Use of In-Season Reflectance for Predicting Yield Potential in Bermudagrass
J. Mosali1, Kefyalew Girma2, R. K. Teal2, K. W. Freeman2, and W. R. Raun2*
Contribution from the Oklahoma Agricultural Experiment Station and The Samuel Roberts Noble Foundation
*Correspondence: William R. Raun, 044 North Ag Hall, Department of Plant and Soil Sciences, Oklahoma State University, OK 74078; Fax: (405) 774 5269; E-mail: email@example.com
AbstractSpatial variability of soil nutrients is known to exist at distances less than 1 meter. Recently, an on-the-go system for application of N fertilizer based on spectral measurements known as in-season estimated yield (INSEY) improved N use efficiency by as much as 17% in winter wheat. Six trials were conducted in 2001,2002 and 2003 at Ardmore and Burneyville, OK with an objective to develop an index similar to INSEY for use in predicting yield potential in bermudagrass and that can be used for adjusting fertilizer N rates. Initial results indicate that 55% of variation in predicted bermudagrass forage yield was explained by a Bermudagrass –INSEY (B-INSEY) index and where 54% of the variation in forage N uptake was explained using normalized difference vegetative index (NDVI). The remaining challenge is to develop appropriate N fertilizer rates based on this information and apply these rates using on-the-go technology.
Keywords: in-season estimated yield (INSEY), bermudagrass, Normalized difference vegetation index (NDVI), Spectral indices, NUE
During the past few decades, the largest increase in the use of agricultural inputs has been fertilizer N (Johnston, 2000). Because many plant nutrients are non-renewable and depleting rapidly, efficient use of applied fertilizers is important in these times of high production costs and environmental concern. Currently, nitrogen use efficiency (NUE) for worldwide cereal production is estimated to be 33% (Raun and Johnson, 1999) and for forage production, around 45%. The general production practice is to apply most of the N based on a yield goal early in the spring. Johnson (1991) suggested that in order to take advantage of the above average growing conditions in dryland agriculture, it is better to set the yield goal above that of average yields. Yield goal is the “yield per acre you hope to grow” clearly indicating the risk the farmer is taking when he calculates the amount of fertilizer for the crop before production (Dahnke et al., 1988). Usually, fertilizer rates are defined by a specified yield goal, taking into account available soil N (Raun et al., 2001).
Osborne et al. (1999) reported that though yield increased with increasing rates of N fertilizer, N fertilizer recovery levels in bermudagrass were greatest (85%) at N rates less than 224 kg N ha-1, and recovery was less than 20% when 1344 kg N ha-1 was applied. Mathias et al. (1978) reported that bermudagrass yields and N concentration increased while percent recovery decreased with rising N applications up to 448 kg N ha-1.
The presence of spatial variability in agricultural landscapes is an issue demanding careful consideration for efficient use of fertilizers. One approach to increase fertilizer use efficiency is variable rate technology (VRT). Carr et al. (1991) investigated economic efficiency of uniform fertilizer rates for the whole field versus variable rates for dryland wheat in accordance with soil units that had different crop yield potential. They showed positive returns of $53.57- $58.10 kg-1 when optimum treatments for a specific soil were applied rather than uniform rates for the whole field. Although soil units and satellite images distinguish field elements by nutrient availability, their separation is rather poor (coarse scale), which results in low efficiency of variable versus uniform application.
NUE is also complicated by cropland spatial variability that is known to exist at resolutions smaller than 1 m2 (Solie et al., 1996; Raun et al., 1998). Raun et al. (1998) and Solie et al. (1999) reported that variability exists in 0.3 m by 0.3 m bermudagrass plots with regard to the availability of nutrients. The same work reported that variable fertilizer treatment of crops, where each field element is treated separately, can be an effective alternative to the existing uniform fertilizer application practices. Nitrogen fertilizer requirements depend on the potential N uptake by the crop and are related to the overall yield potential. Potential yield is the yield that can be produced on a specific soil under specific weather conditions and that changes with time (Raun et al., 2001).
