Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture

https://doi.org/10.1016/j.rse.2003.12.013Get rights and content

Abstract

A growing number of studies have focused on evaluating spectral indices in terms of their sensitivity to vegetation biophysical parameters, as well as to external factors affecting canopy reflectance. In this context, leaf and canopy radiative transfer models are valuable for modeling and understanding the behavior of such indices. In the present work, PROSPECT and SAILH models have been used to simulate a wide range of crop canopy reflectances in an attempt to study the sensitivity of a set of vegetation indices to green leaf area index (LAI), and to modify some of them in order to enhance their responsivity to LAI variations. The aim of the paper was to present a method for minimizing the effect of leaf chlorophyll content on the prediction of green LAI, and to develop new algorithms that adequately predict the LAI of crop canopies. Analyses based on both simulated and real hyperspectral data were carried out to compare performances of existing vegetation indices (Normalized Difference Vegetation Index [NDVI], Renormalized Difference Vegetation Index [RDVI], Modified Simple Ratio [MSR], Soil-Adjusted Vegetation Index [SAVI], Soil and Atmospherically Resistant Vegetation Index [SARVI], MSAVI, Triangular Vegetation Index [TVI], and Modified Chlorophyll Absorption Ratio Index [MCARI]) and to design new ones (MTVI1, MCARI1, MTVI2, and MCARI2) that are both less sensitive to chlorophyll content variations and linearly related to green LAI. Thorough analyses showed that the above existing vegetation indices were either sensitive to chlorophyll concentration changes or affected by saturation at high LAI levels. Conversely, two of the spectral indices developed as a part of this study, a modified triangular vegetation index (MTVI2) and a modified chlorophyll absorption ratio index (MCARI2), proved to be the best predictors of green LAI. Related predictive algorithms were tested on CASI (Compact Airborne Spectrographic Imager) hyperspectral images and, then, validated using ground truth measurements. The latter were collected simultaneously with image acquisition for different crop types (soybean, corn, and wheat), at different growth stages, and under various fertilization treatments. Prediction power analysis of proposed algorithms based on MCARI2 and MTVI2 resulted in agreements between modeled and ground measurement of non-destructive LAI, with coefficients of determination (r2) being 0.98 for soybean, 0.89 for corn, and 0.74 for wheat. The corresponding RMSE for LAI were estimated at 0.28, 0.46, and 0.85, respectively.

Introduction

Green leaf area index (LAI) is a key variable used by crop physiologists and modelers for estimating foliage cover, as well as forecasting crop growth and yield. Its determination is critical for understanding biophysical processes of forest and crop canopies and for predicting their growth and productivity Daughtry et al., 1992, Goetz & Prince, 1996, Liu et al., 1997, Moran et al., 1995, Moran et al., 1997, Tucker et al., 1980. The expression “green LAI”, used in this paper, represents the LAI of living leaves regardless to their photosynthetic capacity. Living leaves can have similar structural characteristics but various pigment contents (i.e., chlorophyll). The exposed area of living leaves plays a key role in various biophysical processes such as plant transpiration and CO2 exchange. These functions are important for understanding exchanges between the vegetation and the atmosphere. Because LAI is functionally linked to the canopy spectral reflectance, its retrieval from remote sensing data has prompted many investigations and studies in recent years. This has led to the development of different techniques aiming to improve its estimation over large areas, mainly through the use of spectral indices, model inversions, and spectral mixture analysis. The latter has been successfully used to define an end-member of interest, determine its relative abundance, then correlate it with ground-measured LAI; thus, more or less significant correlation levels were found between LAI and sunlit fraction (Hu et al., 2004), shadow fraction (Peddle & Johnson, 2000), and crop fraction (Pacheco et al., 2001). Only very few studies have focused on inversion of sophisticated radiative transfer models to retrieve green LAI of crop canopies (Jacquemoud et al., 2000). The common and widely used approach has been to develop relationships between ground-measured LAI and vegetation indices Chen & Cihlar, 1996, Fassnacht et al., 1997, Spanner et al., 1990. Consequently, a large number of relationships have been established, and a wide range of determination coefficients (0.05<r2<0.66) between satellite-derived spectral indices and LAI were found Baret & Guyot, 1991, Brown et al., 2000, Chen, 1996.

