Crop Science Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in Crop Science
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (22)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Agricola
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Related Collections
Right arrow Remote Sensing
Crop Science 42:1547-1555 (2002)
© 2002 Crop Science Society of America

CROP PHYSIOLOGY & METABOLISM

Relationship between Growth Traits and Spectral Vegetation Indices in Durum Wheat

N. Aparicioa, D. Villegasa, J. L. Arausb, J. Casadesúsb and C. Royo*,a

a Area de Cultius Extensius, Centre UdL-IRTA, Rovira Roure, 177, 25198 Lleida, Spain
b Unitat de Fisiologia Vegetal, Facultat de Biologia, Univ. de Barcelona, Diagonal 645, 08028 Barcelona, Spain

* Corresponding author (conxita.royo{at}irta.es)


   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Future wheat yield improvements may be gained by increasing total dry matter (TDM) production. Vegetation indices (VI) based on spectral reflectance ratios have been proposed as an appropriate method to assess TDM and leaf area index (LAI) in wheat. This study was undertaken to determine whether VI could accurately identify TDM and LAI in durum wheat {Triticum turgidum var. durum (Desf.) Bowden [= T. turgidum subsp. durum (Desf.) Husn.]} and serve as indirect selection criteria in breeding programs. Total dry matter and LAI were determined from destructive sampling from booting to milk-grain in seven field experiments conducted under Mediterranean conditions. Each experiment included one of two sets of 20 or 25 genotypes. Field reflectance values were collected using a portable field spectroradiometer. Two VI, the normalized difference vegetation index (NDVI) and the simple ratio (SR), were derived from spectral measurements and their predictive value for TDM and LAI was evaluated. The best stages for growth trait appraisal were Stages 65 and 75 of the Zadoks scale. The power of VI for assessing TDM was lower than their predictive value for LAI. The suitability of VI for the assessment of growth traits depended on the range of variability existing within the experimental data. Vegetation indices accurately tracked changes in LAI when data were analyzed across a broad range of different growth stages, environments, and genotypes. However, their value as indirect genotypic selection criteria for TDM or LAI was limited, since they lacked predictive ability for specific environment/growth stage combinations.

Abbreviations: LAD, leaf area duration • LAI, leaf area index • NDVI, normalized difference vegetation index • NSS, number of spikes per square meter • SR, simple ratio • TDM, total aboveground dry matter (g m-2) • VI, vegetation indices


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
GRAIN YIELD may be defined as a product of the TDM of the crop and its harvest index (Donald and Hamblin, 1976). In the past, increases in bread wheat yields around the world have largely been associated with changes in harvest index, whereas increases in TDM have been small or negligible (Cox et al., 1988; Austin et al., 1989; Slafer and Andrade, 1993). However, a study considering durum wheat bred by the International Maize and Wheat Improvement Center (CIMMYT) and released between 1960 and 1984 concluded that yield gains were mainly due to increased TDM with negligible increases in harvest index (Waddington et al., 1987). Several authors suggest that future wheat yields may be gained by increasing TDM production while maintaining harvest index, since further improvements in this index may be difficult to achieve (Donald and Hamblin, 1976; Richards, 1987, 2000). Moreover, in Mediterranean environments, final yield often depends largely on the retranslocation of assimilates from the leaves and stems to the grain, so the higher the TDM around anthesis, the higher the yield attained (Turner, 1982).

Destructive sampling for TDM assessment is tedious and time-consuming, and reduces the area available for determining final grain yield in small research plots. Also, sampling errors often hinder the detection of genotypic differences (Whan et al., 1991). In the last years, VI based on spectral reflectance measurements have been proposed as a reliable nondestructive method for quickly estimating TDM (Tucker, 1979; Asrar et al., 1984) and LAI in wheat and barley (Elliott and Regan, 1993). Total dry matter is mainly determined by two processes: (i) the interception by the canopy of incident solar irradiance, which depends on the photosynthetic area of the canopy; and (ii) the conversion of the intercepted radiant energy to potential chemical energy, which relies on the overall photosynthetic efficiency of the crop (Hay and Walker, 1989). A rapid, nondestructive simultaneous estimation of both processes may be provided by VI based on reflectance values at specific wavelengths or formulations using simple operations between reflectances at given wavelengths (Field et al., 1994).

