Improving Nitrogen Use Efficiency in Cereal Grain Production with Optical Sensing and Variable Rate Application (Agron. J. 94:815-820)   PDF version from Agron. J.

Introduction
Consumption of N Fertilizer 
World consumption of fertilizer N was 85,529,551 metric tons in 1999 (FAO, 2001). Of the total fertilizer N consumed, cereal production accounts for 60%, or 51,317,730 metric tons (FAO, 1995). Only 33% of the total N applied for cereal production in the world is actually removed in the grain (Raun and Johnson, 1999), much less than that generally reported (Hardy and Havelka, 1975). In 1999, the unaccounted 67% represented a $15.9 billion annual loss of N fertilizer (Raun and Johnson, 1999). With the increasing costs of N fertilizer due to natural gas shortages, the unaccounted 67% is now estimated to be worth more than $20 billion dollars annually. Considering these poor use efficiencies and the associated costs of improper management, technological advances are needed to reduce excess nutrient applications. 

Low N Use Efficiency 
Nitrogen use efficiency (NUE) in cereal grain production is low for a variety of reasons. Plant N losses as NH3 have accounted for 52 to 73% of labeled N (15N) in corn (Francis et al., 1993), and over 21% in winter wheat (Harper et al., 1987; Daigger et al., 1976). Fertilizer N losses via denitrification have been estimated at 9.5% in winter wheat (Aulakh et al., 1982), 10% in rice (DeDatta et al., 1991), and more than 10% in corn (Hilton et al., 1984). Fertilizer N losses due to surface runoff range between 1% and 13% (Blevins et al., 1996; Chichester and Richardson, 1992). Urea fertilizers applied to the surface without incorporation can result in ammonia volatilization losses in excess of 40% (Fowler and Brydon, 1989; Hargrove et al., 1977). In cooler temperate climates, nitrate losses through tile drainage have approached 26 kg N/ha/yr or 23% of the total N applied (Drury et al., 1996). In general, loss of N only occurs when mineral N (NH4 and NO3) are present in excess of plant needs (Johnson and Raun, 1995). 

Spatial Scale of N Availability 
Conventional N fertilization practices apply a single rate over areas of tens to hundreds of hectares before the crop is planted. Following extensive soil sampling, optical sensor measurements of plants, and geostatistical analyses, we determined that the spatial scale of N availability was at 1m2 and that each square meter needed to be treated independently (Raun et al., 1998; Solie et al., 1999). This contrasts with the 1 hectare grid soil sampling currently promoted in precision agriculture. At a typical cost of $10.00 per sample for soil analyses, soil sampling to manage at the meter level is impractical. 

Response Index (RI) 
Evaluation of grain yield response to N fertilization in 15-yr maize and 30-yr wheat experiments has shown that check plots where no N has been applied exhibit wide variation in the supply of soil N from year to year (Johnson and Raun, 2001). This temporal dependence of N availability reinforces the need for mid-season measurements that account for N supplied through mineralization. Raun et al. (2001) developed an index to predict potential grain yields with no added fertilization (YP0). However, it was necessary to determine the potential yield increase that could be achieved from in-season applications of fertilizer N. This work led to the development of a fertilizer response index (RI) that was calculated by dividing average NDVI from a non-N limiting strip (created in each field by fertilizing a strip at a rate where N would not be limiting throughout the season) by the average NDVI in a parallel strip that is representative of the N availability across the field as affected by N fertilizer applied by the farmer (Johnson and Raun, 2001). Farmer N fertilization practices could result in 0-N availability to non-N limiting conditions, thus, the initial preplant non-N-limiting strip would likely range anywhere from 20 to 100 kg N ha-1. Thus, computing RI will thus require the addition of a non-N limiting strip in each field where NDVI from that strip will be divided by NDVI from any strip in the rest of the field receiving the fixed farmer preplant N rate. This quantitative response index is in turn multiplied by the predicted yield with no added N (YP0) to determine the potential yield with added N fertilization (YPN). 

