A temperature-controlled spectrometer system for continuous and unattended measurements of canopy spectral radiance and reflectance

This paper describes the development of a fully automated system for collecting high-resolution spectral data over a forested footprint. The system comprises a pair of off-the-shelf spectrometers in a custom-built thermal enclosure with a fixed off-nadir downward (target)-pointing fibre and upward-pointing fibre for irradiance measurement. Both instruments sample simultaneously via custom-written and user-controlled software during all weathers and sky conditions. The system is mounted on a 25 m eddy covariance scaffolding tower, approximately 7 m from a Scots pine forest canopy. The system was installed at the University of Helsinki’s SMEAR-II Field Station in Hyytiälä in March 2010 and has been operating continuously through a joint programme between the Universities of Edinburgh and Helsinki. The system was designed to capture diurnal and seasonal variation in vegetation light-use efficiency and fluorescence through the capture and analysis of well-defined narrow spectral features, but its implementation would permit the extraction of further optical signals linked to vegetation biophysical variables, and provide a continuous data stream with which to validate satellite data products including vegetation indices such as the photochemical reflectance index (PRI) as well as spectral indicators of solar induced fluorescence.     

A path radiance estimation algorithm using reflectance measurements in radiometric control areas

A new algorithm of path radiance estimation based on measurements of surface reflectance at radiometric control areas (RCAs) is proposed. Path radiance estimates of the proposed RCA-based method were compared against estimates of other methods including the dark object subtraction (DOS) method, the multi-band regression (MBR) technique and the covariance matrix method (CMM). The RCA-based method is superior to other methods based on three qualitative assessment criteria and a quantitative assessment based on measurements of molecule and aerosol optical depths (AODs) available from the Aerosol Robotic Network (AERONET) and Moderate Resolution Imaging Spectroradiometer (MODIS) data archive. Asphalt-paved surface, which could be easily identified in most images, was also found to be a good choice for RCAs. The DOS method and the CMM tend to overestimate path radiances. Although in our study the MBR technique and the RCA-based method seem to perform equally well, estimates of the MBR technique may be less reliable.     

Estimating forest variables from top-of-atmosphere radiance satellite measurements using coupled radiative transfer models

Traditionally, it is necessary to pre-process remote sensing data to obtain top of canopy (TOC) reflectances before applying physically-based model inversion techniques to estimate forest variables. Corrections for atmospheric, adjacency, topography, and surface directional effects are applied sequentially and independently, accumulating errors into the TOC reflectance data, which are then further used in the inversion process. This paper presents a proof of concept for demonstrating the direct use of measured top-of-atmosphere (TOA) radiance data to estimate forest biophysical and biochemical variables, by using a coupled canopy-atmosphere radiative transfer model. Advantages of this approach are that no atmospheric correction is needed and that atmospheric, adjacency, topography, and surface directional effects can be directly and more accurately included in the forward modelling.In the case study, we applied both TOC and TOA approaches to three Norway spruce stands in Eastern Czech Republic. We used the SLC soil-leaf-canopy model and the MODTRAN4 atmosphere model. For the TOA approach, the physical coupling between canopy and atmosphere was performed using a generic method based on the 4-stream radiative transfer theory which enables full use of the directional reflectance components provided by SLC. The method uses three runs of the atmosphere model for Lambertian surfaces, and thus avoids running the atmosphere model for each new simulation. We used local sensitivity analysis and singular value decomposition to determine which variables could be estimated, namely: canopy cover, fraction of bark, needle chlorophyll, and dry matter content. TOC and TOA approaches resulted in different sets of estimates, but had comparable performance. The TOC approach, however, was at its best potential because of the flatness and homogeneity of the area. On the contrary, the capacities of the TOA approach would be better exploited in heterogeneous rugged areas. We conclude that, having similar performance, the TOA approach should be preferred in situations where minimizing the pre-processing is important, such as in data assimilation and multi-sensor studies.     

