Effect of orbital drift and sensor changes on the time series of AVHRR vegetation index data

This paper assesses the effect of changes in solar zenith angle (SZA) and sensor changes on reflectances in channel 1, channel 2, and normalized difference vegetation index (NDVI) from the advanced very high resolution radiometer (AVHRR) Pathfinder land data set for the period July 1981 through September 1994. First, the effect of changes in SZA on channel reflectances and NDVI is derived from equations of radiative transfer in vegetation media. Starting from first principles, it is rigorously shown that the NDVI of a vegetated surface is a function of the maximum positive eigenvalue of the radiative transfer equation within the framework of the theory used and its assumptions. A sensitivity analysis of this relation indicates that NDVI is minimally sensitive to SZA changes, and this sensitivity decreases as leaf area increases. Second, statistical methods are used to analyze the relationship between SZA and channel reflectances or NDVI. It is shown that the use of ordinary least squares can generate spurious regressions because of the nonstationary property of time series. To avoid such a confusion, the authors use the notion of cointegration to analyze the relation between SZA and AVHRR data. Results are consistent with the conclusion of theoretical analysis from equations of radiative transfer. NDVI is not related to SZA in a statistically significant manner except for biomes with relatively low leaf area. From the theoretical and empirical analysis, they conclude that the NDVI data generated from the AVHRR Pathfinder land data set are not contaminated by trends introduced from changes in solar zenith angle due to orbital decay and changes in satellites (NOAA-7, 9, 11). As such, the NDVI data can be used to analyze interannual variability of global vegetation activity.     

Analysis of interannual changes in northern vegetation activityobserved in AVHRR data from 1981 to 1994

This paper reports on the analysis of Pathfinder AVHRR land (PAL) data set that spans the period July 1981 to September 1994. The time series of normalized difference vegetation index (NDVI) data for land areas north of 45° N assembled by correcting the PAL data with spectral methods confirms the northerly greening trend and extension of the photosynthetically active growing season. Analysis of the channel reflectance data indicates that the interannual changes in red and near-infrared reflectances are similar to seasonal changes in the spring time period when green leaf area increases and photosynthetic activity ramps up. Model calculations and theoretical analysis of the sensitivity of NDVI to background reflectance variations confirm the hypothesis that warming driven reductions in snow cover extent and earlier onset of greening are responsible for the observed changes in spectral reflectances over vegetated land areas     

Estimation of the ratio of sensor degradation between NOAA AVHRRchannels 1 and 2 from monthly NDVI composites

Vegetation monitoring, based on the normalized difference vegetation index (NDVI) calculated from the Advanced Very High Resolution Radiometer (AVHRR) channels 1 and 2 data, requires continuous updates of calibration coefficients to correct for the sensor degradation in these channels. A method was developed to estimate calibration coefficients with monthly composited NDVI data from desert targets without recourse to the component channels 1 and 2 data. The method was tested on NDVI data from the AVHRR onboard the NOAA-7, -9, and -11 satellites for the period from 1982 until 1993. The results of the method outlined in this paper correlated high r, between 0.94 and 0.95, with the results from other studies that estimated sensor degradation for the individual AVHRR bands     

Detection of interannual vegetation responses to climaticvariability using AVIRIS data in a coastal savanna in California

Ecosystem responses to interannual weather variability are large and superimposed over any long-term directional climatic responses making it difficult to assign causal relationships to vegetation change. Better understanding of ecosystem responses to interannual climatic variability is crucial to predicting long-term functioning and stability. Hyperspectral data have the potential to detect ecosystem responses that are undetected by broadband sensors and can be used to scale to coarser resolution global mapping sensors, e.g., advanced very high resolution radiometer (AVHRR) and MODIS. This research focused on detecting vegetation responses to interannual climate using the airborne visible-infrared imaging spectrometer (AVIRIS) data over a natural savanna in the Central Coast Range in California. Results of linear spectral mixture analysis and assessment of the model errors were compared for two AVIRIS images acquired in spring of a dry and a wet year. The results show that mean unmixed fractions for these vegetation types were not significantly different between years due to the high spatial variability within the landscape. However, significant community differences were found between years on a pixel basis, underlying the importance of site-specific analysis. Multitemporal hyperspectral coverage is necessary to understand vegetation dynamics     

