Simulation of hyperspectral and directional radiance images using coupled biophysical and atmospheric radiative transfer models

Future remote sensing satellite missions exploring the earth will feature advanced hyperspectral and directional optical imaging instruments. Given the complex nature of the data to be expected from these missions, a thorough preparation for them is essential and this can be accomplished by realistic simulation of the imagery data, years before the actual launch. Based on given spectral and directional capabilities of the instrument, and in combination with biophysical land surface properties obtained from existing imagery, the spectral and directional responses of several types of vegetation and bare soil have been simulated pixel by pixel using the radiative transfer models PROSPECT (for hyperspectral leaf reflectance and transmittance), GeoSAIL (for two-layer canopy bidirectional spectral reflectance), and MODTRAN4 (for atmospheric hyperspectral and directional effects). In this way, one obtains realistically simulated hyperspectral and directional top-of-atmosphere spectral radiance images, with all major effects included, such as heterogeneity of the landscape, non-Lambertian reflectance of the land surface, the atmospheric adjacency effect, and the limited spatial resolution of the instrument. The output of the image simulations can be used to demonstrate the capabilities of future earth observation missions. In addition, instrument specifications and image acquisition strategies might be optimized on the basis of simulated image analysis results, and new advanced data assimilation procedures could be validated with realistic inputs under controlled circumstances. This paper describes the applied methodology, the study area with the input images, the set-up of the actual image simulations, and discusses the final results obtained.     

Aspects of operational atmospheric correction of hyperspectral imagery

The large number of spectral bands of hyperspectral instruments and the time required for the calculation of atmospheric look-up tables and the reflectance image cube pose very challenging requirements on an operational processing facility. This contribution presents some aspects and suggestions to reduce the processing time. Essential components are a precalculated database with a reduced number of spectral bands, an interactive phase to determine the appropriate atmospheric parameters, and a choice between medium and high accuracy levels for the atmospheric correction. The medium accuracy levels work with look-up tables for a reduced number of spectral bands employing interpolation for the channels omitted in the look-up tables. The high accuracy level uses tables for all channels and includes the scan angle dependence of the atmospheric radiance and transmittance functions. These ideas were successfully implemented and tested during several airborne hyperspectral campaigns resulting in an estimated time saving of a factor 3-7. The deviations of field measured reflectance spectra and spectra retrieved from airborne HyMap imagery are in the range of 2-3% or better.     

Evaluation of supervised classification algorithms for identifying crops using airborne hyperspectral data

Sufficient training data must be acquired to classify areas of interest using a supervised classification method and hyperspectral data. However, the relatively small size of agricultural plots in Japan means that there is no training area large enough to represent a feature of interest. In this study, a new method for identifying crops using hyperspectral remotely sensed data has been proposed in order to resolve the problem of identifying training areas in agricultural crops. This method was then compared with conventional methods. The proposed method was found to be most effective for identifying crops using hyperspectral data in an agricultural land area.     

Daytime fire detection using airborne hyperspectral data

The shortwave infrared region of the electromagnetic spectrum, covering wavelengths from 1400 to 2500 nm, can include significant emitted radiance from fire. There have been relatively few evaluations of the utility of shortwave infrared remote sensing data, and in particular hyperspectral remote sensing data, for fire detection. We used an Airborne Visible InfraRed Imaging Spectrometer (AVIRIS) scene acquired over the 2003 Simi Fire to identify the hyperspectral index that was able to most accurately detect pixels containing fire. All AVIRIS band combinations were used to calculate normalized difference indices, and kappa was used to compare classification ability of these indices for three different fire temperature ranges. The most accurate index was named the Hyperspectral Fire Detection Index (HFDI). The HFDI uses shortwave infrared bands centered at 2061 and 2429 nm. These bands are sensitive to atmospheric attenuation, so the impacts of variable elevation, solar zenith angle, and atmospheric water vapor concentration on HFDI were assessed using radiative transfer modeling. While varying these conditions did affect HFDI values, relative differences between background HFDI and HFDI for 1% fire pixel coverage were maintained. HFDI is most appropriate for detection of flaming combustion, and may miss lower temperature smoldering combustion at low percent pixel coverage due to low emitted radiance in the shortwave infrared. HFDI, two previously proposed hyperspectral fire detection indices, and a broadband shortwave infrared-based fire detection index were applied to AVIRIS scenes acquired over the 2007 Zaca Fire and 2008 Indians Fire. A qualitative comparison of the indices demonstrated that HFDI provides improved detection of fire with less variability in background index values.     