N Fertilization Optimization Algorithm 
By knowing the quantitative response to fertilizer N achievable for a given area, the N fertilization optimization algorithm (NFOA) was developed to determine the prescribed N rate needed for each 1m2 based on predicted yield (YP0) and the specific response index (RI) for each field. The NFOA accounts for spatially variable potential yield, early season N uptake, and responsiveness of the crop to N input. Stepwise algorithm calculations are as follows: 1. Predict potential grain yield achievable with no additional N fertilization (YP0) from the grain yield-INSEY equation, where INSEY = NDVI (Feekes 4 to 6)/days from planting where GDD>0 [GDD=(Tmin + Tmax)/2 - 4.4°C, where Tmin and Tmax represent daily ambient low and high temperatures]. Lukina et al. (2001) showed that a single equation could be used to predict grain yield over a wide production range (0.5 to 6.0 Mg/ha), diverse sites, and with differing planting and harvest dates. 2. Predict the magnitude of response to N fertilization (in-season-response-index, or RINDVI), RINDVI, computed as: NDVI collected from growing winter wheat anytime from Feekes 4 to Feekes 6 in non-N-limiting fertilized plots divided by NDVI Feekes 4 to Feekes 6 in a parallel strip receiving the farmer preplant N rate. RINDVI has been found to be highly correlated with the response index at harvest, RIHARVEST, which is similarly computed by dividing the grain yield from the non-N-limiting fertilized plots by the yield from plots receiving the farmer preplant N rate (Mullen et al., 2001). The farmer preplant N rate could range anywhere from 0 to a rate applied for non-N limiting conditions. 3. Determine the predicted yield that can be attained with added N (YPN) based both on the in-season response index (RINDVI) and the potential yield achievable with no added N as follows: YPN = YP0 * RINDVI RINDVI was limited so as not to exceed 2.0, and YPN was similarly limited not to exceed the maximum obtainable yield (YPMAX). YPMAX was determined by the farmer, or previously defined as a biological maximum for a specific cereal crop, grown within a specific region, and under defined management practices (e.g., YPMAX for dryland winter wheat produced in central Oklahoma would be 7.0 Mg/ha). RINDVI was capped at 2.0, as in-season applications of N would unlikely lead to greater than 2X yields of wheat for YPN compared to baseline YP0. 4. Predict percent N in the grain (PNG) based on YPN that includes inverse relation to yield level: PNG = b0 + b1*YPN 5. Calculate predicted grain N uptake (GNUP), predicted percent N in the grain multiplied times YPN: GNUP = YPN * PNG 6. Calculate predicted forage N uptake (FNUP) from NDVI: FNUP = b0 +b1e b2NDVI 7. Determine in-season fertilizer N requirement (FNR) FNR = (GNUP - FNUP)/0.70 A divisor of 0.70 is used since the theoretical maximum N use efficiency of an in season N application is approximately 70%. The use of active growing days from planting and NDVI (estimate of total N uptake and/or biomass) in the computation of INSEY allows integration of the effects of both winter and spring growing conditions and date of planting. The INSEY index is essentially the rate of N uptake (kg forage N assimilated/day) by the plant. This approach is consistent with work showing the relationship between above ground plant dry weight and cumulative growing degree days (Rickman et al., 1996). The objective of this work was to determine the validity of the use of INSEY (Lukina et al., 2001; Raun et al., 2001) and NFOA to modulate N mid-season at 1m2 spatial resolution.

Materials and Methods 
Early on in this project, we noted the need to develop a sensing system capable of predicting potential forage N uptake (Stone et al., 1996), and wheat grain yields at meter level spatial resolution. The strategy we investigated relies on remotely sensed spectral reflectance measurements to estimate plant N uptake and eventual yield. These estimates are used to modulate the addition of N fertilizer during early growth stages of the plant (between Feekes 4 and 6(Large, 1954)). Our initial index for in-season estimated yield (INSEY) was computed by dividing the normalized difference vegetation index (NDVI = [(NIRref/NIRinc)-(Redref/Redinc)] / [(NIRref/NIRinc)+(Redref/Redinc)], where NIRref and Redref = magnitude of reflected light, and NIRinc and Redinc = magnitude of the incident light) by the number of days from planting to sensing (Raun et al., 2001). This index was shown to be a reliable mid-season predictor of final grain yield over 16 locations in Oklahoma and 7 in Virginia. Four winter wheat experiments were established in the fall of 2000. Locations and associated soils were: Chickasha, OK, Dale silt loam (fine-silty, mixed, superactive, thermic Pachic Haplustoll); Perkins, OK, Teller sandy loam (fine-loamy, mixed, thermic Udic Argiustolls); Covington, OK, Renfrow silt loam (fine, mixed, thermic Vertic Paleustolls); and Lahoma, OK, Grant silt loam (fine-silty, mixed, thermic Udic Argiustolls). Treatment structure is reported in Table 1. All field experiments used a randomized complete block design with four replications. Plot size was 6 m x 4 m. For treatments 1-5, the entire plot area (24m2) was treated with a uniform N rate. For treatments 6-8, each 1m2 area within the 24m2 main plot was sensed and treated independently. Field plot activities and initial composite soil test levels are reported in Table 2. Collection of spectral reflectance from each 1m2, computation of NDVI, and optical sensors used were consistent with past work (Raun et al., 2001). All NDVI calculations were made with measurements taken using a hand-held multispectral reflectance optical sensor designed and fabricated at Oklahoma State University. The optical sensor simultaneously measured incident and reflected light from the plant at 671±6 and 780±6 nm. NDVI calculations based on reflectance levels have been shown to be an excellent predictor of total plant N uptake (Feekes growth stages 4 to 9). Varietal differences were not targeted because of limited differences in post-dormancy NDVI readings for common varieties grown in this region (Sembiring et al., 2000). Reflectance readings from all experiments were collected between February and April, and ranged from 136 to 153 days after planting. At all locations, winter wheat was optically sensed between Feekes physiological growth stage 4 (leaf sheaths beginning to lengthen) and 6 (first node of stem visible) (Large, 1954). Ammonium nitrate was applied within 7 days of sensing for treatments 2-4, and 6-8 (Table 1). The NFOA was used to determine N rates for each 1m2 for treatments 6, 7, and 8. Wheat was harvested in early June at all locations. Grain subsamples from each plot were ground to pass a 140 mesh screen and total N in grain was analyzed using a Carlo Erba NA-1500 dry combustion analyzer (Schepers et al., 1989).