Research to develop applications for VI has focused on wide bandwidths (Blackmer et al., 1996). High spectral resolution devices have recently improved in sensitivity, decreased in cost, and increased in availability. Improved technology and increased interest in the application of remote sensing techniques to agricultural studies (Moran et al., 1997) encourages a close examination of optimal wavelengths and designs of appropriate sensors to monitor crop changes. Vegetation indices have been proposed as proper estimators of LAI and TDM through the contrasting reflectances in the red and near infrared regions of the spectrum. The most widespread VI are the NDVI and the SR. Both are related to TDM and LAI, and have been used to indirectly estimate photosynthetic capacity and net primary productivity (Sellers, 1987), as well as crop yield, including that of wheat (Pinter et al., 1981; Aparicio et al., 2000).

Numerous studies have examined the correlation between VI and diverse measures of canopy structure and plant composition, such as TDM, LAI, green area index, chlorophyll concentration, and N concentration of leaves (Tucker, 1979; Asrar et al., 1984; Wiegand et al., 1991; Elliot and Regan, 1993; Bellairs et al., 1996). It is well known that the relationships between these traits and VI are influenced by factors such as plant growth stage or canopy structure (Bellairs et al., 1996). Therefore, comprehensive studies involving contrasting environments, wide genetic backgrounds, and measurements at different plant growth stages are required to verify the usefulness of VI as indicators of LAI, TDM, and yield. There is a considerable body of literature concerning the relationship between VI and growth parameters in bread wheat. However, studies investigating the usefulness of VI to discriminate between durum wheat genotypes for LAI, TDM, and yield differences in Mediterranean environments are lacking. The objective of this study was to determine whether VI assessed across a broad genotypic/environmental range at different growth stages could be useful indirect selection tools to complement empirical selection in a durum wheat breeding program.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Culture
Seven field experiments, each one including one of two sets of 25 and 20 durum wheat genotypes, were carried out at different sites in northeast Spain in 1998 and 1999 (Table 1) . Both sets included commercial cultivars and inbred genotypes of different origins in order to incorporate a wide range of genetic variability (see Table 2 for genotype identification) and had 10 common genotypes.


View this table:
[in this window]
[in a new window]
  		Table 1. Location, site description, and growth stages where measurements were made.

 

View this table:
[in this window]
[in a new window]
  		Table 2. Genotype means (±SE) for leaf area index (LAI), total dry matter (TDM), Normalized Difference Vegetation Index (NDVI), and Simple Ratio (SR), determined at midanthesis across experiments, of the two sets of 25 and 20 genotypes.

 
The sites were chosen to widely represent the soil and climatic conditions prevalent in the region. Each experiment (combination of site and year, according to Table 1) was sown in a randomized complete block design with four replicates, in 10-m-long six-row plots with a 20-cm row spacing. Seeding rates were adjusted for each genotype to achieve a density of 550 viable seeds m-2. Soil analyses were done prior to sowing, and fertilizers were applied as recommended. Weeds and diseases were controlled, when necessary, using appropriate chemicals. Experiments 1 to 4 (see Table 1 for description) were carried out under rainfed conditions, whereas Exp. 5 and 6 were flood-irrigated twice, and Exp. 7 was irrigated three times at monthly intervals during spring. The total amount of water received by the crop ranged from 179 mm in Exp. 1 to 407 mm in Exp. 7 (Table 1).

Data Recorded
Each plot was divided into two 5-m-long subplots. One of them was used for successive destructive biomass sampling, whereas the other remained intact for reflectance measurements and grain yield determinations. Total dry matter sampling and spectral reflectance measurements were made at booting, heading, midanthesis, and medium milk-grain, corresponding to Stages 45, 55, 65, and 75 of the Zadoks scale (Zadoks et al., 1974) respectively. Growth stages described in the remainder of the paper are based on Zadoks scale. Details of the measurements made in each experiment are shown in Table 1.

