SPOT‐4 VEGETATION multi‐temporal compositing for land cover change studies over tropical regions

Multi‐temporal compositing of SPOT‐4 VEGETATION imagery over tropical regions was tested to produce spatially coherent monthly composite images with reduced cloud contamination, for the year 2000. Monthly composite images generated from daily images (S1 product, 1‐km) encompassing different land cover types of the state of Mato Grosso, Brazil, were evaluated in terms of cloud contamination and spatial consistency. A new multi‐temporal compositing algorithm was tested which uses different criteria for vegetated and non‐vegetated or sparsely vegetated land cover types. Furthermore, a principal components transformation that rescales the noise in the image—Maximum Noise Fraction (MNF)—was applied to a multi‐temporal dataset of monthly composite images and tested as a method of additional signal‐to‐noise ratio improvement. The back‐transformed dataset using the first 12 MNF eigenimages yielded an accurate reconstruction of monthly composite images from the dry season (May to September) and enhanced spatial coherence from wet season images (October to April), as evaluated by the Moran's I index of spatial autocorrelation. This approach is useful for land cover change studies in the tropics, where it is difficult to obtain cloud‐free optical remote sensing imagery. In Mato Grosso, wet season composite images are important for monitoring agricultural crop cycles.     

Categorization of land‐cover change processes based on phenological indicators extracted from time series of vegetation index data

A straightforward method for categorizing temporal patterns of land‐cover change is presented. Two successive years of enhanced vegetation index (EVI) data derived from the Moderate Resolution Imagery Spectrometer (MODIS) were analysed. Five phenological indicators were extracted. Based on the inter‐annual difference of each of the five indicators, indices of change in phenology were calculated. An unsupervised classification of these five indices of change applied to pixels characterized by a high change magnitude led to the identification of seven categories of land‐cover change patterns. Thirty‐one per cent of the change pixels could clearly be explained by a difference in only one or two phenological indicators, e.g. a shift in the start of the growing season or an interruption of the growing season due to floods. The remaining change pixels were explained by a combination of more than two indices of change. The output of this analysis is an allocation of change pixels to broad categories of land‐cover change as a preliminary step for finer resolution analyses.     

Understanding precipitation patterns and land use interaction in Tibet using harmonic analysis of SPOT VGT‐S10 NDVI time series

Time series analysis of Normalized Difference Vegetation Index (NDVI) imagery is a powerful tool in studying land use and precipitation interaction in data‐scarce and inaccessible areas. The Fast Fourier Transform (FFT) was applied to the annual time series of 36 average dekadal NDVI images. The dekadal annual average pattern was calculated from 189 NDVI images from April 1998 to June 2003 acquired with the VEGETATION instruments of the SPOT‐4 and SPOT‐5 satellites in Tibet. It is shown that the first two harmonic terms of a Fourier series suffice to distinguish between land use classes. The results indicate that the highest biomass production occurs before the monsoon peak. Regression analysis with 15 meteorological stations has shown that the total amount of precipitation during the growing season shows the strongest relation with the sum of the amplitudes of the first two harmonic terms (R ²?=?0.72). Inter‐annual NDVI variation based on Fourier‐transformed time series was studied and it was shown that, early in the season, the expected NDVI behaviour of the up‐coming season could be forecast; if linked to food production this might provide a robust early warning system. The most important conclusion from this work is that harmonic time series analysis yields more reliable results than ordinary time series analysis.     

Spatio-temporal errors in land–cover change analysis: implications for accuracy assessment

This research examined the spatial and temporal patterns of error in time-series classified maps as a first step to creating a model to propagate error in post-classification change analysis. Two Landsat images were acquired for Pittsfield Township, MI, USA, classified, and overlaid to produce a map of change. Error variables were created for the classified maps. Hypotheses were proposed describing the spatial and temporal structures of error in the classified maps, and evaluated using geostatistics and point pattern analysis. Results showed that the spatial error structure for all three classified maps was composed of both a small-scale, local error interactions, and a large-scale trend. Errors in the individual land-cover (LC) maps interacted as a result of temporal dependence, which increased the expected accuracy of the map of change. These findings have important implications on change accuracy assessment, particularly affecting how sampling schemes are defined and how accuracy information is reported.     

View angle effects on the discrimination of selected Amazonian land cover types from a principal‐component analysis of MISR spectra

Multi‐angle Imaging Spectroradiometer (MISR) data, collected in four bands and at nine view angles in the Brazilian Amazon region, were used to describe view‐angle effects on the spectral response and discrimination of three forest types; close and open lowland forests, open submontane forest and green/emerging pastures. A principal‐component analysis (PCA) was applied over 450 bidirectional reflectance factor (BRF) MISR spectra (10 pixels, five land covers and nine view angles) to characterize the spectral‐angular variability in the dataset and to identify the best view direction to enhance land cover discrimination. The analysis was extended into the images of the different cameras, which were classified for the presence of the forest covers using the minimum distance of the pixels to the average PC1 and PC2 scores of each forest class calculated from spectra analysis. Results showed an increase in the mean reflectance over the spectral bands (brightness) of the land covers from nadir to extreme viewing, as indicated by the first principal component, especially in the backward direction due to the predominance of sunlit view vegetation components. The transition from the backward (sunlit view surface components) to the forward (shaded view surface components) scattering directions was also characterized by changes in the shape of the BRF spectra, as indicated by decreasing PC2 score or near‐infrared/blue ratio values. The variations in the MISR BRF followed the regularities expected from theory. PCA results also indicated that the best viewing to discriminate the forest types was the backward scattering direction (?26.1° view angle), whereas the less favourable viewing was the forward scattering direction under the view shading condition (e.g. +45.6° view angle). The overall classification accuracy for the three forest types increased from 52.4% at +45.6° view angle to 78.7% at nadir, and to 95.0% at a ?26.1° view angle. From nadir to extreme view angles, directional effects produced a NDVI decrease for the forest types and an NDVI increase for the green and especially emerging pastures. Results demonstrated that data acquisition in off‐nadir viewing may improve the discrimination and mapping of the Amazonian land cover types.     

Characterizing implications of two-dimensional space–time orientations for principal component analysis of geographic time series

This article introduces four new modes of principal component analysis (PCA) to investigate space–time variability in an image time series. Using the concept of tensors, an image time series can be understood as a space–time cube and can be analysed using six different orientations by grouping the basic elements (voxels) of the cube across different dimensions. Voxels grouped across columns or rows of the cube to produce vectors result in profiles. Voxels grouped across different planes to produce matrices result in slices. The traditional S-mode and T-mode PCA are thus the profile modes and slice modes across time and across space, respectively. This research introduces two profile-mode orientations across longitude and latitude and two slice-mode orientations across longitude–time and latitude–time. The research shows that a more complete understanding of the spatio-temporal variability in the data set can be achieved by investigating these different orientation modes, as individual modes have the capability of capturing variability in a particular dimension of a spatio-temporal data set. A case study was carried out using weekly anomalies of the AVISO (Archiving, Validation and Interpretation of Satellite Oceanographic data) sea surface height product filtered for tropical instability waves (TIWs) for a three-year time period from 1997 to 1999 in the tropical Pacific region. The results show that PCA with longitude as the dimension of variability and latitude–time as the dimension of variability were able to capture the TIW and barotropic Rossby wave propagation across the equatorial Pacific. The other two orientation modes were able to detect dominant latitudinal locations for TIW.     

Time–space radiometric normalization of TM/ETM+ images for land cover change detection

A novel approach to image radiometric normalization for change detection is presented. The approach referred to as stratified relative radiometric normalization (SRRN) uses a time-series of imagery to stratify the landscape for localized radiometric normalization. The goal is to improve the detection accuracy of abrupt land cover changes (human-induced, natural disaster, etc.) while decreasing false detection of natural vegetation changes that are not of interest. These vegetation changes may be associated with such phenomena as phenology, growth and stress (e.g. drought), which occur at varying spatial and temporal scales, depending on landscape position, vegetation type, season, precipitation history and historic episodes of local disturbance. The SRRN approach was tested for a study area on the Californian border between the USA and Mexico using Landsat Thematic Mapper and Enhanced Thematic Mapper Plus satellite imagery. Change products were generated from imagery radiometrically normalized using the SRRN procedure and with imagery normalized using a traditional empirical line technique. Reference data derived from high spatial resolution airborne imagery were utilized to validate the two change products. The SRRN procedure provided several benefits and was found to improve the overall accuracy of detecting abrupt land cover changes by nearly 20%.     

Monitoring Sahelian floodplains using Fourier analysis of MODIS time‐series data and artificial neural networks

Fourier analysis of Moderate Resolution Image Spectrometer (MODIS) time‐series data was applied to monitor the flooding extent of the Waza‐Logone floodplain, located in the north of Cameroon. Fourier transform (FT) enabled quantification of the temporal distribution of the MIR band and three different indices: the Normalized Difference Vegetation Index (NDVI), the Normalized Difference Water Index (NDWI), and the Enhanced Vegetation Index (EVI). The resulting amplitude, phase, and amplitude variance images for harmonics 0 to 3 were used as inputs for an artificial neural network (ANN) to differentiate between the different land cover/land use classes: flooded land, dry land, and irrigated rice cultivation. Different combinations of input variables were evaluated by calculating the Kappa Index of Agreement (KIA) of the resulting classification maps. The combinations MIR/NDVI and MIR/EVI resulted in the highest KIA values. When the ANN was trained on pixels from different years, a more robust classifier was obtained, which could consistently separate flooded land from dry land for each year.     

Evaluation of AVHRR PAL and GIMMS 10‐day composite NDVI time series products using SPOT‐4 vegetation data for the African continent

Global 8 km resolution AVHRR (advanced very high resolution radiometer) NDVI (normalized difference vegetation index) 10‐day composite data sets have been used for numerous local to global scale vegetation time series studies during recent years. AVHRR Pathfinder (PAL) NDVI was available from 1981 until 2001, and the new AVHRR GIMMS NDVI was available from 1981 to the present time. A number of aspects potentially introduce noise in the NDVI data set due to the AVHRR sensor design and data processing. NDVI from SPOT‐4 VGT data is considered an improvement over AVHRR, and for this reason it is important to examine how and if the differences in sensor design and processing influence continental scale NDVI composite products. In this study, the quality of these AVHRR NDVI time series are evaluated by the continental scale 1 km resolution SPOT‐4 vegetation (VGT) 10‐day composite (S10) NDVI data. Three years of AVHRR PAL (1998–2000) and seven years of GIMMS (1998–2004) have been compared to 8 km resampled SPOT‐4 VGT (1998–2004) data. The dynamic range of SPOT‐4 VGT NDVI tends to be higher than the AVHRR PAL NDVI, whereas there is an exact match between AVHRR GIMMS NDVI and SPOT‐4 VGT NDVI. Ortho‐regression analysis on annually integrated values of AVHRR PAL/GIMMS and SPOT‐4 VGT on a continental scale reveals high correlations amongst the AVHRR and the SPOT data set, with lowest RMSE (root mean square error) on the GIMMS/SPOT‐4 VGT compared to the PAL/SPOT‐4 VGT.

Analyses on decade data likewise show that a linear relation exists between Spot‐4 VGT NDVI and the two AVHRR composite products; GIMMS explaining most of the Spot‐4 VGT NDVI variance compared to PAL. These results show that the AVHRR GIMMS NDVI is more consistent with Spot‐4 VGT NDVI compared to AVHRR PAL versus Spot‐4 VGT NDVI (in terms of RMSE and dynamic range) and can therefore be considered the more accurate long time AVHRR data record. Analyses performed on monthly maximum composites and decade composite data, however, reveal intra‐annual variations in the correlation between SPOT‐4 VGT and the two AVHRR data sets, which are attributed to different cloud masking algorithms. The SPOT‐4 VGT cloud‐screening algorithm is insufficient, thereby suppressing the rainy season NDVI.     

Assessing the impacts of the 2003 hot and dry spell with SPOT HRVIR images time series over south‐western France

A severe heatwave affected southern Europe and, in particular, south‐western France during the summer of 2003. The area was subjected to a severe dry spell with high temperatures and very little precipitation during a nearly 4‐month‐long period. A series of monthly 20‐m spatial resolution images, acquired between 2002 and 2003 by the SPOT HRVIR (High Resolution Visible–Infrared) was used to examine the impact of the dry and hot spell over a 50?km×50?km area of south‐western France. The use of the high spatial resolution data permitted the distinction of the various land surface types, and facilitated the diagnosis of the hot and dry spell effects for each class. The Normalized Difference Vegetation Index (NDVI) time series of the four principal vegetated surfaces (meadow, deciduous forest, wheat and maize) reveal different responses to the dryness and hot spell. A significant shortening of the 2003 phenological cycle was observed for the meadows, but the response was not as clear for the deciduous forests. For the crop surfaces, a shortening of the cycle is observed, although the impact of the drought was translated differently as a function of the crop type.     

Recent land‐cover/use change associated with land degradation in the Lake Baringo catchment,Kenya, East Africa: evidence from Landsat TM and ETM+

Many parts of East Africa are experiencing dramatic changes in land‐cover/use at a variety of spatial and temporal scales, due to both climatic variability and human activities. Information about such changes is often required for planning, management, and conservation of natural resources. Several methods for land cover/change detection using Landsat TM/ETM+ imagery were employed for Lake Baringo catchment in Kenya, East Africa. The Lake Baringo catchment presents a good example of environments experiencing remarkable land cover change due to multiple causes. Both the NDVI differencing and post‐classification comparison effectively depicted the hotspots of land degradation and land cover/use change in the Lake Baringo catchment. Change‐detection analysis showed that the forest cover was the most affected, in some sections recording reductions of over 40% in a 14‐year period. Deforestation and subsequent land degradation have increased the sediment yield in the lake resulting in reduction in lake surface area by over 10% and increased turbidity confirmed by the statistically significant increase (t = ?84.699, p<0.001) in the albedo between 1986 and 2000. Although climatic variations may account for some of the changes in the lake catchment, most of the changes in land cover are inherently linked to mounting human and livestock population in the Lake Baringo catchment.     

Land‐cover change and environmental impact analysis in the Greater Mankato area of Minnesota using remote sensing and GIS modelling

Land use and land‐cover (LULC) data provide essential information for environmental management and planning. This research evaluates the land‐cover change dynamics and their effects for the Greater Mankato Area of Minnesota using image classification and Geographic Information Systems (GIS) modelling in high‐resolution aerial photography and QuickBird imagery. Results show that from 1971 to 2003, urban impervious surfaces increased from 18.3% to 32.6%, while cropland and grassland decreased from 54.2% to 39.1%. The dramatic urbanization caused evident environmental impacts in terms of runoff and water quality, whereas the annual air pollution removal rate and carbon storage/sequestration remained consistent since urban forests were steady over the 32‐year span. The results also indicate that highly accurate land‐cover features can be extracted effectively from high‐resolution imagery by incorporating both spectral and spatial information, applying an image‐fusion technique, and utilizing the hierarchical machine‐learning Feature Analyst classifier. This research fills the high‐resolution LULC data gap for the Greater Mankato Area. The findings of the study also provide valuable inputs for local decision‐makers and urban planners.     

Assessing spatio‐temporal variations in plant phenology using Fourier analysis on NDVI time series: results from a dry savannah environment in Namibia

Time series of Normalized Difference Vegetation Index (NDVI) from the Advanced Very High Resolution Radiometer (AVHRR) were used to capture plant phenology in Etosha National Park, a dry savannah environment in Namibia. Data from two consecutive growing periods with different precipitation conditions were included to study impacts of inter‐seasonal rainfall variations on a highly water‐limited ecosystem. Additionally, a contemporary reference map with four major vegetation units was used to compare phenology between plant formations. Phenological attributes were acquired for both seasons using Fourier analysis. Parameters were calculated for the entire study area and further stratified with respect to the mapping units of the reference. Vegetation growth was found to vary significantly between the two periods in accordance with available rainfall data. Additionally, separability of vegetation entities based on Fourier parameters was weak due to within‐class scattering and was commonly outranged by inter‐seasonal variations. Finally, discrimination of cover types was tested by combining selected Fourier parameters in a clustering procedure. Spatial class distribution was compared to the reference statistically and only a moderate correspondence was discovered. We conclude that Fourier‐based NDVI attributes are limited for cover‐type discrimination across space and time, as they only quantify certain aspects of plant phenology and seem to be largely altered by the actual rainfall situation.