A statistical complement to deterministic algorithms for the retrieval of aerosol optical thickness from radiance data

As a complement to the conventional deterministic geophysical algorithms, we consider a faster, but less accurate approach: training regression models to predict aerosol optical thickness (AOT) from radiance data. In our study, neural networks trained on a global data set are employed as a global retrieval method. Inverse distance spatial interpolation and region-specific neural networks trained on restricted, localized areas provide local models. We then develop two integrated statistical methods: local error correction of global retrievals and an optimal weighted average of global and local components. The algorithms are evaluated on the problem of deriving AOT from raw radiances observed by the Multi-angle Imaging SpectroRadiometer (MISR) instrument onboard NASA's Terra satellite. Integrated statistical approaches were clearly superior to global and local models alone. The best compromise between speed and accuracy was obtained through the weighted averaging of global neural networks and spatial interpolation. The results show that, while much faster, statistical retrievals can be quite comparable in accuracy to the far more computationally demanding deterministic methods. Differences in quality vary with season and model complexity.     

Correcting atmospheric masking to retrieve the spectral albedo of land surfaces from satellite measurements

A radiative transfer approach to the problem of atmospheric correction of satellite images in the solar spectral range is presented which includes all multiple scattering processes without any approximation. The numerical solution is accepted as satisfying, if the numerical accuracy is better than I per cent. This means that the accuracy of the atmospheric correction depends almost exclusively on the quality of the auxiliary data on the atmospheric state and the surface reflection indicatrix. Byextensivemodel calculations these parameter driven error bounds have been quantified. Thus the calculation results in a corrected albedo image with specified error bounds. This seems to be the first algorithm available for atmospheric correction of real imagery data which relies on a numerical exact treatment of multiple scattering. The program EXACT (EXact Atmospheric Correction Technique) has so far been used with Landsat Thematic Mapper (TM), NOAA AVHRR (National Oceanic and Atmospheric Administration Advanced Very High Resolution Radiometer) and also with airborne Daedalus ATM images. The algorithm has been validated by comparison of satellite data to ground measurements and between different sensors. Errors of the derived albedos were found to remain below 0·01 for visible and near-infrared sensor channels of this set of radiometers.     

Modeling spectral reflectance of optically complex waters using bio-optical measurements from Tokyo Bay

This study presents an approach for optimally parameterizing a reflectance model. A parameterization scheme is realized based on a comprehensive bio-optical data set, including subsurface downwelling and upwelling irradiance spectra, absorption spectra of particle and dissolved substances, as well as chlorophyll and total suspended matter concentrations at 45 stations near Tokyo Bay between 1982 and 1984. The irradiance reflectance model is implemented with three-component inherent optical property submodels.In this parameterization scheme, an unsupervised classification was applied in the hyper-spectral space of reflectance, leading to three spectrally distinct optical water types. The reflectance model was parameterized for the entire data set, and then parameterized for each of the water types. The three sets of type-specific model parameters, which define corresponding IOP submodels, are believed to accommodate differences in the optical properties of the in-water constituents. The parameterized reflectance model was evaluated by both reconstructing measured reflectance spectra and solving for the nonlinear inverse problem to retrieve in-water constituent concentrations. The model accuracy was significantly improved in the forward direction for classified waters over that of non-classified waters, but no significant improvement was achieved in the retrieval accuracy (inverse direction). A larger data set with greater resolution of constituent inherent optical properties would likely improve the modeling results.     