For growth trait determination, on each sampling date, a 50-cm row length (40–50 plants, 0.10 m-2) from a central row in each plot was randomly selected and the plants harvested, following the methodology described in Aparicio et al. (2000). In the laboratory, the number of plants per sample was recorded and the number of spikes per square meter (NSS) was calculated. Five representative plants per plot were separated into leaves, culms, and spikes. Green area projection (one side) of the different plant parts was measured using a leaf-area meter (AT Dias II, Delta-T Devices, Cambridge) and LAI was computed as the ratio of green leaf area per sample area. Yellow and dry leaves were not considered. The samples were oven-dried at 80°C for 48 h, weighed, and TDM (g m-2) determined. Approximating the area from a trapezium to the curve of LAI against time, following Ramos et al. (1982), leaf area duration (LAD) was calculated and was determined in growing degree-days.

Spectral measurements were made with a portable spectroradiometer fitted with an 18° field-of-view optics (model FieldSpec UV/VNIR, Analytical Spectral Devices, Boulder, CO) as described in Aparicio et al. (2000). This unit detects radiation in 512 contiguous bands, with a 1.4-nm-bandwidth maximum response, from 350 to 1050 nm, covering the visible and near-infrared portion of the spectrum. The sensor was placed on a vertical rod to take readings from a nadir position, with the sensor raised 2 m above the ground. Readings were made at midday on cloud-free days. Three readings (1–2 s each), each being the average of five scans, were made on three different portions of each plot. The reflectance spectrum was calculated in real time as the ratio between the reflected and the incident spectra of the canopy. The incident spectrum was obtained every five plots (every minute approximately), from the light reflected by a white reference panel with a very close to Lambertian surface (Spectralon, Labsphere, North Sutton, NH). The VI SR and NDVI were calculated using narrow-band reflectance values as follows:

where R indicates reflectance and the subindex indicates the wavelength (in nm) (Peñuelas et al., 1993).

Total dry matter was assessed in three replicates on each sampling occasion, whereas reflectance and grain yield were measured in all four replications. Grain yield (kg ha-1) was determined on a plot basis of 12 m2 and is reported at a 10% moisture level.

Statistical Analyses
ANOVA was performed across experiments but independently for each of the two sets, consisting of 20 and 25 genotypes. Type III SS was used in calculations, given the unbalanced nature of the model. All statistical analyses were carried out using standard SAS-STAT procedures (SAS Institute, 1987). Table Curve 2D (Jandel Corporation, 1994) was used to fit the best equation to the relationship between two variables.


   RESULTS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Growth Traits
The analyses of variance shown in Table 3 , determined for the two sets of genotypes measured at different growth stages across several experiments, revealed that the main factors considered in the analyses were significant in most cases for growth traits. The percentage of the sum of squares explained by genotype effect was much lower in all cases than the variability explained by the experiment or growth stage (deduced from Table 3). The seven experiments showed great differences in grain yield. Average yield of Exp. 7 was {approx}15 times higher than that of Exp. 1, with values of 7009 and 486 kg ha-1, respectively. Across experiments, genotype mean yield ranged between 2097 and 7973 kg ha-1 for the set of 25 genotypes, and between 375 and 6169 kg ha-1 for the set of 20 genotypes.


View this table:
[in this window]
[in a new window]
  		Table 3. Mean squares of the analyses of variance for leaf area index (LAI), total dry matter (TDM), number of spikes per square meter (NSS), leaf area duration (LAD), and the vegetation indices normalized difference vegetation index (NDVI) and simple ratio (SR) across experiments and growth stages.

 
Total dry matter tended to be maximum at Stage 75 (Table 4) in all experiments. On the other hand, greatest LAI was reached at Stage 45 in Exp. 3 and 6, and then started to decrease. The lowest NSS (414 spikes m-2) was recorded in Exp. 1, whereas Exp. 6 and 4 had 681 and 701 spikes m-2, respectively. Exp. 3, 5, and 7 had 612, 512, and 614 spikes m-2, respectively. Leaf area duration increased consistently from Exp. 3 (643 growing-degree days) to Exp. 7 (1862 growing-degree days). Exp. 4, 5, and 6 had LAD of 766, 1070, and 1436 growing-degree days, respectively.


View this table:
[in this window]
[in a new window]
  		Table 4. Means (±SE) of leaf area index (LAI), total dry matter (TDM), normalized difference vegetation index (NDVI), and simple ratio (SR) determined at different growth stages for each experiment. See Table 1 for experiment description.

