Francisco Javier García-Haro

A GENERALIZED SOIL-ADJUSTED VEGETATION INDEX

Operational monitoring of vegetative cover by remote sensing currently involves the utilisation of vegetation indices (VIs), most of them being functions of the reflectance in red (R) and near-infrared (NIR) spectral bands. A generalized soil-adjusted vegetation index (GESAVI), theoretically based on a simple vegetation canopy model, is introduced. It is defined in terms of the soil line parameters (A and B) as: GESAVI=(NIRBRA)/(R+Z), where Z is related to the red reflectance at the cross point between the soil line and vegetation isolines. As Z is a soil adjustment coefficient, this new index can be considered as belonging to the SAVI family. In order to analyze the GESAVI sensitivity to soil brightness and soil color, both high resolution reflectance data from two laboratory experiments and data obtained by applying a radiosity model to simulate heterogeneous vegetation canopy scenes were used. VIs (including GESAVI, NDVI, PVI and SAVI family indices) were computed and their correlation with LAI for the different soil backgrounds was analyzed. Results confirmed the lower sensitivity of GESAVI to soil background in most of the cases, thus becoming a very efficient index. This good index performance results from the fact that the isolines in the NIR-R plane are neither parallel to the soil line (as required by the PVI) nor convergent at the origin (as required by the NDVI) but they converge somewhere between the origin and infinity in the region of negative values of both NIR and R. This convergence point is not necessarily situated on the bisectrix, as required by other SAVI family indices. D 2002 Published by Elsevier Science Inc.

Linear spectral mixture modelling to estimate vegetation amount from optical spectral data

Abstract Spectral mixture modelling has developed in recent years as a suitable remote sensing tool for analysing the biophysical and compositional character of ground surfaces. In this paper the potentiality of the linear spectral mixture model to extract vegetation related parameters from 0·4-2·5 μm reflectance data has been tested. High spectral resolution reflectance measurements of soil-plant mixtures with different soil colour and plant densities were carried out in a laboratory experiment. The constrained least-squares and the factor analysis unmixing procedures were applied to generate endmember fractions of the components present in the mixtures and to test the validity of the model. It is concluded that the derived fraction of the vegetation endmember is less sensitive to soil background than the NDV[. The accuracy attainable by this modelling approach can be considered sufficient for many practical purposes, being operational in the monitoring of vegetation from satellite data.

Monitoring fire-affected areas using Thematic Mapper data

In this paper three methods for updating inventories of burned areas have been presented and examined. They include Multitemporal Principal Component Analysis (MPCA), Change Vector Analysis (CVA) and Multitemporal NDVI Classification (MNC). First, 11 Landsat-5 Thematic Mapper (TM) images of a forest area were radiometrically corrected to derive a multitemporal series of intercomparable images for each spring from 1984 to 1994. Then, in order to check the feasibility of the three approaches, they were used for mapping fire burns that occurred during 1992. The various procedures yielded different maps of burned areas; the MNC method seemed to be more reliable than the others, because it merges spectral data corresponding not only to 1992 (pre-fire) and 1993 (post-fire) but also to 1994 (the second year after the fires), which is key in the vegetation regeneration. Finally, this methodology was automated to yield an inventory of burned areas for each year during the period of study.

A Mixture Modeling Approach to Estimate Vegetation Parameters for Heterogeneous Canopies in Remote Sensing

In this article, we describe a reflectance model which parametrizes the reflectance of vegetation canopies from optical properties of leaves and soil, and dominant canopy structural parameters. The model assumes certain principles of geometric models, for example, that sensor integrates the radiance reflected from three components, plant, shaded soil, and illuminated soil. Its inversion provides compositional information of the ground surface that is linked with the interpretation of the linear spectral mixture modeling (LSMM). This model also offers the potential for retrieving other meaningful biophysical properties such as LAI. The model has been tested on simulated spectra of spectral mixtures in presence of significant multiple scattering. Results indicate that this modeling approach is a suitable remote sensing tool for retrieving the vegetation abundance in heterogeneous canopies, which is interpreted as the fractional cover of plants with well-defined structural parameters.

Vegetation cover seasonal changes assessment from TM imagery in a semi-arid landscape

This work evaluates the suitability of spectral mixture analysis (SMA) methods to assess vegetation cover seasonal changes in a desertification context. Our main interest is to produce remotely sensed derived maps, sensitive to vegetation activity and quite independent of the soil background. A further aim is to analyse the inter-annual variations of this magnitude for different natural vegetation species, in response to seasonal and climatic changes. Fractional vegetation cover (FVC) was obtained using a Variable Endmember Spectral Mixture Analysis (VESMA) technique. The aim is to identify the main vegetation cover and lithological units and decompose them in separate stages. The use of specific spectral signatures for each pixel allows for a better adaptation of the endmembers to local conditions, which is an important prerequisite to ensure the accuracy of fractions. The method has been tested on a well documented area, the Guadalentin river basin, located in south-eastern Spain. Unlike pine forest and stipa classes, rosmarinus, sparse shrubs and seasonal grasses classes displayed larger inter-annual variability, showing higher stress in response to water availability. A comparative analysis between FVC and the Normalized Difference Vegetation Index (NDVI) was also conducted. Average values were used as indicators of the dynamics of the vegetation cover, with the variance of each vegetation class giving similar results. The correlation between both magnitudes varied from 55% for the class with least coverage to 90% for the densest vegetation class. Regarding seasonal evolution, the average values and standard deviations of the changes in each vegetation class in specific periods were related to seasonal changes and the effects of the rainfall pattern. Significant differences were found between the two methods, with FVC showing a higher coherence.

Extraction of endmembers from spectral mixtures

Abstract Linear spectral mixture modeling (LSMM) divides each ground resolution element into its constituent materials using endmembers which represent the spectral characteristics of the cover types. However, it is difficult to identify and estimate the spectral signature of pure components or endmembers which form the scene, since they vary with the scale and purpose of the study. We propose three different methods to estimate the spectra of pure components from a set of unknown mixture spectra. Two of the methods consist in different optimization procedures based on objective functions defined from the coordinate axes of the dominant factors. The third one consists in the design of a neural network whose architecture implements the LSMM principles. The different procedures have been tested for the case of three endmembers. First, were used simulated and real data corresponding to mixtures of vegetation and soil. Factors that limit the accuracy of the results, such as the number of channels and the level of data noise have been analyzed. Results have indicated that the three methods provide accurate estimations of the spectral endmembers, especially the third one. Moreover, the second method, that is based on the exploration of the mixture positions in the factor space, has demonstrated to be the most appropriate when the dimensionality of the data is reduced. Finally, this procedure was applied on a Landsat-5 TM scene.

Intercomparison and quality assessment of MERIS, MODIS and SEVIRI FAPAR products over the Iberian Peninsula

Abstract The fraction of absorbed photosynthetically active radiation (FAPAR) is a key variable in productivity and carbon cycle models. The variety of available FAPAR satellite products from different space agencies leads to the necessity of assessing the existing differences between them before using into models. Discrepancies of four FAPAR products derived from MODIS, SEVIRI and MERIS (TOAVEG and MGVI algorithms), covering the Iberian Peninsula from July 2006 to June 2007 are here analyzed. The assessment is based on an intercomparison involving the spatial and temporal consistency between products and a statistical analysis across land cover types. In general, significant differences are found over the Iberian Peninsula concentrated on the temporal variation and absolute values. The MODIS and MERIS/MGVI FAPAR products clearly show the highest and lowest absolute values, respectively, along with the lowest intra-annual variation. When considering individual land cover types, the largest FAPAR disagreements among the analyzed products were found between MODIS-MERI/MGVI and MERIS/TOAVEG-MERIS/MGVI over broadleaf and needleaf forests, with discrepancies quantified by RMSE higher than 0.30 and absolute bias higher than 0.25. These discrepancies can lead to relative gross primary production differences up to 65%.

A methodology to generate a synergetic land-cover map by fusion of different land-cover products

Abstract The main goal of this study is to develop a general framework for building a hybrid land-cover map by the synergistic combination of a number of land-cover classifications with different legends and spatial resolutions. The proposed approach assesses class-specific accuracies of datasets and establishes affinity between thematic legends using a common land-cover language such as the UN Land-Cover Classification System (LCCS). The approach is illustrated over a large region in Europe using four land-cover datasets (CORINE, GLC2000, MODIS and GlobCover), but it can be applied to any set of existing products. The multi-classification map is expected to improve the performance of individual classifications by reconciling their best characteristics while avoiding their main weaknesses. The intermap comparison reveals improved agreement of the hybrid map with all other land-cover products and therefore indicates the successful exploration of synergies between the different products. The approach offers also estimates for the classification confidence associated with the pixel label and flexibility to shift the balance between commission and omission errors, which are critical in order to obtain a desired reliable map.

Accuracy assessment of fraction of vegetation cover and leaf area index estimates from pragmatic methods in a cropland area

The fraction of vegetation cover (FVC) and the leaf area index (LAI) are important parameters for many agronomic, ecological and meteorological applications. Several in‐situ and remote sensing techniques for estimating FVC and LAI have been developed in recent years. In this paper, the uncertainty of in‐situ FVC and LAI measurements was evaluated by comparing estimates from LAI‐2000 and digital hemispherical photography (DHP). The accuracy achieved with a spectral mixture analysis algorithm and two vegetation indices‐based methods was assessed using atmospherically corrected Landsat Thematic Mapper (TM) data over the Barrax cropland area where the European Space Agency (ESA) SENtinel‐2 and FLuorescence EXperiment (SEN2FLEX) field campaign was carried out in July 2005. The results indicate that LAI‐2000 and DHP performances are comparable, with uncertainties of 5% for FVC and 15% for effective LAI. The selected remote sensing methods are shown to be consistent, with a notable overall accuracy (root mean square error, RMSE) of 0.07 (10% in relative terms) for FVC and 0.8 (30%) for LAI. Similar bounds were found on upscaling in‐situ measurements with empirical transfer functions (TFs). These results suggest that the pragmatic methods considered applied at high resolution with minimum calibration data could be useful for mapping FVC and LAI in the study area, reducing in‐situ labour‐intensive characterization necessities for validation studies.

Multitemporal Monitoring of Plant Area Index in the Valencia Rice District with PocketLAI

Leaf area index (LAI) is a key biophysical parameter used to determine foliage cover and crop growth in environmental studies in order to assess crop yield. Frequently, plant canopy analyzers (LAI-2000) and digital cameras for hemispherical photography (DHP) are used for indirect effective plant area index (PAIeff) estimates. Nevertheless, these instruments are expensive and have the disadvantages of low portability and maintenance. Recently, a smartphone app called PocketLAI was presented and tested for acquiring PAIeff measurements. It was used during an entire rice season for indirect PAIeff estimations and for deriving reference high-resolution PAIeff maps. Ground PAIeff values acquired with PocketLAI, LAI-2000, and DHP were well correlated (R2 = 0.95, RMSE = 0.21 m2/m2 for Licor-2000, and R2 = 0.94, RMSE = 0.6 m2/m2 for DHP). Complementary data such as phenology and leaf chlorophyll content were acquired to complement seasonal rice plant information provided by PAIeff. High-resolution PAIeff maps, which can be used for the validation of remote sensing products, have been derived using a global transfer function (TF) made of several measuring dates and their associated satellite radiances.