Results 
Grain yield means, RINDVI and RIHARVEST are reported by treatment and location in Table 3. The standard error of the difference (SED) between two equally replicated means is reported by site. Wheat grain yield levels at the four sites included in this work were close to the state average of 2016 kg ha-1 (30 bu ac-1) over the past 5 years. The four locations (over 1 year) represented different environments where NFOA was being tested, thus, allowing us to test whether or not the concept was unique to a single environment or transcending environments. 

At Lahoma and Perkins, wheat was planted late due to dry fall conditions. At all sites, severe cold was encountered in December and January, thus restricting winter growth. Spring growing conditions were good, characterized by adequate and timely rainfall, limited disease, and no frost damage. 

Large differences in forage N uptake (accurately predicted using NDVI, Lukina et al., 2001) were noted at all sites, and these differences in N uptake produced large disparity in the minimum and maximum N rates applied determined using the NFOA (treatments 6, 7, and 8, Table 4). For treatment 6 (all fertilizer applied mid-season, variable rate), at Covington, the minimum was 32.4 and the maximum 102.8 kg N ha-1. This is a broad range considering that it comes from 96 1m2 sub plots (4 reps, 24m2 plot size). Similarly, a wide range was noted at the other sites, indicative of large spatial variability within relatively small areas. 

At three of the four sites, (exception was Perkins), a significant response in grain yield was observed as a result of applying N (Table 3). The importance of applying preplant fertilizer in order to maximize yields was evident when comparing results from the 45 kg N ha-1 preplant + mid-season-NFOA (treatment 8) to those where all N was applied mid-season (treatments 2 and 3, Table 3). 

Results from the four sites confirmed previous work showing that yield potential could be accurately predicted (Raun et al., 2001). At Chickasha, low yield potential (YP0), and a limited response to N were projected. As a result, NFOA predicted that yields would be maximized at low mid-season N rates, which was in fact observed (Table 3). Yields were maximized for treatment 8 (45 kg N ha-1 preplant + mid-season N, variable applied, 1784 kg ha-1) compared to treatment 4 (45 kg N ha-1 preplant + 45 kg N ha-1 mid-season, 1677 kg ha-1), where an additional 29 kg N ha-1 was applied with no associated yield increase. Similarly, comparing the yields obtained from mid-season-only treatments, it is apparent that treatment 6 (all fertilizer applied mid-season, variable rate average of 19.8 kg N ha-1) was equal in yield to that obtained when either 45 or 90 kg N ha-1 as a fixed rate was applied mid-season (treatments 2 and 3). 

At the Perkins site, the sandy loam soil dries out quickly without timely rain, and lower soil moisture storage becomes more yield limiting than the silt loam soils at the other sites, thus, measured grain yields were lower than predicted. This anomaly has been confirmed by other studies at this site (Raun et al., 2001). In addition, predicted response to applied N from in-season NDVI measurements was overestimated by RINDVI at this site, likely due to limiting moisture at anthesis that restricted response to other adequately supplied growth factors. Because no yield response to N was noted, it was not included in the average estimates of revenue and NUE in Table 3. 

