoos continues to yield new insights regarding soft‐stone vessel production, but the source of raw material used to craft these objects initially remained unknown as no natural chloritite deposits had been found in the immediate vicinity of the site. We present results of hyperspectral satellite imagery target detection used to search for natural surface occurrences of chloritite near Aqir al‐Shamoos. Despite limitations in carbonate and phyllosilicate mineral discrimination, field survey of the satellite detection results revealed chloritite outcrops in the ophiolite formation mountains 1.4 km northwest of Aqir al‐Shamoos near the newly discovered archaeological site of ‘Waby al‐Zady. Reflectance spectroscopy (0.35–2.5 μm) of material from multiple chloritite sources indicates that chloritite from ‘Waby al‐Zady is spectrally very similar to material at Aqir al‐Shamoos but quite different in comparison with material from other locations. These findings indicate that: (1) hyperspectral satellite imagery target detection can be useful for chloritite prospection, (2) ‘Waby al‐Zady (or somewhere very nearby) was probably the source of chloritite used at Aqir al‐Shamoos, and (3) chloritite reflection spectra may be useful for distinguishing chloritite from different sources.

Estimation of suspended particulate matter in turbid coastal waters: application to hyperspectral satellite imagery.

An empirical algorithm is proposed to estimate suspended particulate matter (SPM) ranging from 0.675 to 25.7 mg L-1 in the turbid Pearl River estuary (PRE). Comparisons between model predicted and in situ measured SPM resulted in R2s of 0.97 and 0.88 and mean absolute percentage errors (MAPEs) of 23.96% and 29.69% by using the calibration and validation data sets, respectively. The developed algorithm demonstrated the highest accuracy when compared with existing ones for turbid coastal waters. The diurnal dynamics of SPM was revealed by applying the proposed algorithm to reflectance data collected by a moored buoy in the PRE. The established algorithm was implemented to Hyperspectral Imager for the Coastal Ocean (HICO) data and the distribution pattern of SPM in the PRE was elucidated. Validation of HICO-derived reflectance data by using concurrent MODIS/Aqua data as a benchmark indicated their reliability. Factors influencing variability of SPM in the PRE were analyzed, which implicated the combined effects of wind, tide, rainfall, and circulation as the cause.

Mapping of forest alliances with simulated multi-seasonal hyperspectral satellite imagery

A consistent and hierarchical classification of vegetation, such as the U.S. National Vegetation Classification (NVC) system, supports comprehensive conservation and management of natural ecosystems. At a detailed level, the NVC alliance is defined by diagnostic species and composition. Maps at this level of classification are often produced at local to regional scales (areas <25,000 km2) with costly manual to semi-automated interpretation of high resolution imagery. The main objective of this study was to assess the effectiveness of machine learning for automated, per-pixel (30 m) mapping of forest alliances with multi-seasonal hyperspectral imagery from a future satellite mission (HyspIRI), as simulated from Airborne Visible/Infrared Imaging Spectrometer Classic (AVIRIS-C) data. The study area was the San Francisco (S.F.) Bay Area, California, where we mapped forest alliances at regional and county scales. We implemented the Support Vector Machine (SVM) classifier in a two-stage approach, first mapping regional land cover followed by forest alliances in closed-canopy tree pixels. Predictor variables were reflectance bands and hyperspectral metrics based on indices, derivatives and absorption-fitting techniques applied to reflectance spectra, with data grouped into summer and three-season (spring, summer, fall) sets. For forest alliances, hyperspectral metrics improved overall accuracy of classifications by 2.9 to 6.4% relative to classifications based on the original reflectance bands. Multi-seasonal data improved overall accuracy by 1.3 to 6.2% relative to summer-only data. Using multi-seasonal metrics, the S.F. Bay Area regional classification with 21 alliances had an overall accuracy of 65.7% (Kappa 0.63), while the Sonoma County classification with 16 alliances had an accuracy of 75.9% (Kappa 0.72). Most forest alliances had internal variation in lifeform, species and structural properties that increased within-class spectral-temporal variation and complicated discrimination. Despite this challenge, classification accuracies were similar to regional NVC alliance reference data. We conclude that a hyperspectral satellite, with its repeat and global image acquisitions, has strong potential for accurate and economical mapping and monitoring the Earth's vegetation communities.