Land cover dependence in the detection of contaminated pixels insatellite optical data

The detection of partially contaminated pixels over land is necessary for quantitative applications of satellite optical measurements to estimate surface biophysical parameters such as leaf area index or vegetation composition. Threshold-based algorithms suffer from the heterogeneity of land cover and the seasonal variability of the radiation reflected and emitted by the land surface. As an alternative, a method based on a Fourier series approximation to the seasonal trajectory of the normalized difference vegetation index (NDVI) had been previously developed (Cihlar 1996). In this paper, we introduce modifications to the basic algorithm to more closely represent NDVI seasonal trends for different land cover types, as well as a simplified way to determine the time- and pixel-specific contamination thresholds. Based on the tests with 1993-1996 Advanced Very High Resolution Radiometer (AVHRR) data over Canada, the modified procedure effectively detects contaminated pixels for boreal ecosystems after the growing season of interest. The modifications also improved its performance while the growing season is in progress; in this case, at least one complete previous growing season coverage is required to provide the temporal series needed to establish the thresholds. The modified procedure also yields a contamination parameter that may be used to estimate the most likely value for NDVI or other variables for each pixel. It is concluded that the procedure would perform effectively in other areas, provided that the NDVI temporal trajectories of the cover types of interest can he represented by a mathematical model     

Assessment of regional biomass-soil relationships using vegetationindexes

The development of photosynthetic active biomass in different ecological conditions, as indicated by normalized difference vegetation indices (NDVIs) is compared by performing a stratified sampling (based on soil associations) on data acquired over Indiana. Data from the NOAA-10 Advanced Very High Resolution Radiometer (AVHRR) were collected for the 1987 and 1988 growing seasons. An NDVI transformation was performed using the two optical bands of the sensor (0.58-0.68 μm and 0.72-1.10 μm). The NDVI is related to the amount of active photosynthetic biomass present on the ground. Statistical analysis of results indicate that land-cover types (forest, forest/pasture, and crops), soil texture, and soil water-holding capacity have an important effect on vegetation biomass changes as measured by AVHRR data     

Hidden Markov models applied to vegetation dynamics analysis usingsatellite remote sensing

Details hidden Markov models (HMM) with respect to their ability to represent time series of remotely sensed data as well as to analyze vegetation dynamics at large scales. The present approach is shown to be a powerful way to classify and extract various dynamics parameters as well as to detect phenological anomalies. The methodology is applied and validated using the Advanced Very High Resolution Radiometer (AVHRR) normalized difference vegetation index (NDVI) time series. The model is then used to determine vegetation active cycle and the length of the growing season in the West African savanna     

Directional effects in a daily AVHRR land surface temperature dataset over Africa

Land surface temperature (LST) is a key indicator of the land surface state and can provide information on surface-atmosphere heat and mass fluxes, vegetation water stress, and soil moisture. Split-window algorithms have been used with National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) data to estimate instantaneous LST for nearly 20 years. However, the low accuracy of LST retrievals associated with intractable variability has often hindered its wide use. In this study, we developed a six-year daily (day and night) NOAA-14 AVHRR LST dataset over continental Africa. By combining vegetation structural data available in the literature and a geometric optics model, we estimated the fractions of sunlit and shaded endmembers observed by AVHRR for each pixel of each overpass. Although our simplistic approach requires many assumptions (e.g., only four endmember types per scene), we demonstrate through correlation that some of the AVHRR LST variability can be attributed to angular effects imposed by AVHRR orbit and sensor characteristics, in combination with vegetation structure. These angular effects lead to systematic LST biases, including "hot spot" effects when no shadows are observed. For example, a woodland case showed that LST measurements within the "hot-spot" geometry were about 9 K higher than those at other geometries. We describe the general patterns of these biases as a function of tree cover fraction, season, and satellite drift (time past launch). In general, effects are most pronounced over relatively sparse canopies (tree cover <60%), at wet season sun-view angle geometries (principal plane viewing) and early in the satellite lifetime. These results suggest that noise in LST time series may be strongly reduced for some locations and years, and that long-term LST climate data records should be normalized to a single sun-view geometry, if possible. However, much work remains before these can be accomplished.     