Higher yields and response to mid-season N were predicted and observed at Covington. At this site, a higher N need was calculated (104.3 kg N ha-1, treatment 8) than what would normally be applied mid-season by farmers. It was therefore encouraging to find that this added N resulted in increased grain yield (3269 kg ha-1, treatment 8, versus 2744 kg ha-1, treatment 4). Projecting whether or not a response to applied N could be achieved is critical to this work. Excluding Perkins, the predicted response to applied N using optical sensor measurements (RINDVI) in early spring was positively correlated with grain yield response that could be attributed to applied N in the harvested grain (RIHARVEST). 

For the four sites evaluated, the largest difference in plant growth due to preplant N nutrition was predicted to take place at Lahoma from in-season NDVI measurements and that was confirmed at harvest, two months later (0 N versus 90 N preplant). Wheat growth in treatments 2, 3, 6, and 7 was similar, and notably poor in early April when yield potential was sensed, since none of these treatments received preplant N. The response index predicted the magnitude of an achievable N response, since nearly double the yields were produced from mid-season applied N (RINDVI of 2.22 and an RIHARVEST of 2.19). Having the ability to predict that yields can be doubled if mid-season N is applied is in itself a powerful tool. Furthermore, it is equally important to know how much N to apply to achieve that doubling of yields. At the Lahoma site, 50.9 kg N ha-1 (spatially applied) was needed to produce yields projected with RINDVI, equal to 90 kg N ha-1 applied mid-season (treatment 6 versus treatment 3, Table 3). Applying the NFOA enables the determination of yield increases possible via mid-season application of N and it allows us to estimate how much N is needed to obtain that projected yield. Although applying all of the N preplant (treatment 5) produced maximum yields at this site, this management practice requires that farmers take more risk. Once a good plant stand is secured (dryland wheat production is highly dependent upon rainfall soon after planting), added fertilizer inputs can be tailored to what is made possible by the growing environment. 

Averaged over the 3 sites with N response, when all N was applied mid-season based on NFOA (treatment 6), grain yields were increased (+273 kg ha-1) compared to a similar single rate, using similar fertilizer N rates (43.1 versus 45 kg N ha-1, treatment 2). At $0.10 per kg of wheat grain, this would have a value of $27.30 per hectare. When comparing treatment 6 (all fertilizer applied mid-season, variable rate) to a much higher single N rate of 90 kg N ha-1 applied mid-season (treatment 3), the same amount of grain with the variable rates was produced, but with 46.9 kg less N ha-1. At $0.55 per kg N, the savings in fertilizer N would have a value of $25.79 per hectare. Similar results were noted when ½ of the N rate (22.6 kg N ha-1) predicted using NFOA was applied, producing 1619 kg grain ha-1, contrasted with a grain yield of 1562 kg ha-1 and 45 kg N ha-1 applied at a single rate (treatment 7 versus 2). 

Simple estimates of revenue (averaged over the three sites where significant differences due to treatment were observed) for all treatments are reported in Table 3 (grain revenue minus fertilizer costs). Using the same values for grain and fertilizer previously reported, treatment 8 (45 kg N ha-1 preplant + mid-season N variably applied) increased revenue by more than $9.00 over all other treatments, but required 17.5 kg N ha-1 (45 + 62.5=107.5) more N when compared to an average N rate of 90 (applied preplant, split, or all mid-season). Similar benefits of treatment 6 which used NFOA can be seen over both the 45 and 90 kg N ha-1 mid-season single rates (treatments 2 and 3), increasing revenue by more than $28.00 ha-1 while using less fertilizer N. Treatments 2, 3, and 6 received all N mid-season, the only difference being that treatment 6 received N spatially applied to each 1 m2. In either scenario this increased income will more than cover the increased technology costs, expected to be somewhere between $4.00 and $5.00 ha-1. We expect the greatest economic benefit for this practice to occur under conditions of high and spatially varying N stress. 

Estimates of NUE were determined by subtracting N removed (grain yield times total N) in the grain of 0-N plots from that found in plots receiving added N, divided by the rate of N applied. Averaged over locations, NUE was improved by more than 15% when comparing treatment 2 with treatment 6 where similar rates were applied. All of the treatments that employed NFOA (treatments 6, 7, and 8) resulted in equal or increased NUE when compared to any of the single rate combinations (treatments 2-5). The environmental benefit of this increased NUE cannot be determined, but is considered important.

Discussion 
Placing a biological limit on the maximum obtainable yield (YPMAX) is necessary if a similar NFOA will be applied in other regions, with other crops, soils, and differing management (tillage, irrigation, etc.) practices. For example, maximum yields for hard red winter wheat under dryland production in central Oklahoma will seldom exceed 7.0 Mg ha-1 (103 bu ac-1). Alternatively, winter wheat grain yields under irrigation in western Oklahoma can reach 11 Mg ha-1 (162 bu ac-1). Because the NFOA depends on predicted yield, it is critical that we apply reasonable agronomic limits on what would be a likely result under optimum management. 

The Response Index (RINDVI) accounts for both the likelihood of obtaining a response to in-season applied N, and the magnitude of the response to applied N at a given level of potential yield with no additional fertilizer (YP0). The predicted yield that can be achieved with added N fertilization or YPN = (YP0)*RINDVI, will not generally be more than double YP0. Because it would be unlikely to double yields (YP0) from in-season applied N (YPN), we placed a limit of 2.0 on RINDVI. In this regard, YPMAX is needed to place limits on YPN in those cases where YPN may exceed the biological limits previously documented for specific environments. An exception to the RI limit of 2 would be expected in environments conducive to high N immobilization (e.g., no-till) or small contributions from N mineralization (e.g., irrigated desert soils). 

A prototype of a commercial scale variable N rate applicator that employs the concepts discussed in this paper has been developed (www.ntechindustries.com). Implementation of the NFOA concept requires collection of mid-season NDVI measurements from optical sensors mounted ahead of each fertilization nozzle, and prescribing fertilizer rates computed on-the-go for each 1m2 area. The optical sensor based N fertilizer applicator is equipped with a GPS receiver for post processed geo-referencing of all optical sensor data. For each field, farmers will provide the date of planting in order to compute INSEY (NDVI/days from planting where GDD>0) on-the-go. Growing degree day data is available to growers through various means. Just prior to planting, a non-N-limiting strip (NLS) will be applied in each field. If farmers apply preplant N at a lower rate or if they do not apply fertilizer at all, the NLS will be used to later establish a field specific Response Index (RINDVI). Prior to applying mid-season fertilizer, the non-N-limiting strip will be optically sensed adjacent to the farmer practice in order to determine the field specific RINDVI. 

Improvement in fertilizer N use efficiency beyond the promising results of these experiments may be possible from foliar applications of urea ammonium nitrate solutions (common liquid N fertilizer used for in-season applications) and by variable N rate application. Granular ammonium nitrate fertilizer applied to each 1m2 reported here likely would have decreased NUE, since, unlike foliar applied N it would be subject to surface runoff, microbial immobilization, volatilization, and denitrification prior to being absorbed by plant roots. 

This study demonstrates that crop reflectance measurements using optical sensors can be used to set more efficient and profitable fertilization levels. The techniques that have been developed are appropriately applied at spatial scales of 1 m2 and will require optical sensor-equipped variable rate applicators. The techniques rely on non-N limiting test strips in fields which allow an in-season estimate of fertilizer response. The use of NFOA may eventually replace N fertilization rates determined using production history (yield goals), provided that the production system allows for in-season application of fertilizer N. Fertilizing each 1m2 area based on mid-season estimates of grain yield and the likelihood of achieving a response to added fertilizer could lead to improved NUE in cereal grain crops.

References 

Aulakh, M.S., D.A. Rennie, and E.A. Paul. 1982. Gaseous nitrogen losses from cropped and summer fallowed soils. Can. J. Soil Sci. 62:187-195.

Blevins, D.W., D.H. Wilkison, B.P. Kelly, and S.R. Silva. 1996. Movement of nitrate fertilizer to glacial till and runoff from a claypan soil. J. Environ. Qual. 25:584-593.

Chichester, F.W., and C.W. Richardson. 1992. Sediment and nutrient loss from clay soils as affected by tillage. J. Env. Qual. 21:587-590.

Daigger, L.A., D.H. Sander, and G.A. Peterson. 1976. Nitrogen content of winter wheat during growth and maturation. Agron. J. 68:815-818.

DeDatta, S.K., R.J. Buresh, M.I. Samsom, W.N. Obcemea, and J.G. Real. 1991. Direct measurement of ammonia and denitrification fluxes from urea applied to rice. Soil Sci. Soc. Am. J. 55:543-548.

Drury, C.F., C.S. Tan, J.D. Gaynor, T.O. Oloya, and T.W. Welacky. 1996. Influence of controlled drainage-subirrigation on surface and tile drainage nitrate loss. J. Environ. Qual. 25:317-324.

FAO. 2001. FAOSTAT. www.fao.org.

FAO. 1995. World agriculture: towards 2010 an FAO study. Nikos Alexandratos (ed.). Food and Agriculture Organization of the United Nations and John Wily & Sons, West Sussex, England. p. 190.

Fowler, D.B., and J. Brydon. 1989. No-till winter wheat production on the Canadian prairies: placement of urea and ammonium nitrate fertilizers. Agron. J. 81:518-524.

Francis, D.D., J.S. Schepers, and M.F. Vigil. 1993. Post-anthesis nitrogen loss from corn. Agron. J. 85:659-663.

Hardy, R.W.F., and U.D. Havelka. 1975. Nitrogen fixation research: a key to world food? Science:188:633-643.

Hargrove, W.L., D.E. Kissel, and L.B. Fenn. 1977. Field measurements of ammonia volatilization from surface applications of ammonium salts to a calcareous soil. Agron. J. 69:473-476.

Harper, L.A., R.R. Sharpe, G.W. Langdale, and J.E. Giddens. 1987. Nitrogen cycling in a wheat crop: soil, plant, and aerial nitrogen transport. Agron. J. 79:965-973.

Hilton, B.R., P.E. Fixen, and H.J. Woodward. 1994. Effects of tillage, nitrogen placement, and wheel compaction on denitrification rates in the corn cycle of a corn-oats rotation. J. Plant Nutr. 17:1341-1357.

Johnson, G.V., and W.R. Raun. 1995. Nitrate leaching in continuous winter wheat: use of a soil-plant buffering concept to account for fertilizer nitrogen. J. Prod. Agric. 8:486-491.

Johnson, G.V., and W.R. Raun. 2001. Nitrogen response index as a guide to fertilizer management. Agron. J. (in review)

Large, E.C. 1954. Growth stages in cereals. Illustration of the Feekes Scale. Plant Pathol. 3:128-129.

Lukina, E.V., K.W. Freeman, K.J. Wynn, W.E. Thomason, R.W. Mullen, A.R. Klatt, G.V. Johnson, R.L. Elliott, M.L. Stone, J.B. Solie, and W.R. Raun. 2001. Nitrogen fertilization optimization algorithm based on in-season estimates of yield and plant nitrogen uptake. J. Plant Nutr. 24:885-898.

Mullen, R.W., W.E. Thomason, G.V. Johnson, K.W. Freeman, M.L. Stone, J.B. Solie, and W.R. Raun. 2001. Use of an in-season response index to predict potential yield increases from applied nitrogen. In Agronomy abstracts. ASA, Madison, WI.

Raun, W.R., J.B. Solie, G.V Johnson, M.L. Stone, R.W. Whitney, H.L. Lees, H. Sembiring and S.B. Phillips. 1998. Micro-variability in soil test, plant nutrient, and yield parameters in bermudagrass. Soil Sci. Soc. Am. J. 62:683-690.

Raun, W.R., and G.V. Johnson. 1999. Improving nitrogen use efficiency for cereal production. Agron. J. 91:357-363.

Raun, W.R., G.V. Johnson, M.L. Stone, J.B. Solie, E.V. Lukina, W.E. Thomason and J.S. Schepers. 2001. In-season prediction of potential grain yield in winter wheat using canopy reflectance. Agron. J. 93:131-138.

Rickman, R.W., Sue E. Waldman, and Betty Klepper. 1996. MODWht3: A development-driven wheat growth simulation. Agron J. 88:176-185.

Schepers, J. S., D.D. Francis, and M.T. Thompson. 1989. Simultaneous determination of total C, total N, and 15N on soil and plant material. Commun. in Soil Sci. Plant Anal., 20:949-959.

Sembiring, H., H.L. Lees, W.R. Raun, G.V. Johnson, J.B. Solie, M.L. Stone, M.J. DeLeon, E.V. Lukina, D.A. Cossey, J.M. LaRuffa, C.W. Woolfolk, S.B. Phillips, and W.E. Thomason. 2000. Effect of growth stage and variety on spectral radiance in winter wheat. J. Plant Nutr. 23:141-149.

Solie, J.B., W.R. Raun and M.L. Stone. 1999. Submeter spatial variability of selected soil and bermudagrass production variables. Soil Sci. Soc. Am. J. 63:1724-1733.

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.