AVIRIS Research Articles

A novel method to simulate AVIRIS-NG hyperspectral image from Sentinel-2 image for improved vegetation/wildfire fuel mapping, boreal Alaska

• - Vegetation mapped from simulated AVIRIS-NG image offers greater details and accuracy. • − 33% improvement in the new vegetation map accuracy. • - A novel approach to generate improved vegetation maps for Alaska’s boreal domain. Detailed vegetation maps are one of the primary inputs for forest and wildfire management. Hyperspectral remote sensing is a proven technique for detailed and accurate vegetation mapping. However, the availability of recent hyperspectral imagery in Alaska is limited because of the logistics and high cost involved in its acquisition. In this study, we simulated AVIRIS-NG (Airborne Visible InfraRed Imaging Spectrometer - Next Generation) hyperspectral data from widely available Sentinel-2 multispectral data using the Universal Pattern Decomposition Method (UPDM). The UPDM is a spectral unmixing technique that uses detailed ground spectra of vegetation classes and the Spectral Response Functions of AVIRIS-NG and Sentinel-2 sensors to simulate imagery with the same number of bands and spectral resolution as an AVIRIS-NG image. We simulated three images (each covering an area of 100 km × 100 km) from two ecoregions to test portability of the approach. We collected ground spectra of vegetation and bare ground during summers (2019–2021) using a PSR+ 3500 hand-held spectroradiometer and created a spectral library for this study. The Iterative Endmember Selection (IES) algorithm was used to optimize the spectral library and to select the most representative endmembers for simulation: birch, spruce, and gravel. We validated the simulated hyperspectral imagery by comparing it with available AVIRIS-NG images. The simulated image was visually and spectrally similar to the AVIRIS-NG image (RMSE of 0.03 and 0.02 for birch and spruce spectra, respectively). We applied the Random Forest image classification model to derive detailed vegetation maps from the simulated images. Our vegetation map showed an improvement of 33% in the map accuracy compared to the LANDFIRE EVT map. This study demonstrated an efficient and cost-effective approach to derive detailed vegetation maps at the Sentinel scene scale by simulating hyperspectral images in Google’s cloud environment. It offers a novel pathway to generate detailed vegetation and fuel maps for the whole boreal region of Alaska to aid effective forest and fire management.

A non-destructive estimation of chlorophyll-a and -b over different crops using airborne imaging spectroscopy observations

Non-destructive estimation of chlorophyll-a (Chl-a) and chlorophyll-b (Chl-b), is required for assessment and understanding of vegetation structural and functional dynamics. We have developed an empirical model for estimating Chl-a and Chl-b for different crops using Airborne Visible-Infrared Imaging Spectrometer Next Generation (AVIRIS-NG) data. Second derivative reflectance at 672 nm and 587 nm has shown maximum sensitivity towards Chl-a and Chl-b content, respectively. Separate linear models within a bootstrapped resampling framework were developed. During calibration, mean R2 of 0.46 and 0.51 were obtained for Chl-a and -b respectively. While during model validation, in both the cases mean R2 was found to be 0.3. This work shows the capability of airborne hyperspectral imaging data in segregating chlorophyll types using a simplistic modeling framework.

A Hybrid Clustering Method with a Filter Feature Selection for Hyperspectral Image Classification

Hyperspectral images (HSI) provide ample spectral information of land cover. The hybrid classification method works well for HSI; however, how to select the suitable similarity measures as kernels with the appropriate weights of hybrid classification for HSI is still under investigation. In this paper, a filter feature selection was designed to select the most representative features based on similarity measures. Then, the weights of applicable similarity measures were computed based on coefficients of variation (CVs) of similarity measures. Implementing the similarity measures as the kernels with weights into the K-means algorithm, a new hybrid changing-weight classification method with a filter feature selection (HCW-SSC) was developed. Standard spectral libraries, operative modular imaging spectrometer (OMIS) airborne HSI, airborne visible/infrared imaging spectrometer (AVIRIS) HSI, and Hyperion satellite HSI were selected to inspect the HCW-SSC method. The results showed that the HCW-SSC method has the highest overall accuracy and kappa coefficient (or F1 score) in all experiments (97.5% and 0.974 for standard spectral libraries, 93.21% and 0.9245 for OMIS, 79.24% and 0.8044 for AVIRIS, and 81.23% and 0.7234 for Hyperion) compared to the classification methods (93.75% and 0.958 for standard spectral libraries, 88.27% and 0.8698 for OMIS, 73.12% and 0.7225 for AVIRIS, and 56.34% and 0.3623 for Hyperion) without feature selection and the machine-learning method (68.27% and 0.6628 for AVIRIS, and 51.21% and 0.4255 for Hyperion). The experimental results demonstrate that the new hybrid method performs more effectively than the traditional hybrid method. This also shed a light on the importance of feature selection in HSI classification.