 
Total dry matter and LAI were closely associated. The correlation coefficients between TDM and LAI determined at Stages 65 and 75 across experiments were r = 0.87 and r = 0.78 (P < 0.05), respectively. Genotypic means (Table 2) across experiments for LAI at Stage 65 ranged from 2.7 (‘Altar-Aos’ and Moulchahba-1) to 3.9 (Zeina-1) for the set of 25 genotypes, whereas TDM, in the same stage, ranged between 1088 g m-2 (Chacan) to 1420 g m-2 (Stojocri-3). For the set of 20 genotypes, the LAI ranged from 1.0 (‘Mexa’ and D. Pedro) to 2.0 (Valira) and the TDM between 417 g m-2 for D. Pedro to 781 g m-2 for Anton.

Spectral Vegetation Indices
Spectral VI differed between genotypes, experiments, and growth stages (Table 3). The genotype effect explained a low percentage of the total variability ({approx}3%), whereas the experiment and growth stage explained {approx}40 and 20%, respectively. The mean values of the indices for each experiment and growth stage are summarized in Table 4. Normalized difference vegetation difference and SR tended to be maximum around booting and decreased from this stage to the end of the cycle, coinciding with crop senescence. Normalized difference vegetation index determined at Stage 65 was almost three times higher in the most productive experiment (Exp. 7) than in the less productive one (Exp. 1), and SR determined at Stage 65 in Exp. 7 was >12 times higher than the value obtained in Exp. 1. Normalized difference vegetation index ranged from 0.80 (Altar-Aos and Lahn/Haucan) to 0.86 (‘Awalbit’, ‘Bicrecham-1’, Chacan, and Zeina-2), and SR ranged from 12.4 (‘Sebah’) to 17.8 (Zeina-2) across the experiments conducted with the set of 25 genotypes (Table 2). For the set of 20 genotypes, NDVI at anthesis ranged from 0.39 (Sebah) to 0.45 (Anton), and SR from 4.6 (Borlí and Sebah) to 7.2 (Antón, Bolo, and Lagost-3).

Relationship between Spectral Vegetation Indices and Growth Traits
The predictive value of the spectral VI for the assessment of growth traits was studied at different levels. When the whole data across experiments, growth stages, and genotypes were considered together, the relationships tended to be highly significant (Fig. 1) . However, NDVI and SR were better predictors for LAI than TDM, since the best equations fitted to the relationship between TDM and VI reached coefficients of determination of r2 = 0.49 and r2 = 0.51 (P < 0.05) for NDVI and SR, respectively. In contrast, r2 values between LAI vs. NDVI and LAI vs. SR were 0.75 and 0.81 (P < 0.05), respectively (Fig. 1). The best equations to fit the relationships between growth traits and VI were nonlinear in all cases.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 1. Patterns of the relationships between leaf area index (LAI) and vegetation indices across experiments and growth stages. Vegetation indices studied were the normalized difference vegetation index (NDVI) and the simple ratio (SR). Each point represents the mean value of a genotype in a given experiment and growth stage, from Stage 45 to Stage 75 (n = 400).

 
Significant associations were also found between VI and growth traits at each growth stage from Stage 45 to Stage 75, when data across experiments were considered (Table 5) . However, the predictive value of both NDVI and SR tended to increase at Stages 65 and 75 compared with earlier developmental stages. Both the NSS and LAD could be properly assessed at Stage 75.


View this table:
[in this window]
[in a new window]
  		Table 5. Coefficients of determination of the relationships between growth traits and spectral vegetation indices determined at different growth stages across experiments. LAD, leaf area duration; LAI, leaf area index; NDVI, normalized difference vegetation index; NSS, number of spikes per square meter; SR, simple ratio; TDM, total aboveground dry matter (g m-2).

 
The relationship between VI and growth traits were studied at each experiment and developmental stage by fitting the best equation to the experimental data (Table 6) . All the equations were nonlinear. In this case, the adequacy of VI for the assessment of growth traits varied widely depending on trait, growth stage, and experiment. No significant relationships were found between VI and TDM for any experiment at any stage, with the exception of the relationship between SR and TDM at Stage 65 for Exp. 6 which did not have large predictability (r2 = 0.22, P < 0.05). In contrast, LAI could be properly assessed by NDVI and SR from Stages 55 to 75 in some experiments, but not in others (Table 6). The r2 for the relationship between LAI and VI were significant for Exp. 1, 3, and 5 at Stages 65 and 75, and for Exp. 3 also at Stage 55, whereas the relationships between VI and LAI were not significant at any growth stage in Exp. 4 in which the number of spikes was the highest of all experiments. However, in all cases r2 values were low, indicating that the VI could not explain much of the variation occurring in LAI. Also, NDVI or SR were not suitable for the assessment of LAI at Stage 45. Both NDVI and SR were suitable for assessing LAD when determined at Stages 65 or 75 in Exp. 3, 5, and 7, but not in Exp. 4 and 6 (Table 7) . However, r2 values (even though significant) were all low, indicating that the VI explained little of the variation occurring for LAD.


View this table:
[in this window]
[in a new window]
  		Table 6. Equations that best fitted to the relationship between spectral vegetation indices (independent variable) and leaf area index (dependent variable) in each experiment and growth stage.

 

View this table:
[in this window]
[in a new window]
  		Table 7. Equations that best fitted to the relationship between leaf area duration (dependent variable) and the spectral vegetation indices (independent variable) determined at each experiment at Stages 65 and 75 of Zadoks scale (Zadoks et al., 1974).

 
In order to study the implications of experimental conditions for associations between growth traits and VI, the r2 for the best-fit equations were plotted against the range for LAI. For both NDVI (Fig. 2a) and SR (Fig. 2b) the r2 increased when the variability in the experimental data was high; for example, when the range in LAI values was 4.0 or greater. However, when the LAI range was less, r2 were much less (Fig. 2). The tendency was similar for TDM, since significant r2 appeared when all data or data across experiments within different growth stages were considered, but nonsignificant when data were analyzed within particular experiments and growth stages.



View larger version (15K):
[in this window]
[in a new window]
  		Fig. 2. Relationship between the coefficients of determination of the best-fit equations for the relationships between leaf area index (LAI) and vegetation indices [normalized difference vegetation index (NDVI) and simple ratio (SR)], and the range in LAI between genotypes in each case.

 

   DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
CONCLUSIONS
 REFERENCES
 
Studies on the usefulness of VI for agronomic and plant breeding purposes are frequently carried out on a small number of genotypes, in few environments, and at a specific crop growth stage. This limits the applicability of the results obtained to cereal breeding programs. In this study, the VI were determined in seven experiments, involving five geographical areas, 2 yr, and a large number of genotypes. Growth traits were calculated from direct measurements on harvested plants since studies using allometric methods can overestimate biomass traits (Spanner et al., 1990). Moreover, four of the main stages of crop development, from booting to milk-grain, were used to assess growth traits and VI.

A wide range of variability was observed in the growth traits and the VI determined at different developmental stages. This variability was mainly explained by the environmental conditions associated with the experiment, whereas genotype had a lesser influence on it. Differences between extreme genotypes across experiments in LAI determined at anthesis were on average 1.2 and 1.0 for the sets of 25 and 20 genotypes, respectively, whereas differences in TDM between extreme genotypes were 332 g m-2 and 364 g m-2 for the sets of 25 and 20 genotypes, respectively (Table 2). These results suggest a relatively small genetic variability for growth traits among the durum wheat genotypes involved in this study.

The usefulness of VI to monitor changes in the canopy caused by the growth of the plant has been investigated in durum wheat (Aparicio et al., 2000) and other crops (Gausman et al., 1971; Lukina et al., 2000; Sembiring et al., 2000). Normalized difference vegetation index and SR have been strongly related to TDM and LAI (Tucker, 1979; Peñuelas et al., 1993, 1996, 1997; Gamon et al., 1995; Aparicio et al., 2000). In our study, VI followed changes in canopy structure from Stage 45 to 75. Thus, they tended to be maximum at booting, when the crop reached the highest leaf and green areas, as has been described in other cereals (Royo and Tribó, 1997). The reliability of NDVI and SR to monitor changes in durum wheat growth is also supported by the highest values of both indices in the most productive environments, with more LAI and TDM (Table 4). Thus, in Exp. 1 and 2, those with less canopy coverage, the values of NDVI at milk-grain were around 0.3, which is characteristic of a light yellow-brown canopy or a stressed crop with a reduced capacity to absorb photosynthetically active radiation (Gamon et al., 1995). In contrast, in the most productive environments having greatest TDM and LAI, NDVI reached values close to 0.9, which is characteristic of vigorous wheat canopies with dark green foliage (Fernández et al., 1994). A comparison of the performance of NDVI and SR showed that SR was a more sensitive index compared with NDVI. For example, decreases in LAI and SR between anthesis and milk-grain were close to 40% (Table 4), whereas NDVI declined by <13% in the same comparison. This probably occurred because the SR is not a normalized index.

The relationship between VI and growth traits in our study was generally curvilinear, as also reported by other authors (Asrar et al., 1984; Wiegand et al., 1992). However, a different pattern was found for the relationship of NDVI and SR with LAI. Thus, SR increased consistently with increases in LAI, even being sensitive to high levels of LAI, which agrees with previous reports (Wanjura and Hatfield, 1987). In contrast, NDVI lost its discrimination power for LAI values higher than 3 to 4 (Fig. 1), the same range defined by Curran (1983) for crops such as wheat, corn (Zea mays L.), sorghum [Sorghum bicolor (L.) Moench], and several grasses. It has already been reported that for LAI beyond 3, the addition of more leaf layers to the canopy does not entail great changes in NDVI (Sellers, 1987; Aparicio et al., 2000). Carlson and Ripley (1997) explained that the decrease in the sensitivity of NDVI at high LAI values occurs because the reflectance of solar radiation from the underlying soil surface or lower leaf layers is largely attenuated when the ground surface is completely obscured by the leaves. When canopy cover is achieved (LAI > 3), red reflectance reaches a minimum of 0.03 to 0.04, since most light is absorbed in the upper leaves. However, near infrared reflectance continues increasing even after the canopy is closed, since 40 to 50% of incident infrared radiation is typically transmitted through the upper leaves. This radiation next hits the second layer of leaves and about half is again transmitted and the other half reflected. Of the reflected near infrared radiation, 50% is again absorbed by the upper leaves and will contribute to 12.5% of the near infrared radiation reflected back to the sensor. The sensor will only receive about 3% of incident radiation, due to the third layer of leaves. Thus, the radiation reflected by the fourth layer of leaves reaching the sensor is often within the signal to noise ratio of the sensor. In our study, the vertical asymptote of the curve was reached for NDVI values of {approx}0.9, in agreement with the results obtained by Bellairs et al. (1996) for wheat and barley (Hordeum vulgare L.), but higher than the values reported by Curran (1983) for dense vegetation. In fact, the threshold value of LAI above which NDVI approaches an asymptotic limit would rely on the vegetation type, age, and leaf water content (Paltridge and Barber, 1988).

The analysis of the suitability of spectral VI for the assessment of growth traits was carried out at different levels. The results showed that their predictive value depended on the range of variability existing in the experimental data. Thus, when data of different growth stages and/or experiments were analyzed together, and differences in LAI were 4 or greater, NDVI and SR accurately tracked changes in LAI as indicated by r2 values of 0.57 or greater, and significant in all cases (Fig. 2). However, when the range of variability in LAI was smaller, such as when data within a specific growth stage or experiment were being analyzed independently, the predictive value of VI were much less. One factor contributing to this low predictability was the number of spikes per unit area and the leaf area of the canopy. When the NSS was higher than 650 spikes per m-2 (as in Exp. 4 and 6), the coefficients of determination of the equations fitted to the relationships between LAI and VI decreased greatly, reaching values close to zero in most cases (Table 6). The reflectance of the spikes probably caused some distortion in measurements made in the visible and near infrared ranges (Shibayama et. al., 1986). Our results indicated that NDVI lost its predictive power for LAI at values >3 to 4, whereas SR was a useful indicator of LAI at values above this level, assuming that a wide range of LAI was available. This explains the low r2 values between NDVI and LAI at Stage 45 in Exp. 6 and at Stage 65 in Exp. 7.

Significant relationships between VI and growth traits are reported in the literature when a wide range of variability is present in the experimental data. However, in many studies variability does not come from the genetic background, but is induced by experimental treatments, such as rates of fertilizers (Raun et al., 2001; Flowers et al., 2001), water availability (Fernández et al., 1994; Mogensen et al., 1996; Peñuelas et al., 1994), or varying levels of soil salinity (Peñuelas et al., 1997). The relatively narrow genetic background existing at present in durum breeding programs may be a constraint to the use of VI as tools to complement empirical selection.

