Airborne Imaging Spectrometer Research Articles

Detecting high-temperature anomalies from Sentinel-2 MSI images

High-temperature anomalies (HTAs) of the earth's surface, such as fires, volcanic activities, and industrial heat sources, have a profound impact on Earth's system. Sentinel-2 Multispectral Instrument (MSI) provides spatially-specific information for precisely measuring the location and extent of HTAs at a fine scale. However, detecting HTAs from MSI images remains challenging because the emitted radiance of an HTA in the short-wave infrared (SWIR) bands can be easily mixed with the reflected solar radiance background in the daytime; and an increasing number of atypical cases in MSI images need to be treated with the enhanced spatial resolution. A generic HTA detection approach that handles both anthropogenic and natural HTAs will broaden the scope of MSI applications. In this study, (i) we highlight two spectral characteristics of HTAs in the far-SWIR, near-SWIR, and NIR bands (i.e., (ρfar-SWIR - ρnear-SWIR)/ρNIR ≥ 0.45 and (ρfar-SWIR -ρnear-SWIR) ≥ ρnear-SWIR - ρNIR) that can effectively enhance HTAs from background geo-features, based on the reflectance spectra in airborne imaging spectrometer data. (ii) We propose a tri-spectral thermal anomaly index (TAI) that jointly uses the two high-temperature-sensitive SWIR bands and the high-temperature-insensitive NIR band to enhance HTAs, based on the above characteristics and a comprehensive sampling of different types of HTAs from 1,974 MSI images. (iii) We develop a TAI-based approach for MSI images to detect HTAs in general. The proposed approach was applied to detect different types of HTAs, including different biomass burnings, active volcanoes, and industrial HTAs, over a wide range of land-cover scenarios. Validations and comparisons demonstrate the proposed approach is reliable and performs better than the existing state-of-the-art HTA detection approaches. Evaluations on two types of small industrial HTAs, including operating kilns and enclosed landfill gas flares, show that the HTA detection probability of the TAI-based approach from time-series MSI images is ~ 84.91% and 88.23%, respectively. Further investigations show that the TAI-based approach also has good transferability in detecting HTAs from multispectral images acquired by Landsat-family satellites.

Correction of Sunglint Effects in High Spatial Resolution Hyperspectral Imagery Using SWIR or NIR Bands and Taking Account of Spectral Variation of Refractive Index of Water

Passive high spatial resolution hyperspectral and multispectral imaging systems in the solar spectral region from aircraft and satellite platforms are being increasingly used for remote sensing of coastal waters and inland lakes. However, the remotely sensed data are often contaminated by the specular reflection of solar radiation at the air/water interface. Thus, the effects of sunglint need to be corrected. The purpose of sunglint correction is to remove the undesired specular reflected portion of radiance that does not carry any information about the water body and its bottom surface properties and to obtain the water-leaving radiance from the measured total radiance. In general, adequate treatment of the spectral variation of the sunglint effect with due consideration of the physics and the spectroscopy principles has not been done in previously reported hyperspectral sunglint removing techniques. In this study, a practical sunglint removing technique using either a shortwave infrared (SWIR) band or a near-infrared (NIR) band has been developed, where the liquid water absorption is sufficiently large, and the water-leaving radiance (in units of reflectance) is negligible. In the proposed method, proper consideration has been taken for the spectral Fresnel reflection in the wavelength range of 0.35–2.5 mm. Four cases of sunglint removal from hyperspectral imaging data acquired from two airborne imaging spectrometers have been presented.

Spectroscopic imaging of sub-kilometer spatial structure in lower-tropospheric water vapor

Abstract. The subgrid spatial variability of water vapor is an important geophysical parameter for modeling tropical convention and cloud processes in atmospheric models. This study maps sub-kilometer spatial structures in total atmospheric column water vapor with visible to shortwave infrared (VSWIR) imaging spectroscopy. We describe our inversion approach and validate its accuracy with coincident measurements by airborne imaging spectrometers and the AERONET ground-based observation network. Next, data from NASA's AVIRIS-NG spectrometer enable the highest-resolution measurement to date of water vapor's spatial variability and scaling properties. We find second-order structure function scaling exponents consistent with prior studies of convective atmospheres. Airborne lidar data show that this total column measurement provides information about variability in the lower troposphere. We conclude by discussing the implications of these measurements and paths toward future campaigns to build upon these results.