Coupled soil-leaf-canopy and atmosphere radiative transfer modeling to simulate hyperspectral multi-angular surface reflectance and TOA radiance data

Coupling radiative transfer models for the soil background and vegetation canopy layers is facilitated by means of the four-stream flux interaction concept and use of the adding method. Also the coupling to a state-of-the-art atmospheric radiative transfer model like MODTRAN4 can be established in this way, thus enabling the realistic simulation of top-of-atmosphere radiances detected by space-borne remote sensing instruments. Possible applications of coupled modeling vary from mission design to parameter retrieval and data assimilation. This paper introduces a modified Hapke soil BRDF model, a robust version of the PROSPECT leaf model, and a modernized canopy radiative transfer model called 4SAIL2. The latter is a hybrid two-layer version of SAIL accommodating horizontal and vertical heterogeneities, featuring improved modeling of the hot spot effect and output of canopy absorptances. The integrated model is simply called SLC (soil-leaf-canopy) and has been implemented as a speed-optimized Windows DLL which allows efficient use of computer resources even when simulating massive amounts of hyperspectral multi-angular observations. In this paper various examples of possible model output are shown, including simulated satellite image products. First validation results have been obtained from atmospherically corrected hyperspectral multi-angular CHRIS-PROBA data of the Upper Rhine Valley in Germany.     

Directional reflectance properties determined by analysis of airborne multispectral scanner data and atmospheric correction

Directional reflectance properties of natural surfaces are very important in the interpretation of remotely sensed data. The analysis of multispectral scanner data shows a distinct dependence on scan angle, wavelength (0.4–1.1 μm), and classes (e.g., bare soil, vegetation). This relationship can be described by polynomials determined by regression methods. Atmospheric effects are computed with a simple model by parametrization of the multiple scattered skylight. The model permits a quick and sufficient estimation of the airlight, depending on the data collection conditions. Comparisons of the scanner data with the corresponding model yield the following results: The differences between airborne and ground measurements are due to atmospheric effects. The directional variation in brightness is mainly caused by the object itself with the exception of short wavelengths and/or a very low albedo. The hue shift of vegetation is essentially produced by the object and modified by the atmosphere. Taking into account the directional reflectance properties in a direction-dependent classification procedure results in an improvement of up to 20% in comparison with algorithms used so far. Quantitative relationships between ground measurements and radiation measurements by airborne sensors including atmospheric effects can be determined with the proposed methods.     

A comparison between two radiative transfer models for atmospheric correction over a wide range of wavelengths

Correction of atmospheric effects caused by molecular and particulate scattering and absorption from hyperspectral measurements is desirable whenever comparisons are to be made with data acquired under different atmospheric or viewing conditions. Absolute atmospheric correction of optical remotely sensed data relies on Radiative Transfer Codes (RTCs) capable of simulating interactions between the radiation and the medium. Several RTCs are available with different features, but the two most widely used nowadays are the Second Simulation of Satellite Signal in the Solar Spectrum (6S) and the MODerate-Resolution TRANsmittance Code (MODTRAN). In this paper, we perform a comparison between these two models and report the results obtained, specifically investigating the differences that emerged during the various test stages.     

Comparison of two atmospheric correction methods for the classification of spaceborne urban hyperspectral data depending on the spatial resolution

For remote-sensing applications such as spectra classification or identification, atmospheric correction constitutes a very important pre-processing step, especially in complex urban environments where a lot of phenomenons alter the shape of the signal. The objective of this article is to compare the efficiency of two atmospheric correction algorithms, COCHISE (atmospheric COrrection Code for Hyperspectral Images of remote-sensing SEnsors) and an empirical method, on hyperspectral data and for classification applications. Classification is carried out on several simulated spaceborne data sets with different spatial resolutions (from 1.6 to 9.6 m). Four classifiers are considered in the study: a k-means, a Support Vector Machine (SVM), and a sun/shadow version of each of them, which processes sunlit and shadowed pixels separately. Results show that the most relevant atmospheric method for classification depends on the spatial resolution of the processed data set. Indeed, if the empirical method performs better on high-resolution data sets (up to 4%), its superiority fades out as the spatial resolution decreases, especially with the lower spatial resolution where COCHISE can be 10% more accurate than the empirical method.     

Evaluation of MODIS SWIR and NIR-SWIR atmospheric correction algorithms using SeaBASS data

Using the NASA maintained ocean optical and biological in situ data that were collected during 2002-2005, we have evaluated the performance of atmospheric correction algorithms for the ocean color products from the Moderate Resolution Imaging Spectroradiometer (MODIS) on Aqua. Specifically, algorithms using the MODIS shortwave infrared (SWIR) bands and an approach using the near-infrared (NIR) and SWIR combined method are evaluated, compared to the match-up results from the NASA standard algorithm (using the NIR bands). The in situ data for the match-up analyses were collected mostly from non-turbid ocean waters. It is critical to assess and understand the algorithm performance for deriving MODIS ocean color products, providing science and user communities with the important data quality information. Results show that, although the SWIR method for data processing has generally reduced the bias errors, the noise errors are increased due mainly to significantly lower sensor signal-noise ratio (SNR) values for the MODIS SWIR bands, as well as the increased uncertainties using the SWIR method for the atmospheric correction. This has further demonstrated that future ocean color satellite sensors will require significantly improved sensor SNR performance for the SWIR bands. The NIR-SWIR combined method, for which the non-turbid and turbid ocean waters are processed using the NIR and SWIR method, respectively, has been shown to produce improved ocean color products.     

Application of underwater hyperspectral data for color correction purposes

Color correction of underwater images has been considered a difficult task for a number of reasons. Those include severe absorption of the water column, the unpredictable behavior of light under the water surface, limited access to reliable data for correction purposes, and the fact that we are only able to process three spectral channels, which is insufficient for most color correction applications. Here, the authors present a method to estimate a hyperspectral image from an RGB image and pointwise hyperspectral data. This is then used to color correct the hyperspectral underwater image and transform it back into RGB color space.     

The influence of urban structures on impervious surface maps from airborne hyperspectral data

Airborne hyperspectral data fulfills the high spectral and spatial resolution requirements of urban remote sensing applications. Its high spectral information content enables delineating impervious areas including the separation of built-up and non built-up surfaces, thus being of high relevance for many urban environmental applications. However, two phenomena related to surface structure negatively impact the accuracy of maps from such airborne data sets: (1) displaced buildings that lead to confusion between the class built-up and adjacent non built-up areas as a function of building height and view-angle; (2) urban street trees obscuring impervious surface underneath. Both effects have so far not been investigated from airborne hyperspectral data and potential sources of inaccuracy are usually not differentiated in analysis utilizing such data. Thus, the positive influence of hyperspectral information might have been undervalued in many cases. We set up an analysis scheme that allows for separately quantifying sources of error when producing land cover maps from urban areas. Given reliable cadastral information on building extent and street network, a detailed analysis for a relatively large Hyperspectral Mapper data set acquired over Berlin, Germany, was performed. Results show that both building displacement and impervious surface obscured by tree crowns are of great impact: at large view-angles, building displacement adds up to 16% error compared to nadir regions; more than 30% of the street area is classified as vegetation. Moreover, both effects show irregularities that prohibit empirical correction: misclassification due to building displacement also depends on view-direction, i.e. illumination properties and shadow, while the influence of trees differs significantly along streets and inside residential areas. Results from this work underline the necessity to consider all image processing steps when evaluating the accuracy and reliability of remote sensing products and they depict directions for future methodological development.     

An optimization approach to estimate and calibrate column water vapour for hyperspectral airborne data

The article describes a novel approach to estimate and calibrate column water vapour (CWV), a key parameter for atmospheric correction of remote-sensing data. CWV is spatially and temporally variable, and image-based methods are used for its inference. This inference, however, is affected by methodological and numeric limitations, which likely propagate to reflectance estimates. In this article, a method is proposed to estimate CWV iteratively from target surface reflectances. The method is free from assumptions for at sensor radiance-based CWV estimation methods. We consider two cases: (a) CWV is incorrectly estimated in a processing chain and (b) CWV is not estimated in a processing chain. To solve (a) we use the incorrect estimations as initial values to the proposed method during calibration. In (b), CWV is estimated without initial information. Next, we combined the two scenarios, resulting in a generic method to calibrate and estimate CWV. We utilized the hyperspectral mapper (HyMap) and airborne prism experiment (APEX) instruments for the synthetic and real data experiments, respectively. Noise levels were added to the synthetic data to simulate real imaging conditions. The real data used in this research are cloud-free scenes acquired from the airborne campaigns. For performance assessment, we compared the proposed method with two state-of-the-art methods. Our method performed better as it minimizes the absolute error close to zero, only within 8–10 iterations. It thus suits existing operational chains where the number of iterations is considerable. Finally, the method is simple to implement and can be extended to address other atmospheric trace gases.     

Detection of initial damage in Norway spruce canopies using hyperspectral airborne data

Current broadband sensors are not capable of separating the initial stages of forest damage. The current investigation evaluates the potential of hyperspectral data for detecting the initial stages of forest damage at the canopy level in the Norway spruce (Picea abies (L.) Karst) forests of Czech Republic. Hyperspectral canopy reflectance imagery and foliar samples were acquired contemporaneously for 23 study sites in August 1998. The sites were selected along an air pollution gradient to represent the full range of damage conditions in even-aged spruce forests. The changes in canopy and foliar reflectance, chemistry and pigments associated with forest damage were established. The potential of a large number of spectral indices to identify initial forest damage was determined. Canopy hyperspectral data were able to separate healthy from initially damaged canopies, and therefore provided an improved capability for assessment of forest physiology as compared to broadband systems. The 673-724 nm region exhibited maximum sensitivity to initial damage. The nine spectral indices having the highest potential as indicators of the initial damage included: three simple band ratios, two derivative indices, two modelled red-edge parameters and two normalized bands. The sensitivity of these indices to damage was explained primarily by their relationship to foliar structural chemical compounds, which differed significantly by damage class.     

On the application of the MODTRAN4 atmospheric radiative transfer code to optical remote sensing

The quantification of atmospheric effects on the solar radiation measured by a spaceborne or airborne optical sensor is required for some key tasks in remote sensing, such as atmospheric correction, simulation of realistic scenarios or retrieval of atmospheric parameters. The MODTRAN4 code is an example of state‐of‐the‐art atmospheric radiative transfer code, as it provides very accurate calculations by means of a rigorous mathematical formulation and a very fine spectral resolution. However, the application of MODTRAN4 to remote sensing is not straightforward for the average user for a number of reasons: the provided output parameters do not exactly correspond to those necessary for the construction of the at‐sensor signal by combination with the surface reflectance, an advanced knowledge of radiative transfer theory and atmospheric physics is needed for the understanding of the input parameters and all their possible combinations, and the computation time may be too high for many practical applications. This work is intended to give explicit solutions to those problems. MODTRAN4 has been modified so that the proper atmospheric parameters are calculated and delivered as output. In addition, the most important execution options are investigated, and the compromise between accuracy and computation time is analysed. The performance of the proposed methodology is demonstrated by generating a look‐up table (LUT) enabling fast but accurate radiative transfer calculations for the atmospheric correction of data acquired by the Compact High Resolution Imaging Spectrometer (CHRIS) on board the Project for On‐Board Autonomy (PROBA).     

Semi-automatic geometric correction of airborne hyperspectral push-broom images using ground control points and linear features

A semi-automatic geometric correction method is presented for high-resolution airborne hyperspectral push-broom images in this paper. The method mainly consists of a correction model based on ground control points and linear features as well as a semi-automatic extraction procedure for linear features. With a panchromatic image as the reference map, the distorted hyperspectral image can be calibrated using ground control points and linear features digitized on both the map and the distorted image. In particular, the ground control points are chosen manually, and the linear features can be extracted semi-automatically through the combination of region segmentation based on the region-scalable fitting energy model and a further correction procedure. The experimental results of two hyperspectral images demonstrate that our method can achieve visually well-rectified images and high calibration accuracy.     

Improving within-genus tree species discrimination using the discrete wavelet transform applied to airborne hyperspectral data

Discrete wavelet analysis was assessed for its utility in aiding discrimination of three pine species (Pinus spp.) using airborne hyperspectral data (AVIRIS). Two different sets of Haar wavelet features were compared to each other and to calibrated radiance, as follows: (1) all combinations of detail and final level approximation coefficients and (2) wavelet energy features rather than individual coefficients. We applied stepwise discriminant techniques to reduce data dimensionality, followed by discriminant techniques to determine separability. Leave-one-out cross validation was used to measure the classification accuracy. The most accurate (74.2%) classification used all combinations of detail and approximation coefficients, followed by the original radiance (66.7%) and wavelet energy features (55.1%). These results indicate that application of the discrete wavelet transform can improve species discrimination within the Pinus genus.     

Spectral unmixing of airborne hyperspectral data for baseline mapping of mine tailings areas

The Kam Kotia mine tailings areas near Timmins in Ontario, Canada have been generating and discharging acidic mine drainage (AMD) into the surrounding areas for more than 35 years, killing large areas of forest and polluting the local water system. This paper presents results from the remote sensing monitoring programme in the Kam Kotia mine. Hyperspectral TRW (Thompson Ramo Wooldridge Inc.) Imaging Spectrometer III data were acquired over the Kam Kotia mine and tailings areas. This paper describes (1) the data pre‐processing (noise removal, atmospheric correction, spectral smile correction, scene‐based calibration) needed to radiometrically calibrate the images and (2) a novel procedure which combines constrained spectral mixture analysis and threshold‐based classification. With this developed procedure one can retrieve fraction maps of major mine tailings‐related surface materials and hence generate a surface map separating green vegetation, transition zones, dead vegetation, and oxidized tailings, and calculate the extent (surficial area) of each of the zones. The four zones are correlated with the extent and degree of vegetation cover affected by tailings material and are interpreted to span respectively from very low to medium, high, and very high AMD pollution. This procedure can be used to monitor changes in the course of the boundary between affected zones and finally quantify the rehabilitation process in mine tailings areas with high vegetation cover.     

Water correction for improved benthic vegetation signal using satellite-borne hyperspectral data

Benthic mapping employs field surveys, hydroacoustic measurements, aerial photography, and satellite imagery. Effective benthic mapping involves removing overlying water effects from atmospherically corrected remotely sensed data to enhance signals from the seafloor. Our previous water correction algorithms depend on controlled laboratory measurements of substrates in clear water, which had challenges for replication. A more simplified water correction algorithm is presented, which uses bathymetry and only a few pixels from the image. Spectral profiles were extracted from four pixels in a Hyperspectral Imager for Coastal Oceans (HICO) image that was acquired in February 2014 over Indian River Lagoon, Florida. The four locations were chosen based on the assumption there were two types of homogeneous substrates at two depths. Our new algorithm calculates water column reflectance and water absorption at the instance of image data acquisition directly from the four pixel values. Water correction demonstrates improved benthic feature depiction including the near-infrared signals for benthic vegetation. A simple ratio was applied to the corrected image and demonstrates restored submerged vegetation signals.     

Simulation of atmospheric profile retrieval from hyperspectral infrared data under cloudy conditions

In this paper, simulated space-based high spectral resolution AIRS (Atmospheric Infrared Sounder) infrared radiances with different cloud top heights and effective cloud fractions are used to demonstrate measurement sensitivity and atmospheric profile retrieval performance. Simulated cloudy retrievals of atmospheric temperature and moisture derived from the statistical eigenvector regression algorithm are analysed with different effective cloud fractions and different cloud heights. The results show that knowledge of cloud height is critical to sounding retrieval performance and the root mean square error of retrieved temperature and the mixed ratio of water vapour below the cloud top increases with effective cloud fraction. When there is 50 hPa error in the cloud height the retrieval accuracy of temperature and humidity decrease, compared with when the cloud height is known perfectly; the temperature retrieval is more sensitive to cloud height error than humidity retrieval. Collocated cloudy AIRS and the associated clear MODIS (Moderate Resolution Imaging Spectroradiometer) infrared observations within the AIRS field of view (FOV) are also used to demonstrate profile retrieval improvement below the cloud layer. It is demonstrated that using collocated clear MODIS multispectral imager data along with AIRS high spectral resolution infrared radiances can greatly improve the single FOV cloudy retrieval even under opaque cloudy conditions.     

Atmospheric correction of hyperspectral airborne GCAS measurements over the Louisiana Shelf using a cloud shadow approach

As an image-driven method to correct for atmospheric effects, the cloud shadow (CS) approach does not require accurate radiometric calibration of the sensor, making it feasible to process remotely sensed data when radiometric calibration may contain non-negligible uncertainties. Using measurements from the Geostationary Coastal and Air Pollution Events Airborne Simulator and from the Moderate Resolution Imaging Spectroradiometer over the Louisiana Shelf, we evaluate the CS approach to airplane measurements in turbid-water environments. The original CS approach somehow produced remote-sensing reflectance (R_rs, sr^?1) with an abnormal spectral shape, likely a result of the assumption of identical path radiance for the pair of pixels in and out of the shadow, which is not exactly valid for measurements made from a low-altitude airplane. To overcome this limitation, an empirical scheme using an effective wavelength-dependent radiance reflectance for the cloud (γ, sr^?1) was developed and reasonable GCAS R_rs retrievals are then generated, which were further validated against in situ R_rs. Issues and challenges in applying CS to measurements of low-altitude airplanes are discussed.