Assessment of spectral reduction techniques for endmember extraction in unmixing of hyperspectral images

Spectral mixture modelling is one of the most important techniques for classifying hyperspectral data at sub-pixel resolution. The identification of spectrally pure endmembers for estimating their corresponding abundances is an important step in spectral unmixing. The application of spectral reduction prior to endmember extraction would optimize the process by increasing the sensitivity of the algorithms to the most distinctive and informative features of the dataset. The objective of this study is to compare different spectral reduction techniques prior to endmember extraction on six real hyperspectral datasets, including an Airborne Visible InfraRed Imaging Spectrometer- Next Generation (AVIRIS-NG) image over Indian sub-continent. The endmembers identified from different combinations of spectral reduction and endmember extraction techniques are used for linear spectral unmixing on the original datasets. The performance of such combinations after unmixing were compared in terms of pixel reconstruction error and also the computation time for each dataset. Spectral reduction by both feature extraction techniques like Principal Component Analysis (PCA), Independent Component Analysis (ICA), Minimum Noise Fraction (MNF), and a feature selection technique based on Partial Informational Correlation (PIC) measure were analysed. The PIC based spectral reduction was found to perform well in terms of reconstruction error and computation time when combined with N-FINDR endmember algorithm. This approach could be adopted for spectral reduction in unmixing of datasets with similar endmember classes.

Carbon dioxide and water vapour mapping over tropical Indian atmosphere retrieved from AVIRIS-NG hyperspectral images

This work estimates the spatial distributions of atmospheric column averages of carbon dioxide (CO2) and water vapour (H2O) at the tropical atmosphere of India using the hyperspectral images of Airborne Visible InfraRed Imaging Spectrometer - Next Generation (AVIRIS-NG) flown over India for the first time in 2015–2016 and 2018. The retrieval algorithm accounts for the effect of altitude variation of the airborne sensor. A significant usefulness of AVIRIS-NG images is established by propounding a methodology for assessing the spatial distribution of atmospheric columnar H2O concentration, an important component of tropical climate, from the comparison of the two consecutive CO2 absorption bands. The spatial disparity of air-mixed CO2 over the places of widely varying natural and man-made features reflects the equilibrium of emission and sequestration in different extents. The gross long term CO2 change over the major portion of Indian atmosphere studied with NASA Giovanni open source data reveals a linear increase of the average CO2 concentration by about 2 ppm/year during years 2003–2016, almost up to the time of AVIRIS-NG first image procurement. The recent CO2 changes in India derived from the database of Orbiting Carbon Observatory-2 (OCO-2) are around 1.5–2.3 ppm/year increase for years 2016–2019 for different Indian sites. The new features of the CO2 trends over India are the extent of temporal increase and the inadequacy of the surrounding marine sequestration. The proportional spatial variations of H2O retrieved by the present method are consistent with the original H2O distributions supplied with the AVIRIS-NG images.

Hyperspectral image classification using Hyb-3D convolution neural network spectral partitioning

Hyperspectral image classification (HSIC) on remote sensing imaging has brought immersive achievement using artificial intelligence technology. In deep learning convolution neural networks (CNN), 2D-CNN, and 3D-CNN methods are widely used to classify the spectral-spatial bands of hyperspectral images (HSI). The proposed Hybrid 3D-CNN (H3D-CNN) model framework for deeper features extraction predicts classification accuracy in supervised learning. The model reduces the narrow gap between supervised and unsupervised learning and the complexity and cost of the previous models. The HSI classification analysis is carried out on real-world data sets of Indian pines Salinas datasets captured by Airborne visible, infrared imaging spectrometer (AVIRIS) sensors that performed superior classification accuracy results.