Radiative transfer equation with surface scattering for ocean and atmospheric parameters retrieval from radiometric measurements

Microwave radiometers have been used with success in the past for the retrieval of ocean surface and atmospheric parameters. However,for the most effective use of the measurements provided by these instruments, it is neeessary that performant inversion algorithms be available. Existing algorithms have already proved their usefulness but are often based on statistical regressions on the measurements from some particular experiment or on artifical data sets. Progress towards better performances still remains possible. We have developed an iterative retrieval procedure based on a physical scattering and propagation model for observing the ocean surface and the intervening atmosphere. A correction for scattering by the roughened sea surface is proposed and introduced in a previouslydeveloped contracted form of the radiative transfer equation. This correction is modelled for a range of frequencies from 5 to 40GHz, and represented in the form of a polynomial expansion depending on two parameters only: the total atmospheric attenuation, and the wind friction velocity. Results of inversions on simulated measurements obtained with this approximate model are presented and discussed.     

The error in satellite retrieved temperature profiles due to the effects of atmospheric aerosol particles

Radiances were computed for the eight temperature sounding infrared channels of the SIRS-B instrument for a clear model and for two very hazy models of the atmosphere. Temperature profiles were retrieved from these radiances using the minimum-information technique. For the retrieval, it was assumed that the computed radiances were measured by a noise-free instrument and the atmosphere was clear. Also, the true model temperature profile was used as the initial profile. As a result, deviations of the retrieved profile from the true one occurred only for haze models and were due solely to the neglect of aerosol-particle absorption and scattering in the retrieval process. Temperature errors were greatest for the haze model with large stratospheric aerosol-particle concentration. The rms temperature error between 100 and 1000 mbar for this model was 0.7°C. The largest deviations occurred at 966 mbar (1.5°C warmer) and 673 mbar (1.2°C cooler). Since the atmosphere is typically much less hazy than these models represent, we have concluded that the error caused by the neglect of aerosol effects is a small part of the rms error computed from radiosonde intercomparison analyses ($\text{≈2.5°}\text{C}$).     

A parallel spectral model for atmospheric transport processes

The paper describes a parallel implementation of a grand challenge problem: global atmospheric modeling. The novel contributions of our work include (1) a detailed investigation of opportunities for parallelism in atmospheric global modeling based on spectral solution methods, (2) the experimental evaluation of overheads arising from load imbalances and data movement for alternative parallelization methods, and (3) the development of a parallel code that can be monitored and steered interactively based on output data visualizations and animations of program functionality or performance. Code parallelization takes advantage of the relative independence of computations at different levels in the earth's atmosphere, resulting in parallelism of up to 40 processors, each independently performing computations for different atmospheric levels and requiring few communications between different levels across model time steps. Next, additional parallelism is attained within each level by taking advantage of the natural parallelism offered by the spectral computations being performed (e.g. taking advantage of independently computable terms in equations). Performance measurements are performed on a 64-node KSR2 supercomputer. However, the parallel code has been ported to several shared memory parallel machines, including SGI multiprocessors, and has also been ported to distributed memory platforms like the IBM SP-2.     

A spectral directional reflectance model of row crops

A computationally efficient reflectance model for row planted canopies is developed in this paper through separating the contributions of incident direct and diffuse radiation scattered by row canopies. The row model allows calculating the reflectance spectrum in any given direction for the optical spectral region. The performance of the model is evaluated through comparisons with field measurements of winter wheat as well as with an established 3D computer simulation model. Especially the systematic comparisons with the computer simulation model demonstrate that the model can adequately simulate the characteristic distribution of directional reflectance factors of row canopies, which is shown in the polar map of reflectance as a high or low value stripe approximately parallel to the row orientation, besides the hotspot effect. Physical mechanisms causing the dynamics were proposed and supported by comparison studies. The features of reflectance distributions of row canopies, which are distinctively different from those of homogeneous canopy, imply that it is problematic to use one-dimensional radiation transfer model to interpret radiation data and estimate the structural or spectral parameters of row canopies from reflectance measurements. Finally, further improvements needed for the current model are briefly discussed.     

