Nitrogen assessment in rice using hyperspectral reflectance.

Abstract

A completely randomized greenhouse project was performed to determine if hyperspectral reflectance curves could differentiate rice tissue grown on three soil types and with three different commonly used nitrogen fertilizers, hyperspectral reflectance is a remote sensing technology that, in this case, measures the light distribution reflected from rice leaves and allows estimates of the distribution of various pigments. The concentrations of these pigments (chlorophylls, carotinoids, phycobilins, etc) in rice are partially based on soil fertility, especially nitrogen. Thus, hyperspectral reflectance could become a major diagnostic tool to monitor rice production in the field. The results of the greenhouse project demonstrated that (1) soil type differences were important to the overall plant biomass, (2) nitrogen sources promoted biomass production compared to the zero-N control, (3) the three N-sources were equally effective in promoting plant biomass. The hyperspectral reflectance curves were able to discriminate between the zero-N control and the N-sources; however, the hyperspectral reflectance curves were not able to discriminate among the three N-sources. Elemental analysis of the rice tissue demonstrated that the nitrogen concentrations were relatively uniform and possibly deficient. Field trials with rice tissue having higher nitrogen levels may reveal the ability to hyperspectral reflectance curves to discriminate between nitrogen sources.

Key Words: Nitrogen, Rice, Hyperspectral Reflectance

1. Introduction.

Remote sensing is a collection of technologies to detect an attribute of a system without interacting with that system. Hyperspectral reflectance (HR) is a remote sensing technology that measures the light distribution reflected from an object and analyzes the light distribution to infer the composition of the object or some quality parameter derivative from the composition. In plant agriculture, HR has been extensively investigated to infer plant health or nutrient status (Table 1).

HR technologies generally operate within the realm of the visible (380 to 720 nm) and infrared (720 to 106 nm) bandwidths. Most commonly, plant agriculture studies use the blue (380 to 500 nm), green (500 to 600 nm), red (600 to 720 nm) and NIR (720 to 1300 nm) bandwidths; whereas soil studies may use the entire visible and infrared regions. Because the total reflectance is partially dependent on climatic and plant canopy parameters, wavelength ratios provide useful information related to plant pigment contents and nutrient concentrations. Numerous studies have focused on ratios involving the NIR and either the green or red bandwidths. The green and red bandwidths are substantially influenced by chlorophyll and carotinoid pigment concentrations, which are primarily influenced by the leaf nitrogen content (Bronson et al., 2003; Shanahan et al., 2001; Stone et al., 1996). Phosphorus concentrations may also be important towards chlorophyll and carotinoid content expression (Osbourne et al., 2002; Milton et al., 1991). The N1R region is more responsive to the presence of cellular structures created by air--cell wall--protoplasm--chloroplast interfaces (Ma et al., 2001; Bronson et al., 2003).

Inplantagriculture, leafhyperspectralreflectance curves are influenced by crop selection, variety, growth stage, soil water availability, nutrient status, plant pathogens, insect damage, crop protection chemicals, and other environmental constraints (Table 1). Thus, proper experimental protocol requires standardizing conditions to minimize the variance of genetic, cultural and environmental conditions not associated with the main treatment to be investigated.

A typical normalized reflectance curve demonstrates two minima, located near 490 nm and 670 nm, and one maximum, located near 550 nm. The NIR region typically shows a broad, relatively flat, plateau between 750 and 950 nm. Two vegetative indices (GNDVI and GVI) have been proposed using the NIR and green portions of the visible spectrum and two vegetative indices have been proposed using the NIR and red portions (RNDVI and RVI) of the visible spectrum (Blackmer et al., 1996; Bronson et al., 2003). These vegetative indices are calculated as:

(1) GNDVI = [R.sub.820] - [R.sub.550]/[R.sub.820] + [R.sub.550]

(2) GVI = [R.sub.820]/[R.sub.550]

(3) RNDVI = [R.sub.820] - [R.sub.670]/[R.sub.820] + [R.sub.670]

(4) RVI = [R.sub.820]/[R.sub.670],

where R is the relative reflectance.