Sensitivity analysis of pansharpening in hyperspectral change detection

Change detection (CD) is one of the most important uses of remote sensing, and it plays a key role in many applications. Satellite hyperspectral imagery has a high spectral resolution but low spatial resolution, which results in images with mixed pixels. To improve spatial resolution in hyperspectral images, multiresolution fusion techniques must be used, one which is called pansharpening (PS). This paper investigates the sensitivity and performance of CD methods by fusing Advanced Land Imager and Hyperion datasets based on a PS algorithm. Three different CD algorithms are used here for that purpose: cross-covariance (CC), cross equalization (CE), and principal component analysis (PCA). In addition, Gram-Schmidt (GS), HySure, and PCA are utilized as the PS methods of choice. The CD results obtained from both the original hyperspectral data and from the spatially fused data are compared to reveal the potential of PS in CD applications. Furthermore, the presented procedure also shows that the HySure method in particular yields good results for the CD.

Comparison of simulated hyperspectral HyspIRI and multispectral Landsat 8 and Sentinel-2 imagery for multi-seasonal, regional land-cover mapping

This study used simulated hyperspectral (HyspIRI) and multispectral (Landsat 8 OLI, Sentinel-2 MSI) satellite imagery to compare regional land-cover mapping capabilities (San Francisco Bay Area, California) within an analytical framework that included consistent reference data and classification rules. Imagery had the pixel resolution (30m) and extent (30,000km2) of a Landsat scene, with multi-seasonal (spring, summer, fall) acquisitions from year 2013. Primary study objectives were to assess differences in map accuracy related to Multiple Endmember Spectral Mixture Analysis (MESMA) and Random Forests (RF) classifiers, spectral resolution (hyperspectral vs. multispectral), and temporal resolution (multi-seasonal vs. summer). The RF classifier generally outperformed MESMA by 1.1 to 9.0% overall accuracy, with the exception of summer HyspIRI reflectance data. There were no clear patterns in accuracy when comparing HyspIRI and simulated multispectral reflectance bands with RF and MESMA classifiers. With summer data, HyspIRI had significantly higher accuracy for MESMA (+5.5 to +8.7%) and significantly lower accuracy for RF (−9.7 to −16.4%). There were no significant differences in accuracy when using multi-seasonal HyspIRI and multispectral data with RF or MESMA (<1.2% differences). There were highly significant improvements in overall accuracy (1.7 to 20.9%) with multi-season over summer-only images for all sensors, sample scales and classifiers. This result indicates that repeat image acquisitions from satellite sensors are important for land-cover classification, irrespective of sensor spectral resolution. A companion study (Clark & Kilham, 2016) that used RF with hyperspectral metrics derived from HyspIRI reflectance bands (from indices, derivatives, and absorption fitting) showed significant improvements in overall accuracy relative to map classifications in this study, which all used independent reflectance bands. These findings point to the need of additional land-cover mapping research with machine learning and hyperspectral data that span the spatial and temporal scales afforded by a satellite.

Quantifying land use/land cover spatio-temporal landscape pattern dynamics from Hyperion using SVMs classifier and FRAGSTATS®

This study aims to quantify the landscape spatio-temporal dynamics including Land Use/Land Cover (LULC) changes occurred in a typical Mediterranean ecosystem of high ecological and cultural significance in central Greece covering a period of 9 years (2001–2009). Herein, we examined the synergistic operation among Hyperion hyperspectral satellite imagery with Support Vector Machines, the FRAGSTATS® landscape spatial analysis programme and Principal Component Analysis (PCA) for this purpose. The change analysis showed that notable changes reported in the experimental region during the studied period, particularly for certain LULC classes. The analysis of accuracy indices suggested that all the three classification techniques are performing satisfactorily with overall accuracy of 86.62, 91.67 and 89.26% in years 2001, 2004 and 2009, respectively. Results evidenced the requirement for taking measures to conserve this forest-dominated natural ecosystem from human-induced pressures and/or natural hazards occurred in the area. To our knowledge, this is the first study of its kind, demonstrating the Hyperion capability in quantifying LULC changes with landscape metrics using FRAGSTATS® programme and PCA for understanding the land surface fragmentation characteristics and their changes. The suggested approach is robust and flexible enough to be expanded further to other regions. Findings of this research can be of special importance in the context of the launch of spaceborne hyperspectral sensors that are already planned to be placed in orbit as the NASA’s HyspIRI sensor and EnMAP.