Atmospheric correction against algorithm for NOAA-AVHRR products:theory and application

Investigation of the effect of atmospheric constituents on NOAA Advanced Very High Resolution Radiometer (AVHRR) visible and near-infrared data is presented. The general remote sensing equation, including scattering, absorption, and bidirectional reflectance effects for the AVHRR solar bands, is described. The magnitude of the atmospheric effects for AVHRR solar bands with respect to their impact on the normalized difference vegetation index (NDVI) and the surface bidirection reflectance is examined. Possible approaches for acquiring atmospheric information are discussed, and examples of atmospheric correction of surface reflectance and NDVI are given. Invariant effects (ozone absorption and molecular scattering) and variant effects (water vapor absorption and aerosol scattering) are shown to dominate the atmospheric effects in the AVHRR solar bands     

Designing optimal spectral indexes for remote sensing applications

Satellite remote sensing data constitute a significant potential source of information on our environment, provided they can be adequately interpreted. Vegetation indexes, a subset of the class of spectral indexes, represent one of the most commonly used approaches to analyze data in the optical domain. An optimal spectral index is very sensitive to the desired information (e.g. the amount of vegetation), and as insensitive as possible to perturbing factors (such as soil color changes or atmospheric effects). Since both the desired signal and the perturbing factors vary spectrally, and since the instruments themselves only provide data for particular spectral bands, optimal indexes should be designed for specific applications and particular instruments. This paper describes a rational approach to the design of an optimal index to estimate vegetation properties on the basis of the red and near-infrared reflectances of the AVHRR instrument, taking into account the perturbing effects of soil brightness changes, atmospheric absorption and scattering. The rationale behind the Global Environment Monitoring index (GEMI) is explained, and this index is proposed as an alternative to the Normalized Difference Vegetation Index (NDVI) for global applications. The techniques described here are generally applicable to any multispectral sensor and application     

Upscale integration of normalized difference vegetation index: theproblem of spatial heterogeneity

The correspondence between the normalized difference vegetation index (NDVI) calculated from average reflectances, M_NDVI , and NDVI integrated from individual NDVIs, I_NDVI , by simulating AVHRR data from high spatial resolution SPOT 1 Haute Resolution Visible radiometer and Landsat Thematic Mapper data is analyzed. For the considered sites, located in tropical West Africa and temperate France, and the scales analyzed, 300-1000 m, a strong correlation exists between the two types of index. The relationship is almost perfectly linear, with a slope depending slightly on the variability of the vegetation cover. Effecting the scale change using M_NDVI instead of I_NDVI does not introduce significant errors     

A Gaussian adaptive resonance theory neural network classificationalgorithm applied to supervised land cover mapping using multitemporalvegetation index data

Neural network classifiers have been shown to provide supervised classification results that significantly improve on traditional classification algorithms such as the Bayesian (maximum likelihood [ML]) classifier. While the predominant neural network architecture has been the feedforward multilayer perceptron known as backpropagation. Adaptive resonance theory (ART) neural networks offer advantages to the classification of optical remote sensing data for vegetation and land cover mapping. A significant advantage is that it does not require prior specification of the neural net structure, creating as many internal nodes as are needed to represent the calibration (training) data. The Gaussian ARTMAP classification algorithm bases the probability that input training samples belong to specific classes on the parameters of its Gaussian distributions: the means, standard deviations, and a priori probabilities. The performance of the Gaussian ARTMAP classification algorithm in terms of classification accuracy using independent validation data indicated was over 70% accurate when applied to an annual series of monthly 1-km advanced very high resolution radiometer (AVHRR) satellite normalized difference vegetation index (NDVI) data. The accuracies were comparable to those of fuzzy ARTMAP and a univariate decision tree, and significantly higher than a Bayesian classification algorithm. Algorithm testing is based on calibration and validation data developed and applied to Central America to map the International Geosphere-Biosphere Programme (IGBP) land cover classification system     

Seasonality extraction by function fitting to time-series of satellite sensor data

A new method for extracting seasonality information from time-series of satellite sensor data is presented. The method is based on nonlinear least squares fits of asymmetric Gaussian model functions to the time-series. The smooth model functions are then used for defining key seasonality parameters, such as the number of growing seasons, the beginning and end of the seasons, and the rates of growth and decline. The method is implemented in a computer program TIMESAT and tested on Advanced Very High Resolution Radiometer (AVHRR) normalized difference vegetation index (NDVI) data over Africa. Ancillary cloud data [clouds from AVHRR (CLAVR)] are used as estimates of the uncertainty levels of the data values. Being general in nature, the proposed method can be applied also to new types of satellite-derived time-series data.     

Influence of near-surface soil moisture on regional scale heatfluxes: model results using microwave remote sensing data fromSGP97

During the 1997 Southern Great Plains Hydrology Experiment (SGP97), passive microwave observations using the L-band electronically scanned thinned array radiometer (ESTAR) were used to extend surface soil moisture retrieval algorithms to coarser resolutions and larger regions with more diverse conditions. This near-surface soil moisture product (W) at 800 m pixel resolution together with land use and fractional vegetation cover (f_c) estimated from normalized difference vegetation index (NDVI) was used for computing spatially distributed sensible (H) and latent (LE) heat fluxes over the SGP97 domain (an area ~40×260 km) using a remote sensing model (called the two-source energy Balance-soil moisture, TSEB_SM, model). With regional maps of W and the heat fluxes, spatial correlations were computed to evaluate the influence of W on H and LE. For the whole SGP97 domain and full range in f_c, correlations (R) between W and LE varied from 0.4 to 0.6 (R~0.5 on average), while correlations between W and H varied from -0.3 to -0.7 (R~-0.6 on average). The W-LE and W-H correlations were dramatically higher when variability due to f_c was considered by using NDVI as a surrogate for f_c and computing R between heat fluxes and corresponding W values under similar fractional vegetation cover conditions. The results showed a steady decline in correlation with increasing NDVI or f_c. Typically, |R|≳0.9 for data sorted by NDVI having values ≲0.5 or f_c ≲0.5, while |R|≲0.5 for the data sorted under high canopy cover where NDVI≳0.6 or f_c≳0.7     

Integration of high- and low-resolution satellite data to estimate pine forest productivity in a Mediterranean coastal area

The estimation of vegetation primary productivity is particularly important in fragile Mediterranean environments that are vulnerable to both natural and human-induced perturbations. The current work was aimed at using remotely sensed data taken by various sensors to infer information about a protected coastal pine forest in Tuscany (Central Italy), which could serve for driving a simplified model of carbon fluxes, C-Fix. Being based on the direct relationship between normalized difference vegetation index (NDVI) and fraction of absorbed photosynthetically active radiation (FAPAR), C-Fix uses satellite and standard meteorological data to simulate gross (GPP) and net (NPP) primary productivity of forest ecosystems. Due to the limited size of the study area, a major difficulty was in creating an NDVI dataset with suitable spatial and temporal resolutions, which was essential for the model functioning. To reach this objective, eight Thematic Mapper (TM)/Enhanced Thematic Mapper Plus (ETM+) images of two years (2000 and 2001) were merged to low-resolution NDVI estimates taken by both the Advanced Very High Resolution Radiometer (AVHRR) and VEGETATION (VGT) sensors. The C-Fix outputs for representative pine forest sites were evaluated by comparison to accurate estimates derived from a model of forest ecosystem processes previously calibrated in a similar environment (Forest-BGC). This analysis showed the potential of C-Fix for rapidly estimating GPP over wide forest areas when suitable NDVI inputs are provided. In particular, a slight superiority of VGT over AVHRR data was demonstrated, which could be reasonably attributed to the relevant higher radiometric and geometric properties. The estimation of NPP was instead quite inaccurate, due to the problematic simulation of forest respiration, which should necessarily rely on more complete modeling operations.     

Estimating leaf area index from satellite data

A method for estimating leaf area index from visible and near infrared measurements of vegetation above a soil background is applied to a Landsat Thematic Mapper data set. Some constants required for the procedure are inferred from the scattergram of data values. The resulting image illustrates variability of leaf area index over an agricultural area. The mixed-pixel case, corresponding to low-resolution data from the NOAA Advanced Very High Resolution Radiometer (AVHRR) is also discussed, and a vegetation index is suggested for both high- and low-resolution data. Consideration of the two types of data leads to the suggestion that a sampled high spatial resolution sensor (50-100 m) be added to the AVHRR in order to permit accurate inference of vegetation conditions over agricultural areas     

Early Warning and Crop Condition Assessment Research

The Early Warning Crop Condition Assessment Project of AgRISTARS was a multiagency and multidisciplinary effort. Its mission and objectives were centered around development and testing of remote-sensing techniques that enhance operational methodologies for global crop-condition assessments. The project developed crop stress indicator models that provide data filter and alert capabilities for monitoring global agricultural conditions. The project developed a technique for using NOAA-n satellite advanced very-high-resolution radiometer (AVHRR) data for operational crop-condition assessments. This technology was transferred to the Foreign Agricultural Service of the USDA. The project developed a U. S. Great Plains data base that contains various meteorological parameters and vegetative index numbers (VIN) derived from AVHRR satellite data. It developed cloud screening techniques and scan angle correction models for AVHRR data. It also developed technology for using remotely acquired thermal data for crop water stress indicator modeling. The project provided basic technology including spectral characteristics of soils, water, stressed and nonstressed crop and range vegetation, solar zenith angle, and atmospheric and canopy structure effects.     

Atmospheric corrections in the thermal infrared: global and watervapor dependent split-window algorithms-applications to ATSR and AVHRRdata

The split-window method is an appropriate way to perform atmospheric corrections of satellite brightness temperatures in order to retrieve the surface temperature. A climatological data set of 1761 different radio soundings, the TIGR database, has been used to develop two different split-window methods. A global quadratic (QUAD) method, with global coefficients to be applied on a worldwide scale, and a water vapor dependent (WVD) algorithm. The first method includes a quadratic term in the split-window equation that roughly accounts for the water vapor amount. The other method explicitly includes the water vapor amount in each split-window coefficient. When applied to the 1761 radio soundings, the latter method gives better results than the global one, especially when the surface emissivity is far from unity (0.95 or less) and when the water vapor reaches great values. Both algorithms have been tested on ATSR/ERSI and AVHRR/NOAA data over sea pixels. The QUAD algorithm gives correct results for simulations (the standard error is 0.2 K) and experimental data (the bias ranges from -0.1 to 0.4 K). The WVD algorithm appears to be more accurate for both simulations (the standard error is less than 0.1 K) and AVHRR experimental data when climatological water vapor contents are used (the bias ranges from -0.2 to 0.1 K)     

Portable Instrument for Normalized Difference Vegetation Index

By using four specially designed narrow bandpass filters and photodetectors in the instrument,the incident and reflected radiances of sun light on the vegetation are optically sensed, at the red and near infrared bands, then the normalized difference vegetation index(NDVI) is processed by a microprocessor.Compared with conventional spectrometer measuring method of NDVI, the instrument is easy to be used,compact, light and low-cost.