During recent decades, substantial efforts were expended in improving the Normalized Difference Vegetation Index (NDVI) and in developing new indices aiming to compensate for soil background influences Bannari et al., 1996, Baret et al., 1989, Huete, 1988, Qi et al., 1994, Rondeaux et al., 1996, as well as for atmospheric effects Karnieli et al., 2001, Kaufman & Tanre, 1992. Even though the external perturbing factors related to changes in soil brightness and atmospheric conditions were taken into account, vegetation indices still have definite intrinsic limitations; they are not a single measure of a specific variable of interest such as pigment content, plant geometry, or canopy architecture. So far, it has not been possible to design an index which is sensitive only to the desired variable and totally insensitive to all other vegetation parameters (Govaerts et al., 1999). Therefore, different indices were defined for different purposes, and optimized to assess a process of interest. For instance, some spectral indices were proposed to capture the photochemical processes associated with photosynthesis activity such as light use efficiency or to estimate leaf pigment content Broge & Leblanc, 2000, Chappelle et al., 1992, Daughtry et al., 2000, Gamon et al., 1992, Haboudane et al., 2002, Kim et al., 1994. Others were designed to retrieve LAI Brown et al., 2000, Chen, 1996 or to quantify vegetation fraction (Gitelson et al., 2001).

A major problem in the use of these indices arises from the fact that canopy reflectance, in the visible and near-infrared, is strongly dependent on both structural (e.g., LAI) and biochemical properties (e.g., chlorophyll) of the canopy Goel, 1988, Jacquemoud et al., 2000, Zarco-Tejada et al., 2001. Moreover, LAI and chlorophyll content have similar effects on canopy reflectance particularly in the spectral region from the green (550 nm) to the red edge (750 nm). To uncouple their combined effect, recent studies Daughtry et al., 2000, Haboudane et al., 2002 have demonstrated that leaf chlorophyll content can be estimated with minimal confounding effects due to LAI through a combination of two kinds of spectral indices: indices sensitive to pigment concentration and indices resistant to soil optical properties influence. Conversely, no studies have focused on the retrieval of LAI without interference of chlorophyll effects. The latter generate a considerable scatter in the relationship between LAI and the vegetation index of choice.

In practice, LAI prediction from remotely sensed data faces two major difficulties: (1) vegetation indices approach a saturation level asymptotically when LAI exceeds 2 to 5, depending on the type of vegetation index; (2) there is no unique relationship between LAI and a vegetation index of choice, but rather a family of relationships, each a function of chlorophyll content and/or other canopy characteristics. To address these issues, a few studies have been carried out to assess and compare various vegetation indices in terms of their stability and their prediction power of LAI Baret & Guyot, 1991, Broge & Leblanc, 2000 while others have dealt with modifying some vegetation indices to improve their linearity with, and increase their sensitivity to, LAI Chen, 1996, Brown et al., 2000, Nemani et al., 1993. Consequently, some indices have been identified as best estimators of LAI because they are less sensitive to the variation of external parameters affecting the spectral reflectance of the canopy, namely soil optical properties, illumination geometry, and atmospheric conditions. However, the effect of leaf chlorophyll variations on the LAI–vegetation index relationship remains an unsolved problem. How does chlorophyll concentration influence the behavior of a vegetation index of choice? Which of these indices, suitable to LAI prediction, is least sensitive to chlorophyll changes? Is there a single LAI versus vegetation index curve for a wide range of leaf chlorophyll content?

The need for an answer to these questions has inspired the present study as a contribution to improving the use of hyperspectral remote sensing to predict LAI in the context of precision farming. The main purpose is to suggest a spectral index that is suitable to simply, and yet accurately, determine LAI of crop canopies for agriculture management purposes. The research focuses on reducing the variability in LAI estimates due to changes in leaf chlorophyll concentration. To achieve these objectives, PROSPECT and SAILH radiative transfer models were used to simulate crop canopy reflectance for various biochemical, structural and observation conditions, then a set of indices that have proven to be resistant to atmospheric and soil brightness effects were assessed in terms of their responsivity/resistance to chlorophyll content variability. Another objective was to validate the modeling approach through the application to airborne hyperspectral data and the comparison with field measurements carried out simultaneously with image acquisition.

