Water content estimation in vegetation with MODIS reflectance data and model inversion methods

https://doi.org/10.1016/S0034-4257(02)00197-9Get rights and content

Abstract

Statistical and radiative-transfer physically based studies have previously demonstrated the relationship between leaf water content and leaf-level reflectance in the near-infrared spectral region. The successful scaling up of such methods to the canopy level requires modeling the effect of canopy structure and viewing geometry on reflectance bands and optical indices used for estimation of water content, such as normalized difference water index (NDWI), simple ratio water index (SRWI) and plant water index (PWI). This study conducts a radiative transfer simulation, linking leaf and canopy models, to study the effects of leaf structure, dry matter content, leaf area index (LAI), and the viewing geometry, on the estimation of leaf equivalent water thickness from canopy-level reflectance. The applicability of radiative transfer model inversion methods to MODIS is studied, investigating its spectral capability for water content estimation. A modeling study is conducted, simulating leaf and canopy MODIS-equivalent synthetic spectra with random input variables to test different inversion assumptions. A field sampling campaign to assess the investigated simulation methods was undertaken for analysis of leaf water content from leaf samples in 10 study sites of chaparral vegetation in California, USA, between March and September 2000. MODIS reflectance data were processed from the same period for equivalent water thickness estimation by model inversion linking the PROSPECT leaf model and SAILH canopy reflectance model. MODIS reflectance data, viewing geometry values, and LAI were used as inputs in the model inversion for estimation of leaf equivalent water thickness, dry matter, and leaf structure. Results showed good correlation between the time series of MODIS-estimated equivalent water thickness and ground measured leaf fuel moisture (LFM) content (r2=0.7), demonstrating that these inversion methods could potentially be used for global monitoring of leaf water content in vegetation.

Introduction

Quantitative estimation of leaf biochemical and canopy biophysical variables is a key element to the successful application of remote sensing in vegetation monitoring, a major goal in terrestrial ecology and a long-term research objective given the complexity of the vegetation canopies and phenomena Goetz et al., 1992, Verstraete et al., 1994. Accurate estimates of leaf pigments, nitrogen, dry matter, water content, and leaf area index (LAI) from remote sensing can assist in determining vegetation physiological status Carter, 1994, Peñuelas et al., 1994, the study of species and seasonal dependence (e.g., Belanger, Miller, & Boyer, 1995), and may serve as bioindicators of vegetation stress (e.g., Luther & Carroll, 1999, Zarco-Tejada et al., 2001). The remote determination of one of these biochemical constituents, vegetation water content, has important implications in agriculture and forestry (Gao & Goetz, 1995), it is essential for drought assessment in natural vegetation (Peñuelas, Filella, Biel, Serrano, & Savé, 1993), and it is a major driver in predicting the susceptibility to fire Chandler et al., 1983, Pyne et al., 1996, Ustin et al., 1998.

Several studies demonstrate the existing link between leaf-level reflectance in the 400–2500 nm spectral region and the amount of water in the leaf through optical indices, regression analysis and radiative transfer modeling Aldakheel & Danson, 1997, Allen et al., 1971, Carter, 1991, Carter, 1993, Ceccato et al., 2001, Danson et al., 1992, Gausman et al., 1970, Hunt et al., 1987, Jacquemoud & Baret, 1990. The primary and secondary effects of water content on leaf reflectance were studied by Carter (1991) showing that sensitivity of leaf reflectance to water content was greatest in spectral bands centered at 1450, 1940, and 2500 nm. Indirect effects of water content on reflectance were also found at 400 nm, in the red edge at 700 nm (Filella & Peñuelas, 1994) and on vegetation indices such as NDVI (Roberts, Green, & Adams, 1997). The effects of leaf structure on the water absorption bands were studied by Danson et al. (1992) showing that derivative reflectance calculated in the water absorption features minimized the effects due to leaf structure, therefore maximizing the sensitivity to leaf water content. Lately, the broad use of leaf radiative transfer models such as PROSPECT (Jacquemoud & Baret, 1990) for broadleaf species, LIBERTY (Dawson, Curran, & Plummer, 1998) based on radiative transfer characterization in needles, and LEAFMOD (Ganapol, Johnson, Hammer, Hlavka, & Peterson, 1998) among others, enable the simulation of the leaf optical properties as a function of leaf structural and biochemical constituents such as chlorophyll a+b (Ca+b), dry matter (Cm), and leaf equivalent water thickness (Cw).