Assessing the accuracy of hyperspectral and multispectral satellite imagery for categorical and Quantitative mapping of salinity stress in sugarcane fields

This study evaluates the feasibility of hyperspectral and multispectral satellite imagery for categorical and quantitative mapping of salinity stress in sugarcane fields located in the southwest of Iran. For this purpose a Hyperion image acquired on September 2, 2010 and a Landsat7 ETM+ image acquired on September 7, 2010 were used as hyperspectral and multispectral satellite imagery. Field data including soil salinity in the sugarcane root zone was collected at 191 locations in 25 fields during September 2010. In the first section of the paper, based on the yield potential of sugarcane as influenced by different soil salinity levels provided by FAO, soil salinity was classified into three classes, low salinity (1.7–3.4dS/m), moderate salinity (3.5–5.9dS/m) and high salinity (6–9.5) by applying different classification methods including Support Vector Machine (SVM), Spectral Angle Mapper (SAM), Minimum Distance (MD) and Maximum Likelihood (ML) on Hyperion and Landsat images. In the second part of the paper the performance of nine vegetation indices (eight indices from literature and a new developed index in this study) extracted from Hyperion and Landsat data was evaluated for quantitative mapping of salinity stress. The experimental results indicated that for categorical classification of salinity stress, Landsat data resulted in a higher overall accuracy (OA) and Kappa coefficient (KC) than Hyperion, of which the MD classifier using all bands or PCA (1–5) as an input performed best with an overall accuracy and kappa coefficient of 84.84% and 0.77 respectively. Vice versa for the quantitative estimation of salinity stress, Hyperion outperformed Landsat. In this case, the salinity and water stress index (SWSI) has the best prediction of salinity stress with an R2 of 0.68 and RMSE of 1.15dS/m for Hyperion followed by Landsat data with an R2 and RMSE of 0.56 and 1.75dS/m respectively. It was concluded that categorical mapping of salinity stress is the best option for monitoring agricultural fields and for this purpose Landsat data are most suitable.

Mapping of land cover in northern California with simulated hyperspectral satellite imagery

Land-cover maps are important science products needed for natural resource and ecosystem service management, biodiversity conservation planning, and assessing human-induced and natural drivers of land change. Analysis of hyperspectral, or imaging spectrometer, imagery has shown an impressive capacity to map a wide range of natural and anthropogenic land cover. Applications have been mostly with single-date imagery from relatively small spatial extents. Future hyperspectral satellites will provide imagery at greater spatial and temporal scales, and there is a need to assess techniques for mapping land cover with these data. Here we used simulated multi-temporal HyspIRI satellite imagery over a 30,000km2 area in the San Francisco Bay Area, California to assess its capabilities for mapping classes defined by the international Land Cover Classification System (LCCS). We employed a mapping methodology and analysis framework that is applicable to regional and global scales. We used the Random Forests classifier with three sets of predictor variables (reflectance, MNF, hyperspectral metrics), two temporal resolutions (summer, spring-summer-fall), two sample scales (pixel, polygon) and two levels of classification complexity (12, 20 classes). Hyperspectral metrics provided a 16.4–21.8% and 3.1–6.7% increase in overall accuracy relative to MNF and reflectance bands, respectively, depending on pixel or polygon scales of analysis. Multi-temporal metrics improved overall accuracy by 0.9–3.1% over summer metrics, yet increases were only significant at the pixel scale of analysis. Overall accuracy at pixel scales was 72.2% (Kappa 0.70) with three seasons of metrics. Anthropogenic and homogenous natural vegetation classes had relatively high confidence and producer and user accuracies were over 70%; in comparison, woodland and forest classes had considerable confusion. We next focused on plant functional types with relatively pure spectra by removing open-canopy shrublands, woodlands and mixed forests from the classification. This 12-class map had significantly improved accuracy of 85.1% (Kappa 0.83) and most classes had over 70% producer and user accuracies. Finally, we summarized important metrics from the multi-temporal Random Forests to infer the underlying chemical and structural properties that best discriminated our land-cover classes across seasons.