Section snippets

The study area

The study area is located at the former Greenbelt Farm of Agriculture and Agri-Food Canada (45°18′N, 75°45′W, Ottawa, Canada). Over three successive years, different crops (corn, wheat, soybean) were grown on a 30-ha field with a drained clay loam soil as well as on adjacent fields managed by private producers. Prior knowledge of the field management and plant stress patterns helped in selecting ground truth sites of contrasting productivity in order to test the performance of predictive

Sensitivity to chlorophyll effects and saturation level: simulated data

The relationships between vegetation indices and green LAI are not unique, they exhibit a considerable scatter caused by chlorophyll content variation and/or the influence of other canopy characteristics. In fact, the indices are designed to measure vegetation greenness in which chlorophyll content plays a major role beside the amount of green leaves. To understand this influence, vegetation indices selected for this study are plotted against green LAI as a function of chlorophyll

Conclusions

Quantification of the canopy leaf area index (LAI) and its spatial distribution provides an avenue to improve the interpretation of remotely sensed data over vegetated areas, as well as valuable information to aid the development of approaches for ecosystem functioning and ecosystem productivity. Due to the complexity of the relationship between canopy reflectance and its biochemical components and structural descriptors, the present study used the radiative transfer model, PROSPECT and SAILH,

Acknowledgements

The authors are very grateful for the financial support provided by GEOmatics for Informed Decisions (GEOIDE), the Canadian Space Agency (CSA) and Agriculture and Agri-Food Canada. We thank Lawrence Gray, Phil Brasher and Heidi Beck of Aviation International for making CASI airborne field campaigns work efficiently. The field support of Lynda Blackburn, Dave Dow, Mathew Hinther and Dave Meridith was greatly appreciated. We also thank anonymous reviewers for valuable suggestions and criticism.

References (56)

  • D. Haboudane et al.

    Integrated narrow-band vegetation indices for prediction of crop chlorophyll content for application to precision agriculture

    Remote Sens. Environ.

    (2002)
  • B. Hu et al.

    Retrieval of the canopy leaf area index in the BOREAS flux tower sites using linear spectral mixture analysis

    Remote Sens. Environ.

    (2004)
  • A.R. Huete

    A soil vegetation adjusted index (SAVI)

    Remote Sens. Environ.

    (1988)
  • S. Jacquemoud et al.

    Prospect: A model for leaf optical properties spectra

    Remote Sens. Environ.

    (1990)
  • S. Jacquemoud et al.

    Estimating leaf biochemistry using the PROSPECT leaf optical properties model

    Remote Sens. Environ.

    (1996)
  • A. Karnieli et al.

    AFRI—aerosol free vegetation index

    Remote Sens. Environ.

    (2001)
  • O. Lillesaeter

    Spectral reflectance of partly transmitting leaves: Laboratory measurements and mathematical modeling

    Remote Sens. Environ.

    (1982)
  • J. Liu et al.

    A process-based boreal ecosystem productivity simulator using remote sensing inputs

    Remote Sens. Environ.

    (1997)
  • B. Matsushita et al.

    Integrating remotely sensed data with an ecosystem model to estimate net primary productivity in East Asia

    Remote Sens. Environ.

    (2002)
  • M.S. Moran et al.

    Opportunities and limitations for image-based remote sensing in precision crop management

    Remote Sens. Environ.

    (1997)
  • E. Pattey et al.

    Effects of nitrogen application rate and weather on corn using micrometeorological and hyperspectral reflectance measurements

    Agric. For. Meteorol.

    (2001)
  • B.C. Pengelly et al.

    Radiation interception and the accumulation of biomass and nitrogen by soybean and three tropical annual forage legumes

    Field Crops Res.

    (1999)
  • J. Qi et al.

    A modified soil vegetation adjusted index

    Remote Sens. Environ.

    (1994)
  • J. Qi et al.

    Leaf area index estimates using remotely sensed data and BRDF models in a semiarid region

    Remote Sens. Environ.

    (2000)
  • G. Rondeaux et al.

    Optimization of soil-adjusted vegetation indices

    Remote Sens. Environ.

    (1996)
  • I.B. Strachan et al.

    Impact of nitrogen and environmental conditions on corn as detected by hyperspectral reflectance

    Remote Sens. Environ.

    (2002)
  • W. Verhoef

    Light scattering by leaf layers with application to canopy reflectance modeling: The SAIL model

    Remote Sens. Environ.

    (1984)
  • A. Bannari et al.

    Effets de la couleur et de la brillance du sol sur les indices de végétation

    Int. J. Remote Sens.

    (1996)
  • Cited by (1968)

    View all citing articles on Scopus
    View full text