Derivative vegetation indices as a new approach in remote sensing of vegetation

Abstract

This paper focuses on the advantages of derivative vegetation indices over simple reflectance-based indices that are traditionally used for remote sensing of vegetation. The idea of using reflectance derivatives instead of simple reflectance spectra was proposed several decades ago. Despite this, it has not been widely used in monitoring systems because the derivatives lack reliable parameters. In addition, most satellite monitoring systems are not equipped with hyperspectral sensors, which are considered necessary for operating with the reflectance derivatives. Here, we present original data indicating that the chlorophyll-related derivative index D₇₂₅/D₇₀₂ we derived can be accurately estimated from a reflectance spectrum of 10 nm resolution that would be suitable for most satellite-based sensors. Furthermore, the index is not sensitive to soil reflectance and can therefore be used for testing of open crops. Presence of blanc reflectance is also unnecessary. Preliminary results of index testing are presented. Perspectives on using this and other derivative indices are discussed.

Keywords

derivative vegetation indices / remote sensing / vegetation status

Introduction

Vegetation is one of the most important elements that must undergo remote monitoring, as it is a necessary component of the living world. The photosynthetic apparatus of plants is a sensitive indicator of their state due to the rapid changes in chlorophyll content induced by variation of environmental conditions. Chlorophyll presence causes the specific shape of reflectance spectrum of vegetation that allows for the use of the spectra for remote assessment of plant status. Chlorophyll-related spectral indices are widely proposed for assessing various characteristics of vegetation as well as yield prognosis (Sid'ko, 2004; Wang et al., 2004). Some ecological stresses can be also detected; for example, leaks of gas or oil from pipelines (Smith et al., 2004).

Spectral methods of vegetation testing are widely applied in systems of large-scale remote monitoring. For example, the MARS system was developed for versatile control of agricultural crops in Europe (http://mars.jrc.it). One of its main approaches is based on the use of normalized difference vegetation index (NDVI) and its modifications. NDVI is a combination of spectral coefficients of brightness. The last ones must be estimated as a ratio of spectral brightness of tested vegetation and blank white surface. By definition, spectral brightness is related to a rather wide spectral interval of about 50-30 nm. There are at least two main factors that cause errors which can essentially deform the spectral data: 1) difficulty in obtaining the satisfactory blank surface from satellite measurements; 2) coefficients of brightness are extremely influenced by the measurement timing of areas with incomplete projective covering of soil by vegetation.

An intensive development of airborne and satellite hyperspectral equipment permits the use of another, more effective, approach to chlorophyll estimation. It is based on application of derivative vegetation indices.

Derivative vegetation indices for chlorophyll estimation

The main features in reflectance spectrum of vegetation are caused by presence of a red edge region at 680–750 nm and of a green maximum at about 550 nm. The features are caused by specificity of chlorophyll absorption spectrum. It was shown by many researchers that variations of chlorophyll content induce changes in a shape of spectrum in red edge region (Collins, 1978; Horler et al., 1983; Ferns et al., 1984; Baret et al., 1987; Kochubey et al., 1987; Rock et al., 1988; Milton and Monat, 1989; Miller et al., 1990; Filella and Penuelas, 1994). Therefore, a quantitative parameter of a shape of the spectral curve has to be informative in relation to pigment concentration. A sensitive parameter has been found, which is a ratio of extremums in the 1st derivative plot from a reflectance spectrum of leaf in the red edge region (Kochubey et al., 1988; Zarco-Tejada et al., 2002; 2003; 2004; Kochubey and Bidyuk, 2003; Smith et al., 2004). Linear regressions have been proposed for estimation of chlorophyll in leaves using a derivative vegetation index D_λ1/D_λ2, which is the ratio of intensities between short-wavelength and long-wavelength extremums in the 1st derivative plot (Kochubey et al., 1988; Kochubey and Bidyuk, 2003). Recently, we proposed another derivative index D₇₂₅/D₇₀₂ that, as opposed to the previous ones, is calculated as the ratio of intensities at two fixed wavelengths, 725 and 702 nm, although these points do not always coincide with maxima in the derivative plot (Kochubey and Kazantsev, 2007). It was shown that the regression describes the most reliable interrelation between chlorophyll contents and the derivative index. As the regression was the same for four plant species, it may be assumed a universal one and applicable for various plants. One more important feature of D₇₂₅/D₇₀₂ index simplifies an algorithm for computing. The procedure for searching extremums in the 1st derivative plot is excluded (Kochubey and Bidyuk, 2003). Thus, the index may be proposed as the fittest for use in hardware-software complexes for remote sensing of vegetation.