Normalizing the spectral reflectance curves is a common practice. Normalizing spectral reflectance curves in laboratory situations is usually performed using commercially prepared ceramic plates, whereas field normalization typically involves optimally fertilized plots prepared similarly to the canopy to be surveyed (Blackmer et al., 1996; Bronson et al., 2003; Huete, 1987; Stone et al, 1996; Ruan et al., 2002).

In the United States, rice acreage is less extensive than that of many other crops and the amount of research is correspondingly smaller. HR presents an attractive option to monitor early season nitrogen sufficiency and to predict the timing of midseason nitrogen applications. Nitrogen sources for rice production commonly include urea and ammonium sulfate; however, Japanese rice production relies almost exclusively on ammonium chloride (Tisdale et al., 1985). Thus, HR may become a useful nitrogen monitoring tool provided the influences of soil type, variety type, N-source, N-application options, and other soil fertility parameters are understood.

The purpose of this investigation is to determine if HR may differentiate among a popular rice variety cultured to three different soil types and fertilized with three different nitrogen sources.

2. Materials and Methods

A greenhouse experiment involving rice (Oryza sativa, var. cocodrie) was performed to determine if early-season nitrogen levels in the plant tissue could be correlated with hyperspectral reflectance. A completely randomized block design was developed having three soil types as the main treatment and having sub-treatments consisting of three nitrogen sources. All combinations of the factorial design contained three replications.

Three soils were collected from production fields by excavating the A horizon and were considered soils representative of the (1) Commerce series (Fine-silty, mixed, nonacid, active, thermic Aeric Fluvaquents), (2) Sharkey series (Very-fine, montmorillonitic, nonacid, thermic Vertic Haplaquepts), and the (3) Crowley series (Fine, montmorillonitic, thermic Typic Albaqualfs). Soil samples were air-dried and gently crushed to pass a 2mm screen. Routine soil tests were performed by the University Missouri Soil Testing Laboratory.

Three nitrogen sources were selected: (1) ammonium sulfate (21-0-0), (2) ammonium chloride (25-0-0) and (3) urea (45-0-0). In addition, a zero nitrogen rate served as a control. Air-dried soil was added to plastic lined greenhouse pots to provide 5.3 lbs (2.4 kg) and planted with the variety 'Cocodrie'. For each pot, a nitrogen source was added at a rate equivalent to 120 lbs N Acre-1 (134 kg ha-1). Fertilizer additions were accomplished by dissolving sufficient fertilizer in distilled water; such that, a 0.025 L aliquot added to standing rice (4 inches, 10 cm) provided the indicated nitrogen. Immediately, a standing pond of water (simulated paddy) was imposed and continuously maintained for the duration of the experiment at a depth of approximately 1 inch (2.5 cm). Harvest of plant material, 0.5 cm above the soil surface, occurred approximately 8 weeks after plant emergence. For each pot, the number of plants was counted. Plant tissue analysis was performed by a laboratory specializing in plant tissue analysis.

A Spectron Engineering SE-590 spectroradiometer was operated indoors in a darkened room to obtain spectral measurements. The instrument was placed in a stand and mounted a measured distance above a photolab layout table to give a sensor field of six degrees. Samples were consistently placed in a circular mound (about 0.5 cm thick and 15 cm diameter), flattened to totally fill the viewfield and illuminated from opposite sides with four quartz halogen lamps. The spectrometer was first calibrated using a Spectralon reflectance panel, then two measurements per sample were averaged to generate a hyperspectral reflectance curve. The hyperspectral response curves were ratioed with the white-plate reflectance curve to generate a normalized hyperspectral reflectance curve (NHRC).

NHRCs were prepared by plotting relative amplitude on the ordinate and wavelength on the abscissa. In-addition, the mean vegetative index values (GNDVI, GVI, RNDVI, and RVI) were calculated from the three replications of each experimental unit. Statistical evaluation of plant biomass, elemental concentration and the various hyperspectral features involved analysis of variance, Duncan's multiple range test, linear regression and Pearson correlation (Hoshmand, 1994) using a Microsoft-Excel format.

3. Results And Discussion

a. Plant Biomass and Nutrient Concentrations

Analysis of variance and multiple range testing demonstrated that biomass production (top growth normalized to 100 plants) was significantly different (P = 0.02) between the Sharkey and Crowley soils, with the Crowley soil having the greatest biomass (Table 2). All other soil-biomass comparisons were not significant.

