Improving the estimation of noise from NOAA AVHRR NDVI for Africa using geostatistics

The accuracy of NOAA AVHRR NDVI data can be poor because of interference from several sources, including cloud cover. A parameter of the variogram model can be used to estimate the contribution of noise from the total variation in an image. However, remotely sensed information over large areas incorporates non-stationary (regional) trend and directional effects, resulting in violation of the assumptions for noise estimation. These assumptions were investigated at five sites across Africa for a range of ecological environments over several seasons. An unsupervised spectral classification of multi-temporal NDVI data partially resolved the problem of non-stationarity. Quadratic polynomials removed the remaining regional trend and directional effects. Isotropic variograms were used to estimate the noise contributing variation to the image. Standardized estimates of noise ranged from a minimum of 18.5% in west Zambia to 68.2% in northern Congo. Cloud cover and atmospheric particulates (e.g. dust) explained some of the regional and seasonal variations in noise levels. Image artifacts were also thought to contribute noise to image variation. The magnitude of the noise levels and its temporal variation appears to seriously constrain the use of AVHRR NDVI data.     

Controls over intra-seasonal dynamics of AVHRR NDVI for the Arctic tundra in northern Alaska

We analysed the Normalized Difference Vegetation Index (NDVI), calculated from biweekly NOAA Advanced Very High Resolution Radiometer (AVHRR) images for northern Alaska at both regional (latitudinal gradients) and site scales. Our objectives were to determine if tundra types and arctic subzones could be differentiated in terms of intra-seasonal patterns of greenness, and to construct the relationships between NDVI and air and soil temperatures. There were common intra-seasonal patterns of NDVI along two latitudinal transects, and a general latitudinal gradient of time of greenness onset and length of growing season was observed. At the site scale, in most cases, wet tundra (WT) had the lowest NDVI values throughout the year, while shrub tundra (ST) had the highest values. The peak NDVI appeared in the period of 22 July to 4 August, with mean values of 0.552 for ST, 0.495 for moist acidic tundra (MAT), 0.434 for sandy tundra (Sandy), 0.426 for moist non-acidic tundra (MNT) and 0.343 for WT. The earliest onset of greenness occurred in ST, followed by MAT, Sandy and MNT, while WT had the latest onset. There were positive linear relationships between bi-weekly NDVI anomalies and air temperature, soil surface temperature, and 20?cm depth soil temperature anomalies in the region. Plant functional type abundances, tundra type, air and soil temperatures all appeared to influence the seasonal dynamics of NDVI.     

Early detection system of drought in East Asia using NDVI from NOAA/AVHRR data

Abstract

The influence of surface bidirectional reflectance factors, including shadowing, and of atmospheric aerosol variability are modelled for their effects on the remote sensing of desert targets from space in the 0·?μm region at high spatial resolution. The white sand reflectance data of Salomonson are used as the basis for the simulation. The effects of the surface bi-directional reflectance and atmospheric aerosol on the nadir-normalized reflectance measured at the satellite are discussed individually and jointly. The net influence of these two factors is shown to depend on the magnitude of other parameters, such as the surface reflectance and solar zenith angle.     

A flexible approach for woody cover assessment from SPOT HRV XS data in semi-arid West Africa. Application in northern Burkina Faso

Spectral mixture analysis was applied to woody cover appraisal in a semi-arid landscape of West Africa. The spectral response of a SPOT HRV XS image was modelled from signatures and relative abundance of four pure basic components: green tree foliage, associated shadow and two types of bare soil surfaces (clear vs. dark petroferric). Signatures were determined from the image by interpreting a bidimensional scattergram of pixels in the brightness/greenness spectral space. Estimation of subpixel abundances was made possible by prior observation that dense woody cover cannot occur on dark petroferric soils. Consequently, the scattergram had a peculiar shape enabling the distinction between a 'soil ridge' consisting of pixels bearing no woody cover, and a 'vegetation ridge' expressing the variation of woody cover on clear soils. The relationship between canopy surface and associated shadowing was considered through Li-Strahler geometric-optical canopy reflectance models, with the main parameters estimated from field measurements of individual trees. Subpixel abundance of canopy cover proved highly correlated with an independent estimate obtained from large-scale aerial photographs. The present approach could hence yield relevant estimates of woody cover, even in the presence of varying soil surface conditions, and with no prerequisite of characterizing spatial patterns of trees.     

