R.K. Teal, B. Tubana, K. Girma, K. Freeman, Brian Arnall, O. Walsh and W.R. Raun

ABSTRACT

Drastic increases in the cost of nitrogen (N) fertilizer and increased public scrutiny has encouraged development and implementation of improved N management practices. This study evaluated the relationship between corn grain yield and early season normalized difference vegetation index (NDVI) sensor readings to determine if in-season grain yield potential prediction was achievable. Linear and non-linear regression models were used to determine the relationships present between grain yield and several predictor  variables using Procedures in SAS. Categorizing NDVI measurement by leaf stage indicated that growth stage was critical for predicting grain yield potential. Regression analysis showed poor exponential relationships occurred between NDVI from early sensor measurements (V6 to V7 leaf stage) and grain yield. By the V8 stage a strong relationship was achieved between NDVI and grain yield accounting for 77% of the variation. Later sensor measurements (V9 and later) failed to distinguish variation in green biomass as a result of canopy closure. Normalizing the NDVI with GDD (using GDD INSEY) did not significantly improve yield potential prediction (R² = 0.73), but broadens the yield potential prediction equation to include temperature and allows for adaptation into various climates.  Similar to categorizing sensor measurements by leaf stage, weak exponential relationships occurred between NDVI and grain yield when sensor measurements were categorized by GDD ranging from 500 to 799. Sensor measurements at the range of 800-1000 GDD resulted in a significant exponential relationship between grain yield and NDVI (R² = 0.76) to equal that of the V8 leaf stage. Categorizing NDVI by GDD (800-1000 GDD) extended the sensing time two additional leaf stages (V7 to V9) to allow a practical window of opportunity for sidedress N applications. This study confirms that yield potential in corn could be accurately predicted in-season with NDVI for determination of N need.

INTRODUCTION

Nitrogen (N) is well documented as a limiting nutrient in crop production and considered one of the best inputs a producer can invest in to make a profit under an appropriate management system. In 2003, about 11.5 million metric tons of N fertilizer was applied to 96% of the total corn (Zea mays L.) acreage in the Great Plains (USDA-NASS, 2005). With the present 33% average nitrogen use efficiency (NUE) in world cereal crop production (Raun and Johnson, 1999), over 6.7 million metric tons of N fertilizer was lost to the environment in the Great Plains for the 2003 crop year at a cost of 2.3 billion dollars.  Improved N management is essential to maintain producers income and diminish environmental degradation. 

Traditionally, N application rates have been made based on grain yield goals determined from a recent 5-year crop yield average increased typically by 10 to 30% to assure adequate N for above average growing conditions (Johnson, 1991; Dahnke et al., 1988). Dahnke et al. (1988) defined yield goal as the yield per acre you hope to grow. However, setting unrealistic yield goals and not accounting for yield variation between fields and within a field has led to consistent, excessive N application. As a result, some fields have enough inorganic N in the soil in semi-arid regions to supply adequate N for multiple years of cereal crop production.  Given the fluctuation of growing conditions annually, yield goal may vary from past average yield to potential yield (Dhanke et al., 1988).

Several studies improved the use of yield goal in N decision management by taking into account the soil NO₃^- level (Johnson et al., 1997). Producers are recommended to apply 33 kg N ha^-1 for every 1 Mg of wheat (Johnson et al., 1997) and 20 kg N ha^-1 for every 1 Mg of corn (Schmitt et al., 1998), subtracting the soil NO₃^- level. Further research showed that the percent increase in grain yield goal over the 5-year average should be based on the amount of moisture available to the crop either at planting (Rehm and Schmitt, 1989), or the amount of stored soil water available at depths up to 1.5 m in the spring and the anticipated amount of growing season precipitation in areas where water is limiting (Black and Bauer, 1988). However, the environment is not controlled by a single growth factor but rather a compounded effect of soil fertility, climate and inputs.

Additional research has focused on determining N need in-season by either adjusting yield goal or N availability. Other research has used presidedress NO₃^-N soil tests (PSNT) as an indication of available mineralized N and setting N application limits based on NO₃^-N levels in the soil (Magdoff et al., 1984; Durieux et al., 1995; Spellman et al., 1996). While the PSNT method increased NUE, consistent results were not obtained since sidedress N applications were not adjusted for fluctuating environmental conditions effecting yield goals. In-season N management was developed using chlorophyll meters (SPAD meters) to determine N need based on an N sufficiency index [(as-needed treatment/ well-fertilized treatment) * 100] where N application was recommended when the index value dropped under 95% (Blackmer and Schepers, 1995; Varvel et al., 1997). Indirect chlorophyll content measurement in-season has been successful in determining N need due to the highly correlated relationship between chlorophyll content and leaf N concentration (Wolfe et al., 1988; Schepers et al., 1992). Drawbacks of chlorophyll meter sampling for determining N need are that plant to plant variation can range up to 15% (Peterson et al., 1993), requiring many measurements to obtain a representative average, and where N recommendations are similar in principle to the PSNT method where N recommendations are based on N concentration alone.

Another approach to improving NUE is to adjust yield goal mid-season by determining yield potential. Johnson (1991) defined yield potential of a crop as a function of the growing condition in the field. Raun et al. (2001) established a non-destructive estimation of yield potential using spectral measurements in winter wheat based on the concept developed by Tucker (1979) that normalized difference vegetative index (NDVI) is highly correlated with total above ground biomass. In their work, critical growth stages between Feekes 4 to 6 were identified at which yield potential could be predicted as a result of the strong relationship between NDVI and actual grain yield. The relationship between NDVI and grain yield was further improved when NDVI was normalized by dividing it by the number of days from planting to sensing where the growing degree days (GDD = [(Tmin + Tmax)/2 - 4.4C], (Tmin and Tmax recorded from daily data) were greater than zero (Lukina et al., 2001). Positive GDD are days where growth is possible and in-season estimated yield (INSEY), the insignia given to NDVI normalized by ambient temperature, measures biomass produced per day of positive growth. In six out of nine locations over a two-year period, a strong relationship existed between wheat grain yield and INSEY with a coefficient of determination of 83% (Raun et al., 2001).

In developing an algorithm for topdress N application in wheat, Raun et al. (2002) combined mid-season yield potential prediction and N response using INSEY and a response index (RI). Developed to measure N response in-season with NDVI, RI is calculated by dividing NDVI from a non-limiting N strip by NDVI from a parallel strip that represents N availability across the field (Mullen et al., 2003). When combining both mid-season yield potential and N response, NUE of wheat was increased by more than 15% (Raun et al., 2002). After evaluating continued wheat research, Raun et al. (2005) recognized that while their yield potential prediction model was a good predictor of grain yield potential, the opportunity still existed for the model to over or underestimate yield potential. They further stated that to correctly predict yield potential, models should be fitted to grain yields not influenced by adverse conditions from sensing to harvest. Therefore, Raun et al. (2005) adjusted the constant a within their exponential model (y = ae^bx) so that the number of observations above the curve was 32% of the total data points. Using the adjusted curve (YP₀ + 1 standard deviation) represented the attainable yield potential in rainfed winter wheat from mid-season (February) to harvest based on the six years of data collected over 30 sites at the time of publication.

Success of this technology initiated the development of algorithms for other equally important crops such as corn. This study was conducted to determine the most effective growth stage or range to predict grain yield potential and establish an equation to predict corn yield potential generated from actual yield and early season spectral field measurements.

MATERIALS AND METHODS

Statistical analysis was conducted on experimental data from 21 existing field trials throughout Oklahoma spanning from 2002 to 2005. Soil descriptions for each location are as follow: Eastern Oklahoma Research Station (near Haskell, OK), Taloka silt loam (fine, mixed, thermic Mollic Albaqiustoll); Efaw Research Agronomy Research Farm (Stillwater, OK), Easpur loam (fine-loamy, mixed, superactive, thermic Fluventic Haplustol); Lake Carl Blackwell (LCB) Agronomy Research Farm (near Stillwater, OK), Pulaski fine sandy loam (course-loamy, mixed, nonacid, thermic Typic Ustifluvent); and Perkins, Teller sandy loam (fine-loamy, mixed, thermic Udic Argiustoll).

All field trials employed a randomized complete block design with three replications except for the By-row experiments, which consisted of four random rows, 30 meters in length. The other trials plot size measured 3.0 x 6.1 m and 3.0 x 9.1 m, however all trials were planted in conventional tillage with 76.2 cm row spacing. Each experiment had different specific goals, but in the same course of improving NUE. Plots that received no sidedress N were selected from the different trials for yield potential prediction. Additional experiment information is reported in Table 1.

Using the GreenSeeker Hand Held Sensor (Ntech Industries, Ukiah, CA), NDVI values were collected from growth stages V6 to V11 with the sensor nadir to the ground and approximately 70 cm above the crop canopy. Leaf stage with the corresponding days from planting (DFP) and growing degree days (GDD) are summarized for all sites in Table 2. As an index used to estimate green biomass (Tucker 1979), NDVI was computed as:

Where:  fraction of emitted NIR radiation returned from the sensed area (reflectance)

 fraction of emitted red radiation returned from the sensed area (reflectance)

Grain harvest method varied from trial to trial, the majority of the trials were mechanically harvested with a Massey Ferguson 8XP experimental combine and the rest were harvested by hand (Table 1). The experimental combine method consisted of harvesting the two center rows from each plot using a yield-monitoring computer (Harvest Master) installed on the combine to record grain weight and moisture levels. Corn grain harvested by hand consisted of picking and shucking the two center rows of each plot separately (or single row for the by-row trials) and recording the total ear weight for each row. From each row four random ears were collectively weighed, dried in a forced air oven at 66^oC, and weighed again to determine moisture levels. The four ears were then shelled using a Root-Healey Manufacturing Company (Plymouth, OH) hand-crank corn sheller and the grain weight was taken to determine an average percent grain weight for each row. Finally, grain yield for both harvest methods was determined by adjusting grain weight to 15.5 percent moisture.  

Linear and non-linear regression models were used to determine the relationships present between grain yield and NDVI using Procedures in SAS (SAS, 2002). In addition, an in-season estimated yield (INSEY) equation for yield potential prediction was established similar to Raun et al. (2002). Several indices were evaluated, however only two had a high combined R² when compared to the other indices tested. The days from planting to sensing (DFP) INSEY (Raun et al., 2004, unpublished data) was calculated as:

Where:  days from planting to sensing

In addition, the cumulative growing degree days (GDD) INSEY was calculated as:

            Where:  cumulative growing degree days (GDD) from planting to sensing and calculated using the optimum day method (Barger, 1969)

            Where: bases of 10 C and 30 C minimum and maximum temperatures

The equation resulting from the best line that described the relationship between actual corn grain yield and INSEY (both DFP INSEY and GDD INSEY) was fitted and the equation was used for predicting yield potential for corn. Also, the yield potential + one standard deviation method (Raun et al., 2005) was utilized to develop an accurate measurement of yield potential. 

RESULTS AND DISCUSSION

Growth stage was a major factor in predicting yield potential. Regression analysis showed that weak exponential relationships occurred between NDVI and grain yield when sensor measurements were categorized by leaf stage from 6 to 7 leaves (Table 3), likely due to the yield potential still developing after NDVI measurement. However, a strong relationship between NDVI was achieved at V8 (Figure 1) with a coefficient of determination (R²) value of 0.77. Later sensor measurement (V9 and later) relationships with grain yield were similar to earlier (before V8) comparisons where yield potential was not accurately determined (Table 3). As noted by Teal et al. (2006), the later NDVI readings were unable to distinguish variation, where as yield potential had not yet completely developed in the early measurements.

Varvel et al. (1997) reported that maximum grain yields in corn could not be realized when severe N deficiencies occurred at V8 due to lost yield potential. However, Scharf et al. (2002) found that N applications (0 preplant) could be delayed as late as V11 with no yield loss and only minor yield loss (about 3%) when N sidedress was delayed until V12 to V16. Other recent work showed that when preplant N applications (90 kg N ha^-1) were applied, maximum yield levels could still be obtained when sidedress was delayed until V10, but delaying sidedress N applications further (VT) or not applying preplant N to responsive sites reduced grain yield (Olga Walsh, personal communication, March 2006). The effects of delayed N application on grain yield is highly dependent on available mineralized N and plant need, but this research clearly indicates that predicting yield potential at V8 is highly desirable for maximum effectiveness of sidedress N application. 

As a method of normalizing NDVI measurements over various environmental conditions, the DFP INSEY was used in the initial corn yield prediction equation. The DFP INSEY estimated average biomass produced per day as the determinant of yield prediction. Normalizing the NDVI by DFP GDD generally improved the yield potential prediction model at most leaf stages (Table 3), but not at V8 (Figures 2 and 3). Generally, the GGD INSEY model had higher coefficient of determinations than the DFP INSEY model, although not significant at most leaf stages (Table 3). 

As shown above, sensing time is critical in predicting yield potential. However, trials evaluated in this study showed DFP to the V8 growth stage and the corresponding GDD varied considerably (Table 2). As a result, setting ranges to identify the leaf stage with these values may not be viable. Temperature has been well documented as the primary factor governing the rate of leaf appearance in corn (Berbecel and Eftimescu, 1972; Coelheo and Dale, 1980; Warrington and Kanemasu, 1983). Berbecel and Eftimescu (1972) went further to state that moisture stress reduced the length of the internodes between leaves, but that it had little effect on the number of leaves set. Swan et al. (1987) reported that no-till production requires greater GDD to reach V6 than conventional tillage, indicating that additional variables are present that influence crop growth and the consistency of GDD at a growth stage. Although the NDVI and DFP INSEY yield potential prediction equations were effective, given that GDD is a measurement of temperature and NDVI is a measurement of green biomass, the use of GDD INSEY should normalize NDVI more consistently over various field conditions and climates.

The equation from the GDD INSEY relationship at V8 can be used to predict corn yield potential, however, there is narrow range of sensing time at this particular leaf stage.  According to our results, sensing time range at V8 leaf stage falls within 3 to 8 days before reaching leaf stage V9.  The duration of V8 leaf stage varies depending on the growing condition and there will be risk of failing to measure NDVI within this leaf stage as a result of unavoidable circumstances such as bad weather.  To address this limitation, sensor measurements were categorized by GDD.  This approach used absolute values to create the category before identifying at which GDD range the best relationship between actual yield and GDD INSEY could be drawn.  Further, GDD can be predetermined by DFP from the date of planting would provide a rough indication that sensing time based on GDD is approaching.

Similar to categorizing sensor measurements by leaf stage, regression analysis showed weak exponential relationships between NDVI and grain yield when sensor measurements were categorized by GDD ranging from 500 to 799 (Table 4).  The NDVI measurements that fall between 800 and 1000 GDD and corresponding grain yields had a significant exponential relationship (Figure 4), similar to V8 from the leaf stage category (Figure 3). Note that this GDD range includes data from V7 to V9 growth stage (Table 2). The exponential relationship implies that 76% of the variation in actual grain yield can be explained by NDVI measured and grain yield then can be computed using the equation: grain yield = 0.76e^3.2498(NDVI). Later sensor measurement (> 1000 GDD) relationships with grain yield were similar to the earlier (<800 GDD) comparisons where yield potential was not accurately determined (Table 4). Using GDD INSEY in the yield prediction model did not significantly improve the relationship (R² = 0.75) with grain yield (Figure 5), but gives an additional factor to normalize the data for ambient temperature. Categorizing the sensor measurements by GDD actually improved the relationship between grain yield and GDD INSEY (R² = 0.73 to 0.75) over the leaf stage method, but moreover extended viable yield potential prediction two leaf stages, broadening the critical sensing window to a practical time frame.

CONCLUSIONS

            Corn grain yield potential was accurately predicted with NDVI at the V8 growth stage over four years of data, explaining 77% of the variability. However, categorizing sensor data by GDD, while not improving the accuracy of yield potential prediction with NDVI, extended the critical sensing window two leaf stages. Normalizing NDVI with GDD (using GDD INSEY) did not significantly improve yield potential prediction, but broadens the yield potential prediction equation to include temperature and allows for adaptation into various climates. Exponential equations accurately defining the relationship between GDD INSEY and actual grain yield were established at V8 (R² = 0.73) and for GDD ranging from 800 to 1000 GDD (R² = 0.75). Both equations are capable of approximating corn yield potential at the same level of accuracy except that categorizing sensor data by GDD offers an advantage by extending the critical sensing window two additional leaf stages (V7-V9) and giving an absolute value to determine proper sensing time. 

REFERENCES

Barger, G.L. 1969. Total growing degree days. Weekly Weather Crop Bull. 56 (18, May

5):10.

Berbecel, O. and M. Eftimescu. 1972. Effect of agrometeorological conditions on maize

growth and development. P. 45-50. In Part 1. Meteorology and hydrology no. 2. Institute of Meteorology and Hydrology, Bucharest, Rumania (English translation).

Black, A.L. and A. Bauer. 1988. Setting winter wheat yield goals. P. 24-34. In

J.L. Havlin (ed) Proc. Workshop Central Great Plains Profitable Wheat Management, Wichita, KS. 17-20 Aug. 1988. Potash and Posphate Inst., Altanta, GA.

Blackmer, T.M. and J.S. Schepers. 1995. Use of a chlorophyll meter to monitor

nitrogen status and schedule fertigation for corn. J. Prod. Agric. 8(1):56-60.

Coelho, D.T. and R.F. Dale. 1980. An energy-crop growth variable and temperature

function for predicting corn growth and development: planting and silking. Agron. J. 72:503-510.

Dahnke, W.C., L.J. Swenson, R.J. Goos, and A.G. Leholm. 1988. Choosing a

crop yield goal. SF-822. North Dakota State Ext. Serv., Fargo.

Durieux, R.P., H.J. Brown, E.J. Stewart, J.Q. Zhao, W.E. Jokela, and F.R.

Magdoff. 1995. Implications of nitrogen management strategies for nitrate leaching potential: Roles of nitrogen source and fertilizer recommendations system.  Agron. J. 87:884-887.

Johnson, G.V. 1991. General model for predicting crop response to fertilizer.

Agron. J. 83:367-373.

Johnson, G.V., W.R. Raun, H. Zhang, and J.A. Hattey. 1997. Soil fertility handbook.

Okla. Agric. Exp. Sta. Stillwater, Ok.

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.

Magdoff, F.R., D. Ross, and J. Amadon. 1984. A soil test for nitrogen availability

to corn. Soil Sci. Soc. Am. J. 48:1301-1304.

Mullen, R.W., K.W. Freeman, W.R. Raun, G.V. Johnson, M.L. Stone, and J.B.

Solie.  2003.  Identifying an in-season response index and the potential to increase wheat yield with nitrogen.  Agron. J. 95:347-351.

Peterson, T.A., T.M. Blackmer, D.D. Francis, and J.S. Schepers. 1993. Using a

chlorophyll meter to improve N management. NebGuide G93-1171-A.  Cooperative Extension, Institute of Agriculture and Natural Resources, Univ. of Nebraska-Lincoln.

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. Sollie, 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.

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.

Raun, W.R., J.B. Solie, M.L. Stone, K.L. Martin, K.W. Freeman, R.W. Mullen, H.

Zhang, J.S. Schepers, and G.V. Johnson. 2005. Optical sensor-based algorithm for crop nitrogen fertilization. Commun. Soil Sci. Plant Anal. 36:2759-2781.

Rehm, G. and M. Schmitt. 1989. Setting realistic crop yield goals. Minnesota

Ext. Serv. Ag-FS-3873. Univ. of Minnesota.

SAS Institute. 2002. The SAS system for windows version 8.0. SAS Institute Inc., Cary,

NC.

Scharf, P.C., W.J. Wiebold, and J.A. Lory. 2002. Corn yield response to nitrogen

fertilizer timing and deficiency level. Agron. J. 94:435-441.

Schepers, J.S., D.D. Francis, M. Vigil, and F.E. Below. 1992. Comparison of

corn leaf nitrogen concentration and chlorophyll meter readings. Commun. Soil Sci. Plant Anal. 23(17-20):2173-2187.

Spellman, D.E., A. Rongni, D.G. Westfall, R.M. Waskom, and P.N. Soltanpour.

1996. Pre-sidedress nitrate soil testing to manage nitrogen fertility in irrigated corn in a semi-arid environment. Commun. Soil Sci. Plant Anal. 27:561-574.

Swan, J.B., E.C. Schneider, J.F. Moncrief, W.H. Paulson, and A.E. Peterson. 1987.

Estimating corn growth, yield, and grain moisture from air growing degree days and residue cover. Agron. J. 79:53-60.

Teal, R.K., K. Girma, K.W. Freeman, K.L. Martin, D.B. Arnall, J.W. Lawles, and W.R.

Raun. 2006. Influence of hybrid, population, and nitrogen rate on spectral prediction of corn grain yield. Agron. J. (In review)

Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring

vegetation. Remote Sensing of Environment 8:127-150.

USDA-NASS Agricultural Statistics. 2005. Accessed last January 21, 2006. Available at

http://www.usda.gov/nass/pubs/agr05/acro05.htm

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. Soc. Am. J. 61:1233-1239.

Warrington, I.J. and E.T. Kanemasu. 1983. Corn growth response to temperature and

photoperiod. II. Leaf initiation and leaf appearance rates. Agron. J. 75:755-761.

Wolfe, D.W., D.W. Henderson, T.C. Hsiao, and A. Alvino. 1988. Interactive

water and nitrogen effects on senescence of maize: II. Photosynthetic decline and longevity of individual leaves. Agron. J. 80:865-870.

Figure 5.  Relationship between grain yield and GDD INSEY from sensor measurements ranging 800-1000 GDD over four years and 17 locations, 2002-2005.  Where YP₀ = yield potential; YP₀ calculated = the mean + one standard deviation.