Analysis of variance demonstrated that within soil type differences were apparent (P = 0.0001) and that N-sources were a dominant factor in the production of biomass. The ammonium sulfate and ammonium chloride sources were consistently superior in the production of biomass (Table 2). Urea was less effective in encouraging plant growth in the Commerce and Crowley soil systems and as effective as the other N-sources in the Sharkey soil. The No-Nitrogen (control) treatment produced dramatically less biomass than the fertilizer N-sources. Plant height generally correlated (r2=0.93) with biomass production (Table 2).

Soil type had no significant influence on the nitrogen content of the rice tissue. Within soil type comparisons of the various N-sources also produced insignificant N concentration differences. Nitrogen concentrations for all treatments are appreciably smaller than the 3.9 to 4.2 sufficiently range established by the California Plant Health Association (CPHA, 2002). We suspect that the 2.5 cm level of standing water (typically rice fields maintain 10 to 15 cm standing water) promoted nitrification and subsequent denitrification because of low-level O2 diffusion into the soil surface. We also suggest that the additional growth resulting from the nitrogen fertilizer sources promoted biomass production and effectively "diluted" the nitrogen. The well-known "dilution effect" is discussed in CPHA (2002) and Tisdale et al. (1985).

Phosphorus concentrations are adequate (0.1 to 0.5 % P are commonly accepted phosphorus sufficiency levels; CPHA, 2002) and soil differences are not evident; however, within individual soil types the P tissue concentrations are generally greater in the control units when compared to the N-source units. We also partially attribute the slightly diminished P concentrations associated with the nitrogen source treatments to the "dilution effect". The ammonium sulfate treatment appreciably enhanced the sulfur tissue content, a reasonably expected result.

b. Normalized Hyperspectral Reflectance Curves

The NHRC's typically span from 350 to 1150 nm; however, the various green and red vegetative indices only require spectral values ranging from 500 to 850 nm. The curves are similar in showing a minimum near 490 nm and 670 nm, with a broad plateau from 770 nm to 850 nm. A reflectance maximum generally occurs between the two reflectance minima, near 550 nm (Fig. 1 to 3).

[FIGURES 1-3 OMITTED]

The NHRC's from each experimental unit were used to calculate individual vegetative indices, which were averaged to obtain vegetative indices for each N-source treatment. Analysis of variance demonstrated no soil influence on the vegetative indices (Table 3). Analysis of variance and multiple range testing showed that the GNDVI had smaller reflectance values for the control, with no significant differences consistently demonstrated among the N-sources. Plant biomass and the GNDVI did reveal a reasonable correlation, a feature mostly attributed to the small GNDVI and biomass values for the no-nitrogen control (GNDVI = 0.0015 * plant biomass + 0.435 r2 = 0.75). Nitrogen contents were not significant in influencing the vegetative indices. The total uptake of nitrogen (biomass * %N) also correlated to the GNDVI; however, this relationship may be viewed as a re-expression of the GNDVI-Biomass relationship, given the lack of significance in the nitrogen concentrations. Other vegetative indices were less significant in predicting plant performance.

4. Conclusions

This greenhouse study attempted to determine if NHRC of rice tissue could differentiate the effects of soil type and nitrogen source. NHRC were not able to discriminate any influence of soil type or N-source. The GNDVI was able to discern spectral differences between the N-source and the zero-nitrogen control.

We speculate that the greenhouse environment did not permit the establishment of sufficient water ponding to effectively reduce the nitrification-denitrification processes, leading to marginal nitrogen contents in the tissues. We further speculate that a field trial having a range of nitrogen rates, resulting in a range of N tissue concentrations from 1 to 4% may allow NHRC to discern these N concentration differences.

Table 1. Significant investigations using hyperspectral technologies to
monitor field crop production.