Decomposition of vegetation cover into woody and herbaceous components using AVHRR NDVI time series

A method is developed to separate Normalised Difference Vegetation Index (NDVI) time series data into contributions from woody (perennial) and herbaceous (annual) vegetation, and thereby to infer their separate leaf area indices and cover fractions. The method is formally consistent with fundamental linearity requirements for such a decomposition, and is capable of rejecting contaminated NDVI data. In this study, estimates of annual averaged woody cover and monthly averaged herbaceous cover over Australia are determined using Pathfinder AVHRR Land series (PAL) Global Area Coverage (GAC) Advanced Very High Resolution Radiometer (AVHRR) NDVI data from 1981 to 1994, together with ground-based measurements of leaf area index (LAI) and foliage projective cover (FPC).     

Estimating spatio-temporal patterns of agricultural productivity in fragmented landscapes using AVHRR NDVI time series

The characteristics of Normalized Difference Vegetation Index (NDVI) time series can be disaggregated into a set of quantitative metrics that may be used to derive information about vegetation phenology and land cover. In this paper, we examine the patterns observed in metrics calculated for a time series of 8 years over the southwest of Western Australia—an important crop and animal production area of Australia. Four analytical approaches were used; calculation of temporal mean and standard deviation layers for selected metrics showing significant spatial variability; classification based on temporal and spatial patterns of key NDVI metrics; metrics were analyzed for eight areas typical of climatic and production systems across the agricultural zone; and relationships between total production and productivity measured by dry sheep equivalents were developed with time integrated NDVI (TINDVI). Two metrics showed clear spatial patterns; the season duration based on the smooth curve produced seven zones based on increasing length of growing season; and TINDVI provided a set of classes characterized by differences in overall magnitude of response, and differences in response in particular years. Frequency histograms of TINDVI could be grouped on the basis of a simple shape classification: tall and narrow with high, medium or low mean indicating most land is responsive agricultural cover with uniform seasonal conditions; broad and short indicating that land is of mixed cover type or seasonal conditions are not spatially uniform. TINDVI showed a relationship to agricultural productivity that is dependent on the extent to which crop or total agricultural production was directly reduced by rainfall deficiency. TINDVI proved most sensitive to crop productivity for Statistical Local Areas (SLAs) having rainfall less than 600 mm, and in years when rainfall and crop production were highly correlated. It is concluded that metrics from standardized NDVI time series could be routinely and transparently used for retrospective assessment of seasonal conditions and changes in vegetation responses and cover.     

Assessing the effects of human-induced land degradation in the former homelands of northern South Africa with a 1 km AVHRR NDVI time-series

There is a pressing need for an objective, repeatable, systematic and spatially explicit measure of land degradation. In northeastern South Africa (SA), there are large areas of the former homelands that are widely regarded as degraded. A time-series of seasonally integrated 1 km, Advanced Very High Resolution Radiometer (AVHRR) normalized difference vegetation index (NDVI) data was used to compare degraded rangelands [mapped by the National Land Cover (NLC) using Landsat Thematic Mapper (TM) imagery] to nondegraded rangelands within the same land capability units (LCUs). Nondegraded and degraded areas in the same LCU (paired areas) were compared by: (i) testing for differences in spatial mean ∑NDVI values, (ii) calculating the relative degradation impact (RDI) as the difference between the spatial mean ∑NDVI values of paired areas expressed as a percentage of nondegraded mean value, (iii) investigating the relationship between RDI and rainfall and (iv) comparing the resilience and stability of paired areas in response to natural variations in rainfall. The ∑NDVI of degraded areas was significantly lower for most of the LCUs. Relative degradation impacts (RDI) across all LCUs ranged from 1% to 20% with an average of 9%. Although ∑NDVI was related to rainfall, RDI was not. Degraded areas were no less stable or resilient than nondegraded. However, the productivity of degraded areas, i.e., the forage production per unit rainfall, was consistently lower than nondegraded areas, even within years of above normal rainfall. The results indicate that there has not been a catastrophic reduction in ecosystem function within degraded areas. Instead, degradation impacts were reflected as reductions in productivity that varied along a continuum from slight to severe, depending on the specific LCU.     