Resistance of the derivative vegetation indices to contribution of soil reflectance

Reflectance spectra of vegetation are subjected to essential distortions due to contribution of background (soil) reflectance in the cases when areas with incomplete projective covering of soil are measured remotely. Distortions of the spectra depend on optical characteristics of the background and on chlorophyll content of plants in tested areas (Fig. 1). A detailed analysis of the results obtained with reflectance spectra of model systems “plant-soil” with various projective covering of soil has shown that the maximal difference in value of D₇₂₅/D₇₀₂ does not exceed 10% in comparison with variant of 100% projective covering (Table 1).

The analogous calculation has been performed for NDVI (Table 1). Comparison of the results presented in Table 1 displays more essential influence of soil contribution on NDVI.

Possibility of application of the derivative vegetation indices for remote sensing

Some problems arise when derivative vegetation indices are used for remote (airborne or satellite) testing of phytocenoses. One problem relates to decoding of information contained in the reflectance spectrum of remotely measured vegetation. For example, if a crop consists of plants with multi-layer leaf arrangements, it needs to determine the share of each leaf layer’s contribution to the reflectance spectrum. Another problem is low spectral resolution (no more than 10 nm) in the majority of satellite-based devices. Thus, it is necessary to clarify if it is possible to apply the derivative indices method for chlorophyll estimation in each case.

To solve the first problem we conducted remote measurements on wheat crops containing plants with 4-6 leaf layers. The crops were grown under various levels of mineral nutrition and were tested at several growth stages, in order to obtain greater diversity in chlorophyll content. Reflectance measurements were performed with a field spectrometer we developed (Yatsenko et al., 2005; www.vegetation.kiev.ua) in 500-800 nm range with 1.5 nm sampling. The device was set at a height of 1 m above plant tops and caught light signals from a 40 cm × 40 cm fragment of a crop. Five to nine plants from each tested fragment were sampled for laboratory analysis. Chlorophyll content per leaf area was estimated separately for leaves of each layer using a standard chemical method (Wellburn, 1994). Multiple regression analysis showed that the upper layer of leaves gave a preferable contribution, about 85%, to the reflectance spectrum. Thus, chlorophyll estimations made on a basis of reflectance spectra of the crops consisting of multi-layer plants relate mainly to pigment contents in the upper leaf layer. This fact needs to be taken into account when developing algorithms for yield prognosis.

The possibility of using derivative vegetation index under lower spectral resolution has been analyzed with simulation approaches. Reflectance spectra of wheat leaves and crops were measured with high resolution 1 nm (leaves) and 1.5 nm (crops). Then the spectra were transformed to spectral curves with 5 and 10 nm resolutions. Magnitudes of D₇₂₅/D₇₀₂ index were calculated using spectral curves of different spectral resolution. Figure 2 shows the 1st derivative plots of reflectance spectra of winter wheat leaves with different spectral resolutions. For all cases, a resolution decrease evokes changes in the shape of the 1st derivatives. It is mainly related to smoothing of the fine details of the structure (Fig. 2), but only slight changes of D₇₂₅/D₇₀₂ index are observed (Table 2). The same result has been obtained for field measured reflectance spectra of winter wheat crops (Table 3).

Tables 2 and 3 show that the difference between D₇₂₅/D₇₀₂ index computed with spectral curves of high and low resolution is limited to 8% for different chlorophyll contents. Decreased spectral resolution is maximally effective when chlorophyll concentration is medium. The changes are similar at both leaf and canopy levels.

It is important to note that a slight sensitivity of D₇₂₅/D₇₀₂ index to spectral resolution in intervals up to 10 nm is probably caused by a high informative potential of intervals closest to 702 and 725 nm. Earlier, we investigated the functional capability of the indices formed from various spectral points in the red edge region of the reflectance spectrum (Kochubey and Kazantsev, 2007). It was shown that it is only possible to obtain the linear regression for the D₇₂₅/D₇₀₂ index with a constant term equal to zero, as it is required when chlorophyll content tends to zero. The regression caused the lowest error in chlorophyll estimation and was applicable to various plant species. Broadening a spectral interval to 10nm creates blurring of information in each spectral channel. However, these are not overlapped and there is a wide spectral interval between their limits, from 707 to 720 nm. It is possible that only these points promote maintenance of D₇₂₅/D₇₀₂ ratio for information potential.