Authors (date)              Crop            Remote Sensing Technology

Bausch et al. 1996          Corn            Spectral Reflectance
Blackmer, Schepers, 1994    Corn            Chlorophyll meter
Blackmer, Schepers, 1996    Corn            Spectral Reflectance
Cassanova et al. 1998       Rice            Spectral Reflectance
Huete et al. 1885           Cotton          Spectral Reflectance
Huete 1987                  Cotton          Spectral Reflectance
Hussain et al. 2000         Rice            Chlorophyll meter
Wood et al., 1992           Cotton          Chlorophyll meter
Li et al. 2001              Cotton          Spectral Reflectance
Ma et al. 2001              Soybean         Spectral Reflectance
Saranga et al. 1998         Cotton          Near-infrared Analysis
Varvel et al. 1997          Corn            Chlorophyll meter
Thomas and Gausman, 1977    Corn, etc       Spectral Reflectance
Shanahan et al. 2001        Corn            Spectral Reflectance
Bronson et al. 2003         Cotton          Spectral Reflectance
Osborne et al. 2002         Corn            Spectral Reflectance
Masoni et al. 1996          Corn, wheat     Spectral Reflectance
Milton et al. 1991          Soybean         Spectral Reflectance
Stone et al. 1996           Winter Wheat    Spectral Reflectance
Maas. 1998                  Cotton          Spectral Reflectance
Raun et al. 2002.           Wheat           Spectral Reflectance
Takebe et al. 1990          Rice            Spectral Reflectance
Tarpley et al. 2000         Cotton          Spectral Reflectance
Wu et al. 1998              Cotton          Spectral Reflectance

Authors (date)              Important Applications

Bausch et al. 1996          GVI for N in predicting growth/yield
Blackmer, Schepers, 1994    Monitor N status
Blackmer, Schepers, 1996    Wavelengths at 550 and 710nm for N
                              deficiencies
Cassanova et al. 1998       Effective for biomass and leaf area index
Huete et al. 1885           Separate soil/plant reflectance
Huete 1987                  Separate soil/plant reflectance
Hussain et al. 2000         Monitor N status
Wood et al., 1992           Monitor N status
Li et al. 2001              Plant growth-soil influences assessed by
                              RNDVI
Ma et al. 2001              NDVI-yield relations, screening genotypes
Saranga et al. 1998         NIRA analysis for leaf nitrogen content
Varvel et al. 1997          Determine N sufficiency levels
Thomas and Gausman, 1977    Leaf chlorophyll / carotenoid content
                              estimation
Shanahan et al. 2001        GNDV I at mid-grain fill correlated yield
Bronson et al. 2003         GNDVI predicted leaf N, RNDVI predicted
                              biomass
Osborne et al. 2002         NIR and blue wavebands estimated phosphorus
                              stress
Masoni et al. 1996          Fe, S, Mg, Mn deficiencies decreased
                              adsorption
Milton et al. 1991          P-deficiency related to higher reflectance
                              in green-yellow
Stone et al. 1996           log (1/NIR reflectance) indicated N,
                              Green--chlorophyll c
Maas. 1998                  Estimate cotton canopy ground cover
Raun et al. 2002.           Variable rate fertilization practices
Takebe et al. 1990          Reflectance ratios to estimate rice canopy
                              N
Tarpley et al. 2000         NDVI correlated with leaf N
Wu et al. 1998              correlated band ratios with N and
                              chlorophyll

Table 2. Plant troductivity and nitrogen, phosphorus, sulfur
composition in a factorial nitrogen design in involving rice.

                                         Plant Tissue

                             weight x 100
Treatment                   [plants.sup.-1]    Height      N

                                 gram            cm        %

Commerce
No-Nitrogen                      12.5a           --      1.82a
Ammonium Sulfate                 55.0b           --      1.76a
Ammonium Chloride                55.0b           --      1.94a
Urea                             25.8c           --      1.79a
Mean Across N treatments       37.1 -,I          --      1.83 I

Sharkey
No-Nitrogen                      11.4a         21.6a     1.89a
Ammonium Sulfate                 33.0b         35.1b     1.66a
Ammonium Chloride                39.7b         34.7b     1.82a
Urea                             27.6b         28.7b     1.69a
Mean Across N treatments       27.9 I, -        30.1     1.77 I

Crowley
No-Nitrogen                      18.4a         25.4a     1.99a
Ammonium Sulfate                 68.0b         42.0b     1.82a
Ammonium Chloride                63.4b         42.3b     2.15a
Urea                             37.4c         35.1b     1.81a
Mean Across N treatments      46.8 II, I        36.2     1.94 I