Several research efforts focus on the application of leaf-level indices calculated from water-absorption bands, statistical relationships between leaf reflectance and leaf water content, and scaling-up methods to canopy level through radiative transfer simulation. As an example, airborne visible infrared imaging spectrometer (AVIRIS) imagery was used to derive equivalent water thickness in vegetation using nonlinear and linear least squares spectral matching techniques, achieving good agreements with ground measured leaf fuel moisture (LFM) content (Gao & Goetz, 1995). In other studies, Pinzón, Ustin, Castañeda, and Smith (1998) and Ustin et al. (1998) using hierarchical foreground/background analysis (HFBA) showed general success retrieving equivalent water thickness by radiative transfer modeling using the 960-nm absorption band, but breaking down for low LAI canopies due to the increasing effects of soil background and shadows on sparse vegetation. Other ratios, such as the plant water index (PWI, R970/R900) (Peñuelas, Piñol, Ogaya, & Filella, 1997) was used to map vegetation water content with AVIRIS imagery, but found to be affected not only by water content, but also by canopy structure and viewing geometry (Serrano, Ustin, Roberts, Gamon, & Peñuelas, 2000) therefore highly dependent on bi-directional and geometrical effects of the vegetation canopy. The normalized difference water index (NDWI) calculated as (R860R1240)/(R860+R1240) was suggested by Gao (1996) in a theoretical study, demonstrating its potential applicability for canopy-level water content estimation based on the liquid water absorption band centered at 1240 nm enhanced by canopy scattering. Nevertheless, Zarco-Tejada and Ustin (2001) showed in a simulation study the dependency of NDWI and the simple ratio water index (SRWI, R858/R1240) on leaf-level variables such as leaf structure and dry matter content, and most importantly, on canopy LAI.

These partially successful results, obtained when estimating vegetation water content, demonstrate the need for modeling efforts to account for leaf and canopy variables and the viewing geometry. Efforts to investigate the successful scaling up of optical indices from leaf to canopy level through radiative transfer simulation Haboudane et al., 2002, Zarco-Tejada et al., 1999, Zarco-Tejada et al., 2001 show the importance of these methods when estimating leaf biochemical constituents from canopy reflectance data. Nevertheless, few investigations are found which apply these radiative transfer methods and model inversion techniques for leaf water content estimation from canopy reflectance imagery. Examples are the simulation studies to estimate equivalent water thickness by model inversion by Jacquemoud, Baret, Andrieu, Danson, and Jaggard (1995), Jacquemoud et al. (1996) and Fourty and Baret (1997) linking the PROSPECT leaf model (Jacquemoud & Baret, 1990) and SAIL canopy model (Verhoef, 1984), showing successful retrievals of Cw from continuous hyperspectral (AVIRIS-simulated) and multispectral (TM-simulated) spectra taking into account leaf- and canopy-level variables. Recently, Ceccato, Flasse, and Grégoire (2002) and Ceccato, Gobron, Flasse, Pinty, and Tarantola (2002) developed an optical index for SPOT-VEGETATION sensor based on global sensitivity analysis using radiative transfer models.

The work presented here investigates the applicability of these radiative transfer techniques to MODIS reflectance data for vegetation water content estimation. A simulation study with synthetic spectra and MODIS-equivalent reflectance is presented, investigating the spectral capabilities of MODIS for estimating leaf equivalent water thickness by inversion of a linked leaf–canopy model. Further, a seasonal field study was undertaken for leaf sampling and analysis of LFM content from 10 study sites, measuring fresh and dry weight of leaf samples. Time-series of MODIS reflectance spectra from the study sites during the period of the field experiment were used for model inversion to estimate equivalent water thickness by radiative transfer simulation.

Section snippets

Modeling the effects of water content on MODIS-equivalent reflectance

Willstatter and Stoll (1913) presented the earliest description of a theory to explain the leaf optical properties, with subsequent improvements and development of new leaf models: Allen and Richardson (1968); Breece and Holmes (1971); Woolley (1971); Allen, Gausman, Richardson, and Thomas (1969), Allen, Gayle and Richarson (1970). Allen and Richardson (1968) applied the Kubelka–Munk (K–M) theory (Kubelka & Munk, 1931) to study the interaction of light with stacked plant leaves, relating leaf

Field sampling study and results for Cw estimation from MODIS data

A field sampling campaign was conducted for the analysis of LFM content, measuring fresh and dry weight from leaf samples collected in 10 study sites of chaparral vegetation in California (USA) between March and September, 2000. Data collection was conducted as part of research to study spatially explicit models of fire spread through chaparral fuels (Morais, 2001). MODIS reflectance data were obtained for the same period of field data acquisition, and inversion methods conducted linking

Conclusion

Simulation methods and results from the modeling and field studies described in this article demonstrate that leaf water content can be globally monitored with MODIS reflectance data by radiative transfer modeling. Optical indices previously suggested in the literature as potential indicators of vegetation water content, such as SRWI and NDWI, were used in a simulation study with linked leaf–canopy models. Leaf- and canopy-level variables such as leaf structure, dry matter content, soil

Acknowledgements

This work was supported by NASA contract NAG5-9360 Incorporating New EOS Data Products into Models to Improve Estimates of Biogeochemical Processes. The authors gratefully acknowledge M. Marais and D. Roberts (University of California, Santa Barbara) and C. Lee (California State University, Long Beach) for their contribution in the leaf data used in this study. W. Verhoef and S. Jacquemoud are gratefully acknowledged for providing computer code for leaf and canopy reflectance models; B.-C. Gao

References (55)

  • S Jacquemoud et al.

    Comparison of four radiative transfer models to simulate plant canopies reflectance—direct and inverse mode

    Remote Sens. Environ.

    (2000)
  • S Jacquemoud et al.

    Prospect: a model of leaf optical properties spectra

    Remote Sens. Environ.

    (1990)
  • S Jacquemoud et al.

    Extraction of vegetation biophysical parameters by inversion of the PROSPECT+SAIL models on sugar beet canopy reflectance data. Application to TM and AVIRIS sensors

    Remote Sens. Environ.

    (1995)
  • S Jacquemoud et al.

    Estimating leaf biochemistry using the PROSPECT leaf optical properties model

    Remote Sens. Environ.

    (1996)
  • J.E Luther et al.

    Development of an index of Balsam Fir vigor by foliar spectral reflectance

    Remote Sens. Environ.

    (1999)
  • J Peñuelas et al.

    Reflectance indices associated with physiological changes in nitrogen- and water-limited sunflower leaves

    Remote Sens. Environ.

    (1994)
  • D.A Roberts et al.

    Temporal and spatial patterns in vegetation and atmospheric properties from AVIRIS

    Remote Sens. Environ.

    (1997)
  • L Serrano et al.

    Deriving water content of chaparral vegetation from AVIRIS data

    Remote Sens. Environ.

    (2000)
  • S.L Ustin et al.

    Estimating canopy water content of chaparral shrubs using optical methods

    Remote Sens. Environ.

    (1998)
  • W Verhoef

    Light scattering by leaf layers with application to canopy reflectance modelling: the SAIL model

    Remote Sens. Environ.

    (1984)
  • Y.Y Aldakheel et al.

    Spectral reflectance of dehydrating leaves: measurements and modeling

    Int. J. Remote Sens.

    (1997)
  • W.A Allen et al.

    Water and air changes in grapefruit, corn and cotton leaves with maturation

    Agron. J.

    (1971)
  • W.A Allen et al.

    Interaction of isotropic light with compact plan leaf

    J. Opt. Soc. Am.

    (1969)
  • W.A Allen et al.

    Plant canopy irradiance specified by the Duntley equations

    J. Opt. Soc. Am.

    (1970)
  • W.A Allen et al.

    Interaction of light with a plant canopy

    J. Opt. Soc. Am.

    (1968)
  • M.J Belanger et al.

    Comparative relationships between some red edge parameters and seasonal leaf chlorophyll concentrations

    Can. J. Remote Sens.

    (1995)
  • H.T Breece et al.

    Bidirectional scattering characteristics of healthy green soybean and corn leaves vivo

    Appl. Opt.

    (1971)
  • Cited by (444)

    • Integrated wildfire danger models and factors: A review

      2023, Science of the Total Environment
    View all citing articles on Scopus
    View full text