Models for surface reflection of radiance and polarized radiance: Comparison with airborne multi-angle photopolarimetric measurements and implications for modeling top-of-atmosphere measurements

In this paper, we investigate the surface-atmosphere radiative interaction in application to the problem of aerosol satellite remote sensing over land. First, we test different models of the Bidirectional Reflectance and Polarization Distribution Function (BRDF and BPDF) for bare soil and vegetation surfaces using multi-angle, multi-spectral photopolarimetric airborne measurements of the Research Scanning Polarimeter (RSP). Then, we investigate the performance of different models of BRDF and BPDF for modeling top-of-atmosphere measurements. We have found that different BRDF models can describe the RSP measurements equally well. However, for soil surfaces, the different BRDF models show a different dependence on illumination geometry (solar zenith and azimuth angles), as well as a different dependence on viewing angle outside the range of RSP measurements. This implies that different models describe the surface-atmosphere interaction differently, leading for soil surfaces to differences in the top-of-atmosphere reflectance up to 4-5%, whereas at surface level the models agree within 2% for RSP illumination and measurement geometry. For vegetation, the different BRDF models show more similar dependence on illumination geometry, meaning that, in general, the differences in top-of-atmosphere reflectances are smaller than the differences in surface total reflectances. For the BPDF, we compare the empirical model of Nadal and Breon (1999) and the model developed by Maignan et al. (2009) with a newly developed model. The latter model compares better with RSP measurements. It was shown that, though all models have essentially different angular profiles at different illumination and viewing geometries, the difference of the top-of-atmosphere degree of linear polarization is less or is of the same order as the degree of linear polarization difference at the surface level taken at RSP illumination and measurement geometry. For the considered models, it can be up to 0.015 but is mostly below 0.005.     

Retrieval of surface directional reflectance properties using ground level multiangle measurements

Knowledge of the directional reflectance properties of natural surfaces such as soils and vegetation canopies is essential for classification studies and canopy model inversion. Atmospheric correction schemes, using various levels of approximation, are described to retrieve surface bidirectional reflectance factors (BRFs) and directionalhemispherical reflectances (albedos) from multiangle radiance measurements taken at ground level. The retrieval schemes are tested on simulated data incorporating realistic surface BRFs and atmospheric models containing aerosols. Sensitivity of the atmospherically corrected BRFs and associated directional-hemispherical reflectances to various aerosol properties and the sun-view geometry is illustrated. A measurement strategy for obtaining highly accurate surface reflectance properties also is examined in the context of instrument radiometric calibration, knowledge of the atmospheric properties, and sun-view angular coverage.     

Coupled retrieval of aerosol optical thickness, columnar water vapor and surface reflectance maps from ENVISAT/MERIS data over land

An algorithm for the derivation of atmospheric parameters and surface reflectance data from MEdium Resolution Imaging Specrometer Instrument (MERIS) on board ENVIronmental SATellite (ENVISAT) images has been developed. Geo-rectified aerosol optical thickness (AOT), columnar water vapor (CWV) and spectral surface reflectance maps are generated from MERIS Level-1b data over land. The algorithm has been implemented so that AOT, CWV and reflectance products are provided on an operational manner, making no use of ancillary parameters apart from those attached to MERIS products. For this reason, it has been named Self-Contained Atmospheric Parameters Estimation from MERIS data (SCAPE-M). The fundamental basis of the algorithm and applicable error figures are presented in the first part of this paper. In particular, errors of ± 0.03, ± 4% and ± 8% have been estimated for AOT, CWV and surface reflectance retrievals, respectively, by means of a sensitivity analysis based on a synthetic data set simulated under a usual MERIS scene configuration over land targets. The assumption of a fixed aerosol model, the coarse spatial resolution of the AOT product and the neglection of surface reflectance directional effects were also identified as limitations of SCAPE-M. Validation results are detailed in the second part of the paper. Comparison of SCAPE-M AOT retrievals with data from AErosol RObotic NETwork (AERONET) stations showed an average Root Mean Square Error (RMSE) of 0.05, and an average correlation coefficient R² of about 0.7-0.8. R² values grew up to more than 0.9 in the case of CWV after comparison with the same stations. A good correlation is also found with the MERIS Level-2 ESA CWV product. Retrieved surface reflectance maps have been successfully compared with reflectance data derived from the Compact High Resolution Imaging Spectrometer (CHRIS) on board the PRoject for On-Board Autonomy (PROBA) in the first place. Reflectance retrievals have also been compared with reflectance data derived from MERIS images by the Bremen AErosol Retrieval (BAER) method. A good correlation in the red and near-infrared bands was found, although a considerably higher proportion of pixels was successfully processed by SCAPE-M.     

Structured neural network modelling of multi-valued functions for wind vector retrieval from satellite scatterometer measurements

A conventional neural network approach to regression problems approximates the conditional mean of the output vector. For mappings which are multi-valued this approach breaks down, since the average of two solutions is not necessarily a valid solution. In this article mixture density networks, a principled method for modelling conditional probability density functions, are applied to retrieving Cartesian wind vector components from satellite scatterometer data. A hybrid mixture density network is implemented to incorporate prior knowledge of the predominantly bimodal function branches. An advantage of a fully probabilistic model is that more sophisticated and principled methods can be used to resolve ambiguities.     

Algorithm for retrieval of the effective snow grain size and pollution amount from satellite measurements

We present an algorithm for retrieval of the effective Snow Grain Size and Pollution amount (SGSP) from satellite measurements. As well as our previous version (Zege et al., 2008, 1998), the new algorithm is based on the analytical solution for snow reflectance within the asymptotic radiative transfer theory. The SGSP algorithm does not use any assumptions on snow grain shape and allows for the snow pack bidirectional reflectance distribution function (BRDF). The algorithm includes a new atmospheric correction procedure that allows for snow BRDF. This SGSP algorithm has been thoroughly validated with computer simulations. Its sensitivity to the atmosphere model has been investigated. It is shown that the inaccuracy of the snow characteristic retrieval due to the uncertainty in the aerosol and molecular atmosphere model is negligible, as compared to that due to the measurement errors at least for aerosol loads typical for polar regions. The significant advantage of the SGSP over conventional algorithms, which use a priori assumptions about particle shape and (or) not allow for the BRDF of the individual snow pack, is that the developed retrieval still works at low sun elevations, which are typical for polar regions.     

Evaluation of measurement errors in ground surface reflectance for satellite calibration

Abstract

One of the more efficient methods used for in-flight calibration of Earth resource satellites is based on measurements performed at ground level on a test site. An experimental study has been conducted in La Crau Sèche (south east France), where a calibration site for SPOT satellites is intended. The accuracy of the calibration depends, critically, on the accuracy of ground bidirectional reflectance factor (BDRF) measurements.

All of the different sources of error are analysed. These are due to two series of factors depending on the characteristics of the radiometer (electronic charac teristics, absolute calibration, angular setting of the radiometer) and of the ground surface (the spectral, spatial, angular and temporal variability of the BDRF). The relative weight of these different causes of error is determined from experimental data. This analysis shows that, besides the well-known disturbing factors such as the calibration of the radiometer and the spatial variability of the BDRF, two other factors can introduce large measurement errors: the spectral and angular variability of reflectance of the site.

This detailed analysis of the different causes of error is not only valid for the calibration of a satellite, but it can also be used to draw up guidelines for performing accurate BDRF measurements in natural conditions for any application.     

Estimating surface temperatures from satellite thermal infrared data—A simple formulation for the atmospheric effect

In recent years the availability of high spatial resolution thermal infrared data from satellites has prompted the use of energy budget models relating satellite-derived surface temperatures to surface moisture, near-surface thermal inertia, energy exchange with the atmosphere, etc. However, correction of the high spatial resolution satellite data for atmospheric water vapor effects can represent a substantial computational burden unless simplifying assumptions are utilized. A simple formulation is developed and its applicability tested by application to standard meteorological soundings at a time near the overpass of an NOAA operational satellite. It appears that reasonable estimates of surface temperature (±2–3°C) are readily obtained for areas of order 100–300 km².     

On determining appropriate aerosol optical depth values for atmospheric correction of satellite imagery for biophysical parameter retrieval: requirements and limitations under Australian conditions

Atmospheric correction of high spatial resolution (10–30 m pixel sizes) satellite imagery for use in large-area land-cover monitoring is difficult due to the lack of aerosol optical depth (AOD) estimates made coincident with image acquisition. We present a methodology to determine the upper and lower bounds of AOD estimates that allow the subsequent calculation of a biophysical variable of interest to a pre-determined precision. Knowledge of that range can be used to identify an appropriate method for estimating AOD. We applied the methodology to Landsat 5 Thematic Mapper data in Queensland (QLD) and New South Wales (NSW), Australia, and determined that AOD must be estimated within approximately 0.05 of actual AOD for retrieval of foliage projective cover (FPC) to a precision of 10%. That knowledge was then used to determine the relative merit of using a fixed constant, Aerosol Robotic Network (AERONET) climatology, or dense dark vegetation (DDV) method for estimating AOD in QLD and NSW. It was found that using a fixed AOD of 0.05 allows estimates of FPC within 10% of their true value when the true value of AOD is less than 0.1. Such AOD values account for approximately 90% of all inland observations and 65% of coastal observations as determined by analysis of data obtained from AERONET. Using an AERONET climatology to estimate AOD was found to increase the likelihood of accurate FPC retrieval in coastal locations to 83%, although it should be noted that AERONET data are very sparse. DDV has potential in eastern and central areas for retrieving AOD observations with greater precision than fixed values or climatologies. However, more work is needed to understand the temporal variation of vegetation reflectance before the DDV method can be used operationally.     

An atmospheric correction method for the automatic retrieval of surface reflectances from TM images

Most of the atmospheric correction methods proposed in the literature are not easily applicable in reaJ cases. The most sophisticated models frequently require inputs which are not commonly available, whilst traditional simple dark object subtraction techniques do not generally give real reflectance values. In the present work an atmospheric correction method applicable to Landsat-TM data is described, which requires only inputs that are commonly available and the presence in the imaged scenes of some dark surfaces in TM bands 1 (blue) and 3 (red). The method consists of an inversion algorithm based on a simplified radiative transfer model in which the characteristics of atmospheric aerosols are estimated by the use of the path radiance in two TM bands rather than a priori assumed. On the basis of this information, which is crucial for determining the atmospheric properties, the retrieval of real reflectances from TM images is possible. The method can be applied to all TM scenes in which some dark points can be realistically supposed to be present, which is particularly advantageous in retrospective studies. Several TM scenes taken from different landscapes and in different seasons were corrected using the model. The reflectance values found were tested against ground measurements and compared with data from the literature. The results show a substantial improvement in the accuracy of the reflectance estimates with respect to estimates without atmospheric correction. Given some care in the identification of dark values, the relative error in actual reflectance retrieval is always rather low (?10–20 per cent); this error can be considered acceptable for most practical applications.     

A simple geometrical model for analysing the spectral response of a citrus canopy using satellite images †

Abstract

A simple geometrical model has been proposed for a citrus canopy. We assume the citrus orchard to be a lattice structure, with the trees positioned at its points and where the composite-scene reflectance is the sum of the reflectance of its individual components as weighted by their respective surfaces within a unit area. The model has been used to analyse the citrus spectral response obtained from Landsat-5 TM images for winter and summer, where the status of the orchard is different. The correlations between spectral and geometrical data show the influence of per cent crop cover, shadows and background in the composite scene reflectance. We conclude that the summer images could be more useful than the winter ones for parcel classification according to per cent crop cover.