                                        Plant Tissue

Treatment                     P         S        N / 100 plants

                              %         %             gram

Commerce
No-Nitrogen                 0.45a     0.18a          0.23a
Ammonium Sulfate            0.37b     0.50b          0.97b
Ammonium Chloride           0.37b     0.19a          1.07b
Urea                        0.37b     0.18a          0.46c
Mean Across N treatments    0.39 I    0.26 I        0.68-,I

Sharkey
No-Nitrogen                 0.39a     0.26a          0.22a
Ammonium Sulfate            0.31b     0.32b          0.55b
Ammonium Chloride           0.33b     0.19b          0.72c
Urea                        0.33b     0.22b          0.46b
Mean Across N treatments    0.34 I    0.26 I        0.49 I,-

Crowley
No-Nitrogen                 0.38a     0.17a          0.37a
Ammonium Sulfate            0.29b     0.42b          1.23b
Ammonium Chloride           0.31b     0.17a          1.36b
Urea                        0.36a     0.19a          0.68c
Mean Across N treatments    0.34 I    0.24 I       0.91 II, I

Identical arabic letters within columns for individual soils indicate
N-sources are not significantly different at P = 0.05% probability.
Identical Roman letters within columns indicate that soil differences
are not significantly different at P = 0.05% probability

Table 3. Hyperspectral reflectance ratios for rice cultured under
different N treatments.

Treatment            GNDVI     GVI     RNDVI     RVI

                                 Commerce

No Nitrogen          0.47a    2.77a    0.76a     7.27a
Ammonium Sulfate     0.52b    3.14b    0.78a     8.16b
Ammonium Chloride    0.53b    3.26b    0.79a     8.45b
Urea                 0.50b    3.00b    0.77a     7.54a

                                 Sharkey

No Nitrogen          0.44a    2.55a    0.72a     6.26a
Ammonium Sulfate     0.51a    3.05b    0.76a     7.35b
Ammonium Chloride    0.48a    2.88b    0.73a     6.49a
Urea                 0.46a    2.67a    0.69a     5.40c

                                 Crowley

No Nitrogen          0.45a    2.56a    0.73a     6.49a
Ammonium Sulfate     0.52b    3.20b    0.79b     8.40b
Ammonium Chloride    0.55b    3.48b    0.82b    10.10c
Urea                 0.48a    2.84a    0.73a     6.61a

Identical letters within columns indicate not significantly different
at P = 0.05% probability.

5. Literature Cited

Bausch, W, C., and H.R. Duke. 1996. Remote sensing of plant nitrogen status in corn. Trans. ASAE. 39:1869-1875

Blackmer, T.M., and J.S. Schepers. 1994. Techniques for monitoring crop nitrogen status in corn. Common. Soil Sci. Plant Anal. 25:1791-1800.

Blackmer, T.M., J.S. Schepers, G.E. Varvel, and E.A. Walter-Shea. 1996. Nitrogen deficiency detection capabilities of light reflection from irrigated corn canopies. Agron. J. 88:1-5.

Bronson, K.F., T.T. Chua, J.D. Brooker, J.W. Keeling, and R.J. Lascano. 2003. In-season nitrogen status sensing in irrigated cotton: II. Leaf nitrogen and biomass. Soil Sci. Soc. Am. J. 67:1439-1448.

Cassanova, D., G.F. Epema, and J. Goudriaan. 1998. Monitoring rice reflectance at field level for estimating biomass and LAI. Field Crops Res. 55:83-92.

California Plant Health Association. 2002. Western fertilizer handbook. Interstate Pub., Danville, Il.

Hoshmand, A.R. 1994. Experimental research design and analysis. CRC Press, Boca Raton, FL.

Huete, A.R. 1987. soil-dependent spectral response in developing plant canopy. Agron. J. 79:61-68.

Huete, A.R., R.D. Jackson, and D.F. Post. 1985. Spectral response of a plant canopy with different soil backgrounds. Remote Sens. Environ. 17:37-53.

Hussain, F., K.F. Bronson, Yadvinder-Singh, Bijay-Singh, and S. Peng. 2000. Use of chlorophyll meter sufficiency indices for Nitrogen management in irrigated rice in Asia. Agron. J. 92:875-879.