The results obtained indicate that 10 nm resolution, which as a rule is inherent in satellite sensors, is quite applicable for derivative vegetation indices use for remote chlorophyll estimation.

Influence of incomplete projective covering of soil on D₇₂₅/D₇₀₂ index calculated with spectra of low spectral resolution

We analyzed how decreased soil projective covering influences D₇₂₅/D₇₀₂ index calculated from reflectance spectra of different spectral resolution. Reflectance spectra of “plant-soil” systems with various projective covering were simulated with winter wheat leaves on a background of light sand or dark soil. The spectra of 1 nm resolution were recorded with laboratory spectrophotometer and then modified to 10 nm resolution.

The results (Table 4) indicate that even at 25% projective covering of light background (sand) by leaves with low chlorophyll content, there was only a 16% difference in D₇₂₅/D₇₀₂ index when a reflectance spectrum of 10 nm resolution was used.

Thus, our analysis shows that reflectance spectra recorded by a sensor with 10 nm resolution can be used for successful chlorophyll estimation by using the derivative indices method for various objects, including those for which it is practically impossible to obtain such information by using coefficients of brightness.

Practice of using derivative vegetation indices for testing vegetation

Agricultural aspect. Estimation of D₇₂₅/D₇₀₂ was applied for control of growth process in wheat crops. Two wheat cultivars grown under varied nutrition supply were tested during their growth period. The method revealed heterogeneity of chlorophyll content in different areas of one crop, while no visual differences were observed (Fig. 3). The measurements also permitted detection of differences in sensitivity to fertilization of two tested cultivars.

Environmental aspect. We studied the possibility of using reflectance measurements for stress detection in plants. Stress status can be revealed not only with chlorophyll estimation but also with detection of stress pigments of plants, called anthocyanins. Numerous papers report accumulation of these pigments in plant leaves under various extreme conditions of growth such as chilling temperatures, nutrition deficiency, salinity, plant diseases, contamination of soil by oil products, and presence of heavy metals (Mark Hodges and Nozzolillo, 1995; Hipskind et al., 1996; Krupa et al., 1996; Chupakhina and Maslennikov, 2004; Eryılmaz, 2006). Many studies, including ours (Feild et al., 2001; Gitelson et al., 2001; Neill et al., 2002; Kazantsev, 2007), indicate that the appearance of anthocyanins causes essential transformations in leaf reflectance spectrum (Fig. 4), which suggested that an anthocyanin-related derivative index might be developed and may be (individually or together with chlorophyll-related index) used for determining plant stresses. We ultimately managed to develop a system of three derivative indices, which enable estimation of anthocyanin concentration in different plant species in a wide range of concentrations (Kazantsev, 2007). Applicability of the indices was tested on maize and vine plants grown under oil contamination. Crude or machine oil was added into pots with soil where the plants were grown in 1%–5% vol concentration. Presence of the pollutant could be detected by magnitude changes of the indices (Fig. 5).

Derivative vegetation indices as a basis for development monitoring systems for agriculture and ecology assignment

It is demonstrated above that the vegetation derivative index we proposed can be successfully estimated with spectral devices with a spectral resolution of up to 10 nm. Therefore, satellite spectral devices may use the index for remote sensing of vegetation. Earlier, we proposed a hardware-software complex for remote aboveground chlorophyll estimation. Our original software is adjusted for chlorophyll estimation by the original algorithm, but also can be added by other algorithms for detection of important characteristics of vegetation. There are data for testing a set of chlorophyll-based crop characteristics, such as content of total nitrogen (Kochubey et al., 1990; Filella et al., 1995; Moran et al., 2000), extent of ripening (Collins, 1978; González-Sanpedro et al., 2008), yield prognosis (Sid'ko, 2004; Wang et al., 2004), and early detection of some types diseases (Kanemasu et al. 1974; Lorenzen and Jensen, 1989). We obtained preliminary data in relation to the derivative index for estimation of water content in cereals’ leaves with reflectance spectrum in green region (Kazantsev and Kochubey, 2007). We plan to add corresponding algorithms to our software.

Derivative indices may be used to develop some solutions for ecological problems. It is known that many types of ecology stresses influence chlorophyll content, for example drought (Unganai and Kogan, 1998; Maselli, 2004), as well as various contaminations of soil and atmosphere (Milton and Monat, 1989; Smith et al., 2004). Thus, supervision of chlorophyll changes and temporary dynamics may be used to determine stress. The appearance of anthocyanins alongside chlorophyll changes may, in some cases, indicate the presence of a stress characteristic. Thus, derivative indices for chlorophyll and anthocyanins estimation may be taken as a form of vegetation monitoring system. A spectrometer with shorter spectral range, 500-800 nm, may be used as hardware for such a system. Such a hardware-software system specialized for vegetation testing possesses important advantages. In addition to the option of testing a set of important crop characteristics, it is also based on a more simply constructed spectral device. It makes the system more reliable and cheaper, and therefore more commercially attractive.