Cabrera and Kissel (1988) made fertilizer N recommendations based on N mineralized from organic matter. According to Rodriguez and Miller (2000) there was a positive linear relationship between total Kjeldahl nitrogen (TKN) and near infrared reflectance spectroscopy (NIRS). Spectral radiance measurements were evaluated by Sembiring et al. (1998) to identify optimum wavelengths for dual detection of N and P status in bermudagrass (Cynodon dactylon L.) when 0, 112, 224, and 336 kg N ha-1 and 0, 29 or 58 kg P ha-1 were applied in a factorial arrangement of treatments. It was found that biomass, N uptake, P uptake, and N concentration could be predicted using 695/405 nm, with 435 nm as a covariate. Taylor et al. (1998) reported that correlation of forage yield and N removal with red, near infrared (NIR), and normalized difference vegetative index (NDVI) were best with maximum forage production, however, when forage production levels were low, correlation decreased dramatically for the red wavelength compared with NIR and NDVI.
Crawford et al. (1961) reported that the stage of growth, level of N fertilization, plant part, and light intensity all influenced NO3-N concentration, while cultivar, source, time and method of placement had no effect in forages. Kincheloe (1994) reported that field practices should be site specific and the areas within the field to be categorized as best management practices (BMP). He defined BMP’s as those practices that have been tested in research and proven on farmers’ fields as most effective in terms of input efficiency, production potential and environmental protection.
In-season knowledge of potential yield might be the key to successful variable rate fertilizer applications. Raun et al. (2001) demonstrated that the estimated yield (EY) index was a good predictor of grain yield over a wide range of environmental conditions in winter wheat. Raun et al.(2002) later refined this index where only one NDVI reading is taken post dormancy divided by only those days where GDD>0 (including this environmental factor eliminates the days where growth is not possible) from planting to the date of sensing and named it in-season estimate of yield (INSEY) . The same work showed that yield potential based on mid-season estimates increased NUE by 15% when compared to the uniform rates and this was attributed to collecting readings from each 1m2 and fertilizing each 1m2, recognizing that the spatial variability exists at 1m2 resolutions and the potential yield of each 1m2 is different.
The objective of this study is to develop an index similar to INSEY for wheat for use in predicting forage yield potential in bermudagrass which can later be used for adjusting fertilizer N.
MATERIALS AND METHODS
Two field experiments with minimum fertilization located at Burneyville, (Minco silt loam, coarse-silty, mixed, superactive, thermic Udic Haplustolls) and Ardmore, (Wilson silt loam, fine, smectitic, thermic oxyaquic Vertic Haplustalfs) Oklahoma were initiated in April, 2001. These were previously established pastures with “Midland” bermudagrass. The experiments were laid out in a randomized complete block design with eight treatments and three replications. The plots received Urea-N rates of 0, 56, 112, 168 and 224 kg N ha-1 broadcast applied early in the spring at the time of breaking dormancy (last week of March to first week of April).
Plot sizes were 3.04 m x 6.08 m with 3.04 m alleys. Phosphorus and potassium were broadcast applied as per soil test recommendations at both the sites at the initiation of the experiment. During early March of each year, a mix of LoVol 6, Pendimax was used to control weeds. Initial soil test data and dates of activities are reported in Tables 1 and 2, respectively.
Sensor readings were collected for three years at both locations at the time of harvest for each cutting and during in-season growth for most cuttings. In-season readings were collected following at least 10-14 days of active growth (around 3 inches of height). Spectral reflectance measurements during 2001 from the bermudagrass canopy was measured using a handheld sensor that was developed at Oklahoma State University, which included two upward and two downward looking photodiode sensors that collected readings in two bands, red (671± 6 nm) and near infrared (780± 6 nm) bandwidths during 2001(Stone et al., 1996b). The reflectance sensor employed photodiode detectors with inference filters. One pair of filters (up-looking) received incoming light from the sun, and the other pair (down-looking) received light reflected by vegetation and/or soil surface. The instrument used a built-in 16-bit A/D converter that converted the signals from all four photodiode sensors simultaneously. The ratio of readings from down looking to up-looking photodiodes allowed the elimination of fluctuation among readings due to differences in atmospheric conditions, and/or shadows. During 2002 and 2003, sensor readings were taken using a GreenSeeker® Hand Held Optical Sensor (NTech Industries, Inc.) to measure crop reflectance and calculate the NDVI. This sensor is an active sensor (which means it has it’s own self-contained illumination in the both red (650+ 10 nm full width half magnitude) and NIR (770+ 15 nm)) when held approximately at a distance of 60 cm to 100 cm above the crop, it senses an area of 60 cm x 10 cm.
This device measures reflectance which is the fraction of emitted light in the sensed area that is returned to the sensor (Raun et al., 2005). NDVI is calculated based on the following formula
Where: NDVI is as difined above; NIRref and NIRinc are reflected and incident NIR light, respectively ; REDref and REDinc are reflected and incident red light, respectively.
When the bermudagrass was at or near morphological stage of 41 to 49 (anthesis) as defined by West (1990), the forage was harvested. Caution was taken to collect harvest data prior to anthesis since the grass turns a pale color after this stage and there are increased chances to underestimate N uptake thus altering the N content in the grass. Forage was harvested in the center of each plot using a John Deere (GT 262) lawn mower with a cutting width of 96.52 cm which has a forage collection device attached. Forage samples were weighed for fresh weight and sub-sampled for moisture content at the time of harvest. The samples were then dried for 48 hours in a forced air oven at 700C and ground to pass a 0.125 mm (120-mesh) sieve.
The total forage nitrogen content was analyzed using a Carlo-Erba (Milan, Italy) NA-1500 dry combustion analyzer (Schepers et al., 1989). Early-season plant N uptake was determined by multiplying dry matter yield by the total N concentration determined from dry combustion. The difference method (N removed in the check plot subtracted from the N removed in the fertilized plot divided by the amount of fertilizer N applied) was used to determine nitrogen use efficiency.
Data was analyzed using Microsoft Excel and SAS (SAS, 2001). Growing degree days was calculated by subtracting the base temperature from the daily average minimum and maximum temperatures (Table 3). The minimum temperature at which a plant can grow is called the base temperature (Eastin and Sullivan, 1984), which for bermudagrass is 10oc.
Where: NDVI and GDD are as defined above; B-INSEY IS In-season estimate of Bermudagrass forage yield.
At each trial, an N rich strip (N applied at a rate when N would not be limiting through out the growth cycle) was established and 336 kg N ha-1 was applied at the time of breaking of dormancy, followed by 224 kg N ha-1 applied after every harvest until September.
RESULTS AND DISCUSSION
In 2001, NDVI measurements collected at the time of harvest at Burneyville and Ardmore (three and two harvests, respectively) were highly correlated with forage N uptake (Figure 1). This demonstrates that the amount of N present in forage can be predicted using NDVI at the time of sensing, which is consistent with early work by Stone et al. (1996a) who showed that NDVI was highly correlated with wheat forage N uptake. Further, the sensor readings taken 15-20 days after breaking dormancy and after the 1st cutting when the grass was around 7-9 cm high were adjusted for cumulative GDD to determine the B-INSEY index. This index was significantly and positively correlated with forage yield (Figure 2). However, it should be noted that this 2001 database was not robust.
The relationship between NDVI and forage N uptake in 2002 is reported in Figure 3. It should again be noted that these NDVI readings were collected on the same day that harvest data was collected. For this year, the relationship between NDVI and forage N uptake at Ardmore behaved in a different manner due to high weed infestation but still had the same trend only lower yields (data not shown). The first cutting came up very early, so it was not possible to correlate B-INSEY with forage yield using the 1st cutting. B-INSEY was correlated with the second harvest, and 38% of the variation in bermudagrass forage yield was explained (Figure 4).
In 2003, the relationship between NDVI and forge N uptake was strong (R2=0.65) for data over all site-cutting except 1st cutting at Ardmore (Figure 5). Similar to Ardmore in 2002, the first cutting data set behaved a little different than the others, having a lower correlation. Using only data from the first cutting, B-INSEY was highly correlated with forage yield, R2 =0.92 (data not shown).
Combined site years
Over sites and years, these trials demonstrate that spectral reflectance measurements taken mid-season (B-INSEY with forage yield across 7 site years between harvests) coupled with cumulative GDD can be used for predicting the forage yield in bermudagrass (R2=0.55, Figure 6). This tells us that we can predict forage yield potential for each harvest when we sense in-season. Nitrogen forage uptake with NDVI also showed a positive correlation (R2=0.43) across 12 site years (Figure 7). Even when problematic weedy site years (Ardmore) were included, 43% of the variation in forage N uptake was explained (Figure 8). Cumulative growing degree days from dormancy to mid-season and mid-season sensing followed by subsequent harvests provided a reliable estimate for predicting forage yield in bermudagrass after eliminating the problematic 2 site years at Ardmore. Cumulative growing degree days worked in bermudagrass contrary to wheat (Raun et al., 2001) because it is a warm season crop and most of the days are warmer than the temperature growth requirement once it breaks dormancy, and no days are cool enough whereby no growth takes place. Either way, it was difficult to use either the ΣGDD or days where GDD>0 because if moisture became a limiting factor and there is no growth for a long period of time, ΣGDD or days where GDD>0 would not account for the lack of growth when moisture was limiting. Thus moisture availability combined with GDD would likely be ideal. If more in-season sensor readings are available along with rainfall and soil moisture data, the prediction confidence increases using these components. Even without the moisture component and enough readings throughout the growing season, it was exciting to see that most of the variation in forage yield was explained by B-INSEY index.
NDVI was highly correlated with forage N uptake in bermudagrass for most of the harvest dates, excluding the 1st cutting at Ardmore. The B-INSEY (calculated using cumulative GDD’s) index was also highly correlated with final dry matter forage yield when evaluated over locations and years. The problem with this research is determining the correct time to apply fertilizer. The grass should have sufficient growth (at least 2-3 inches of growth) to make accurate recommendations. This research shows potential in managing the temporal variability that occurs from year to year and harvest to harvest variability within a bermudagrass field. It was exciting to find out that prediction of bermudagrass forage yield could be accomplished using a single sensor measurement. Rainfall combined with profile moisture needs to be incorporated into the yield. Also, added work is needed to document the minimum amount of regrowth needed in order to guarantee reliable prediction of yield.
REFERENCES Cabrera, M. L., and D. E. Kissel. 1988. Evaluation of a method to predict nitrogen mineralized from soil organic matter under field conditions. Soil Sci Am J 52:1027-1031. Carr, P. M., G. R. Carlson, J. S. Jacobsen, G. A. Nielsen, and E. O. Skogley. 1991. Farming soils, not fields: a strategy for increasing fertilizer profitability. J Prod Agric 4(1): 57-61. Crawford, R. F., W. K. Kennedy, and W. C. Johnson. 1961. Some factors that affect nitrate accumulation in forages. Agron J 53:159-162. Dahnke, W. C., L. J. Swenson, R. J. Goos, and A. G. Leholm. 1988. Choosing a crop yield goal. North Dakota State Ext. Ser. SF-822. Fargo, North Dakota: North Dakota University. Eastin, J. D., and C. Y. Sullivan. 1984. Environmental stress influences on plant physiology and production. In Physiological basis of crop growth and development, ed. M.B. Tesar, 201-213. Madison, WI: CSSA and ASA. Johnson, G. V. 1991. General model for predicting crop response to fertilizers. Agron J 83:367-373. Johnson, G. V., W. R. Raun, and R. W. Mullen. 2000. Nitrogen use efficiency as influenced by crop response index. p. In Agronomy Abstracts, 291. Madison, WI:ASA. Johnston, A. E. 2000. Efficient use of nutrients in agricultural production systems. Commun Soil Sci Plant Anal 31:1599-1620. Kincheloe, S. 1994. Tools to aid management: The use of site specific management. J Soil Water Conserv 49(2):43-47. Lukina, E. V., K. W. Freeman, K. J. Wynn, W. E. Thomason, R. W. Mullen, M. L. Stone, J. B. Solie, A. R. Klatt, G. V. Johnson, R. L. Elliot 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. Mathias, E. L., O. L. Bennett, and P. E. Lundberg. 1978. Fertilization effects on yield and N concentration of Midland bermudagrass. Agron J 70: 973-976. Osborne S. L., W. R. Raun, G. V. Johnson, J. L. Rogers and W. Altom. 1999. Bermudagrass response to high nitrogen rates, sources, and season of application. Agron J 91:438-444. 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, and W. E. Thomason. 2001. In-season prediction of potential grain yield in winter wheat using canopy reflectance. Agron J 93:131-178. 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, 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., J. B. Solie, K. W. Freeman, M. L. Stone, K. L. Martin, G. V. Johnson, and R. W. Mullen. 2005. Growth Stage, Development, and Spatial Variability in Corn Evaluated Using Optical Sensor Readings. J Plant Nutr 28:173-182. Rodriguez, I. R. and G. L. Miller. 2000. Using Near Infrared Reflectance Spectroscopy to schedule nitrogen applications on dwarf-type bermudagrasses. Agron J 92:423-427. SAS Institute. 2001. SAS/STAT User’s Guide. Release 8.1 ed. SAS Inst., Cary, NC. 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 Soil Sci Plant Anal 20:949-959. Sembiring, H., W. R. Raun, G. V. Johnson, M. L. Stone, J. B. Solie, and S. B. Phillips. 1998. Detection of nitrogen and phosphorus nutrient status in bermudagrass using spectral radiance. J Plant Nutr 21:1189-1206. 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. Solie, J. B., W. R. Raun, R. W. Whitney, M. L. Stone, and J.D. Ringer. 1996. Optical sensor based field element size and sensing strategy for nitrogen application. Trans. ASAE 39 (6):1983-1992. Stone, M. L., J. B. Solie, W. R. Raun, R. W. Whitney, S. L. Taylor, and J. D. Ringer. 1996a. Use of spectral radiance for correcting in-season fertilizer nitrogen deficiencies in winter wheat. Trans. ASAE 39(5):1623-1631. Stone, M. L., J. B. Solie, R. W. Whitney, W. R. Raun, and H. L. Lees. 1996b. Sensors for detection of nitrogen in winter wheat. SAE Technical paper series. SAE Paper No. 961757. Warrendale, PA: SAE. Taylor, S. L., W. R. Raun, J. B. Solie, G. V. Johnson, M. L. Stone, and R. W. Whitney. 1998. Use of spectral radiance for correcting nitrogen deficiencies and estimating soil test variability in an established bermudagrass pasture. J Plant Nutr 21:2287-2302. West, C.P.1990. A proposed growth stage system for bermudagrass. p. 38-42. In Proc. Am. Forage and Grass. Conf., Blacksburg, VA.6-9 June 1990, 38-42. Georgetown, TX: American Forage and Grass Council.
Table 1. Initial surface (0-15 cm) soil chemical characteristics and classification at Ardmore & Burneyville, OK.
NH4-N and NO3-N- 2 M KCl extraction
P and K – Mehlich -III extraction
pH – 1:1 Soil: Water
Table 2. Dates for field activities carried at Ardmore & Burneyville, OK during 2001.
Table 3. Cropping period, GDD, GDD>0 data used at Ardmore
GDD - Cumulative GDD from previous harvest/ breaking dormancy until sensing date
GDD>0-The number of days where GDD>0 until the date of sensing from previous harvest/ breaking dormancy until sensing date
Cropping period- The time between the 2 harvests
Figure 1. Relationship between N uptake and NDVI in bermudagrass forage collected at the time of harvest in 2001 at Burneyville and Ardmore, OK.
Figure 2: Relationship between forage yield and B-INSEY in 2001 at Burneyville and Ardmore, OK.
Figure 3. Relationship between N uptake and NDVI in bermudagrass forage collected at the time of harvest in 2002 at Burneyville and Ardmore, OK.
4. Relationship between forage yield and B-INSEY in bermudagrass forage
collected at the time of harvest in 2002 at Burneyville and Ardmore, OK.
Figure 5. Relationship between N uptake and NDVI in bermudagrass forage collected at the time of harvest without Ardmore 1st cutting in 2003 at Burneyville and Ardmore, OK
Figure 6. Relationship between B-INSEY and forage yield in bermudagrass forage collected at the time of harvest in all site years at Burneyville and Ardmore. OK.
Figure 7. Relationship between N uptake and NDVI in bermudagrass forage collected at the time of harvest in all site years at Burneyville and Ardmore. OK.
Figure 8. Relationship between N uptake and NDVI in bermudagrass forage collected at the time of harvest in all site years (removing the bad sites) at Burneyville and Ardmore. OK.