Li, H., R.J. Lascano, E.M. Barnes, J. Booker, L.T. Wilson, K.F. Bronson, and E. Segrra. 2001. Multispectral reflectance of cotton related to plant growth, soil water and texture, and site elevation. Agron. J. 93:1327-1337.

Ma, B.L., L.M. Dwyer, C. Costa, E.R. Cober, and M.J. Morrison. 2001. Early prediction of soybean yield from canopy reflectance measurements. Agron. J. 93:1227-1234.

Maas, S 1998. Estimating cotton canopy ground cover from remotely sensed scene reflectance. Agron. J. 90:384-388.

Masoni, A., L. Ercoli, and M. Mariotti. 1996. Spectral properties of leaves deficient in iron, sulfur, magnesium and manganese. Agron. J. 88: 937-943.

Milton, N.M., B.A. Eiswerth, and C.M. Ager. 1991. Effect of phosphorus deficiency on spectral reflectance and morphology of soybean plants. Remote Sens. Environ. 36:121-127.

Osborne, S.L., J.S. Schepers, D.D. Francis, and M.R. Schlemmer. 2002. Detection of phosphorus and nitrogen deficiencies in corn using spectral radiance measurements. Agron. J. 94:1215-1221.

Raun, W.R., J.B. Solie, G.V. Johnson, M.L. Stone, R.W. Mullen, K.W.Freeman, W.E. Thomason, and E.V. Lukina. 2002. Improving nitrogen use efficiency in cereal grain production with optical sensing and variable-rate application. Agron. J. 94:815-820.

Saranga, Y., A. Landa, Y. Shekel, A. Bosak, and U. Kafkafi. 1998. Near-infrared analysis of cotton leaves as a guide for nitrogen fertilization. Agron. J. 90:16-21.

Shanahan, J.F., J.S. Schepers, D.D. Francis, G.E. Varvel, W.W. Wilhelm, J.M. Tringe, M.R. Schlemmer, and D.J. Major. 2001. Use of remote sensing imagery to estimate corn grain yield. Agron. J. 93:583-589.

Stone, M.L., J.B. Solie, W.R. Raun, R.W. Whitney, S.L. Taylor, and J.D. Ringer. 1996. Use of spectral radiance for correcting in-season fertilizer nitrogen deficiencies in winter wheat. Trans. ASAE 39:1623-1631.

Takebe, M., T. Yoneyama, K. Inada, and T. Murakami. 1990. Spectral reflectance ratio of rice canopy for estimating crop nitrogen status. Plant Soil 122:295-297.

Tarpley, L., K.R. Reddy, and G.F. Sassenrath-Cole. 2000. Reflectance indices with precision and accuracy in predicting cotton leaf nitrogen concentration. Crop Sci. 40:1814-1819.

Thomas, J.R., and H.W. Gausman. 1977. Leaf reflectance vs leaf chlorophyll and carotenoid concentration for eight crops. Agron. J. 69:799-802.

Tisdale, S.L., W.L. Nelson, and J.D. Beaton. 1985. Soil fertility and fertilizers. Macmillan, NY.

Varvel, G.E., J.S. Schepers, and D.D. Francis. 1997. Ability for in-season correction of nitrogen deficiency in corn using chlorophyll meters. Soil Sci. Am. J. 61:1233-1239.

Wood, C.W., P.W. Tracy, D.W. Reeves, and K.L. Edmisten. 1992. Determination of cotton nitrogen status with a hand-held chlorophyll meter. J. Plant Nutr. 15:1435-1448.

Wu, F., L. Wu, and F. Xu. 1998. Chlorophyll meter to predict nitrogen sidedress requirements for short-season cotton (Gossypium hirsutum L.) Field Crops Res. 56:309-314.

Michael Aide *, Jason Alsup, Nathan Goldschmidt, Braedan Hawkins, Michael Laux, Ryan Layton, Jennifer Murphy, Bree Pearson, Teresa Roy, Bryan Schmid, Jason Wilson. Southeast Missouri State University, Cape Girardeau, MO 63701

* Corresponding Author ([e-mail])