As shown above, the reflectance spectral curves of different spectral resolution can be used for calculation of derivative indices. Therefore our algorithms for computation of chlorophyll and anthocyanins can be applied to aboveground measurements and, with slight modifications, to plane and satellite ones. It creates a basis for an effective system of verification of space measurements.

We also believe that an approach to yield prognosis based on the derivative indices method is more attractive than one based on the NDVI method. This is due in part to the possibility of estimating chlorophyll in the crops of very low projective covering due to the practical insensitivity of the derivative indices to contributions of soil reflection. It is also important to take into account the character of information that is contained in reflectance spectra of the crops of various plant species, which especially relates to the plant crops with multi-layered leaves organization.

References

[1] Baret F, Champion I, Giot G, Podaire A (1987). Monitoring wheat canopies with a high spectral resolution radiometer. Remote Sens Environ, 22(3): 367–378

[2] Chupakhina G N, Maslennikov P V (2004). Plant adaptation to oil stress. Russ J Ecol, 35(5): 290–295

[3] Collins W (1978). Remote sensing of crop type and maturity. Photogramm Eng Remote Sensing, 44(1): 43–55

[4] Eryılmaz F (2006). The relationships between salt stress and anthocyanin content in higher plants. Biotechnol, Biotechnol Equip, 20: 47–52

[5] Feild T S, Lee D W, Holbrook N M (2001). Why leaves turn red in autumn. The role of anthocyanins in senescing leaves of red-osier dogwood. Plant Physiol, 127(2): 566–574

[6] Ferns D C, Zara S J, Barber J (1984). Application pf of high spectral resolution spectroradiometry to vegetation. Photogramm Eng Remote Sensing, 50: 1725–1735

[7] Filella I, Penuelas J (1994). The red edge position and shape as indicators of plant chlorophyll content, biomass and hidric status. Int J Remote Sens, 15(7): 1459–1470

[8] Filella I, Serrano I, Serra J, Penuelas J (1995). Evaluating wheat nitrogen status with canopy reflectance indices and discriminant analysis. Crop Sci, 35(5): 1400–1405

[9] Gitelson A A, Merzlyak M N (1994). Spectral reflectance changes associate with autumn senescence of Aesculus hippocastanum L. and Acer platanoides L. leaves. Spectral features and relation to chlorophyll estimation. J Plant Physiol, 143(3): 286–292

[10] Gitelson A A, Merzlyak M N, Chivkunova O B (2001). Optical properties and nondestructive estimation of anthocyanin content in plant leaves. Photochem Photobiol, 74(1): 38–45

[11] González-Sanpedro M C, Le Toan T, Moreno J, Kergoat L, Rubio E (2008). Seasonal variations of leaf area index of agricultural fields retrieved from Landsat data. Remote Sens Environ, 112(3): 810–824

[12] Hipskind J, Wood K, Nicholson R L (1996). Localised stimulation of anthocyanin accumulation and delineation of pathogen ingress in maize genetically resistant to Biplaris maydis race O. Physiol Mol Plant Pathol, 49(4): 247–256

[13] Horler D N H, Dockrey M, Barber J (1983). The red edge of plant reflectance. Int J Remote Sens, 4(2): 273–288

[14] Kanemasu E T, Niblett C L, Manges H, Lenhert D, Newman M A (1974). Wheat: its growth and disease severity as deduced from ERTS-1. Remote Sens Environ, 3(4): 255–260

[15] Kazantsev T A (2007). Spectral properties and contactless determination of anthocyanins content in leaves of Parthenocissus quinquefolia (L.) Planch. Physiology and Biochemisty of Cultivated Plants (Russian).39: 382–390

[16] Kochubey S M, Bidyuk P I (2003). Novel approach to remote sensing of vegetation. Proc SPIE, 5093: 181–188

[17] Kochubey S M, Kazantsev T A (2007). Changes in the first derivatives of leaf reflectance spectra of various plants induced by variations of chlorophyll content. J Plant Physiol, 164(12): 1648–1655