Coastal shallow waters explorer imaging spectrometer for aerial remote sensing of shallow waters in UV-VIS-NIR broadband.

Advancements in technology have widened the limits of hyperspectral imaging of remote sensing in mapping shallow water benthic habitats and bathymetry. This paper provides a novel airborne imaging spectrometer for the field. The device, composed of an off-axis two-mirror telescope and an advanced Dyson spectrometer working in the ultraviolet-visible near-infrared region, has been designed, fabricated, and evaluated. It maintained high optical performance, with a pixel spatial resolution better than 0.5 mrad, spectral resolution of 3.5 nm, field of view of 28°, and a high numerical aperture for high SNR in 330-900 nm. The imaging spectrometer will help to obtain a combination of the spectral and textural characteristics of bottom features in shallow waters. These features could result in the ability to map habitats over large coastal systems.

Assessing scaling effect in downscaling land surface temperature in a heterogenous urban environment

It is confirmed that there exists a scaling effect in downscaling land surface temperature (DLST) processes. However, a literature review indicates that quantifying and fully understanding scaling effect in DLST processing remains unclear. In this study, a main goal is to quantify, assess and understand scaling effect in downscaling LST product processes at different higher spatial resolutions and spatial extents. A machine leaning model and a traditional multivariate regression model were adopted with corresponding scaling factors extracted from ASTER 15–30 m optical multispectral data and Airborne Imaging Spectrometer for Different Applications (AISA) 2 m hyperspectral visible-near infrared data. MODIS 990 m LST and ASTER 90 m LST products were downscaled to high and very high resolution LST maps. In addition, ETM+ 60 m retrieved LST and Thermal Airborne Broadband Imager (TABI) 2 m retrieved LST and its upscaled LSTs were used to verify higher resolution DLST maps. The experimental results demonstrate that scaling effect in downscaling LST processes is significant, especially downscaling LST to high and very high resolution LST maps. One innovation point derived from findings by assessing the scaling effect in DLST processing is that when DLST processes are at spatial resolutions beyond a range (20–30 m in this study) measured from semivariograms, the processes are safe and their results are reasonable and reliable, and thus their scaling effect may be ignored, but when spatial resolutions and spatial extent lag distance within the range, the DLST processes are not safe, and their results are not reliable and thus the scaling effect has to be considered. Therefore, it is recommended that before conducting a DLST processing project, a range needs to be calculated by plotting semivariograms with high or very high resolution images (better to include three visible and one NIR bands), then DLST processes may be conducted at spatial resolutions lower than and equal to the range.

An evaluation of airborne SWIR imaging spectrometers for CH4 mapping: Implications of band positioning, spectral sampling and noise

The development of instruments and methods to assist in methane (CH4) emissions mapping is essential to the fossil fuel industry. Early identification of local CH4 sources can benefit both petroleum hydrocarbon prospecting and environmental monitoring. Generic airborne imaging spectrometers operating in the shortwave infrared (SWIR) wavelength range (1000−2400 nm) have shown their suitability for this task. However, to date, there is no airborne scanner specifically designed for the detection of CH4 plumes. To overcome current handicaps and achieve better results at local scale, further investigation in sensor design is needed to evaluate which components can be adjusted to improve CH4 mapping with airborne sensors. Here, we focus on the evaluation of currently operational airborne imaging spectrometers for CH4 mapping in the SWIR range. Data acquired over areas with known CH4 emissions by scientific and industry-grade airborne hyperspectral sensors were examined. The research was conducted in three steps: analysis of sensor design, image processing and noise simulation. In the first step, differences in the spectral sampling between the sensors were analyzed. A reference CH4 signature from HITRAN spectral database was resampled to the spectral sampling of each airborne sensor. The center wavelength of diagnostic CH4 absorption features (identified in the convolved signatures) in relation to the position of band centers from each equipment was examined. For the image processing stage, a new CH4 index and a classic matched filtering were used to map CH4 plumes. To assess the impact of the signal-to-noise ratio (SNR) of the airborne sensors on CH4 plume mapping, white noise was added to the data to simulate images with varying SNR levels. Results demonstrated that the wavelength position of band centers is a key for CH4 mapping. The CH4 plumes could be mapped only with scientific-grade sensors, in which the band centers were closer to the center of CH4 features. Simulations with the addition of random noise demonstrated that a noisier signal is probably the reason why the industry-grade sensor tested here failed to map CH4 plumes, given that all instruments have a comparable spectral sampling. Furthermore, the simulations also demonstrated that the density of the plume has also a weight on the mapping of CH4 sources, once the image that captured the densest plume requested a higher addition of noise to be lost. The overall investigation indicates that a hyperspectral airborne sensor with bands properly positioned and scientific-grade SNR would better resolve the narrow CH4 features in the SWIR range.

Evaluating Variable Selection and Machine Learning Algorithms for Estimating Forest Heights by Combining Lidar and Hyperspectral Data

Machine learning has been employed for various mapping and modeling tasks using input variables from different sources of remote sensing data. For feature selection involving high- spatial and spectral dimensionality data, various methods have been developed and incorporated into the machine learning framework to ensure an efficient and optimal computational process. This research aims to assess the accuracy of various feature selection and machine learning methods for estimating forest height using AISA (airborne imaging spectrometer for applications) hyperspectral bands (479 bands) and airborne light detection and ranging (lidar) height metrics (36 metrics), alone and combined. Feature selection and dimensionality reduction using Boruta (BO), principal component analysis (PCA), simulated annealing (SA), and genetic algorithm (GA) in combination with machine learning algorithms such as multivariate adaptive regression spline (MARS), extra trees (ET), support vector regression (SVR) with radial basis function, and extreme gradient boosting (XGB) with trees (XGbtree and XGBdart) and linear (XGBlin) classifiers were evaluated. The results demonstrated that the combinations of BO-XGBdart and BO-SVR delivered the best model performance for estimating tropical forest height by combining lidar and hyperspectral data, with R2 = 0.53 and RMSE = 1.7 m (18.4% of nRMSE and 0.046 m of bias) for BO-XGBdart and R2 = 0.51 and RMSE = 1.8 m (15.8% of nRMSE and −0.244 m of bias) for BO-SVR. Our study also demonstrated the effectiveness of BO for variables selection; it could reduce 95% of the data to select the 29 most important variables from the initial 516 variables from lidar metrics and hyperspectral data.

Three-dimensional radiative transfer effects on airborne and ground-based trace gas remote sensing

Abstract. Air mass factors (AMFs) are used in passive trace gas remote sensing for converting slant column densities (SCDs) to vertical column densities (VCDs). AMFs are traditionally computed with 1D radiative transfer models assuming horizontally homogeneous conditions. However, when observations are made with high spatial resolution in a heterogeneous atmosphere or above a heterogeneous surface, 3D effects may not be negligible. To study the importance of 3D effects on AMFs for different types of trace gas remote sensing, we implemented 1D-layer and 3D-box AMFs into the Monte carlo code for the phYSically correct Tracing of photons In Cloudy atmospheres (MYSTIC), a solver of the libRadtran radiative transfer model (RTM). The 3D-box AMF implementation is fully consistent with 1D-layer AMFs under horizontally homogeneous conditions and agrees very well (<5 % relative error) with 1D-layer AMFs computed by other RTMs for a wide range of scenarios. The 3D-box AMFs make it possible to visualize the 3D spatial distribution of the sensitivity of a trace gas observation, which we demonstrate with two examples. First, we computed 3D-box AMFs for ground-based multi-axis spectrometer (MAX-DOAS) observations for different viewing geometry and aerosol scenarios. The results illustrate how the sensitivity reduces with distance from the instrument and that a non-negligible part of the signal originates from outside the line of sight. Such information is invaluable for interpreting MAX-DOAS observations in heterogeneous environments such as urban areas. Second, 3D-box AMFs were used to generate synthetic nitrogen dioxide (NO2) SCDs for an airborne imaging spectrometer observing the NO2 plume emitted from a tall stack. The plume was imaged under different solar zenith angles and solar azimuth angles. To demonstrate the limitations of classical 1D-layer AMFs, VCDs were then computed assuming horizontal homogeneity. As a result, the imaged NO2 plume was shifted in space, which led to a strong underestimation of the total VCDs in the plume maximum and an underestimation of the integrated line densities that can be used for estimating emissions from NO2 images. The two examples demonstrate the importance of 3D effects for several types of ground-based and airborne remote sensing when the atmosphere cannot be assumed to be horizontally homogeneous, which is typically the case in the vicinity of emission sources or in cities.

Unmixing urban hyperspectral imagery using probability distributions to represent endmember variability

Urban composition can be analyzed through spectral unmixing of images from airborne imaging spectrometers. Unmixing given a spectral library can be accomplished by set-based methods or distribution-based methods. For computational efficiency and optimal accuracy, set-based methods employ a library reduction procedure when applied to large spectral libraries. On the other hand, distribution-based methods model the library by only a few parameters, hence innately accept large libraries. A natural question arises that can distribution-based methods with the original large spectral library achieve comparable performance to set-based methods in urban imagery.In this study, we aim to investigate the unmixing capability of several distribution-based methods, Gaussian mixture model (GMM), normal compositional model (NCM), and Beta compositional model (BCM) by comparing them to set-based methods MESMA and alternate angle minimization (AAM). The data for validation were collected by the AVIRIS sensor over the Santa Barbara region: two 16 m spatial resolution and two 4 m spatial resolution images. 64 validated regions of interest (ROI) (180 m by 180 m) were used to assess estimate accuracy. Ground truth was obtained using 1 m images leading to the following 6 classes: turfgrass, non-photosynthetic vegetation (NPV), paved, roof, soil, and tree. Spectral libraries were built by manually identifying and extracting pure spectra from both resolution images, resulting in 3287 spectra at 16 m and 15,426 spectra at 4 m. The libraries were further reduced to 61 spectra at 16 m and 95 spectra at 4 m for set-based methods. The results show that in terms of mean absolute error (MAE), GMM performed best among the distribution-based methods while MESMA performed best among the set-based methods. For 16 m data, there is no significant difference between GMM and MESMA (MAE = 0.069 vs. MAE = 0.074, p = 0.25). For 4 m data, though GMM is not as accurate as MESMA (MAE = 0.056 vs. MAE = 0.046, p = 7e − 5), it is better than AAM (MAE = 0.056 vs. MAE = 0.065, p = 0.02) which is a re-implementation of MESMA. Further evidence on a reconstructed synthetic dataset implies possible overfitting of the reduced library to the images for MESMA. These findings suggest that the distribution-based method GMM could achieve comparable unmixing accuracy to set-based methods without the need of library reduction, it may also be more stable across datasets, and the current 2-step workflow could be replaced by a single model in applying a universal spectral library.1

Multi-Scale Evaluation of Drone-Based Multispectral Surface Reflectance and Vegetation Indices in Operational Conditions

Compact multi-spectral sensors that can be mounted on lightweight drones are now widely available and applied within the geo- and environmental sciences. However; the spatial consistency and radiometric quality of data from such sensors is relatively poorly explored beyond the lab; in operational settings and against other sensors. This study explores the extent to which accurate hemispherical-conical reflectance factors (HCRF) and vegetation indices (specifically: normalised difference vegetation index (NDVI) and chlorophyll red-edge index (CHL)) can be derived from a low-cost multispectral drone-mounted sensor (Parrot Sequoia). The drone datasets were assessed using reference panels and a high quality 1 m resolution reference dataset collected near-simultaneously by an airborne imaging spectrometer (HyPlant). Relative errors relating to the radiometric calibration to HCRF values were in the 4 to 15% range whereas deviations assessed for a maize field case study were larger (5 to 28%). Drone-derived vegetation indices showed relatively good agreement for NDVI with both HyPlant and Sentinel 2 products (R2 = 0.91). The HCRF; NDVI and CHL products from the Sequoia showed bias for high and low reflective surfaces. The spatial consistency of the products was high with minimal view angle effects in visible bands. In summary; compact multi-spectral sensors such as the Parrot Sequoia show good potential for use in index-based vegetation monitoring studies across scales but care must be taken when assuming derived HCRF to represent the true optical properties of the imaged surface.

California's methane super-emitters.

Methane is a powerful greenhouse gas and is targeted for emissions mitigation by the US state of California and other jurisdictions worldwide1,2. Unique opportunities for mitigation are presented by point-source emitters-surface features or infrastructure components that are typically less than 10 metres in diameter and emit plumes of highly concentrated methane3. However, data on point-source emissions are sparse and typically lack sufficient spatial and temporal resolution to guide their mitigation and to accurately assess their magnitude4. Here we survey more than 272,000 infrastructure elements in California using an airborne imaging spectrometer that can rapidly map methane plumes5-7. We conduct five campaigns over several months from 2016 to 2018, spanning the oil and gas, manure-management and waste-management sectors, resulting in the detection, geolocation and quantification of emissions from 564 strong methane point sources. Our remote sensing approach enables the rapid and repeated assessment of large areas at high spatial resolution for a poorly characterized population of methane emitters that often appear intermittently and stochastically. We estimate net methane point-source emissions in California to be 0.618 teragrams per year (95 per cent confidence interval 0.523-0.725), equivalent to 34-46 per cent of the state's methane inventory8 for 2016. Methane 'super-emitter' activity occurs in every sector surveyed, with 10 per cent of point sources contributing roughly 60 per cent of point-source emissions-consistent with a study of the US Four Corners region that had a different sectoral mix9. The largest methane emitters in California are a subset of landfills, which exhibit persistent anomalous activity. Methane point-source emissions in California are dominated by landfills (41 per cent), followed by dairies (26 per cent) and the oil and gas sector (26 per cent). Our data have enabled the identification of the 0.2 per cent of California's infrastructure that is responsible for these emissions. Sharing these data with collaborating infrastructure operators has led to the mitigation ofanomalous methane-emission activity10.

Remotely Sensing Variation in Ecological Strategies and Plant Traits of Willows in Perialpine Floodplains

AbstractNatural floodplains are characterized by a complex habitat mosaic. However, damming, water storage, and hydropower production affect many floodplains by altering their natural habitat diversity. Field sampling data and imaging spectroscopy are used in combination with statistical models to assess resource allocation strategies of willow stands in perialpine floodplains. Three contrasting floodplain reaches located along two rivers in Switzerland serve as test beds: The Sarine River is partitioned into an upstream and downstream segment under the influence of a dam and a hydropower plant, while the Sense River represents an undisturbed, natural floodplain. Airborne imaging spectrometer data allow mapping of spatially distributed Competitor/Stress tolerator/Ruderal (CSR) strategies using a partial least square modeling approach. Using cross validation, we demonstrate that a statistical modeling approach can reveal variations in CSR scores based on the StrateFy model. Such intraspecific variation of CSR scores cannot be captured by a strategy categorization based solely on the species. Results reveal that willows shifted toward more competition and less stress tolerance along hydrologically altered reaches compared to the willows strategy along the natural control. Moreover, the overall distribution of strategies indicates that stress factors (i.e., limiting growth factors), rather than disturbance (i.e., events leading to partial or total destruction), shape the plant traits of alluvial willow trees. Detailed assessments of resource allocation strategies contribute to a more complete understanding of the continuous and reciprocal shaping between flow regimes, landforms, and alluvial vegetation.

Prospectivity Mapping for Epithermal Deposits of Western Milos Using a Fuzzy Multi Criteria Evaluation Approach Parameterized by Airborne Hyperspectral Remote Sensing Data

A Mineral prospectivity mapping (MPM) approach using a GIS-based weighted linear combination implementation of a Multi-Criteria Evaluation approach utilising a fuzzy Analytical Hierarchy Process to elucidate expert knowledge has been implemented to analyse the spatial distribution of epithermal deposits on the Island of Milos, Greece and model their association with exploration evidence data with the aim of providing insights into the controls on ore deposition. An integrated field and Digital Airborne Imaging Spectrometer (DAIS) hyperspectral and thermal multispectral airborne remote sensing dataset supported by field mapping and laboratory analyses, has been utilised to resolve hydrothermal alteration and parameterise the MPM. This study has highlighted the intimate spatial relationship between topographic highs and locations with high grade silicified alteration at a number of locations. The ability of high spatial resolution multispectral Thermal InfraRed (TIR) remote sensing imagery, integrated with topographic data, to resolve these silicified topographic highs provides an additional tool in the exploration of epithermal deposits. The spatial relationships between silicified lithocaps, high-grade altered rocks, faulting and topographic highs were utilised in the development of the MPM model. A close association between the modelled results and the hydrothermal alteration mapped in the field supports the accuracy of this MPM approach.

A Study on Quality Evaluation Methods of Hyperspectral Image

Because of great theoretical value and application prospects in civil and military affairs, hyperspectral image quality assessment method has been a hot research issue in current target recognition and remote sensing information processing field. In this article, firstly, we explicitly described evaluation methods which affect the quality of hyperspectral image based on two aspects of image quality and spectral quality evaluation. Afterwards, we used data cube provided by Australia Airborne Imaging Spectrometer Hymap as ideal raw spectral data to separately restored the degraded images by using Wiener Recovery, R-L Recovery and CGA Recovery, and then verified the recovery results through image quality assessment methods and spectral assessment methods introduced in this article, and finally made a general judgement on the merits of each method.

Potential use of hyperspectral data to classify forest tree species

BackgroundRemote sensing techniques and data are becoming increasingly popular in forest management, e.g. for change detection and health condition analysis. Tree species recognition is a fundamental issue in taking forest inventories, especially in carbon budget modelling. Hyperspectral imagery provides an accurate classification results for large areas based on a relatively small amount of training data.ResultsA hyperspectral image of a forest stand in north-eastern Poland taken using an AISA (Airborne Imaging Spectrometer for Application) Eagle camera was transformed to extract the most valuable spectral differences and was classified into seven tree types (birch, European beech, oak, hornbeam, European larch, Scots pine, and Norway spruce) using nine classification algorithms. The highest overall accuracy and kappa coefficient were 90.3% and 0.9 respectively using three minimum noise fraction bands and maximum likelihood classifier.ConclusionsHyperspectral imaging of forests can be used to classify major forest tree species with a good degree of accuracy. It is time-efficient and user-friendly; however, the data and software required means that this approach is still expensive at present.

Determining spectral groups to distinguish oil emulsions from Sargassum over the Gulf of Mexico using an airborne imaging spectrometer

During the weathering of marine-spilled oils, various types of oil pollution are formed that can harm marine and coastal environments. Thus, the remote detection, classification and quantification of spilled oils is important in marine environmental monitoring. Although multispectral images can be used to observe various spilled oils, due to confusion between the multispectral backscattered signals, distinguishing spilled oils from floating algae in the same image is challenging. The spectral features of carbon-hydrogen (-C-H) and oxygen-hydrogen (-O-H) groups, and pigments, are diagnostic absorption features and are different from the backscattering signal, they have not been used to improve detection independently. In this study, all the spectral features of the groups were clearly interpreted using reflectance spectra collected from an airborne visible infrared imaging spectrometer (AVIRIS). A reflectance peak-trough detection method to characterize the different spectral groups was used to determine the spectral features of Deepwater Horizon (DWH) oil emulsions and floating Sargassum in the Gulf of Mexico (GOM). The results show that the spilled oils and floating Sargassum can be clearly identified, and the various spilled oils (i.e., different oil emulsions and oil slicks) could also be determined from the differences in the spectral features of the above groups. Finally, we discuss the spectral requirements for the identification of these groups and we conclude that optical remote sensing, including imaging spectrometers, will play an increasingly important role in assessing marine oil spills.

Surface soil clay content mapping at large scales using multispectral (VNIR–SWIR) ASTER data

ABSTRACT The potential of Visible Near-Infrared and Short-Wave Infrared (VNIR–SWIR, 400 nm–2500 nm) hyperspectral imagery for use in multivariate approaches and geostatistical techniques for mapping topsoil properties has been previously demonstrated. However, the use of VNIR–SWIR hyperspectral imagery remains costly, which limits the spatial scales over which it can be applied. This paper aims to evaluate the potential for substituting the more accessible Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) VNIR–SWIR multispectral data for hyperspectral imagery in mapping surface soil clay contents. This study used ASTER multispectral data (nine bands in the VNIR–SWIR spectral domain) acquired over the Cap-Bon region in Tunisia (2000 km2) and 262 surface soil samples collected within the ASTER scene that were subjected to laboratory analysis of the clay fraction (soil particles less than 2 µm). The approach followed two steps: i) estimation of surface soil clay contents for bare soil areas using a Multiple Linear Regression (MLR) model built from the 9 ASTER VNIR–SWIR bands and ii) spatial interpolation (co-kriging) of the soil sampling of measured points and the ASTER-estimates over the whole study area. The MLR model for estimating clay contents using ASTER multispectral data performed correctly ( = 0.60). In addition, this performance is only slightly lower than that obtained using hyperspectral imagery (specifically, an Airborne Imaging Spectrometer for Applications (AISA-DUAL) dual hyperspectral sensor) in a previous study. Moreover, the co-kriging process appeared to yield encouraging results for capturing the large range of variability of clay content values, although it was not able to represent the short scale variability ( = 0.43). Finally, the ASTER multispectral data, despite being underused in the mapping of soil properties, may open up new ways to perform more extensive mapping of surface soil properties in semi-arid contexts characterized by extensive bare and dry soil surfaces.

Minimizing soil moisture variations in multi-temporal airborne imaging spectrometer data for digital soil mapping

Bare soil surfaces vary spatially and temporally as a result of – among other - meteorological conditions and land management practices. Literature shows that most of the existing soil surface corrections are dependent on laboratory measurements. These measurements are time consuming and therefore unsuitable for large amounts of airborne or satellite spectroscopy data. Soil moisture and soil surface roughness present main challenges for in-field and remote sensing application of soil spectroscopy and soil moisture has substantial influence on surface reflectance. Here, we present a straightforward method for soil moisture reduction applicable to large amounts of airborne or satellite spectroscopy data.Multi-temporal airborne imaging spectrometer data was collected with the Airborne Prism EXperiment (APEX) in September 2013, May 2014 and May 2015 on the same pilot site. A new soil moisture index was developed to quantify the soil moisture of the airborne imaging spectroscopy data (R2 of 0.59). The index appeared to be closely related to the existing NINSOL index. A, laboratory based, independent soil moisture spectral dataset was used to quantify the correction factor per wavelength based on the wavelength dependent relative differences.After minimizing the effect of soil moisture variability, the airborne spectrometer data better approximate the reference dry-soil spectra (up to 21.3%). Then, the corrected data were tested for soil properties prediction (soil organic matter (SOM), sand silt and clay). The soil moisture corrected spectra performed marginally better for SOM (RMSE from 3.13 to 3.10) and silt (RMSE from 5.74 to 5.72), although not for clay (RMSE from 9.14 to 9.33) and sand (RMSE from 5.98 to 6.31). The methodology shows that field spectra can be corrected to better match laboratory conditions, focusing on soil moisture as main varying factor. Next step should include the various types of soil surface variation and their interactions. A better understanding of these interactions is needed and can be achieved by field-scale experiments and by modelling the different processes.

The CCB-ID approach to tree species mapping with airborne imaging spectroscopy.

BackgroundBiogeographers assess how species distributions and abundances affect the structure, function, and composition of ecosystems. Yet we face a major challenge: it is difficult to precisely map species across landscapes. Novel Earth observations could overcome this challenge for vegetation mapping. Airborne imaging spectrometers measure plant functional traits at high resolution, and these measurements can be used to identify tree species. In this paper, I describe a trait-based approach to species identification with imaging spectroscopy, the Center for Conservation Biology species identification (CCB-ID) method, which was developed as part of an ecological data science evaluation competition.MethodsThese methods were developed using airborne imaging spectroscopy data from the National Ecological Observatory Network (NEON). CCB-ID classified tree species using trait-based reflectance variation and decision tree-based machine learning models, approximating a morphological trait and dichotomous key method inspired by botanical classification. First, outliers were removed using a spectral variance threshold. The remaining samples were transformed using principal components analysis (PCA) and resampled to reduce common species biases. Gradient boosting and random forest classifiers were trained using the transformed and resampled feature data. Prediction probabilities were calibrated using sigmoid regression, and sample-scale predictions were averaged to the crown scale.ResultsCCB-ID received a rank-1 accuracy score of 0.919, and a cross-entropy cost score of 0.447 on the competition test data. Accuracy and specificity scores were high for all species, but precision and recall scores varied for rare species. PCA transformation improved accuracy scores compared to models trained using reflectance data, but outlier removal and data resampling exacerbated class imbalance problems.DiscussionCCB-ID accurately classified tree species using NEON data, reporting the best scores among participants. However, it failed to overcome several species mapping challenges like precisely identifying rare species. Key takeaways include (1) selecting models using metrics beyond accuracy (e.g., recall) could improve rare species predictions, (2) within-genus trait variation may drive spectral separability, precluding efforts to distinguish between functionally convergent species, (3) outlier removal and data resampling can exacerbate class imbalance problems, and should be carefully implemented, (4) PCA transformation greatly improved model results, and (5) targeted feature selection could further improve species classification models. CCB-ID is open source, designed for use with NEON data, and available to support species mapping efforts.

Subpixel mapping and confusion matrix analysis of plant functional types in peatlands using MESMA applied to AISA Eagle imagery

Quantification of the spatial distribution of fractional cover of plant functional types is important for peatland ecosystems analysis. Using a crisp classification to produce a map of the spatial patterns of plant functional types in peatlands is difficult due to their complex and heterogeneous characteristics. We applied multiple endmember spectral mixture analysis (MESMA) on Airborne Imaging Spectrometer for Applications Eagle data to generate fractional covers of peatland plant functional types and further to assess the accuracy of soft classification at the subpixel scale. This was achieved by constructing the spectral library and selecting the optimal endmembers models (two-, three-, and four-endmember MESMA models) to derive the peatland plant functional types (shrub, graminoid, moss, and forbs) fractional covers. The final fractional cover pattern in each case was produced by joining the results from the multiendmembers models as judged by the root mean squared error thresholds to select the best-fit model for each pixel by minimization of pixel-scale fraction errors. Based on the thresholds, the three- and four-endmember MESMA models and their combination were further tested for their performance using an independent in situ dataset of the four-plant functional types. Their accuracy performance was quantified by a subpixel confusion matrix and residual analysis. The results showed that the combined MESMA model was most accurate in deriving the patterns of fractional cover for shrub, graminoid, and moss compared with the performance of the three- and four-endmember MESMA models. However, all models performed poorly in predicting the fractional cover of the sparsely distributed forbs. The results demonstrate the superiority of MESMA models in extracting the patterns of different plant functional types in peatlands as judged by the accuracy of subpixel fractional extraction.