Responses of Spectral Indices to Variations inVegetation Cover and Soil Background

Stella W. Todd and Roger M. Hoffer

Abstract

The primary objective of this study was to evaluate the effects of variations in soil texture and moisture upon the green vegetation index (GVI)and the normalized difference vegetation index (NDVI) for targets with specific vegetation cover amounts and varying soil backgrounds. The second objective was to understand the difference in information provided between NDVI and GVI relative to estimating vegetation cover. The third objective was to investigate the information contained within the wetnesdbrigh tness plane in relation to soil background characteristics and variations in percent canopy cover. Brightness and wetness were estimated using the Tasseled Cap brightness index (BI) and wetness index (wI]. A simple two-component model of soil and green vegetation reflectance was used to simulate the effects of three soil texture types (sand, silt, and clay] and two soil moisture classes on greenness, brightness, and wetness values. The results indicated that, for the same vegetation percent cover class, targets with more moist soil backgrounds displayed higher NDVI values than did targets with more dry soil backgrounds. In contrast, GVI values were much less influenced by soil background variation. WI values increased as green vegetation cover increased for all soil backgrounds.bThe largest increase was for dry soil backgrounds. BI values either increased or decreased as green vegetation cover increased, depending on soil background brightness. BI and WI provided complimentary spectral information.

Introduction

One of the most promising applications of remote sensing imagery is for the estimation of aboveground plant biomass and/or plant cover across multiple geographic scales. For global or continental scales, the Advanced Very High Resolution Radiometer (AVHRR) is currently the only realistic option for biomass estimation. Its large spatial extent, daily temporal coverage, and low cost facilitate seasonal and yearly estimates of biomass production. Limitations include a relatively large spatial resolution (1 km) and few spectral reflectance channels (one visible and one near-infrared). Biomass across large geographic areas has primarily been estimated using the normalized difference vegetation index (NDVI), which is based on the ratios of red and near-infrared channels (Rouseet al., 1973). The NDVI formulation is identical for AVHRR, the Multispectral Scanner ( ~ s s ) , and the Thematic Mapper (TM) satellite sensors, with the only variation being the width of visible and near-infrared wavebands.

For small geographic regions, sensors other than AVHRR can be employed for biomass estimation, which overcome some of AVHRR'S spatial and spectral resolution limitations. The M ~ S with its smaller spatial resolution (80 m) and increased spectral reflectance resolution (two visible and two near-infrared) expanded the capacity of satellite sensors to

characterize physical scene reflectance characteristics. Vegetation and soil indices utilizing four spectral channels were now possible. The four-band Tasseled Cap orthogonal linear transformation, developed by Kauth and Thomas (1976),produced a plane of data in which the primary axis aligned with soil reflectance variation (brightness) and the secondary axis aligned with green vegetation (greenness] variation. Virtually all of the soil variation observed is arranged in a longcigar-shaped cloud centered along the brightness axis for M s s data.

The TM satellite further increased the spatial (30 m) and spectral resolution (three visible, one near-infrared, and two mid-infrared) for characterizing physical object reflectances. The mid-inhared TM bands are coincident with the water absorption regions of the spectrum (Hoffer, 1978). Mid-infrared band reflectances are incorporated into the Tasseled Cap dimension of total scene reflectance (brightness). They are also contrasted with the near-infrared bands to form the dimension of scene moisture (wetness).In TM space, soil variation is represented by both the brightness and wetness axes, creating a two-dimensional plane (Crist, 1983; Crist and Cicone, 1984).  The wetness axis is sensitive to both soil moisture and vegetation moisture (Crist and Cicone, 1984).

We were interested in determining which spectral indices would be useful for characterizing variation in green vegetation biomass across small regions with heterogenous soil background characteristics. We therefore focused on TM indices for this study. One of the major problems in determining the quantity of green vegetation using satellite sensors is that the spatial resolution of the sensors is generally larger than the vegetation objects. This is true for TM as well as for MSS and AVHRR. Therefore, pixel measurements represent an integration of subpixel reflectance components of soil, vegetation, the reflectance interaction between soil and vegetation, and shadows, all modified by atmosphere (Richardson and Weigand, 1990; Jasinski and Eagleson, 1989).

Several research results have been reported concerning the effects of soil background on ratio-based and orthogonal vegetation index values (Elvidge and Lyon, 1985; Gardner and Blad, 1986; Huete and Jackson, 1987; Huete et al., 1984; Huete et al., 1985; Tueller, 1987). Previous studies showed that backgrounds containing dark colored soils and other low reflecting soils displayed higher TM ratio-based vegetation index values ( N D ~ ) than did light colored soils or other high reflecting soils, given the same vegetation cover (Elvidge and Lyon, 1985; Huete et al., 1985; Huete and Jackson, 1987;Huete and Jackson, 1988; Heilman and Boyd, 1986).

Soil reflectance properties depend on numerous soil characteristics. Field soil reflectance is reduced, particularly in the visible portion of the spectrum, when organic matter, iron oxides, or moisture content is high (Hoffer, 19781. The near-infrared and mid-infrared regions of the spectrum are also affected. Soil has an easily distinguishable characteristic reflectance pattern in the visible, near-infrared and mid-infrared wavelengths. Soil reflectance patterns are generally linear with increasing reflectance as wavelengths increase from visible to mid-infrared.

The characteristic soil reflectance pattern is easily distinguishable from green vegetation (Bartolucci, 1977).Green vegetation reflectance is low for the visible bands (particularly red), with a sharp increase in reflectance in the near-infrared portion of the spectrum. Reflectance is also low in the midinfrared regions associated with water absorption. Physical vegetation properties vary with plant species, environmental stress, and phenology (Hoffer, 1978). Pigmentation and moisture content change as a plant senesces. A loss of chlorophyll pigmentation produces higher visible reflectance, particularly in the red region of the spectrum (Hoffer, 1978). Plant drying also increases visible and mid-inbared reflectance.

Some of the soil induced effects on vegetation indices have been attributed to additional NIR irradiance underneath and in-between canopies due to NIR scattering and transmission properties of the canopy, with intermediate canopies displaying the largest effect (Huete, 1988). Canopy scattering is small with low vegetation cover while the soil signal is small with high vegetation cover. Soil reflects some of the scattered and transmitted NIR flux back toward the sensor,depending upon the soil's reflectance properties.

NIR absorbtance, transmittance, scattering, and reflectance can be modeled based on the physical properties of plant canopies and soil (Suits, 1972; Verhoeff and Bunnik,1981; Verhoef, 1984; Verhoef, 1985). These models were developed for homogeneous plant canopies assuming fixed [Suits modell or arbitrarilv [SAIL model1 inclined leaves (suits, 1972;'verhoeff and ~ u n n i k , l98i). In a comparative study, the Suits and SAIL models were poorly to moderately correlated, respectively, with observed reflectance measurements for homogeneous crops (Badhwar et al., 1985). Modeling heterogeneous canopies would require even greater complexity than homogeneous canopies.

Another approach to modeling canopy reflectance conceptualizes clumps of vegetation as three-dimensional geometric elements against a flat soil background (Jasinskiand Eagleson, 1989).This type of model assumes that soil and vegetation are independent contributors to spectral reflectance. In addition, the effects of canopy shadows can be modeled based on plant size, shape, and geometric distribution. The simplest case is at the nadir view angle where the shadow component drops out. The geometric model can be applied to landscapes using statistical plant distribution patterns (Li and Strahler, 1985).

Previous studies have described methods to minimize soil background effects given a priori knowledge about canopy cover (Huete, 1988) or soil background characteristics (Jackson, 1983). Canopy cover information is often unknown. Realistically, real-time soil background reflectance information is not available. Soil moisture conditions are dynamic and spatially variable even within a small region. Therefore, vegetation indices which are insensitive to differing soil backgrounds are desirable for determining vegetation biomass and/or vegetation cover on small fields or across regions.

While a study by Huete et al. (1985) provided some insights into the limitations of NDVI and GVI for estimating vegetation cover, it focused only on soil brightness variation.

Within TM data, soil reflectance characteristics are distributed in a plane defined by wetness as well as brightness. Therefore, normalization to minimize soil related influences on vegetation indices should be based on the more expanded soil plane information (Huete and Tucker, 1991). Understanding the relationship of both soil type and moisture content with brightness and wetness dimensions and their interaction with greenness values for pixels containing both vegetation and soil components should precede the development of soil normalization applications. This study will investigate wetness as well as brightness characteristics in relation to vegetation index responses to variations in green vegetation canopy cover as well as variations in soil type and moisture content, using a simple two component model of soil and vegetation.

Objectives

The first objective of this study was to evaluate the effects of variations in soil texture and moisture upon the green vegetation index (GVI) and the normalized difference vegetation index (NDVI) for targets with specific vegetation cover amounts and varying soil backgrounds. The second objective was to understand the difference in information provided between NDVI and GVI relative to estimating vegetation cover. The third objective was to investigate the information contained within the wetness and brightness axes in relation to soil background characteristics and variations in percent canopy cover.

Methods

A composite reflectance (soil and green vegetation) was estimated using a simple two-component model, assuming that observations were from nadir, the sun was near zenith, and vegetation and soil components contributed proportionately to the total reflectance. Vegetation and soil components were modeled as linear, non-interacting mixtures. Variation in soil tvDe and moisture content ~roduced variations in soil reflecd 1 tance properties. Vegetation reflectance properties were those of green healthy vegetation only. Variations in reflectances due to vegetation senescence, soil color or organic matter.