FlexBRDF: A Flexible BRDF Correction for Grouped Processing of Airborne Imaging Spectroscopy Flightlines.

Bidirectional reflectance distribution function (BRDF) effects are a persistent issue for the analysis of vegetation in airborne imaging spectroscopy data, especially when mosaicking results from adjacent flightlines. With the advent of large airborne imaging efforts from NASA and the U.S. National Ecological Observatory Network (NEON), there is increasing need for methods that are flexible and automatable across images with diverse land cover. Flexible bidirectional reflectance distribution function (FlexBRDF) is built upon the widely used kernel method, with additional features including stratified random sampling across flightline groups, dynamic land cover stratification by normalized difference vegetation index (NDVI), interpolation of correction coefficients across NDVI bins, and the use of a reference solar zenith angle. We demonstrate FlexBRDF using nine long (150–400 km) airborne visible/infrared imaging spectrometer (AVIRIS)‐Classic flightlines collected on 22 May 2013 over Southern California, where diverse land cover and a wide range of solar illumination yield significant BRDF effects. We further test the approach on additional AVIRIS‐Classic data from California, AVIRIS‐Next Generation data from the Arctic and India, and NEON imagery from Wisconsin. Comparison of overlapping areas of flightlines show that models built from multiple flightlines performed better than those built for single images (root mean square error improved up to 2.3% and mean absolute deviation 2.5%). Standardization to a common solar zenith angle among a flightline group improved performance, and interpolation across bins minimized between‐bin boundaries. While BRDF corrections for individual sites suffice for local studies, FlexBRDF is an open source option that is compatible with bulk processing of large airborne data sets covering diverse land cover needed for calibration/validation of forthcoming spaceborne imaging spectroscopy missions.

FPGA Implementation of a Hardware Optimized Automatic Target Detection and Classification Algorithm for Hyperspectral Image Analysis

In hyperspectral image analysis one of the most important tasks is target detection, requiring the execution of algorithms with high computational complexity. Recently, research efforts have focused on on-board real-time target detection to provide timely responses for swift decisions. Therefore, it is necessary to use a technology that provides the performance needed for real-time target detection, and at the same time meets the satellite payload requirements. Field-programmable gate arrays (FPGAs) have very interesting properties in terms of performance, size and power consumption, which have become the standard option for on-board processing. In this letter, we present a hardware optimized implementation for FPGAs of the automatic target detection and classification algorithm (ATDCA) using the Gram–Schmidt (GS) method for orthogonalization purposes. The ATDCA-GS algorithm is directly coded using VHDL and verified on a Virtex-7 XC7VX690T FPGA using real hyperspectral data (collected by Hyperspectral Digital Imagery Collection Experiment (HYDICE) sensor and by NASA’s Airborne Visible/Infrared Imaging Spectrometer (AVIRIS)) and a synthetic image. Experimental results demonstrate that our hardware version of the ATDCA-GS algorithm outperforms previous implementations (multicore processors, GPUs and accelerators) in both computation time (obtaining real-time performance) and power consumption, demonstrating the suitability of FPGAs for this purpose.

Multitarget Multiple-Instance Learning for Hyperspectral Target Detection

In remote sensing, it is often challenging to acquire or collect a large data set that is accurately labeled. This difficulty is usually due to several issues, including but not limited to the study site’s spatial area and accessibility, errors in the global positioning system (GPS), and mixed pixels caused by an image’s spatial resolution. We propose an approach, with two variations, that estimates multiple-target signatures from training samples with imprecise labels: multitarget multiple-instance adaptive cosine estimator (MTMI-ACE) and multitarget multiple-instance spectral match filter (MTMI-SMF). The proposed methods address the abovementioned problems by directly considering the multiple-instance, imprecisely labeled data set. They learn a dictionary of target signatures that optimizes detection against a background using the adaptive cosine estimator (ACE) and spectral match filter (SMF). Experiments were conducted to test the proposed algorithms using a simulated hyperspectral data set, the MUUFL Gulfport hyperspectral data set collected over the University of Southern Mississippi–Gulfpark Campus, and the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) hyperspectral data set collected over Santa Barbara County, CA, USA. Both simulated and real hyperspectral target detection experiments show that the proposed algorithms are effective at learning target signatures and performing target detection.

Seasonal and interannual drought responses of vegetation in a California urbanized area measured using complementary remote sensing indices

The effects of drought can manifest in vegetation across an array of physiological responses and time scales. In metropolitan areas, vegetation provides shading and cooling during hot and dry conditions, but these benefits can be reduced with drought. While many studies have evaluated interannual vegetation drought responses, these responses to drought can be expressed diversely across seasons, especially in cities that regularly experience seasonal drought (e.g., in Mediterranean climates). Here, we evaluated seasonal and interannual drought responses across the dominant types of urban trees and grasses in the Santa Barbara, California, USA metropolitan area, primarily using Landsat imagery acquired from 2010 to 2019 as well as repeat Airborne Visible Infrared Imaging Spectrometer - Classic (AVIRIS-C) imagery acquired 2013–2015. To track vegetation types, we produced a random forest classification from 4 m AVIRIS-Next Generation (AVIRIS-NG) imagery acquired in June 2014 (overall accuracy = 86%; kappa = 0.85), thresholding to >90% pure pixels for most vegetation types in the coarser time series imagery. We monitored drought response from Landsat imagery using the Normalized Difference Vegetation Index (NDVI) and the difference in land surface temperature (ΔLST) between vegetation and developed/impervious surfaces. We used AVIRIS-C to measure equivalent water thickness (EWT), comparing it with NDVI. During drought years, NDVI was lower and ΔLST was closer to zero. Changes in EWT revealed seasonal adjustments by vegetation that were not readily apparent in the NDVI. To show how drought response expression in vegetation can vary by season, drought duration, and urban vegetation type, we examined the correlations of both NDVI and ΔLST to the Standardized Precipitation Evapotranspiration Index (SPEI) calculated over a range of time spans. For most vegetation types, the strongest correlations of NDVI to SPEI and ΔLST to SPEI were during the summer, except for annual grass and turfgrass NDVI, which had the strongest correlations in the winter. In general, NDVI and ΔLST for annual grass were most often correlated with SPEI at spans <12 months, particularly for NDVI. By contrast, NDVI and ΔLST for trees and turfgrass were commonly also correlated with SPEI at spans ≥12 months, in addition to seasonal time spans of <12 months. This study demonstrates the benefits of using functionally and seasonally distinctive remote sensing variables (NDVI, ΔLST, and EWT) together to quantify changes in vegetation canopy condition during droughts.

Evaluating the Hyperspectral Sensitivity of the Differenced Normalized Burn Ratio for Assessing Fire Severity

Fire severity represents fire-induced environmental changes and is an important variable for modeling fire emissions and planning post-fire rehabilitation. Remotely sensed fire severity is traditionally evaluated using the differenced normalized burn ratio (dNBR) derived from multispectral imagery. This spectral index is based on bi-temporal differenced reflectance changes caused by fires in the near-infrared (NIR) and short-wave infrared (SWIR) spectral regions. Our study aims to evaluate the spectral sensitivity of the dNBR using hyperspectral imagery by identifying the optimal bi-spectral NIR SWIR combination. This assessment made use of a rare opportunity arising from the pre- and post-fire airborne image acquisitions over the 2013 Rim and 2014 King fires in California with the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) sensor. The 224 contiguous bands of this sensor allow for 5760 unique combinations of the dNBR at a high spatial resolution of approximately 15 m. The performance of the hyperspectral dNBR was assessed by comparison against field data and the spectral optimality statistic. The field data is composed of 83 in situ measurements of fire severity using the Geometrically structured Composite Burn Index (GeoCBI) protocol. The optimality statistic ranges between zero and one, with one denoting an optimal measurement of the fire-induced spectral change. We also combined the field and optimality assessments into a combined score. The hyperspectral dNBR combinations demonstrated strong relationships with GeoCBI field data. The best performance of the dNBR combination was derived from bands 63, centered at 0.962 µm, and 218, centered at 2.382 µm. This bi-spectral combination yielded a strong relationship with GeoCBI field data of R2 = 0.70 based on a saturated growth model and a median spectral index optimality statistic of 0.31. Our hyperspectral sensitivity analysis revealed optimal NIR and SWIR bands for the composition of the dNBR that are outside the ranges of the NIR and SWIR bands of the Landsat 8 and Sentinel-2 sensors. With the launch of the Precursore Iperspettrale Della Missione Applicativa (PRISMA) in 2019 and several planned spaceborne hyperspectral missions, such as the Environmental Mapping and Analysis Program (EnMAP) and Surface Biology and Geology (SBG), our study provides a timely assessment of the potential and sensitivity of hyperspectral data for assessing fire severity.