Automatic mapping of surfaces affected by forest fires in Spain using AVHRR NDVI composite image data

In this work, we describe the statistical techniques used to analyze images from the National Oceanic and Atmospheric Administration's advanced very high resolution radiometer for the calculation and mapping of surfaces affected by large forest fires in Spain in 1993 and 1994. Maximum value normalized difference vegetation index (NDVI) composites (MVCs) were generated for every ten-day period over the two years of the study. Two techniques, one regression analysis and the other differencing, were applied to the NDVI-MVCs both before and after each fire event to determine detection thresholds of change and to delineate and objectively evaluate the burned surfaces. The comparison between the single-fires burned areas predicted by the techniques and that provided by the Spanish Forestry Service (ground based) showed that the regression algorithm was more reliable, giving rise to virtually no bias (−0.9%) and a root mean square error (RMS) of 20.3%, both calculated as a percentage of the mean burned area of the whole sample. The technique of differencing provided worse results with a 3.2% bias and a 23.5% RMS error. Likewise, a comparison between. the perimeters of the large fires supplied by official data (GPS-based) and those obtained by the regression method confirmed the validity of the technique not only for calculating fire size, but also for mapping of large forest fires.     

Estimating above-ground net primary productivity of the tallgrass prairie ecosystem of the Central Great Plains using AVHRR NDVI

Above-ground net primary productivity (ANPP) is indicative of an ecosystem's ability to capture solar energy and convert it to organic carbon (or biomass), which may be used by consumers or decomposers, or stored in the form of living and nonliving organic matter. Annual and interannual variation in ANPP is often linked to climate dynamics and anthropogenic influences, such as fertilization, irrigation, above-ground biomass harvest, and so on. The Central Great Plains grasslands occupy over 1.5 million km² and are a primary resource for livestock production in North America. The tallgrass prairies are the most productive grasslands in this region, and the Flint Hills of North America represent the largest contiguous area of unploughed tallgrass prairie (1.6 million ha). Measurements of ANPP are of critical importance to the proper management and understanding of climatic and anthropogenic influences on tallgrass prairie. Yet, accurate, detailed, and systematic measurements of ANPP over large geographic regions do not exist for this ecosystem. For these reasons, this study was conducted to investigate the use of the normalized difference vegetation index (NDVI) to model ANPP of the tallgrass prairie. Many studies have established a positive relationship between the NDVI and ANPP, but the strength of this relationship is influenced by vegetation types and can vary significantly from year to year depending on land use and climatic conditions. The goal of this study was to develop a robust model using the Advanced Very High Resolution Radiometer (AVHRR) biweekly NDVI values to predict tallgrass ANPP. This study was conducted using ANPP measurements from a watershed within the Konza Prairie Biological Station (KPBS) as the primary study area, with additional measurements from the Rannells Flint Hills Prairie Preserve (RFHPP) and biennial ANPP measurements by Kansas State University (KSU) students from tallgrass areas near Manhattan, Kansas. Data from the primary study site covered the period of 1989–2005. The optimal period for estimating ANPP using AVHRR NDVI composite data sets was found to be late July. The Tallgrass ANPP Model (TAM) explained 54% (coefficient of determination, R ² = 0.54, p < 0.001) of the year-to-year variation in ANPP. The creation of 1.0 km × 1.0 km resolution ANPP maps for a four-county (～7000 ha) area for years 1989–2007 showed considerable variation in annual and interannual ANPP spatial patterns, suggesting complex interactions among factors influencing ANPP spatially and temporally.     

Evapotranspiration estimation in the Yellow River Basin,China using integrated NDVI data

It is important to estimate land surface evapotranspiration (ET) for water resources evaluation, drought monitoring and crop production simulation. In this paper, a relationship between annual ET, integrated Advanced Very High Resolution Radiometer (AVHRR) Normalized Difference Vegetation Index (NDVI) and Relative Moisture Index (RMI) was established. Based on this relationship, the spatial distribution and dynamic change of annual ET were estimated for the Yellow River Basin, China from 1982 to 2000. Our analyses involved the use of integrated NDVI data, monthly mean air temperature, and precipitation. Our results showed that the integrated AVHRR NDVI can be used to effectively estimate annual ET in the Yellow River Basin, with an accuracy over 90% for the whole basin.     

Assessing habitat protection regimes in Tanzania using AVHRR NDVI composites: Comparisons at different spatial and temporal scales

Studies assessing temporal changes in vegetation using satellite imagery are complicated by: (1) high interseasonal and interannual variation in phenology that make vegetation comparisons difficult; (2) anthropogenic pressures on the habitats that vary by geographic region and habitat type; and (3) spatial resolution and processing characteristics of available satellite data that differ substantially. This paper addresses these concerns while examining the effects of various forms of protection on different habitat types in Tanzania. First, a long-term (1982-94) vegetation trend was calculated from monthly Normalized Difference Vegetation Index (NDVI) composites to reduce the effect of seasonal fluctuations. Second, we controlled for confounding variables such as habitat type, elevation, aspect and location as well as anthropogenic factors such as fires, roads and refugee camps. Finally, vegetation changes in protected areas and habitat types were examined in order to compare results produced by the different spatial and temporal data (8 km, 7.6 km and 1.1 km). While some results were consistent across spatial and temporal scales, many were not. We therefore recommend that, if possible, analyses of changes in vegetative health be conducted at more than one temporal and spatial scale before management recommendations are put forward.     

Evaluation and comparison of satellite-based rainfall products in Burkina Faso,West Africa

The performance of seven operational high-resolution satellite-based rainfall products – Africa Rainfall Estimate Climatology (ARC 2.0), Climate Hazards Group InfraRed Precipitation with Stations (CHIRPS), Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN), African Rainfall Estimation (RFE 2.0), Tropical Applications of Meteorology using SATellite (TAMSAT), African Rainfall Climatology and Time-series (TARCAT), and Tropical Rainfall Measuring Mission (TRMM) daily and monthly estimates – was investigated for Burkina Faso. These were compared to ground data for 2001–2014 on a point-to-pixel basis at daily to annual time steps. Continuous statistics was used to assess their performance in estimating and reproducing rainfall amounts, and categorical statistics to evaluate rain detection capabilities. The north–south gradient of rainfall was captured by all products, which generally detected heavy rainfall events, but showed low correlation for rainfall amounts. At daily scale they performed poorly. As the time step increased, the performance improved. All (except TARCAT) provided excellent scores for Bias and Nash–Sutcliffe Efficiency coefficients, and overestimated rainfall amounts at the annual scale. RFE performed the best, whereas TARCAT was the weakest. Choice of product depends on the specific application: ARC, RFE, and TARCAT for drought monitoring, and PERSIANN, CHIRPS, and TRMM daily for flood monitoring in Burkina Faso.     

Systematic and random transitions of land-cover types in Burkina Faso,West Africa

In-depth statistical analysis of forest transition between land-cover types over time can reveal the dominant signals of landscape transformation, which are needed in order to develop appropriate land management strategies. We applied a recently developed methodology to analyse the transition matrix of six land-cover classes, derived from 1986 and 2002 Landsat images of an area of 15?675 km² in southern Burkina Faso. Results show that most landscape transformations followed a systematic process. In addition, some transitions occurred as an apparently random process, probably caused by uncommon or sporadic events. Degradation of woodland to shrub-/grassland over 15.7% of the landscape, increases in biomass from woodland to dense forest on 10% of the landscape and conversion of 6% of the landscape from shrub-/grassland to cropland were the dominant signals of forest-cover transitions. From a planning perspective, the dominance of systematic processes should facilitate regional land-use planning and sustainable forest management in a context of immigration and agricultural intensification.     

Persistent changes in NDVI between 1982 and 2003 over India using AVHRR GIMMS (Global Inventory Modeling and Mapping Studies) data

The preliminary results of Normalized Difference Vegetation Index (NDVI) change studies over India using data from Advanced Very High Resolution Radiometer Global Inventory Modeling and Mapping Studies (AVHRR GIMMS) between 1982 and 2003 are presented. The three methodologies of univariate differencing, temporal profiling and anomaly analysis were undertaken. Univariate differencing was used to determine overall NDVI change between 1982 and 2003. A persistence filter was used to filter out ephemeral changes. The temporal profile analyses were carried out over different meteorological subdivisions to compare changes in NDVI with rainfall patterns. In the anomaly analysis, the areas of change were analysed over different land cover categories derived from IRS‐WiFS data. The preliminary results indicate that positive trends in vegetation change occurred over most parts of the country and these changes appear not to be highly correlated with rainfall data, indicating that land cover transformations may be the major driving force behind the changes. The land cover classifications experiencing the greatest increasing NDVI were tropical thorn forests and intensive agriculture and the land cover experiencing very slow growth included current jhum, tropical moist deciduous and temperate evergreen forest. Five‐year moving averages indicate a general increase in NDVI from 1986 to 1998 and then declining thereafter. This is a concern in most of the meteorological subdivisions.     

Identification of land surface temperature and albedo trends in AVHRR Pathfinder data from 1982 to 2005 for northern Siberia

The arctic regions are highly vulnerable to climate change. Climate models predict an increase in global mean temperatures for the upcoming century. The arctic environment is subject to significant changes of the land surface. Especially the changes of vegetation pattern and the phenological cycle in the taiga–tundra transition area are of high importance in climate change research. This study focuses on time series and trend analysis of land surface temperature, albedo, snow water equivalent, and normalized difference vegetation index information in the time period of 1982–2005 for northern Siberia. The findings show strong dependencies between these parameters and their inter-annual dynamics, which indicate changes in vegetation growing period. We found a strong negative correlation between land surface temperature and albedo conditions for the beginning (60–90%) of the growing season for selected hot spot trend regions in northern Siberia.     

Cloud cover and the monitoring of strawburning using AVHRR data

Abstract

We have conducted a cloud-cover analysis of AVHRR scenes of Britain for April-September for the years 1982-1985 and interpreted the results in terms of the usefulness of the data for the studying of agricultural strawburning.     

Investigation of the maximum Normalized Difference Vegetation Index (NDVI) and the maximum surface temperature (Ts) AVHRR compositing procedures for the extraction of NDVI and Ts over forest

An investigation into the impact of the maximum Normalized Difference Vegetation Index (NDVI) and the maximum surface temperature (Ts) compositing procedures (MaN and MaT respectively) upon retrieved NDVI and Ts values extracted from forested areas located across eight months of cloud screened European AVHRR data is described. NDVI values are found to be significantly higher and generally less variable when they are extracted from MaN rather than from MaT composites and Ts values are found to be significantly higher and generally less variable when they are extracted from MaT rather than from MaN composites. The impact of these differences is illustrated within the context of a European forest/non-forest classification that uses both NDVI and Ts data. Higher potential forest/non-forest classification accuracies are found using NDVI data extracted from the MaN composites and Ts data extracted from the MaT composites than from any other combination of composited data. The findings indicate that inappropriate selection of a compositing procedure may have a significant impact upon the subsequent application of NDVI and/or Ts data.     

Evaluating and reducing errors in seasonal profiles of AVHRR vegetation indices over a Canadian northern national park using a cloudiness index

High-temporal coarse resolution remote-sensing data have been widely used for monitoring plant phenology and productivity. Residual errors in pre-processed composite data from these sensors can still be substantial due to cloud contamination and aerosol variations, especially over high cloud-cover areas such as the Arctic. Commonly used smoothing and filtering methods try to reform the often heavily distorted seasonal profiles of vegetation indices one way or another, instead of explicitly dealing with the errors that cause the distortion. As the distortion varies from year to year for a pixel or from pixel to pixel, so does the performance of various smoothing and filtering methods. Consequently, change detection results are likely method dependent. In this study, we investigate alternative methods in order to eliminate bias caused by cloud contamination and reduce random errors due to aerosol variations in the 10 day Advanced Very High Resolution Radiometer (AVHRR) composite data, so that accurate seasonal profiles of vegetation indices can be constructed without the need to apply a smoothing and filtering method. The best alternative method corrects cloud contaminations by spatially pairing averages of simple ratio over cloud-contaminated and clear-sky pixels in a class (SPAC). The SPAC method eliminates bias caused by cloud contamination and reduces the relative random errors to <14% near the start/end of a growing season, and to <8% during the middle growing season for the six treeless wetland and tundra classes in Wapusk National Park. In comparison, with the method whereby all pixels in a class (average all pixels in the class (AAC)) are averaged in a period, the bias could be up to 40% if all the pixels in the composite period are heavily cloud contaminated.     

Monitoring the 1988 severe drought in Indiana,U.S.A. using AVHRR data

Data from the Advanced Very High Resolution Radiometer &lpar;AVHRR&rpar; on board the NOAA&ndash;10 satellite was collected over Indiana for the 1987 and 1988 growing seasons. A Normalized Difference &lpar;NDVI&rpar; transformation was applied to the data. Over 45 fields representing 8 soil associations were sampled to assess the effects of the 1988 drought on the development of natural and cultivated vegetation. The results show the effect of the lack of available moisture to the plants and its effect on the response measured by the AVHRR sensor.