| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Eastern Cereal and Oilseed Res. Cent. (ECORC), Central Exp. Farm, Agric. and Agri-Food Canada, Res. Branch, 960 Carling Ave., Ottawa, ON, K1A 0C6 Canada
b Univ. of Passo Fundo, Passo Fundo, RS, 99001-970, Brazil
* Corresponding author (mab{at}em.agr.ca)
Received for publication January 17, 2001.
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: DOY, day of year • G, green • IR, infrared • LAI, leaf area index • MG, maturity group • NDVI, normalized difference vegetation index • NIR, near infrared • R, red
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Remote-sensing techniques, in particular, multispectral visible and infrared (IR) reflectance, can provide an instantaneous, nondestructive, and quantitative assessment of the crop's ability to intercept radiation and photosynthesize (Ma et al., 1996). The input of reflectance into yield production models has been shown to improve yield estimates (Clevers et al., 1994; Clevers, 1997). Colwell (1956) was the first to use aerial IR photographs to monitor plant disease in the field. The amount of reflectance in the near IR (NIR) range (
= 700–1300 nm) is determined by the optical properties of the leaf tissues: their cellular structure and the air–cell wall–protoplasm–chloroplast interfaces (Kumar and Silva, 1973). These anatomical characteristics are affected in turn by environmental factors such as soil water and/or nutrient status (Gausman et al., 1969; Thomas et al., 1971; Blackmer et al., 1994), soil salinity (Gausman and Cardenas, 1968), and leaf age (Gausman et al., 1970). Reflectance in the visible red (R) range ( = 550–675 nm) has been used to estimate leaf chlorophyll and carotenoid (Benedict and Swidler, 1961; Thomas and Oerther, 1972; Filella et al., 1995) levels and, by extension, the photosynthetic capability of the crop.
The use of NIR or R spectral bands singly does not account for seasonal sun-angle differences and can be affected by atmospheric attenuation in the case of satellite-based (vs. ground-based) measurements. To avoid these problems, a number of indices with reflectance near R and NIR wavelengths have been derived and tested for their ability to accurately predict total wet and/or dry crop biomass, leaf water content, and leaf chlorophyll (Tucker, 1979). Among the best was the simple NIR/R ratio, first used by Rouse et al. (1973), and a weighted difference (NIR - R)/(NIR + R), also termed the normalized difference vegetation index (NDVI).
A number of physical and plant anatomical factors can affect reflectance measurements. When the crop does not cover the entire soil surface, reflectance measured from a certain height above ground level will represent the reflectance of both the canopy and the soil surface, rather than just the crop itself. Colwell (1974) showed that for a 37% soil cover, overall reflectance was close to threefold greater on light-colored soils than on darker soils. The area scanned must be consistently representative of the canopy coverage. Daughtry et al. (1982) showed that the coefficient of variation of reflectance measurements over a soybean crop presenting 71% soil coverage decreased exponentially as sensor height increased. They suggested that the sensor's field of view at the soil level should be several times the row spacing.
A number of studies have related the reflectance of various major field crops to ground cover or leaf area index (LAI) of barley (Hordeum vulgare L.) (Peñuelas et al., 1997), cotton (Gossypium hirsutum L.) (Wiegand and Richardson, 1990), maize (Ma et al., 1996), potato (Solanum tuberosum L.) (Bouman et al., 1992), soybean (Holben et al., 1980), sugarbeet (Beta vulgaris L.) (Clevers, 1997), and wheat (Triticum aestivum L.) (Mahey et al., 1991; Stone et al., 1996). Similar relationships have been developed for leaf chlorophyll concentration (Al-Abbas et al., 1974; Peñuelas et al., 1994; Filella et al., 1995).
Preanthesis NDVI measurements could be used to predict yield or to estimate appropriate midseason fertilizer amendments. While the relationship between reflectance and yield has been extensively studied in wheat and maize, to our knowledge, none of the published studies have included soybean. Therefore, the objectives of this study were to (i) determine whether measured canopy light reflectance can be used to predictively discriminate high from low yield among a large number of historical varieties with known differences in yield potential (Voldeng et al., 1997) and (ii) find the optimum development stage at which yield can be predicted from canopy reflectance measurements. In addition, factors such as soil type and planting density were also examined to determine if they influence canopy reflectance. The overall objective was to determine if canopy reflectance measurements could be used as an accurate, fast, repeatable indicator for screening and ranking soybean genotypes for potential yield.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
A 3-by-42 factorial experiment, arranged in a randomized complete block design with three replications, was used at each site-year. Three population densities (25, 50, and 75 seeds m-2) represented low, optimum, and high seeding rates relative to current recommendations. The final stands were approximately 20, 40, and 60 plants m-2. Of the 42 cultivars representing 58 years (1934–1992) of soybean yield improvement in Canada, 11 were from maturity group (MG) 0, 30 from MG 00, and 1 from MG 000. Detailed genetic background and some agronomic characteristics of these cultivars can be found in Voldeng et al. (1997).
Plot size was 1.6 by 5 m, consisting of four rows spaced 0.4 m apart. Soybean seeds inoculated with Bradyrhizobium japonicum at recommended rates were seeded on 21 and 22 May 1998 and 17 and 20 May 1999. Weeds were controlled by chemical spray. In 1998, a tank mix of Lorox [3-(3,4-dichlorophenyl)-1-methoxy-1-methylurea] at 2 kg ha-1 and Dual II [2-chloro-6'-ethyl-N-(2-methoxy-1-methylethyl)acet-o-toluidide] at 2 L ha-1 was applied at both sites on 27 May, and a second application of Excel Super {(R)-2-[4-(6-chlorobenzoxazol-2-yl-oxy)phenoxy]propionic acid} at 0.67 L ha-1, Basagran Forte [3-isopropyl-1H-2,1,3-benzothiadiazin-4(3H)-one 2,2-dioxide] at 2 L ha-1, and Pinnacle [3-(4-methoxy-6-methyl-1,3,5-triazin-2-yl-carbamoylsulfamoyl)thiophene-2-carboxylic acid] at 8 g ha-1 was used on the clay loam site on 24 June. In 1999, both fields were sprayed with a tank mix of Excel Super at 0.67 L ha-1, Basagran Forte at 2 L ha-1, and Pinnacle at 8 g ha-1 on 10 June and again with Assure II {(R)-2-[4-(6-chloroquinoxalin-2-yl-oxy)phenoxy]propionic acid} at 1.5 L ha-1 on 24 June. Plots were combine-harvested, the seed air-dried, and grain yield reported at 130 g kg-1 moisture.
Reflectance Measurements
Canopy reflectance measurements were made with a hand-held multispectral radiometer (MSR16, CropScan, Rochester, MN), which records incoming radiation and light reflectance from the canopy in eight pass bands (460, 507, 559, 613, 661, 706, 760, and 813 nm). Each band has a half peak band of approximately 5 to 15 nm, depending on the specific pass band. The sensing method used is band-limited optical interference filters and photodiodes. The band-limited optical filters only pass wavelengths of irradiance in the pass-band range to the active surface of the detecting photodiode. The photodiode output current is in direct proportion to the number of photons striking the photodiode. This electrical current was converted to a voltage and amplified by the circuitry in the radiometer. The Data Logger Controller measured and logged these sensor millivolt readings. Data of percent reflectance at each pass band were processed subsequently by a computer program using the calibration and correction constants through a minicomputer connected to the sensor. The sensor head was mounted on an adjustable pole. The sensor receptor, facing the center of the plot (between Rows 2 and 3), was parallel to the ground surface, with a view of 0.8- to 1.0-m diam. At each sampling, duplicate measurements were taken within each plot and averaged. Data collection started when the majority of the MG 00 genotypes were at the R2 stage (Fehr and Caviness, 1977) and was repeated two more times [190, 213, and 229 d of year (DOY) for the sand site and 191, 217, and 231 DOY for the loam site in 1998 and 196, 208, and 228 DOY for the loam site and 197, 209, and 230 DOY for the clay site in 1999]. These dates corresponded to R2, R4, and R5 ±10 d of soybean development stages for all genotypes in the experiment. At each sampling date, reflectance measurements for each site were taken on a sunny day for a total of 6 h. Optimum condition is assumed when canopy reflectance is measured within 2 to 3 h of solar noon. Data collection in the current study may have not occurred under ideal conditions but followed the instrument's recommendations (MSR16, CropScan, Rochester, MN). In all cases, variable weather conditions (clouds and wind) were avoided during data collection. For a typical screening or regional variety performance test, it takes only 1 to 2 h for an experiment with 100 plots or less. Sensor readings from pass bands near 613 and 813 nm were used to derive NDVI (Ma et al., 1996) as follows:
 | [1] |
Data Analysis
Pearson's simple linear correlation coefficients between yield and reflectance at individual wavelength bands or indices were calculated for each site-year and sampling date. Partial correlation coefficients were also obtained using the MANOVA options of SAS (SAS Inst., 1990). Regression was then performed to determine the relationship between reflectance and grain yield when both the simple and partial correlation coefficients for a data set were significant. For the regression model, a linear model (GY = a + bx) was first fit to each data set, and then several curvilinear models were also tested. Of the models fitted, power regression was chosen because it depicted the shape better with larger R2 values than the linear and other curvilinear models. The power function regression model of three parameters was as follows:
 | [2] |
|
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
Choice of Reflectance Indices
Both simple and partial correlation analyses showed that soybean grain yields were negatively correlated (P < 0.01) with canopy reflectance at the 500- to 650-nm wavelengths (r = -0.70 to -0.90) but positively associated with reflectance near 700- to 800-nm wavelengths (r = 0.50–0.80). There was a clear trend for the correlation between yield and reflectance to be greater at the late sampling dates (R4 and R5) than at the early sampling dates (R2). The r values for the correlation between yield and NDVI were among the largest in magnitude and positive. Therefore, it warranted a further quantitative regression analysis.
Our data analyses showed that a power regression of yield as a function of canopy reflectance, expressed as NDVI (GY = a + bNDVIc), was better than linear (GY = a + bNDVI) or exponential (GY = a + beNDVI) functions (data not shown), particularly for the late sampling date (R5). Similarly, Wanjura and Hatfield (1987) reported a larger R2 value for soybean biomass as a power function of NDVI. Table 1 presents the coefficients of determination for a power regression of various reflectance parameters against soybean grain yield for each soil type in both years. Across soil types, the largest R2 values were obtained for IR/R, (IR/R)0.5, and (NDVI + 0.5)0.5, ranging from 0.44 to 0.81, with NDVI showing a greater similarity of R2 values (0.45–0.80) for individual site-year. Although individual R2 values for reflectance near 559 nm [green (G)] and 613 nm (R) were high, the other indices involving G and R reflectance showed much lower R2 (0.13–0.55 for NDVIG; see Table 1). The use of different R and IR wavelengths in the NDVI calculation NDVI2 - NDVI4 (parameters defined in Table 1) did not improve the R2 over that obtained for NDVI. Tucker (1979) found similarly high R2 (0.55–0.92) for IR/R, (IR/R)0.5, NDVI, and (NDVI + 0.5)0.5 for blue grama grass [Bouteloua gracilis (Willd. ex Kunth) Lag. ex Griffiths] biomass and chlorophyll content as well as for IR and IR - R. However, these latter parameters were not normalized and were susceptible to illumination changes. Therefore, NDVI was chosen as the parameter with which to investigate the reflectance–yield relationship in soybean given its common use and the only minor improvement in R2 obtained by the (NDVI + 0.5)0.5 transformation.
|
|
Voldeng et al. (1997) reported that the yearly rate of yield improvement for short-season soybean was significantly greater from 1976 onward than before 1976. Splitting the cultivars based on their year of release (before 1976 vs. 1976 or later) showed no consistent effect on R2 values (data not shown). However, lower yielding cultivars (2250–3100 kg ha-1) showed larger R2 values than higher yielding cultivars (3101–3850 kg ha-1), or all cultivars, for all soil x sampling date comparisons. The better association between yield and canopy reflectance in the low- than high-yielding cultivars was probably related to the fact that older cultivars had greater LAIs than newer ones (Morrison et al., 1999). We hypothesize that the narrow ranges of NDVI for the high-yielding varieties associated with high leaf area densities made it difficult to differentiate among them. Similarly, if R2 values were compared on the basis of the actual measured yield within each year, conditions (i.e., soil type) leading to lower yields showed larger R2 values than conditions leading to higher yields. For example, mean yields for all cultivars for the sandy and loam soils in 1998 were 2060 and 2560 kg ha-1, respectively, while R2 values were 0.80 and 0.65. In 1999, yields on the clay and loam soils were 2990 and 3060 kg ha-1, respectively, while R2 values were 0.70 and 0.45 (Table 1). This relationship held for all of the reflectance parameters presented in Table 1, except G/R, which showed an opposite trend. This was probably due to the fact that reflectance at G and R is negatively corrected to leaf color (greenness), with no relation to leaf area density (Ma et al., 1996). Table 2 (as well as Tables 3 and 4, discussed below) showed that the R2 values were consistently greater for the lower yielding soil each year. These data indicate that measurements of canopy reflectance discriminated high from low potential yields of soybean varieties and that it is possible to rank genotypes based on their potential yields under specific conditions.
|
|
Influence of Sampling Dates
The regression equation parameters and R2 values were generated for all soil x sampling date combinations for both 1998 and 1999 (Table 3). Soybean varieties in this short growing-season region are indeterminate, and yield potential is set at the later stages. As a result, R2 values for the yield–NDVI relationship were increased with later sampling dates on almost all occasions. Mahey et al. (1991) demonstrated that sampling dates had a significant impact on the yield–NDVI relationship in wheat, with the correlation coefficients between NDVI and yield becoming larger after flowering than at booting. However, the relationship broke down during senescence. Stone et al. (1996), also working with wheat, found no such differences in the R2 for 1/NDVI vs. yield among Feekes Development Stages 4 through 6. The improvement in R2 values between yield and NDVI in this study was similar to that observed for maize by Ma et al. (1996) where measurements were taken only within a relatively short period before, during, and after flowering but not extending to near maturity like those of Mahey et al. (1991).
In general, the yield–NDVI relationship was linear (data not shown) at the first two sampling dates (R2 to R4), but fairly clearly a power relationship was required to describe the relationship at the third sampling date (R5) (Fig. 2 and 3) . Changes in equation parameters across year and site indicate that caution must be taken when estimating grain yield under specific conditions. The degree of scattering, especially in the loam site of 1999 (Fig. 3B), was probably due to several factors, namely (i) larger variability in radiation during the day of reflectance measurement associated with large number of plots involved, (ii) weeds in some of the plots, and (iii) different potential yields among genotypes with similar canopy structure. It is relatively easy to alleviate and/or avoid problems no. i and ii because yield–NDVI relationships are applicable for a variety performance test, which usually contains 100 plots or less. Measurement of canopy reflectance would take only 1 to 2 h, with minimum variability in radiation during the measurement, and it is also easier to keep the field weed free. For concern no. iii, as Morrison et al. (1999) noticed, LAI has significantly decreased with year of cultivar release for the last 58 yr and that some of the current soybean cultivars had similar grain yields but with distinct canopy structure (ranges of LAI). The NDVI is strongly related to leaf greenness and LAI or aboveground biomass (Ma et al., 1996). Thus, predicting soybean yield based on the yield–NDVI relationship should take the genotypic canopy structure into account. Fortunately, however, there are generally much smaller ranges in a performance test than our historical variety test.
|
|
|
CONCLUSIONS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Kumudini, J. Omielan, and D. J. Hume Soybean Genetic Improvement in Yield and the Effect of Late-Season Shading and Nitrogen Source and Supply Agron. J., February 29, 2008; 100(2): 400 - 405. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
 |
|
 B. Prasad, B. F. Carver, M. L. Stone, M. A. Babar, W. R. Raun, and A. R. Klatt Genetic Analysis of Indirect Selection for Winter Wheat Grain Yield Using Spectral Reflectance Indices Crop Sci., July 30, 2007; 47(4): 1416 - 1425. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
|
|
J. E. Board, V. Maka, R. Price, D. Knight, and M. E. Baur Development of Vegetation Indices for Identifying Insect Infestations in Soybean Agron. J., April 4, 2007; 99(3): 650 - 656. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. L. Ma, K. D. Subedi, and T. Q. Zhang Pre-Sidedress Nitrate Test and Other Crop-Based Indicators for Fresh Market and Processing Sweet Corn Agron. J., January 1, 2007; 99(1): 174 - 183. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
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]
|
|
|
|
|
|
J. E. Board and H. Modali Dry Matter Accumulation Predictors for Optimal Yield in Soybean Crop Sci., August 1, 2005; 45(5): 1790 - 1799. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. B. Tarr, K. J. Moore, and P. M. Dixon Spectral Reflectance as a Covariate for Estimating Pasture Productivity and Composition Crop Sci., May 6, 2005; 45(3): 996 - 1003. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. F. Bronson, J. D. Booker, J. W. Keeling, R. K. Boman, T. A. Wheeler, R. J. Lascano, and R. L. Nichols Cotton Canopy Reflectance at Landscape Scale as Affected by Nitrogen Fertilization Agron. J., April 27, 2005; 97(3): 654 - 660. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. L. Ma, K. D. Subedi, and C. Costa Comparison of Crop-Based Indicators with Soil Nitrate Test for Corn Nitrogen Requirement Agron. J., March 1, 2005; 97(2): 462 - 471. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. T. Chua, K. F. Bronson, J. D. Booker, J. W. Keeling, A. R. Mosier, J. P. Bordovsky, R. J. Lascano, C. J. Green, and E. Segarra In-Season Nitrogen Status Sensing in Irrigated Cotton: I. Yields and Nitrogen-15 Recovery Soil Sci. Soc. Am. J., September 1, 2003; 67(5): 1428 - 1438. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. F. Bronson, T. T. Chua, J. D. Booker, J. W. Keeling, and R. J. Lascano In-Season Nitrogen Status Sensing in Irrigated Cotton: II. Leaf Nitrogen and Biomass Soil Sci. Soc. Am. J., September 1, 2003; 67(5): 1439 - 1448. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| The SCI Journals | Crop Science | 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 | |||
