Search

We investigate the benefits of assimilating in situ and satellite data of the fraction of photosynthetically active radiation (FAPAR) relative to eddy covariance flux measurements for the optimization of parameters of the ORCHIDEE (Organizing Carbon and Hydrology in Dynamic Ecosystem) biosphere model. We focus on model parameters related to carbon fixation, respiration, and phenology. The study relies on two sites—Fontainebleau (deciduous broadleaf forest) and Puechabon (Mediterranean broadleaf evergreen forest)—where measurements of net carbon exchange (NEE) and latent heat (LE) fluxes are available at the same time as FAPAR products derived from ground measurements or derived from spaceborne observations at high (SPOT (Satellite Pour l′Observation de la Terre)) and medium (MERIS (MEdium Resolution Imaging Spectrometer)) spatial resolutions. We compare the different FAPAR products, analyze their consistency with the in situ fluxes, and then evaluate the potential benefits of jointly assimilating flux and FAPAR data. The assimilation of FAPAR data leads to a degradation of the model‐data agreement with respect to NEE at the two sites. It is caused by the change in leaf area required to fit the magnitude of the various FAPAR products. Assimilating daily NEE and LE fluxes, however, has a marginal impact on the simulated FAPAR. The results suggest that the main advantage of including FAPAR data is the ability to constrain the timing of leaf onset and senescence for deciduous ecosystems, which is best achieved by normalizing FAPAR time series. The joint assimilation of flux and FAPAR data leads to a model‐data improvement across all variables similar to when each data stream is used independently, corresponding, however, to different and likely improved parameter values. Compatibility of in situ NEE and LE data and various FAPAR products through model‐data fusionFAPAR mainly constrains phenology; when assimilated alone, it may degrade the modeled fluxesCombining the two data streams is preferable for improving a process‐based vegetation model

A geometrical model, taking into account the diffuse, as well as the specular component of energy leaving soil surfaces in the visible and near-infrared, is discussed here. The model computes the bidirectional, reflectance of soils illuminated by a single source. A rough soil surface is simulated by equal-sized spheroids regularly spaced on a horizontal surface. The model was tested using soil bidirectional reflectance data obtained in laboratory conditions by Jacquemoud et al in 1992. Two parameters describing soil surface geometry were used for modelling the soil relalive reflectance in laboratory conditions: the relative distance (d/ a) between spheroids (relative to their horizontal radii ( a )and the shape of spheroids (b/ a) ( as proportion of their vertical (b) to horizontal radii (a)) The simulation of reflectance for soil surfaces of pebbles and sand, containing simple dense particles with rounded edges, can be carried out using the d/ a and b/ a ratios which nearly described their aclual geometry. The reflectance of more geometrically complicated soil surfaces, such as clay and peat with irregular secondary porous aggregates, can be simulated by surfaces of effective geometry of vertically elongated spheroids.

This paper describes procedures for characterizing agricultural crops and soils in remote sensing studies. The procedures are based on the accumulated experience of a number of researchers active in this field. Therefore, they represent a compromise between the theoretically desirable and the practically feasible, and should thus be an effective aid in further studies of this type. Although the guidelines were prepared specifically for microwave studies, adjustments were made to render the procedures applicable to optical studies as well. Given the increasing number of research teams involved in remote sensing applied to agriculture, there is an opportunity to acquire a broad data base on soils and crops in various geographic regions. To allow intercomparisons of such data, they must be obtained in a consistent manner. By following the proposed procedures and reporting results using the parameters described here, such intcrcomparisons should be possible on a continental or a global scale.

Improved parameters for a soil-adjusted vegetation index (SAVI) are derived using SAIL model output for simulated wheat canopy reflectance. The SAVI is much less sensitive than the ratio vegetation index to changes in background caused by soil colour or surface soil moisture content. The parameters are developed to minimize variability due to soil brightness over the practical range of solar elevations, lear angle distributions (LAD) and leaf area indices (LAI). The parameters are added to the near-infrared (NIR) and red reflectances before calculating the NIR/red ratio. Three new versions of the SAVI are developed based on the theoretical consideration of the effects of wet and dry soils. All three are superior to the NIR/red ratio, the perpendicular vegetation index and a soil-adjusted vegetation index. Our simplest version requires the addition of a parameter to the red reflectance. The second and third versions require an iterative procedure to find the best parameters that are then added to the red or the NIR and red reflectances. In general, SAVI adjustment parameters required to remove soil-brightness effects decrease as factors that obscure the soil-surface increase. Low solar elevation and high LAI result in increasingly negative values for the parameters. The NIR/red ratio is the vegetation index least influenced by soil brightness at LAI greater than three. Our iterated versions have lowest overall errors but are occasionally subject to higher errors at the combination of low zenith angles and erectophile canopies. The SAVI based on the bare soil reflectance performs best on field data.

A linear regression equation is found relating the photosynthetically active radiation intercepted by the canopy (PARi), measured with hemispherical photographs, and both the normalized difference ND and the ratio NIR/R vegetation indices. On the basis of this equation, NIR/R is used to estimate PARi during the crop cycle. The efficiency with which the PAR absorbed by the crop is transformed into biomass (εc) is calculated for three phenological phases of the crop. Nitrogen fertilization is the main factor affecting light interception. At the booting stage, PARi is about 15 per cent greater for treatments with higher nitrogen levels. εc is influenced by both nitrogen and irrigation levels, and varies with the phenological phases of the crop. For the irrigated plots, εc is higher in the period going from anthesis to soft dough and not in the period from stem elongation to anthesis as most published results indicate. Water stress is the main factor affecting εc. The greatest reductions of ;εc are observed on plots with higher biomass levels when water shortage starts. The results suggest the need for a water stress index for biomass estimations of rain-fed crops in regions susceptible to drought. This would require knowledge of εcc for the crop grown under non-limiting conditions.

Specific absorption coefficients for water and dry matter were estimated using a wide range of variation of fresh leaves. The coefficients were derived from the inversion of the PROSPECT leaf optical property model using reflectance and transmittance spectra measured over the 1 300-2 400-nm domain and the corresponding water content (g.cm-2) and specific leaf weight (mass of dry matter per unit leaf.area, g.cm-2). Results show that the estimated values of the specific absorption coefficient for dry matter were not reliable in the strong water absorption bands, although there was agreement with previous studies in spectral regions where water contributed moderately to leaf absorption. We thus proposed to use the values derived by Fourty et al (1996) for dry leaves for the specific absorption coefficient of dry matter. Estimated values of the specific absorption coefficient of water were slightly higher than the values proposed by Curcio and Petty (1951) for pure water. We then investigated the possibility of estimating leaf water content and specific weight by inverting the PROSPECT model using concurrently or separately reflectance and/or transmittance spectra measured over fresh leaves and the specific absorption coefficients proposed by Fourty et al (1996) for dry matter, and Curcio and Petty (1951) for water. Results obtained on the training data set and on an independent data set show accurate and robust estimates of both water content (RMSE = 0.0025 g.cm-2) and specific leaf weight (RMSE = 0.0016 g.cm-2) when reflectance and transmittance were used concurrently. When either reflectance or transmittance measurements were used, the performances of water and dry matter content estimation decreased because of the relaxation of constraints in the inversion process. Possible applications of these results are discussed. Estimation du contenu en eau et de la masse s?che surfacique des feuilles ? partir de spectres de r?flectance et de transmittance. Les coefficients sp?cifiques de l'eau et de la mati?re s?che sont estim?s en inversant le mod?le Prospect de propri?t?s optiques des feuilles sur une collection vari?e de feuilles fra?ches sur lesquelles les contenus en eau (g.cm-2), les masses s?ches surfacique (masse de mati?re s?che par unit? de surface de feuille, g.cm-2) et les spectres de r?flectance et transmittance dans le domaine 1 300-2 400 nm ont ?t? mesur?s. Les r?sultats montrent que les valeurs estim?es du coefficient d'absorption sp?cifique de la mati?re s?che ne sont pas fiables dans les bandes de forte absorption par l'eau, alors qu'en dehors de ces bandes, un bon accord est observ? avec les r?sultats ant?rieurs. Nous proposons donc d'utiliser les valeurs de coefficient sp?cifique d'absorption de Fourty et al (1996) calcul?es sur des feuilles s?ches. Les valeurs estim?es du coefficient sp?cifique d'absorption de l'eau sont l?g?rement surestim?es par rapport aux valeurs propos?es par Curcio et Petty (1951). Nous avons ensuite ?valu? la possibilit? d'estimer le contenu en eau et la masse s?che surfacique en inversant le mod?le Prospect en utilisant des spectres de r?flectance et/ou de transmittance mesur?s sur des feuilles fra?ches, avec les coefficient sp?cifiques propos?s par Fourty et al pour la mati?re s?che ou par Curcio et Petty pour l'eau. Les r?sultats obtenus sur le jeu d'apprentissage et un jeu de donn?es ind?pendant montrent une bonne estimation du contenu en eau (RMSE = 0,0016 g.cm-2) et de la masse surfacique s?che (RMSE = 0,0016 g.cm-2) quand la r?flectance et la transmittance sont utilis?es simultan?ment. En revanche, quand la r?flectance ou la transmittance est utilis?e seule, la pr?cision des estimations diminue significativement du fait de la r?duction des contraintes impos?es au processus d'inversion. Les applications possibles de ces r?sultats sont discut?es.