Joint VSWIR-TIR retrievals of earth's surface and atmosphere

Remote spectroscopy of terrestrial ecosystems generally falls into one of two categories: UV, Visible and ShortWave InfraRed (VSWIR) methods measuring reflected solar illumination, and Thermal InfraRed (TIR) methods measuring radiation emitted from the Earth's surface. Both interpret the measured radiance with inversion algorithms that estimate surface reflectance, emissivity, surface temperatures, and the atmospheric state. However, the two regimes have traditionally used independent inversion methods. Here we present the first simultaneous inversion of the upwelling radiance spectrum from the UV through the VSWIR and the Thermal Infrared. Maximum A Posteriori estimation determines the surface and atmosphere state that is most consistent with measured radiance across the entire range. We find that the complementary information in the two intervals improves estimates of atmospheric properties, surface emissivity, and surface temperature. Posterior uncertainty is 40% lower in surface temperature and 60% in air temperature. Spurious correlations between atmospheric and surface state are also reduced, with 90% lower correlated uncertainty between total column water vapor and both surface and air temperature. Temperature-emissivity separation also improves, with a 33% reduction in correlated posterior uncertainty between the parameters. We demonstrate the method on coincident data from the Classic Airborne Visible Infrared Imaging Spectrometer (AVIRIS-C) operating in the VSWIR range, and the Hyperspectral Thermal Emission Spectrometer (HyTES) in the TIR range, during flights on NASA's high-altitude ER-2 aircraft. Measurements spanning 0.4 to 12 microns can improve estimates of surface and atmospheric properties, significantly advancing studies in terrestrial ecosystems, geology, agriculture, and urban management.

Multi-season unmixing of vegetation class fractions across diverse Californian ecoregions using simulated spaceborne imaging spectroscopy data

Spaceborne imaging spectrometers are expected to facilitate regional-scale vegetation analyses with multi-season hyperspectral imagery. However, we still lack a better understanding on both whether multi-season hyperspectral approaches are favorable over single-season approaches, as well as on the benefits of hyperspectral compared to multispectral data. Our study investigates the potential of multi-season unmixing of simulated Environmental Mapping and Analysis Program (EnMAP) data for vegetation class fraction mapping across diverse natural and semi-natural ecoregions in California, USA. We utilized spring, summer and fall 2013 simulated EnMAP imagery derived from Airborne Visible / Infrared Imaging Spectrometer (AVIRIS) data covering study sites in the San Francisco Bay Area, Lake Tahoe and Santa Barbara. Regression-based unmixing with synthetic training datasets from spectral libraries was implemented for mapping needleleaf tree, broadleaf tree, shrub, herbaceous and non-vegetation fractions, and independent reference data was used for validation. Multi-season unmixing of simulated EnMAP had average Mean Absolute Errors (MAE) over all classes of 8.7% for the Bay Area, 8.5% for Lake Tahoe and 9.6% for Santa Barbara. However, larger errors in the low and high end of the fraction range remained, particularly in open-canopy woodlands and xeric shrub-dominated regions. Single-season unmixing of simulated EnMAP revealed large seasonal and regional variations within individual vegetation classes. In most cases, the best performing single-season unmixing had similar errors as the multi-season unmixing, i.e., ∆MAEs within ±1.0%. This points to the advantage of the multi-season integration strategy for more robust and generalized mapping independent from season and study site. Relative to EnMAP analyses, multi-season unmixing of multispectral Landsat composites for the same seasons yielded increases in average MAEs of +1.7%, +2.3% and +1.4% for the three study sites. This indicates that the higher spectral resolution of simulated EnMAP provides more relevant discriminative information when comparing contemporary image pairs. Unmixing of seasonal spectral-temporal metrics (STMs) from all available Landsat images for an entire year took advantage of the full temporal detail provided by these ongoing missions. We found Landsat STMs to effectively map vegetation class fractions, with average MAEs of 9.9%, 10.0% and 9.7% for the three study sites. Still, improvements particularly for mapping fractions of the woody vegetation classes through multi-season unmixing of simulated EnMAP point to the benefit of high spectral resolution data, and we assume that a comparable higher temporal resolution of hyperspectral satellites will further positively influence results. Overall, we conclude that multi-season unmixing of spaceborne imaging spectroscopy data holds great potential for advancing vegetation class fraction mapping across natural and semi-natural ecosystems.

Cutting out the middleman: calibrating and validating a dynamic vegetation model (ED2-PROSPECT5) using remotely sensed surface reflectance

Abstract. Canopy radiative transfer is the primary mechanism by which models relate vegetation composition and state to the surface energy balance, which is important to light- and temperature-sensitive plant processes as well as understanding land–atmosphere feedbacks. In addition, certain parameters (e.g., specific leaf area, SLA) that have an outsized influence on vegetation model behavior can be constrained by observations of shortwave reflectance, thus reducing model predictive uncertainty. Importantly, calibrating against radiative transfer outputs allows models to directly use remote sensing reflectance products without relying on highly derived products (such as MODIS leaf area index) whose assumptions may be incompatible with the target vegetation model and whose uncertainties are usually not well quantified. Here, we created the EDR model by coupling the two-stream representation of canopy radiative transfer in the Ecosystem Demography model version 2 (ED2) with a leaf radiative transfer model (PROSPECT-5) and a simple soil reflectance model to predict full-range, high-spectral-resolution surface reflectance that is dependent on the underlying ED2 model state. We then calibrated this model against estimates of hemispherical reflectance (corrected for directional effects) from the NASA Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) and survey data from 54 temperate forest plots in the northeastern United States. The calibration significantly reduced uncertainty in model parameters related to leaf biochemistry and morphology and canopy structure for five plant functional types. Using a single common set of parameters across all sites, the calibrated model was able to accurately reproduce surface reflectance for sites with highly varied forest composition and structure. However, the calibrated model's predictions of leaf area index (LAI) were less robust, capturing only 46 % of the variability in the observations. Comparing the ED2 radiative transfer model with another two-stream soil–leaf–canopy radiative transfer model commonly used in remote sensing studies (PRO4SAIL) illustrated structural errors in the ED2 representation of direct radiation backscatter that resulted in systematic underestimation of reflectance. In addition, we also highlight that, to directly compare with a two-stream radiative transfer model like EDR, we had to perform an additional processing step to convert the directional reflectance estimates of AVIRIS to hemispherical reflectance (also known as “albedo”). In future work, we recommend that vegetation models add the capability to predict directional reflectance, to allow them to more directly assimilate a wide range of airborne and satellite reflectance products. We ultimately conclude that despite these challenges, using dynamic vegetation models to predict surface reflectance is a promising avenue for model calibration and validation using remote sensing data.

Vicarious Calibration of eMAS, AirMSPI, and AVIRIS Sensors During FIREX-AQ

Remote sensing instruments, both aircraft and on-orbit platforms, undergo extensive laboratory calibrations to determine their geometric, spectral, and radiometric responses. Additional in-flight radiometric calibrations can be performed using well-characterized earth targets. The Fire Influence on Regional to Global Environments and Air Quality (FIREX-AQ) campaign provided such an opportunity when the ER-2 aircraft overflew Railroad Valley on August 13 and 15, 2019. Surface reflectances were available from the August 4, 2019 field team and from the Radiometric Calibration Network (RadCalNet) portal, and spectral aerosol optical depths from an on-site AERosol RObotic NETwork (AERONET) sunphotometer. The Enhanced MODIS Airborne Simulator (eMAS), the Airborne Multiangle SpectroPolarimetric Imager (AirMSPI), and the &#x201C;Classic&#x201D; Airborne Visible/Infrared Imaging Spectrometer (AVIRIS-C) sensors individually performed a vicarious calibration using their respective methodologies and selection of input parameters. A comparison of the at-sensor radiances predicted from these independent analyses highlights some of the uncertainties in the inputs, including choice of solar irradiance model. Although good agreement, within 5&#x0025;, is found at visible wavelengths, difference can be as large as 15&#x0025; in the shortwave infrared (SWIR). This highlights the need for the remote sensing community to agree upon a standard solar model, to remove sensor-to-sensor biases derived from in-flight calibrations.

Hyperspectral data compression based upon the principal component analysis

The paper is aimed at developing an algorithm of hyperspectral data compression that combines small losses with high compression rate. The algorithm relies on a principal component analysis and a method of exhaustion. The principal components are singular vectors of an initial signal matrix, which are found by the method of exhaustion. A retrieved signal matrix is formed in parallel. The process continues until a required retrieval error is attained. The algorithm is described in detail and input and output parameters are specified. Testing is performed using AVIRIS data (Airborne Visible-Infrared Imaging Spectrometer). Three images of differently looking sky (clear sky, partly clouded sky, and overcast skies) are analyzed. For each image, testing is performed for all spectral bands and for a set of bands from which high water-vapour absorption bands are excluded. Retrieval errors versus compression rates are presented. The error formulas include the root mean square deviation, the noise-to-signal ratio, the mean structural similarity index, and the mean relative deviation. It is shown that the retrieval errors decrease by more than an order of magnitude if spectral bands with high gas absorption are disregarded. It is shown that the reason is that weak signals in the absorption bands are measured with great errors, leading to a weak dependence between the spectra in different spatial pixels. A mean cosine distance between the spectra in different spatial pixels is suggested to be used to assess the image compressibility.

Identification of minerals in hyperspectral imagery based on the attenuation spectral absorption index vector using a multilayer perceptron

ABSTRACT Various minerals present distinct spectra due to different constituent ions or groups. However, variable imaging conditions of hyperspectral imagery greatly affect the mineral spectra, which makes it difficult to identify minerals. Attenuation spectral absorption index (ASAI) is suggested to extract absorption features from mineral diagnostic bands, which combines spectral absorption index (SAI) with absorption position and resists variations in external factors. To integrate intrinsic features of mineral spectra, the ASAI vector is formed by the ASAIes of diagnostic absorption bands corresponding to all identified minerals, which depicts mineral contents to a certain extent. In this letter, a multilayer perceptron (MLP) model based on attenuation spectral absorption index vector (ASMLP) is proposed to obtain hierarchical features for accurate mineral identification. The hyperspectral data of the Nevada mining area collected by Airborne Visible/Infrared Imaging Spectrometer (AVIRIS) are applied to illuminate the performance of the proposed model. The results show that ASMLP achieves an overall accuracy of 93.62% and outperforms other identification methods.

Detection and quantification of CH&amp;lt;sub&amp;gt;4&amp;lt;/sub&amp;gt; plumes using the WFM-DOAS retrieval on AVIRIS-NG hyperspectral data

Abstract. Methane is the second most important anthropogenic greenhouse gas in the Earth's atmosphere. To effectively reduce these emissions, a good knowledge of source locations and strengths is required. Airborne remote sensing instruments such as the Airborne Visible InfraRed Imaging Spectrometer – Next Generation (AVIRIS-NG) with meter-scale imaging capabilities are able to yield information about the locations and magnitudes of methane sources. In this study, we successfully applied the weighting function modified differential optical absorption spectroscopy (WFM-DOAS) algorithm to AVIRIS-NG data measured in Canada and the Four Corners region. The WFM-DOAS retrieval is conceptually located between the statistical matched filter (MF) and the optimal-estimation-based iterative maximum a posteriori DOAS (IMAP-DOAS) retrieval algorithm, both of which were already applied successfully to AVIRIS-NG data. The WFM-DOAS algorithm is based on a first order Taylor series approximation of the Lambert–Beer law using only one precalculated radiative transfer calculation per scene. This yields the fast quantitative processing of large data sets. We detected several methane plumes in the AVIRIS-NG images recorded during the Arctic-Boreal Vulnerability Experiment (ABoVE) Airborne Campaign and successfully retrieved a coal mine ventilation shaft plume observed during the Four Corners measurement campaign. The comparison between IMAP-DOAS, MF, and WFM-DOAS showed good agreement for the coal mine ventilation shaft plume. An additional comparison between MF and WFM-DOAS for a subset of plumes showed good agreement for one plume and some differences for the others. For five plumes, the emissions were estimated using a simple cross-sectional flux method. The retrieved fluxes originated from well pads, cold vents, and a coal mine ventilation shaft and ranged between (155 ± 71) kg (CH4) h−1 and (1220 ± 450) kg (CH4) h−1. The wind velocity was a significant source of uncertainty in all plumes, followed by the single pixel retrieval noise and the uncertainty due to atmospheric variability. The noise of the retrieved CH4 imagery over bright surfaces (&gt;1 µW cm−2 nm−1 sr−1 at 2140 nm) was typically ±2.3 % of the background total column of CH4 when fitting strong absorption lines around 2300 nm but could reach over ±5 % for darker surfaces (&lt; 0.3 µW cm−2 nm−1 sr−1 at 2140 nm). Additionally, a worst case large-scale bias due to the assumptions made in the WFM-DOAS retrieval was estimated to be ±5.4 %. Radiance and fit quality filters were implemented to exclude the most uncertain results from further analysis mostly due to either dark surfaces or surfaces where the surface spectral reflection structures are similar to CH4 absorption features at the spectral resolution of the AVIRIS-NG instrument.

Combining simulated hyperspectral EnMAP and Landsat time series for forest aboveground biomass mapping

Forest aboveground biomass (AGB) is a critical measure of ecosystem structure and plays a key role in global carbon cycling. Due to its widespread availability, optical remotely sensed data are key for regional- and global-scale AGB assessment, and with the planned and recent launches of spaceborne imaging spectroscopy missions such as the Environmental Mapping and Analysis Program (EnMAP), understanding the benefit of added spectral information for AGB mapping is important. We used simulated EnMAP imagery derived from Airborne Visible InfraRed Imaging Spectrometer (AVIRIS) imagery acquired over Sonoma County, California, USA in combination with Landsat time series to map forest AGB. A Gaussian Process Regression model was implemented to estimate forest AGB from one- and two-date (April and June) EnMAP imagery. A lidar-based reference AGB map was used as base data for training and validation sample extraction. As a comparison, we used corresponding Landsat Best Available Pixel (BAP) composites as well as a year-long 16-day interpolated Landsat time series (TS) for 2013. EnMAP imagery was able to effectively map forest AGB, with the two-date model (RMSE = 97.5 Mg/ha) outperforming the two single date models. All EnMAP models outperformed the corresponding Landsat BAP models, the best of which was the two-date model (RMSE = 108.8 Mg/ha). The added temporal dimension of the Landsat time series resulted in the best Landsat-based AGB map (RMSE = 102.3 Mg/ha). Combining the two datasets further improved AGB mapping efforts, with 2-date EnMAP + 2013 Landsat TS providing the best overall AGB maps (RMSE = 86.0 Mg/ha). This study demonstrates not only the added value of hyperspectral imagery for forest AGB mapping, but also the possible synergies between hyperspectral and multispectral data sources and hence between spectrally and temporally dense information. It can be expected that with the next generation of spaceborne hyperspectral sensors, the combination of dense spectral and temporal data will work to further improve global efforts for mapping forest AGB from optical Earth observation data.

Ecological assessment of Hooghly River considering a few of the more perturbed sites based on some relevant physico-chemical and biological variables—A part of the AVIRIS-NG (NASA-ISRO) ground truth verification

A study was performed on a 32 Km stretch of the River Hooghly as part of a project under the Airborne Visible Infrared Imaging Spectrometer - Next Generation (AVIRIS-NG) scheme between Indian Space Research Organization (ISRO) and National Aeronautics and Space Administration (NASA). It was governed by 33 water quality variables comprising physical, chemical, biological and ecological ones. The selected stations were within the tidal realm of the estuary, constituting the lower-middle stretch of the river that was severely affected by anthropogenic perturbations. Chief biotic components studied were phytoplankton population density, primary and auxiliary phytopigments, and productivity in addition to chromophoric dissolved organic matter (CDOM). Majority of the study sites were observed with very low concentrations of dissolved oxygen along with significant shifts in primary production. Overall prevalence of Cyanophyceae and Bacillariophyceae was indicative of significant eutrophication. Reverse phase high performance liquid chromatography (HPLC) and radiometry revealed the presence of a large array of algal taxonomic groups through the detection of various primary and auxiliary phytopigments. The data and the study itself is of paramount importance being part of the ground truth verification initiative which will be used to render remote sensing techniques more precise in the future to better monitor the estuarine system.