Using coarse remote sensing radar observations to control the trajectory of a simple Sahelian land surface model

In the Sahel, land surface processes are significantly interacting with climate dynamics. In this paper, we present an original method to control a simple Sahelian land surface model coupled to a radiative transfer model (RTM) on the basis of ERS wind scatterometer (WSC) observations. In a first step, a sensitivity study is implemented to identify those parameters of the land surface model that can be estimated through the assimilation of WSC data. The assimilation scheme relies on evolution strategies (ES) algorithm that aims at solving the parameter evaluation problem. These algorithms are particularly well suited for complex (nonlinear) inverse problems. The assimilation scheme is applied to several study sites located in the Sahelian mesoscale site of the African Monsoon Multidisciplinary Analysis Project (Gourma region, Mali). The results are compared with ground observations of herbaceous mass. After the WSC data assimilation, the simulated herbaceous mass curves compare well with observations [187 kilogram of dry matter per hectare (kg DM/ha) of average error]. The simulated water fluxes exhibit a behaviour in agreement with ground measurements performed over similar ecosystems during the Hapex Sahel experiment. The accuracy of estimated herbaceous mass and water fluxes resulting from uncertainties on climatic forcing variable is evaluated using a stochastic approach. The average error on the herbaceous mass values mainly depends on the rainfall estimate accuracy and ranges from 139 to 268 kg DM/ha that compares well with a previous study based on the sole inversion of the radiative transfer model. Finally, this study underlines the need for a multispectral assimilation approach to get a better constraint on water fluxes estimation.     

Analysis of NDVI and scaled difference vegetation index retrievals of vegetation fraction

The normalized difference vegetation index (NDVI) is the most widely used vegetation index for retrieval of vegetation canopy biophysical properties. Several studies have investigated the spatial scale dependencies of NDVI and the relationship between NDVI and fractional vegetation cover, but without any consensus on the two issues. The objectives of this paper are to analyze the spatial scale dependencies of NDVI and to analyze the relationship between NDVI and fractional vegetation cover at different resolutions based on linear spectral mixing models. Our results show strong spatial scale dependencies of NDVI over heterogeneous surfaces, indicating that NDVI values at different resolutions may not be comparable. The nonlinearity of NDVI over partially vegetated surfaces becomes prominent with darker soil backgrounds and with presence of shadow. Thus, the NDVI may not be suitable to infer vegetation fraction because of its nonlinearity and scale effects. We found that the scaled difference vegetation index (SDVI), a scale-invariant index based on linear spectral mixing of red and near-infrared reflectances, is a more suitable and robust approach for retrieval of vegetation fraction with remote sensing data, particularly over heterogeneous surfaces. The proposed method was validated with experimental field data, but further validation at the satellite level would be needed.     

Multi-sensor NDVI data continuity: Uncertainties and implications for vegetation monitoring applications

Consistent NDVI time series are paramount in monitoring ecological resources that are being altered by climate and human impacts. An increasing number of natural resource managers use web-based geospatial decision support tools that integrate time series of both historical and current NDVI data derived from multiple sensors to make better informed planning and management decisions. Representative canopy reflectance and NDVI data were simulated for historical, current and future AVHRR, MODIS and VIIRS land surface monitoring satellites to quantify the differences due to sensor-specific characteristics. Cross-sensor NDVI translation equations were developed for surface conditions. The effect of a range of atmospheric conditions (Rayleigh scattering, ozone, aerosol optical thickness, and water vapor content) on the sensor-specific reflectance and NDVI values were evaluated to quantify the uncertainty in the apparent NDVI for each sensor. MODIS and VIIRS NDVI data are minimally affected by the atmospheric water vapor, while AVHRR NDVI data are substantially reduced by water vapor.Although multi-sensor NDVI continuity can be obtained by using the developed cross-sensor translation equations, the interactions between the spectral characteristics of surface vegetation and soil components, sensor-specific spectral band characteristics and atmospheric scattering and absorption windows will introduce uncertainty due to insufficient knowledge about the atmospheric conditions that affect the signal of the Earth's pixels at the time of data acquisitions. Processing strategies and algorithm preferences among data streams are also hindering cross-sensor NDVI continuity.     

Evaluation of earth observation based long term vegetation trends — Intercomparing NDVI time series trend analysis consistency of Sahel from AVHRR GIMMS, Terra MODIS and SPOT VGT data

AVHRR (Advanced Very High Resolution Radiometer) GIMMS (Global Inventory Modelling and Mapping Studies) NDVI (Normalized Difference vegetation Index) data is available from 1981 to present time. The global coverage 8 km resolution 15-day composite data set has been used for numerous local to global scale vegetation time series studies during recent years. Several aspects however potentially introduce noise in the NDVI data set due to the AVHRR sensor design and data processing. More recent NDVI data sets from both Terra MODIS and SPOT VGT data are considered an improvement over AVHRR and these products in theory provide a possibility to evaluate the accuracy of GIMMS NDVI time series trend analysis for the overlapping period of available data. In this study the accuracy of the GIMMS NDVI time series trend analysis is evaluated by comparison with the 1 km resolution Terra MODIS (MOD13A2) 16-day composite NDVI data, the SPOT Vegetation (VGT) 10-day composite (S10) NDVI data and in situ measurements of a test site in Dahra, Senegal. Linear least squares regression trend analysis on eight years of GIMMS annual average NDVI (2000-2007) has been compared to Terra MODIS (1 km and 8 km resampled) and SPOT VGT NDVI data 1 km (2000-2007). The three data products do not exhibit identical patterns of NDVI trends. SPOT VGT NDVI data are characterised by higher positive regression slopes over the 8-year period as compared to Terra MODIS and AVHRR GIMMS NDVI data, possibly caused by a change in channels 1 and 2 spectral response functions from SPOT VGT1 to SPOT VGT2 in 2003. Trend analysis of AVHRR GIMMS NDVI exhibits a regression slope range in better agreement with Terra MODIS NDVI for semi-arid areas. However, GIMMS NDVI shows a tendency towards higher positive regression slope values than Terra MODIS in more humid areas. Validation of the different NDVI data products against continuous in situ NDVI measurements for the period 2002-2007 in the semi-arid Senegal revealed a good agreement between in situ measurements and all satellite based NDVI products. Using Terra MODIS NDVI as a reference, it is concluded that AVHRR GIMMS coarse resolution NDVI data set is well-suited for long term vegetation studies of the Sahel-Sudanian areas receiving < 1000 mm rainfall, whereas interpretation of GIMMS NDVI trends in more humid areas of the Sudanian-Guinean zones should be done with certain reservations.     

North American vegetation dynamics observed with multi-resolution satellite data

We investigated normalized difference vegetation index data from the NOAA series of Advanced Very High Resolution Radiometers and found regions in North America that experienced marked increases in annual photosynthetic capacity at various times from 1982 to 2005. Inspection of these anomalous areas with multi-resolution data from Landsat, Ikonos, aerial photography, and ancillary data revealed a range of causes for the NDVI increases: climatic influences; severe drought and subsequent recovery; irrigated agriculture expansion; insect outbreaks followed by logging and subsequent regeneration; and forest fires with subsequent regeneration. Vegetation in areas in the high Northern Latitudes appear to be solely impacted by climatic influences. In other areas examined, the impact of anthropogenic effects is more direct. The pattern of NDVI anomalies over longer time periods appear to be driven by long-term climate change but most appear to be associated with climate variability on decadal and shorter time scales along with direct anthropogenic land cover conversions. The local variability of drivers of change demonstrates the difficulty in interpreting changes in NDVI and indicates the complex nature of changes in the carbon cycle within North America. Coarse scale analysis of changes could well fail to identify the important local scale drivers controlling the carbon cycle and to identify the relative roles of disturbance and climate change. Our results document regional land cover land use change and climatic influences that have altered continental scale vegetation dynamics in North America.     

The Yearly Land Cover Dynamics (YLCD) method: An analysis of global vegetation from NDVI and LST parameters

NDVI (Normalized Difference Vegetation Index) has been widely used to monitor vegetation changes since the early eighties. On the other hand, little use has been made of land surface temperatures (LST), due to their sensitivity to the orbital drift which affects the NOAA (National Oceanic and Atmospheric Administration) platforms flying AVHRR sensor. This study presents a new method for monitoring vegetation by using NDVI and LST data, based on an orbital drift corrected dataset derived from data provided by the GIMMS (Global Inventory Modeling and Mapping Studies) group. This method, named Yearly Land Cover Dynamics (YLCD), characterizes NDVI and LST behavior on a yearly basis, through the retrieval of 3 parameters obtained by linear regression between NDVI and normalized LST data. These 3 parameters are the angle between regression line and abscissa axis, the extent of the data projected on the regression line, and the regression coefficient. Such parameters characterize respectively the vegetation type, the annual vegetation cycle length and the difference between real vegetation and ideal cases. Worldwide repartition of these three parameters is shown, and a map integrating these 3 parameters is presented. This map differentiates vegetation in function of climatic constraints, and shows that the presented method has good potential for vegetation monitoring, under the condition of a good filtering of the outliers in the data.     

Quantifying intra-annual persistent behaviour in SPOT-VEGETATION NDVI data for Mediterranean ecosystems of southern Italy

Multi-temporal series of satellite SPOT-VEGETATION Normalized Difference of Vegetation Index (NDVI) data from 1998 to 2003 were exploited for studying persistence in Mediterranean ecosystems of southern Italy. We used Multiple Segmenting Method (MSM), which is well suited to analyze scaling behaviour in short time series, and the Detrended Fluctuation Analysis (DFA), which permits the detection of persistent properties in nonstationary signal fluctuations. Our findings point out to the characterization of Mediterranean ecosystems as governed by persistent mechanisms.     

Assimilation of leaf area index derived from ASAR and MERIS data into CERES-Wheat model to map wheat yield

This study presents a method to assimilate leaf area index retrieved from ENVISAT ASAR and MERIS data into CERES-Wheat crop growth model with the objective to improve the accuracy of the wheat yield predictions at catchment scale. The assimilation method consists in re-initialising the model with optimal input parameters allowing a better temporal agreement between the LAI simulated by the model and the LAI estimated by remote sensing data. A variational assimilation algorithm has been applied to minimise the difference between simulated and remotely-sensed LAI and to determine the optimal set of input parameters. After the re-initialisation, the wheat yield maps have been obtained and their accuracy evaluated.The method has been applied over Matera site located in Southern Italy and validated by using the dataset of an experimental campaign carried out during the 2004 wheat growing season.Results indicate that, LAI maps retrieved from MERIS and ASAR data can be effectively assimilated into CERES-Wheat model thus leading to accuracies of the yield maps ranging from 360 kg/ha to 420 kg/ha.     

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).     

A simple method for integrating a complex model into an ensemble data assimilation system using MPI

This paper details a strategy for modifying the source code of a complex model so that the model may be used in a data assimilation context, and gives the standards for implementing a data assimilation code to use such a model. The strategy relies on keeping the model separate from any data assimilation code, and coupling the two through the use of Message Passing Interface (MPI) functionality. This strategy limits the changes necessary to the model and as such is rapid to program, at the expense of ultimate performance. The implementation technique is applied in different models with state dimension up to .2.7 × 10⁸ The overheads added by using this implementation strategy in a coupled ocean-atmosphere climate model are shown to be an order of magnitude smaller than the addition of correlated stochastic random errors necessary for some nonlinear data assimilation techniques.     

Analysis of teleconnections between AVHRR-based sea surface temperature and vegetation productivity in the semi-arid Sahel

Vegetation productivity across the Sahel is known to be affected by a variety of global sea surface temperature (SST) patterns. Often climate indices are used to relate Sahelian vegetation variability to large-scale ocean-atmosphere phenomena. However, previous research findings reporting on the Sahelian vegetation response to climate indices have been inconsistent and contradictory, which could partly be caused by the variations in spatial extent/definitions of climate indices and size of the region studied. The aim of this study was to analyze the linkage between climate indices, pixel-wise spatio-temporal patterns of global sea surface temperature and the Sahelian vegetation dynamics for 1982-2007. We stratified the Sahel into five subregions to account for the longitudinal variability in rainfall. We found significant correlations between climate indices and the Normalized Difference Vegetation Index (NDVI) in the Sahel, however with different magnitudes in terms of strength for the western, central and eastern Sahel. Also the correlations based on NDVI and global SST anomalies revealed the same East-West gradient, with a stronger association for the western than the eastern Sahel. Warmer than average SSTs throughout the Mediterranean basin seem to be associated with enhanced greenness over the central Sahel whereas colder than average SSTs in the Pacific and warmer than average SSTs in the eastern Atlantic were related to increased greenness in the most western Sahel. Accordingly, we achieved high correlations for SSTs of oceanic basins which are geographically associated to the climate indices yet by far not always these patterns were coherent. The detected SST-NDVI patterns could provide the basis to develop new means for improved forecasts in particular of the western Sahelian vegetation productivity.     

Discrimination of irrigated and rainfed rice in a tropical agricultural system using SPOT VEGETATION NDVI and rainfall data

The classification of irrigated crops by remote sensing requires the use of time series data, since the timing, cropping intensity and duration of cropping is quite variable over the course of a year. Rice is the dominant irrigated crop in tropical and sub‐tropical Asia, where rainfall is high, but is seasonal and often uni‐modal. Existing crop classification methods for rice are not able to distinguish between rainfed and irrigated crops, leading to errors in classification and estimated irrigated area. This paper describes a technique, a ‘peak detector algorithm’, to successfully discriminate between rainfed and irrigated rice crops in Suphanburi province, Thailand. The methodology uses a three‐year time series of Satellite pour l'Observation de la Terre (SPOT) VEGETATION S10 Normalized Difference Vegetation Index (NDVI) data (10 day composites) to identify cropping intensity (number, timing and peak values). Peak NDVI is then lag‐correlated with long term average rainfall data. There is a high correlation at a 40–50 day lag, between a peak rainfall and a ‘single’ peak NDVI of rainfed rice. In irrigated areas, there are multiple peaks, and multiple correlations with low values for at least 90 days after peak rainfall. The methodology currently uses a mask to remove un‐cropped and non‐rice areas, which is derived from existing Geographical Information Systems (GIS). The method achieves a classification accuracy of 89% or better against independent groundtruth data. The procedure is designed as a second level of analysis to refine classifications using other techniques of mapping irrigated area at global and regional scales.     

Comparison of cloud-reconstruction methods for time series of composite NDVI data

Land cover change can be assessed from ground measurements or remotely sensed data. As regards remotely sensed data, such as NDVI (Normalized Difference Vegetation Index) parameter, the presence of atmospherically contaminated data in the time series introduces some noise that may blur the change analysis. Several methods have already been developed to reconstruct NDVI time series, although most methods have been dedicated to reconstruction of acquired time series, while publicly available databases are usually composited over time. This paper presents the IDR (iterative Interpolation for Data Reconstruction) method, a new method designed to approximate the upper envelope of the NDVI time series while conserving as much as possible of the original data. This method is compared quantitatively to two previously applied methods to NDVI time series over different land cover classes. The IDR method provides the best profile reconstruction in most cases. Nevertheless, the IDR method tends to overestimate low NDVI values when high rates of change are present, although this effect can be lowered with shorter compositing periods. This method could also be applied to data before compositing, as well as to reconstruct time series for other biophysical parameters such as land surface temperature, as long as atmospheric contamination affects these parameters negatively.     

The impact of soil reflectance on the quantification of the green vegetation fraction from NDVI

The green vegetation fraction (Fg) is an important climate and hydrologic model parameter. A common method to calculate Fg is to create a simple linear mixing model between two NDVI endmembers: bare soil NDVI (NDVI_o) and full vegetation NDVI (NDVI_∞). Usually it is assumed that NDVI_o is close to zero (NDVI_o ∼ 0.05) and is generally chosen from the lowest observed NDVI values. However, the mean soil NDVI computed from 2906 samples is much larger (NDVI = 0.2) and is highly variable (standard deviation = 0.1). We show that the underestimation of NDVI_o yields overestimations of Fg. The largest errors occur in grassland and shrubland areas. Using parameters for NDVI_o and NDVI_∞ derived from global scenes yields overestimations of Fg that are larger than 0.2 for the majority of U.S. land cover types when pixel NDVI values are 0.2 < NDVI_pixel < 0.4. When using conterminous U.S. scenes to derive NDVI_o and NDVI_∞, the overestimation is less (0.10-0.17 for 0.2 < NDVI_pixel < 0.4). As a result, parts of the conterminous U.S. are affected at different times of the year depending on the local seasonal NDVI cycle. We propose using global databases of NDVI_o along with information on historical NDVI_pixel values to compute a statistically most-likely estimate of Fg. Using in situ measurements made at the Sevilleta LTER, we show that this approach yields better estimates of Fg than using global invariant NDVI_o values estimated from whole scenes. At the two studied sites, the Fg estimate was adjusted by 52% at the grassland and 86% at the shrubland. More significant advances will require information on spatial distribution of soil reflectance.     

Ecosystem mapping at the African continent scale using a hybrid clustering approach based on 1-km resolution multi-annual data from SPOT/VEGETATION

The goal of this study is to propose a new classification of African ecosystems based on an 8-year analysis of Normalized Difference Vegetation Index (NDVI) data sets from SPOT/VEGETATION. We develop two methods of classification. The first method is obtained from a k-nearest neighbour (k-NN) classifier, which represents a simple machine learning algorithm in pattern recognition. The second method is hybrid in that it combines k-NN clustering, hierarchical principles and the Fast Fourier Transform (FFT). The nomenclature of the two classifications relies on three levels of vegetation structural categories based on the Land Cover Classification System (LCCS). The two main outcomes are: (i) The delineation of the spatial distribution of ecosystems into five bioclimatic ecoregions at the African continental scale; (ii) Two ecosystem maps were made sequentially: an initial map with 92 ecosystems from the k-NN, plus a deduced hybrid classification with 73 classes, which better reflects the bio-geographical patterns. The inclusion of bioclimatic information and successive k-NN clustering elements helps to enhance the discrimination of ecosystems. Adopting this hybrid approach makes the ecosystem identification and labelling more flexible and more accurate in comparison to straightforward methods of classification. The validation of the hybrid classification, conducted by crossing-comparisons with validated continental maps, displayed a mapping accuracy of 54% to 61%.     

An empirical investigation of cross-sensor relationships of NDVI and red/near-infrared reflectance using EO-1 Hyperion data

Long term observations of global vegetation from multiple satellites require much effort to ensure continuity and compatibility due to differences in sensor characteristics and product generation algorithms. In this study, we focused on the bandpass filter differences and empirically investigated cross-sensor relationships of the normalized difference vegetation index (NDVI) and reflectance. The specific objectives were: 1) to understand the systematic trends in cross-sensor relationships of the NDVI and reflectance as a function of spectral bandpasses, 2) to examine/identify the relative importance of the spectral features (i.e., the green peak, red edge, and leaf liquid water absorption regions) in and the mechanism(s) of causing the observed systematic trends, and 3) to evaluate the performance of several empirical cross-calibration methods in modeling the observed systematic trends. A Level 1A Hyperion hyperspectral image acquired over a tropical forest—savanna transitional region in Brazil was processed to simulate atmospherically corrected reflectances and NDVI for various bandpasses, including Terra Moderate Resolution Imaging Spectroradiometer (MODIS), NOAA-14 Advanced Very High Resolution Radiometer (AVHRR), and Landsat-7 Enhanced Thematic Mapper Plus (ETM+). Data were extracted from various land cover types typically found in tropical forest and savanna biomes and used for analyses. Both NDVI and reflectance relationships among the sensors were neither linear nor unique and were found to exhibit complex patterns and bandpass dependencies. The reflectance relationships showed strong land cover dependencies. The NDVI relationships, in contrast, did not show land cover dependencies, but resulted in nonlinear forms. From sensitivity analyses, the green peak (∼550 nm) and red-NIR transitional (680-780 nm) features were identified as the key factors in producing the observed land cover dependencies and nonlinearity in cross-sensor relationships. In particular, differences in the extents to which the red and/or NIR bandpasses included these features significantly influenced the forms and degrees of nonlinearity in the relationships. Translation of MODIS NDVI to “AVHRR-like” NDVI using a weighted average of MODIS green and red bands performed very poorly, resulting in no reduction of overall discrepancy between MODIS and AVHRR NDVI. Cross-calibration of NDVI and reflectance using NDVI-based quadratic functions performed well, reducing their differences to ± .025 units for the NDVI and ± .01 units for the reflectances; however, many of the translation results suffered from bias errors. The present results suggest that distinct translation equations and coefficients need to be developed for every sensor pairs and that land cover-dependency need to be explicitly accounted for to reduce bias errors.     

Enhanced biomass prediction by assimilating satellite data into a crop growth model

Complex crop growth models (CGM) require a large number of input parameters, which can cause large errors if they are uncertain. Furthermore, they often lack spatial information. The coupling of a CGM with a radiative transfer model offers the possibility to assimilate remote sensing data while taking into account uncertainties in input parameters. A particle filter was used to assimilate satellite data into a CGM coupled with a leaf-canopy radiative transfer model to update biomass simulations of maize. The synthetic experiment set up to test the reliability of the procedure, highlighted the importance of the acquisition time. The real case study with RapidEye observations confirmed these findings. Data assimilation increased the accuracy of biomass predictions in the majority of the six maize fields where biomass validation data was available, with improvements of up to 15%. The smallest and largest errors in biomass prediction after assimilation were 82 kg/ha and 2116 kg/ha, respectively. Furthermore, data assimilation enabled the production of biomass maps showing detailed spatial variability.     

Assimilating canopy reflectance data into an ecosystem model with an Ensemble Kalman Filter

An Ensemble Kalman Filter (EnKF) is used to assimilate canopy reflectance data into an ecosystem model. We demonstrate the use of an augmented state vector approach to enable a canopy reflectance model to be used as a non-linear observation operator. A key feature of data assimilation (DA) schemes, such as the EnKF, is that they incorporate information on uncertainty in both the model and the observations to provide a best estimate of the true state of a system. In addition, estimates of uncertainty in the model outputs (given the observed data) are calculated, which is crucial in assessing the utility of model predictions.Results are compared against eddy-covariance observations of CO₂ fluxes collected over three years at a pine forest site. The assimilation of 500 m spatial resolution MODIS reflectance data significantly improves estimates of Gross Primary Production (GPP) and Net Ecosystem Productivity (NEP) from the model, with clear reduction in the resulting uncertainty of estimated fluxes. However, foliar biomass tends to be over-estimated compared with measurements. Issues regarding this over-estimate, as well as the various assumptions underlying the assimilation of reflectance data are discussed.     

Assimilation of meteorological and remote sensing data for snowmelt runoff forecasting

A versatile data assimilation scheme for remote sensing snow cover products and meteorological data was developed, aimed at operational use for short-term runoff forecasting. Spatial and temporal homogenisation of the various input data sets is carried out, including meteorological point measurements from stations, numerical weather predictions, and snow maps from satellites. The meteorological data are downscaled to match the scale of the snow products, derived from optical satellite images of MODIS and from radar images of Envisat ASAR. Snow maps from SAR and optical imagery reveal systematic differences which need to be compensated for use in snowmelt models. We applied a semi-distributed model to demonstrate the use of satellite snow cover data for short-term runoff forecasting. During the snowmelt periods 2005 and 2006 daily runoff forecasts were made for the drainage basin Ötztal (Austrian Alps) for time lags up to 6 days. Because satellite images were obtained intermittently, prognostic equations were applied to predict the daily snow cover extent for model update. Runoff forecasting uncertainty is estimated by using not only deterministic meteorological predictions as input, but also 51 ensemble predictions of the EPS system of the European Centre for Medium Range Weather Forecast. This is particularly important for water management tasks, because meteorological forecasts are the main error source for runoff prediction, as confirmed by simulation studies with modified input data from the various sources. Evaluation of the runoff forecasts reveals good agreement with the measurements, confirming the usefulness of the selected data processing and assimilation scheme for operational use.     

Mapping and analysing arctic vegetation: Evaluating a method coupling numerical classification of vegetation data with SPOT satellite data in a probability model

Vegetation and environmental data were collected at 266 sampling points distributed in a regular manner along transects covering the Broggerhalvoya peninsula, on the north-western coast of Spitsbergen. Transects with sampling points were drawn in advance on aerial photographs. The analysis of releves and collection of ground data along transects represent an efficient, representative and precise way of sampling. The vegetation data were classified and 19 plant communities distinguished. The plant communities were subjected to detrended correspondence analysis (DCA). Among the recorded variables, moisture is the one with the highest correlation along axes one and two, and reflects a coincidental moisture and vegetation cover gradient. The vegetation component responsible for this positive correlation is the bryophytes. Likewise, the TWINSPAN classification confirms this gradient in a dendrogram reflecting the hierarchical structure of the plant communities. Plant communities constitute the base of a statistical model that links the communities and the SPOT satellite data. The model then classifies and maps plant communities by means of satellite data, covering the entire Broggerhalvoya peninsula. Satellite data and environmental data were analysed regarding their ability to distinguish the plant communities in a discriminant function analysis (DFA). The results of the DFA indicate that it may be reasonable to include all the information from the different satellite channels when using satellite data for vegetation classification purposes. Among the satellite data the panchromatic channel is the one adding the most unique information to the power of the model in separating plant communities. The classification of satellite data using the probability model indicates that plant communities with less than 30% vegetation cover could be classified with the same degree of confidence or better, as compared with plant communities with more than 30% vegetation cover. The overall percentage of correctly classified releves increased by 13% when using probability level two instead of level one (57.8 to 71.1%). The probability classification model makes it possible to experiment with different probability levels to improve the fit between the vegetation and satellite data classification.