The power of NDVI and SR for estimating TDM was lower than their predictive value for LAI. This result should be compared with the report by Serrano et al. (2000), who in a study involving a winter wheat variety did not find any significant correlation between either SR or NDVI with TDM.


   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
REFERENCES
 
Vegetation indices NDVI and SR were shown to be useful tools to track changes in LAI in durum wheat when a wide range of environments, genotypes, and growth stages are considered. Among the two VI, SR showed predictive ability across a wider range of LAI than did NDVI. However, suitability of these VI as predictive tools for LAI within a group of genotypes being compared at a specific growth stage was poor, probably because genetic variability was not large enough to create wide LAI differences. Number of spikes per unit area may also be a confounding factor reducing the predictive ability of NDVI and SR as predictors of LAI. In conclusion, among the genotypes involved in this study, neither vegetation index showed value as an indirect selection criterion for genotypic performance.


   ACKNOWLEDGMENTS
 
This study was partially funded by CICYT, Spain, under projects AGF96-1137-C02-01 and AGF99-0611-C02 and C03; by INIA under project SC97-039-C02-01; and by ITT- Comissionat per Universitats i Recerca (Generalitat de Catalunya). N. Aparicio and D. Villegas are recipients of Ph.D. grants from from Ministerio de Educación y Cultura and the Comissionat per Universitats i Recerca, respectively.

Received for publication July 17, 2001.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 CONCLUSIONS
 REFERENCES
 


Related articles in Crop Science:

This issue in Crop science

Crop Science 2002 42: 1393-1395. [Full Text]  



This article has been cited by other articles:


Home page
 Agron. J.Home page
D. Inman, R. Khosla, R. Reich, and D. G. Westfall
Normalized Difference Vegetation Index and Soil Color-Based Management Zones in Irrigated Maize
Agron. J., January 11, 2008; 100(1): 60 - 66.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
B. Prasad, B. F. Carver, M. L. Stone, M. A. Babar, W. R. Raun, and A. R. Klatt
Potential Use of Spectral Reflectance Indices as a Selection Tool for Grain Yield in Winter Wheat under Great Plains Conditions
Crop Sci., July 30, 2007; 47(4): 1426 - 1440.
[Abstract] [Full Text] [PDF]

Home page
 J. Environ. Qual.Home page
J. A. Poss, W. B. Russell, and C. M. Grieve
Estimating Yields of Salt- and Water-Stressed Forages with Remote Sensing in the Visible and Near Infrared
J. Environ. Qual., May 31, 2006; 35(4): 1060 - 1071.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
M. A. Babar, M. P. Reynolds, M. van Ginkel, A. R. Klatt, W. R. Raun, and M. L. Stone
Spectral Reflectance to Estimate Genetic Variation for In-Season Biomass, Leaf Chlorophyll, and Canopy Temperature in Wheat
Crop Sci., March 27, 2006; 46(3): 1046 - 1057.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
M. A. Babar, M. P. Reynolds, M. van Ginkel, A. R. Klatt, W. R. Raun, and M. L. Stone
Spectral Reflectance Indices as a Potential Indirect Selection Criteria for Wheat Yield under Irrigation
Crop Sci., February 1, 2006; 46(2): 578 - 588.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
H. Tewolde, K. R. Sistani, D. E. Rowe, A. Adeli, and T. Tsegaye
ESTIMATING COTTON LEAF AREA INDEX NONDESTRUCTIVELY WITH A LIGHT SENSOR
Agron. J., June 17, 2005; 97(4): 1158 - 1163.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
A. D. Baez-Gonzalez, J. R. Kiniry, S. J. Maas, M. L. Tiscareno, J. Macias C., J. L. Mendoza, C. W. Richardson, J. Salinas G., and J. R. Manjarrez
Large-Area Maize Yield Forecasting Using Leaf Area Index Based Yield Model
Agron. J., March 1, 2005; 97(2): 418 - 425.
[Abstract] [Full Text] [PDF]

This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Related articles in Crop Science
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via ISI Web of Science (22)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Agricola
Right arrow Articles by Aparicio, N.
Right arrow Articles by Royo, C.
Related Collections
Right arrow Remote Sensing

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome