Objective assessment of the NOAA global vegetation index data product

Abstract

For the last 10 years the U.S. National Oceanic and Atmospheric Administration has produced an experimental Global Vegetation Index (GVI) data set for terrestrial vegetation research. These data, sampled from advanced very high resolution radiometer (AVHRR) observations, have served as a primary stimulus for global-scale vegetation research but have, so far, not been adequately evaluated. This study reviews the GVI production procedures and compares the resultant observations with a more comprehensive compilation of the AVHRR data being produced at the NASA Goddard Space Flight Center. There are many aspects of the GVI production procedures which could be improved to achieve the desired objectives. In particular, the mapping and sampling procedures employed provide measurements which only approximate the observed GAC measurements. The GVI NDVI record varies more than ±NDVI units (～ 7 per cent of signal) from the GAC record and, in general, seriously underestimates the GAC NDVI measurements. The NDVI portion of the GVI record is compromised through use of digital numbers rather than calibrated reflectance. NDVI measurements from the calibrated channels of the GVI data set produces values that compare favourably with the GAC measurements, but with considerable residual variance. Calculation of a 3 by 3 pixel average of the GVI NDVI measurements reduces residual variance between the data sets to ±0.018 NDVI units (～3 per cent of signal). Decay of sensor calibration and orbital overpass time, experienced by all the AVHRR sensors, as well as differences in these parameters between the sensors are not addressed but the results suggest the importance of accounting for these factors. These results indicate that GVI data sets, following adequate reprocessing, provide reasonable estimates of major regional contrasts in vegetation activity but should not be employed to evaluate local or minor trends.     

The University of Maryland improved Global Vegetation Index product

Staff of the University of Maryland, Laboratory for Global Remote Sensing Studies have reprocessed the National Oceanic and Atmospheric Administration (NOAA) Global Vegetation Index (GVI) data record. Observations from April 1983 to June 1991 were mapped to a consistent projection (Plate Carreé) and radiometrically calibrated to spectral reflectance. Sensor degradation with time was taken into account. The normalized difference vegetation index (NVDI) was computed and bi-weekly composites formed to reduce residual cloud contamination. In addition, a set of data quality indicators were compiled during processing. Inspection of the reprocessed observations indicates that they are a significant improvement over the original GVI data. The temporal patterns in the observations appear consistent over time and between sensor systems. Considerable local variance is still evident in the observations, particularly in humid, cloud-prone regions of the globe. This is indicative of inherent limitations in the GVI data files. The ancillary data files in the reprocessed record may assist in addressing this atmospheric contamination problem. These reprocessed measurements should be of value in current efforts to study biospheric dynamics and in the design of future remote sensing missions to study global change.     

Fourier analysis of multi-temporal AVHRR data applied to a land cover classification

A signal processing technique is presented and applied to annual patterns of the Global Vegetation Index (GVI) derived from the Advanced Very High Resolution Radiometer (AVHRR) to examine the frequency distribution of the multi-temporal signal. It is shown that frequencies of the signal are linked to integrated GVI, seasonal variability and subseasonal variability of the land cover type. These characteristics are used to derive a land cover classification.     

Atmospheric correction of high-spatial-resolution satellite images with adjacency effects: application to EO-1 ALI data

Adjacency effects are an interesting physical phenomenon caused by multiple scattering between the atmosphere and the surface. It is necessary to remove adjacency effects in the surface reflectance retrieved from satellite data at a high spatial resolution. In this study, we propose an atmospheric correction method with adjacency effect correction to derive surface reflectance from Earth Observing-1 (EO-1) Advanced Land Imager (ALI) data. Adjacency effects are corrected using an atmospheric point spread function. An analytical expression of the atmospheric point spread function is presented based on a single scattering approximation. This method was applied to ALI imagery acquired through Watershed Airborne Telemetry Experimental Research (WATER) on 20 May 2008. Compared with the surface reflectance before the adjacency effects were corrected for, the surface reflectance after correction exhibited increased between-pixel contrast. Furthermore, the discrepancies between the surface reflectance before and after corrections decreased from the blue band to the shortwave infrared band.     

Toward aerosol optical depth retrievals over land from GOES visible radiances: determining surface reflectance

Frequent observations of aerosol over land are desirable for aviation, air pollution and health applications. Thus, a method is proposed here to correct surface effects and retrieve aerosol optical depth using visible reflectance measurements from the Geostationary Operational Environmental Satellite (GOES). The surface contribution is determined from temporal compositing of visible imagery, where darker pixels correspond to less atmospheric attenuation and surface reflectance is deduced from the composite using radiative transfer. The method is applied to GOES‐8 imagery over the eastern US. Retrieved surface reflectance is compared with separate retrievals using a priori ground‐based observations of aerosol optical depth. The results suggest that surface reflectances can be determined to within ±0.04. The composite‐derived surface reflectance is further analysed by retrieving aerosol optical depth and validating retrievals with Aerosol Robotic Network (AERONET) observations. This analysis indicates that the retrieved optical depth is least biased, hence the surface reflectance is most accurate, when the composite time period varies seasonally. Aerosol optical depth retrievals from this validation are within ±0.13 of AERONET observations and have a correlation coefficient of 0.72. While aerosol optical depth retrieval noise at low optical depths may be limiting, the retrieval accuracy is adequate for monitoring large outbreaks of aerosol events.     

Parametrized BRDF for atmospheric and topographic correction and albedo estimation in Jiangxi rugged terrain,China

A method is presented for bi‐directional reflectance distribution function (BRDF) parametrization for topographic correction and surface reflectance estimation from Landsat Thematic Mapper (TM) over rugged terrain. Following this reflectance, albedo is calculated accurately. BRDF is parametrized using a land‐cover map and Landsat TM to build a BRDF factor to remove the variation of relative solar incident angle and relative sensor viewing angle per pixel. Based on the BRDF factor and radiative transfer model, solar direct radiance correction, sky diffuse radiance and adjacent terrain reflected radiance correction were introduced into the atmospheric‐topographic correction method. Solar direct radiance, sky diffuse radiance and adjacent terrain reflected radiance, as well as atmospheric transmittance and path radiance, are analysed in detail and calculated per pixel using a look‐up table (LUT) with a digital elevation model (DEM). The method is applied to Landsat TM imagery that covers a rugged area in Jiangxi province, China. Results show that atmospheric and topographic correction based on BRDF gives better surface reflectance compared with sole atmospheric correction and two other useful atmospheric‐topographic correction methods. Finally, surface albedo is calculated based on this topography‐corrected reflectance and shows a reasonable accuracy in albedo estimation.     

Terrain analysis using radar shape-from-shading

This paper develops a maximum a posteriori (MAP) probability estimation framework for shape-from-shading (SFS) from synthetic aperture radar (SAR) images. The aim is to use this method to reconstruct surface topography from a single radar image of relatively complex terrain. Our MAP framework makes explicit how the recovery of local surface orientation depends on the whereabouts of terrain edge features and the available radar reflectance information. To apply the resulting process to real world radar data, we require probabilistic models for the appearance of terrain features and the relationship between the orientation of surface normals and the radar reflectance. We show that the SAR data can be modeled using a Rayleigh-Bessel distribution and use this distribution to develop a maximum likelihood algorithm for detecting and labeling terrain edge features. Moreover, we show how robust statistics can be used to estimate the characteristic parameters of this distribution. We also develop an empirical model for the SAR reflectance function. Using the reflectance model, we perform Lambertian correction so that a conventional SFS algorithm can be applied to the radar data. The initial surface normal direction is constrained to point in the direction of the nearest ridge or ravine feature. Each surface normal must fall within a conical envelope whose axis is in the direction of the radar illuminant. The extent of the envelope depends on the corrected radar reflectance and the variance of the radar signal statistics. We explore various ways of smoothing the field of surface normals using robust statistics. Finally, we show how to reconstruct the terrain surface from the smoothed field of surface normal vectors. The proposed algorithm is applied to various SAR data sets containing relatively complex terrain structure.     

Influence of land surface effects on MODIS aerosol retrieval using the BAER method over Korea

As satellite receiving signals are affected by complex radiative transfer processes in the atmosphere and on land surfaces, aerosol retrieval over land from space requires the ability to determine surface reflectance from the remote measurements. To use the Bremen Aerosol Retrieval (BAER) method for aerosol optical thickness (AOT) retrieval over land at a spatial scale of 1×1 km² from Moderate Resolution Imaging Spectroradiometer (MODIS) data, a linear mixing model with a vegetation index was used to calculate surface reflectances. As the vegetation index is affected by the aerosol present in the atmosphere, an empirical linear relationship between short wavelength infrared (SWIR) channel reflectance and visible reflectance was estimated to calculate a modified aerosol free vegetation index (AFRI) value. Based on a modified AFRI obtained from MODIS SWIR channel reflectance, an improved linear mixing model was applied for aerosol retrieval. A comparison of results between calculated and apparent surface reflectance was satisfactory, with a linear fit slope above 0.94, correlation coefficients above 0.84, and standard deviation below 0.008 for the study area. These results can therefore be used for improved aerosol retrieval over land by the BAER method with MODIS Level 1 data.     

An empirical anisotropy correction model for estimating land surface albedo for radiation budget studies

Land surface albedo is one of the key parameters in the radiation budget, the hydrological cycle and climate modeling studies. It is now widely understood that large errors may occur in the estimation of surface albedo without taking into consideration the anisotropy reflectance effect, which is a general feature of the earth surface. Two major anisotropic correction methods exist for the retrieval of land surface albedo under clear sky conditions. One method involves linearly converting from top of the atmosphere (TOA) albedo to surface albedo, and another is based on the inversion of the Bidirectional Reflectance Distribution Function (BRDF) model of the surface. In the present study, a new approach that utilizes an empirical model for estimating surface albedo has been proposed for snow free land surfaces under clear sky conditions. We analyzed the bidirectional reflectance data set with numerous samples representing various land cover types, which derived from POLDER/ADEOS-1 multi-angle imagery data and distributed by MEDIAS-France. Through the analysis, an empirical relation between bidirectional reflectance and albedo was established and has been discussed in detail. The proposed model can be used for direct estimation of surface albedo from a single BRF observation when the sun-target-sensor geometry is known. No BRDF model inversion scheme is necessary. The present model has no or weak dependence on the existing land surface classifications, and is insensitive to wavelength. The theoretical absolute accuracy of the estimated albedo is approximately 0.010 for visible (0.4-0.7 μm), 0.023 for near infrared (0.7-3.0 μm) and 0.016 for shortwave (0.2-3.0 μm), respectively. Albedo consistency with viewing geometry has been verified, resulting in good agreement for albedo estimated from various viewing directions. Validation of the satellite estimated albedo derived by the proposed method, using field observations were also presented and results show it can give reasonably accurate estimation. The proposed method is expected to be a suitable candidate for practical applications of albedo estimation for sensors that do not perform multi-angle observations.     

Measurements and analysis of in situ multi-angle reflectance of turbid inland water: a case study in Meiliang Bay,Taihu Lake,China

The bidirectional reflectance properties of the anisotropic light field above the water surface are important for a range of applications. The bidirectional reflectance distribution function of oceanic waters has been well characterized but there is a lack of information for turbid inland waters. In addition, there is a lack of bidirectional reflectance data measured in turbid inland waters partially due to the difficulty in collecting in situ water-surface multi-angle remote-sensing reflectance data. To facilitate bidirectional reflectance studies of turbid inland waters using in situ multi-angular reflectance data, we have designed and developed a simple hand-held 3D positioning pole to position the spectrometer optical fibre probe and a specific method to collect the multi-angular reflectance data above the water surface with this pole. Using this device, we collected multi-angular reflectance data in Meiliang Bay, Taihu Lake, China, and analysed the uncertainties in this method. We analysed the bidirectional distribution characteristics of the data, and compared the findings to those in the literature. Both uncertainty analysis and bidirectional distribution characteristics analysis showed that our method is effective in collecting multi-angular reflectance above the water surface and can be applied to validate bidirectional correction models in the future.     

Global vegetation dynamics: satellite observations over Asia

Abstract

The weekly global vegetation index (GVI) derived from the NOAA AVHRR instrument has been analysed for the 1982-1985 period over a wide range of vegetation formations of Asia. Temporal development curves of the index are presented for environments ranging from the desert of central Asia to the tropical forest of Borneo. The paper shows that, despite the coarse resolution of the GVI product, a large set of useful information on ecosystem dynamics and cropping practices can be consistently derived from time series of such data. In addition, it is shown that the impact of the 1982-1983 El Nino Southern Oscillation-related drought can be detected in the GVI data through an analysis of anomalies in the development of selected vegetation formations. The relevance of such analysis for global vegetation monitoring and change detection is then underlined.     

Surface identification using the dichromatic reflection model

The author describes a method based on the dichromatic reflection model for identifying object surfaces. The surface spectral reflectance function of an inhomogeneous object is described as the sum of a constant interface (specular) reflectance and a body (diffuse) reflectance under all illumination geometries. The interface component is used to estimate the spectral power distribution of the illuminant without using a reference white standard, whereas the body component is used as the principal indication of the surface identity. The body reflectance function of each surface is recovered. A method to classify the observed reflectances is developed, and an algorithm to estimate a body reflectance function, unique to each surface, from the classified reflectances is proposed. The author shows the reliability of the surface classification method and the accuracy of estimated body reflectance function     

Spectral indices for lithologic discrimination and mapping by using the ASTER SWIR bands

The Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) is a research facility instrument launched on NASA's Terra spacecraft in December 1999. Spectral indices, a kind of orthogonal transformation in the five-dimensional space formed by the five ASTER short-wave-infrared (SWIR) bands, were proposed for discrimination and mapping of surface rock types. These include Alunite Index, Kaolinite Index, Calcite Index, and Montmorillonite Index, and can be calculated by linear combination of reflectance values of the five SWIR bands. The transform coefficients were determined so as to direct transform axes to the average spectral pattern of the typical minerals. The spectral indices were applied to the simulated ASTER dataset of Cuprite, Nevada, USA after converting its digital numbers to surface reflectance. The resultant spectral index images were useful for lithologic mapping and were easy to interpret geologically. An advantage of this method is that we can use the pre-determined transform coefficients, as long as image data are converted to surface reflectance.     

Cost-effective reflectance calibration method for small UAV images

ABSTRACT

We introduce a cost-effective reflectance calibration method for small unmanned aerial vehicle (sUAV) images using ethylene-vinyl acetate (EVA) greyscale reference panels. The goal is to test if such light-weight and low-cost panels can provide sufficient calibration accuracy to support UAV survey projects. The universal calibration equations to convert red-green-blue (RGB) digital number (DN) values of UAV images to surface reflectance values were constructed based on the relationship between RGB values measured by a colour digitizer and surface reflectance values measured by a spectrometer. We compared the calibration results for UAV ortho-mosaic images acquired at three different illumination conditions in late autumn to the results derived from high-cost commercial panels. The comparison showed high degree of agreement between our method using the EVA panels with the traditional methods using the commercial panels. The Mann–Whitney U test verified our method was statistically more significant at all illumination conditions tests. In addition, the calibration results applied for two different sensors and three different flight altitudes acquired in early summer were satisfactory. This method is transferable to various illumination conditions and flight altitudes as long as the effects of shades and the bidirectional reflectance distribution function (BRDF) are minimal. We expect our research could expedite sUAV image calibration by lowering its cost and levelling its availability.     

Lowest-order correction to GVI data for solar zenith angle effect

Abstract

The aims of this presentation were (i) to simulate the solar zenith angle effect on the Global Vegetation Index (GVI), (ii) to derive an expression for removing such an effect from the GVI data by the above simulation procedure and (iii) to apply this relation to the GVI data obtained from the NOAA-AVHRR imagery.     

Simultaneous determination of aerosol optical thickness and surface reflectance using ASTER visible to near-infrared data over land

An innovative method for the determination of aerosol optical thickness (AOT) and surface reflectance for operational use of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) visible to near-infrared data is presented. This method is designed to obtain the atmospheric parameters needed in the correction of the image. This method is based on a simplified radiative transfer equation describing the relation between the ground surface reflectance, AOT and top-of-atmosphere reflectance. By exploiting the ASTER dual-angle view capabilities in band 3N (Nadir) and band 3B (Backwards), surface reflectance and AOT can be retrieved synchronously. Thus, it solves the problem of separating atmospheric radiance from the transmitted radiance of the surface to some extent. After applying this new atmospheric correction method to three areas of ASTER images, Beijing urban city, the Heihe River Basin and Hong Kong of China, ASTER surface reflectance products (AST07) were obtained. AOT values from in situ measurements of CIMEL Electronique 318 Sun Photometers or AERONET (AErosol RObotic NETwork) and surface reflectance in situ measured using an Analytical Spectral Device (ASD) Field Spec spectral radiometer are used for validation. AOT derived from the new method is consistent with in situ station measurements from CIMEL Electronique 318 Sun Photometer and level 2.0 data from AERONET, with correlation coefficient (R ²) of 0.98 and root mean square error of 0.05, whereas Multi-angle Imaging Spectroradiometer AOT products underestimate AERONET AOT and Moderate-Resolution Imaging Spectroradiometer AOT products overestimate AERONET AOT in these regions. More encouraging is the comparison between the corrected surface reflectance, AST07 and ASD measurements. Root mean square error of AST07 and retrieved surface reflectance are as follows: band 1 (556 nm) = 0.04 and 0.05; band 2 (661 nm) = 0.036 and 0.035; band 3 (807 nm) = 0.056 and 0.038, which suggests that compared with AST07 in bands 2 and 3, retrieved surface reflectance has better agreement with measured reflectance from ASD.     

The synthesis and analysis of color images

I describe a method for performing the synthesis and analysis of digital color images. The method is based on two principles. First, image data are represented with respect to the separate physical factors, surface reflectance and the spectral power distribution of the ambient light, that give rise to the perceived color of an object. Second, the encoding is made efficient by using a basis expansion for the surface spectral reflectance and spectral power distribution of the ambient light that takes advantage of the high degree of correlation across the visible wavelengths normally found in such functions. Within this framework, the same basic methods can be used to synthesize image data for color display monitors and printed materials, and to analyze image data into estimates of the spectral power distribution and surface spectral reflectances. The method can be applied to a variety of tasks. Examples of applications include the color balancing of color images and the identification of material surface spectral reflectance when the lighting cannot be completely controlled.     

Multi-temporal MODIS-Landsat data fusion for relative radiometric normalization, gap filling, and prediction of Landsat data

A semi-physical fusion approach that uses the MODIS BRDF/Albedo land surface characterization product and Landsat ETM+ data to predict ETM+ reflectance on the same, an antecedent, or subsequent date is presented. The method may be used for ETM+ cloud/cloud shadow and SLC-off gap filling and for relative radiometric normalization. It is demonstrated over three study sites, one in Africa and two in the U.S. (Oregon and Idaho) that were selected to encompass a range of land cover land use types and temporal variations in solar illumination, land cover, land use, and phenology. Specifically, the 30 m ETM+ spectral reflectance is predicted for a desired date as the product of observed ETM+ reflectance and the ratio of the 500 m surface reflectance modeled using the MODIS BRDF spectral model parameters and the sun-sensor geometry on the predicted and observed Landsat dates. The difference between the predicted and observed ETM+ reflectance (prediction residual) is compared with the difference between the ETM+ reflectance observed on the two dates (temporal residual) and with respect to the MODIS BRDF model parameter quality. For all three scenes, and all but the shortest wavelength band, the mean prediction residual is smaller than the mean temporal residual, by up to a factor of three. The accuracy is typically higher at ETM+ pixel locations where the MODIS BRDF model parameters are derived using the best quality inversions. The method is most accurate for the ETM+ near-infrared (NIR) band; mean NIR prediction residuals are 9%, 12% and 14% of the mean NIR scene reflectance of the African, Oregon and Idaho sites respectively. The developed fusion approach may be applied to any high spatial resolution satellite data, does not require any tuning parameters and so may be automated, is applied on a per-pixel basis and is unaffected by the presence of missing or contaminated neighboring Landsat pixels, accommodates for temporal variations due to surface changes (e.g., phenological, land cover/land use variations) observable at the 500 m MODIS BRDF/Albedo product resolution, and allows for future improvements through BRDF model refinement and error assessment.     

A comparative evaluation of arid inflow-dependent vegetation maps derived from LANDSAT top-of-atmosphere and surface reflectances

ABSTRACT

In remote sensing, it is commonly accepted that land remote-sensing satellite (LANDSAT) top-of-atmosphere (TOA) reflectance is less accurate than atmospheric correction (AC) reflectance, as the former is not calibrated for possible modifications in the electromagnetic radiation signals due to atmospheric scattering and absorption. This article investigates whether LANDSAT data calibrated for TOA reflectance are an appropriate information source for delineating inflow-dependent vegetation (IDV) in regions with an arid and desert climate, such as the Pilbara region in Western Australia. Knowledge of where IDVs are in the landscape underpins planning their protection and define the baseline for their monitoring when water resource management options are considered. The appropriateness of TOA calibration for the delineation of IDV in the Pilbara was assessed through its comparison with IDV maps derived from AC reflectance. Both radiometric calibration methods (TOA and AC) were applied to a multi-date LANDSAT 5 TM (Thematic Mapper) dataset of 10 images acquired in 2009 and 2010. Two methods based on the application of remote-sensing techniques to identify the extent of temporally invariant vegetation were applied for IDV delineation in the study area. The first method, groundwater-dependent ecosystems mapping (GEM), employs a two-date normalized difference vegetation index (NDVI) dataset for identifying ‘no-change’ clusters of land cover and detecting those related to IDV. The second method applies principal component analysis (PCA) to a multi-date NDVI dataset. The first principal component (PC₁) typically contains features that remain unchanged over time. This includes vegetation with continuous or frequent access to surface and/or groundwater, such as IDV. To delineate the extent of IDV, a thresholding technique was further employed. Spatial similarity between IDV maps produced from TOA and AC reflectance was quantitatively evaluated by the Kappa coefficient. The results showed that TOA and AC IDV maps are in ‘almost perfect’ agreement with the Kappa values above 0.83. This suggests that TOA reflectance is equally appropriate to AC reflectance for mapping in arid and desert climate such as in Pilbara. When the GEM- and PCA-based methods are applied in other study areas with arid or desert climate, the accuracy of the delineated IDV extent may vary. Therefore, the results need to be validated using ground-truth information about known IDV occurrences in the area of interest.     

Algorithm for automatic atmospheric corrections to visible and near-IR satellite imagery

Abstract

An algorithm is developed for automatic atmospheric correction of satellite imagery of the Earth's surface. The algorithm is based solely on the satellite image being corrected and on climatology of the area. It is applicable to low resolution (1 km field of view) and high resolution (10-80m field of view) imagery of land areas for the solar spectrum. The algorithm requires that some pixels in the image will correspond to dense dark vegetation as the surface cover. Once the presence of such pixels is established, the algorithm automatically chooses these pixels, derives the atmospheric optical thickness (a measure of the amount of haze) and corrects the image. The algorithm is sensitive to the assumed reflectance of the dense dark vegetation. As a result, the accuracy of the corrected surface reflectance (p) is expected to be δp-±0.01. It is not very sensitive to the assumed aerosol characteristics, the accuracy of satellite calibration or the knowledge of the exact fraction of the image covered by the dense dark vegetation. The correction algorithm was applied to clear and hazy Landsat Multispectral Scanner images of the same area in the Washington D.C. and the Chesapeake Bay region. The aerosol optical thickness (t_a) derived from the imagery shows a good agreement with simultaneous sunphotometer measurements from the ground within δT_a=±0.20 in band 1 (0.5.0.6) and δt_a=±0.05 in band 2 (0.6-0.7μm). The images in the hazy and clear days were corrected and compared. The comparison shows, for example, that the vegetation index was corrected from 0-39 in the clear day and 0-21 in the hazy day to 0-57± 0.01 in these two days. The algorithm, in its present form, can be applied to satellite imagery that includes at least two channels in the visible part of the spectrum, preferably blue and red. Application to the Advanced Very High Resolution Radiometer type of sensor (with one broad channel in the visible part of the spectrum) would need some modifications.