[18] Kochubey S M, Kobets N I, Shadchina T M (1987). The shape of the reflectance spectra of leaves as an information basis for remote diagnosing of crop state. Physiology and Biochemistry of Cultivated Plants (Russian), 19: 539–545

[19] Kochubey S M, Kobets N I, Shadchina T M (1988). The quantitative analysis of the shape of reflectance spectra of leaves of plants as a method for testing their state. Physiology and Biochemistry of Cultivated Plants (Russian), 20: 535–539

[20] Kochubey S M, Kobets N I, Shadchyna T M (1990). Spectral properties of leaves as a basis for remote diagnostics. Naukova Dumka, Kyiv (in Russian)

[21] Krupa Z, Baranowska M, Orzol D (1996). Can anthocyanins be considered as heavy metal stress indicator in higher plants? Acta Physiol Plant, 18(2): 147–151

[22] Lorenzen B, Jensen A (1989). Changes in leaf spectral properties induced in barley by cereal powdery mildew. Remote Sens Environ, 27(2): 201–209

[23] Maselli F (2004). Monitoring forest conditions in a protected Mediterranean coastal area by the analysis of multiyear NDVI data. Remote Sens Environ, 89(4): 423–433

[24] Mark Hodges D, Nozzolillo C (1995). Anthocyanin and anthocyanoplast content of cruciferous seedlings subjected to mineral nutrient deficiencies. J Plant Physiol, 147(6): 749–754

[25] Miller J R, Hare E W, Wu J (1990). Quantitative characterization of the vegetation red edge reflectance 1. An inverted-Gaussian reflectance model. Int J Remote Sens, 11(10): 1755–1773

[26] Milton N M, Monat D A (1989). Remote sensing of vegetation responses to natural and cultural environment condition. Photogramm Eng Remote Sensing, 55(8): 1167–1173

[27] Moran J A, Mitchell A K, Goodmanson G, Stockburger K A (2000). Differentiation among effects of nitrogen fertilization treatments on conifer seedlings by foliar reflectance: a comparison of methods. Tree Physiol, 20(16): 1113–1120

[28] Neill S O, Gould K S, Kilmartin P A, Mitchell K A, Markham K R (2002). Antioxidant activities of red versus green leaves in Elatostema rugosum. Plant Cell Environ, 25(4): 539–547

[29] Rock B N, Hoshizaki T, Miller J R (1988). Comparison of in situ and airborne spectral measurements of the blue shift associated with forest decline. Remote Sens Environ, 24(1): 109–127

[30] Sid'ko A F (2004). Remote assay for chlorophyll photosynthetic potential of crops on the example of wheat. Biol Bull Russ Acad Sci, 31(5): 450–456

[31] Smith K L, Steven M D, Colls J J (2004). Use of hyperspectral derivative ratios in the red-edge region to identify plant stress responses to gas leaks. Remote Sens Environ, 92(2): 207–217

[32] Unganai L S, Kogan F N (1998). Drought monitoring and corn yield estimation in Southern Africa from AVHRR data. Remote Sens Environ, 63(3): 219–232

[33] Wang Z J, Wang J H, Liu L Y, Huang W J, Zhao C J, Wang C Z (2004). Prediction of grain protein content in winter wheat (Triticum aestivum L.) using plant pigment ratio (PPR). Field Crops Res, 90(2–3): 311–321

[34] Wellburn A R (1994). The spectral determination of chlorophylls a and b as well, as the total carotenoids using various solvents with spectrophotometers of different resolution. J Plant Physiol, 144(3): 307–313

[35] Yatsenko V A, Kochubey S M, Donets V V, Kazantsev T A (2005). Hardware-software complex for chlorophyll estimation in phytocenoses under field conditions. Proc SPIE, 5964: 1–6

[36] Zarco-Tejada P J, Miller J R, Mohammed G H, Noland T L, Sampson P H (2002). Vegetation stress detection through chlorophyll a + b estimation and fluorescence effects on hyperspectral imagery. J Environ Qual, 31(5): 1433–1441

[37] Zarco-Tejada P J, Miller J R, Morales A, Berjón A, Agüera J (2004). Hyperspectral indicies and model simulation for chlorophyll estimation in open-canopy tree crops. Remote Sens Environ, 90(4): 463–476

[38] Zarco-Tejada P J, Pushnik J C, Dobrowski S, Ustin S L (2003). Steady-state chlorophyll a fluorescence detection from canopy derivative reflectance and double-peak red-edge effects. Remote Sens Environ, 84(2): 283–294

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg