S. M. Moges, W.R. Raun, K.
Girma, K. W. Freeman, H. Zhang, D. B. Arnall,
B. Tubana, R. Teal, S. L.
Holtz, O. Walsh and B. Chung
Department of Plant and
Soil Sciences,
Oklahoma State University,
Stillwater, Oklahoma, USA
ABSTRACT
Sensor based nitrogen (N)
management technology has helped to improve fertilizer recommendations for
various crops. The objective of this study was to estimate the yield
potential of grain sorghum (Sorghum bicolor L. Moench) using a hand
held optical sensor. This experiment was conducted with four levels of N
(50,100,150,200 kg ha-1) and three application timing (Preplant,
topdress and split) arranged in a randomized complete block design in three
replications at three locations, in Oklahoma in 2004 and 2005. Sensor
readings were taken using red (650 ± 10 nm) and green (550 + 12.5 nm)
sensors at growth stages 2, 3, 5, 6 and 7. Results from statistical analysis
have shown that 68 and 71% of the variation in sorghum grain yield was
explained by red and green NDVI, respectively. Similarly, grain N content
was correlated to both green (r2= 0.52) and red (r2=0.49)
NDVI readings at growth stage 3. In-season estimated yield (INSEY) was also
found correlated with final grain yield. The results of this experiment
suggest that INSEY can be used as a tool to predict mid-season sorghum grain
yield potential.
Introduction
According to work by
Johnston (2000), N fertilizer has significantly increased yield in the past
few decades as compared to any other agricultural input. Smith et al (1990)
reported that corn and sorghum yield would have dropped by 41 and 19%,
respectively, without N fertilizer application. Due to economic and
environmental reasons (Feinerman et al., 1990), agricultural inputs have to
be managed efficiently specially at periods of high production. Currently,
nitrogen use efficiency (NUE) of grain production is about 33% and about 45%
in forage production (Raun and Johnson 1999).
A major factor limiting
NUE in traditional N management schemes is routine application of large
doses of N early in the season, before the crop can effectively utilize it.
This stored N fertilizer is at considerable risk to environmental losses as
noted in a review by Raun and Johnson (1999). They pointed out that previous
research has shown NUE could be greatly improved by moving away from early
season application and towards a greater emphasis on mid-season applications
of N fertilizer in amounts that better coincide with crop needs.
Most conventional methods of N fertilizer recommendations were
developed on a state or regional scale, so it is questionable
whether these methods can reasonably be used
for
variable-rate N management that attempts to account
for
within-field spatial and temporal variability (Hergert et al.,
1997). Several research studies have found large differences in
crop yield and crop N response within individual fields (Kitchen
et al., 1995; Vetch et al., 1995), confirming the need
for
reliable methods to generate
site-specific
N recommendations (Hergert et al., 1997). Farmers often use
uniform rates for N fertilization based on expected yield (yield goal) that
could be inconsistent from field-to-field and year-to-year. In most
instances expected yield can be higher or lower than the actual yield
depending on factors that are difficult to predict prior to fertilization.
According to Pierce and Nowak (1999) there are
three basic management approaches currently being tested
for
variable-rate N applications. The first involves determining
plant-available N levels from field grid sampling and
interpreting N rates based on current recommendations (i.e., N balance
equation). The second approach bases N rates on observed crop N
responses using replicated strips with varying N rates across the
landscape. The third approach involves determining crop N status
by monitoring (i.e., light reflectance or chlorophyll content).
Currently, the third approach is the most efficient technology used in
site-specific nutrient management programs as it minimizes time, labor and
cost of fertilizer application.
Numerous researchers (Wade et al., 1994;
Ramsey et al., 1995; Roderick et al., 1996) have utilized normalized
difference vegetative index (NDVI), derived from a very high resolution
radiometer collected from satellite platforms,
to
assess the health and condition of crops and natural vegetation
over large geographical regions. Alternatively, Gitelson et al. (1996)
proposed the
use of
the green normalized difference vegetation index (GNDVI) (where
the green band is substituted for the red band in the NDVI
equation), which may prove
to be
more
useful
for assessing canopy variation in green crop biomass. Shanahan et al.
(2001) showed that GNDVI values derived from images acquired
during mid-grain
filling were highly correlated with
grain
yield.
Estimating in-season yield
potential is a key factor for success towards responsible management of N
from environmental and economic perspective. As has been demonstrated by
Raun et al. (2001), estimated yield (EY) was an excellent predictor of grain
yield under diverse environmental conditions. This index was later modified
as in-season estimate of yield (INSEY) and computed as the ratio of NDVI
reading to number of growing degree days from planting to sensing greater
than zero (GDD>0) (Raun et al., 2002). This work has demonstrated 15%
increase in NUE in winter wheat.
Most grain sorghum
producers usually apply N fertilizer pre-plant without considering the
prevailing soil-plant nutrient status within the growing season. Therefore
the objective of this experiment is to estimate sorghum grain yield
potential using INSEY which subsequently be used for appropriate fertilizer
prescription that accounts for N status of the crop up to the sensing date.
Materials and methods
In the summers of 2004 and
2005, a total of five experiments were conducted in Oklahoma at three
locations: Lake Carl Blackwell (Port silt loam-fine-silty, mixed, thermic
Cumulic Haplustolls), Hennessey (Shellabarger sandy loam fine-loamy, mixed,
thermic Udic Argiustioll)
and Efaw
(Easpur loam fine-loamy, mixed, superactive, thermic Fluventic Haplustoll).
Initial soil analyses results are presented in Table 1. Four N fertilizer
rates (50, 100, 150, and 200 kg ha -1) at three application
timings (100% preplant, 100% topdress and 50/50 split applications) were
arranged in a randomized complete block design with three replications on
plot sizes of 6 x 3 m2.
In 2004, sorghum (DK-44
hybrid) was planted on May 6 at all locations at 75 cm row spacing and at a
population of 111,150 plants ha-1. A blanket rate of 24.4 kg P ha-1
was applied pre-plant and incorporated into the soil. Pre-plant N rates were
applied before planting as urea (46% N). Top dress nitrogen was applied as
UAN (28% N) at sorghum growth stage 3 (growing point differentiation)
between June 14 and 16. In 2005, DK-44 hybrid was planted on May 17 at Efaw
and Lake Carl Blackwell and on May 18 at Hennessey and the same management
practices were carried as in 2004. The two middle rows were used for
physical and sensor measurements and harvested for final grain yield.
A GreenSeeker® Hand Held
Optical Sensor (NTech Industries, Inc.) was used to collect normalized
difference vegetative index (NDVI) measurements. This device uses a
patented technique to measure crop reflectance and to calculate NDVI. The
unit senses a 0.6 x 0.01 m area when held at a distance of approximately 0.6
to 1.0 m from the illuminated surface. The sensed dimensions remain
approximately constant over the height range of the sensor. The sensor unit
has self-contained illumination in the red (650 ± 10 nm), green (550 +
12.5) and NIR (770 ± 15 nm) bands. The device measures the fraction of the
emitted light in the sensed area that is returned to the sensor
(reflectance). It samples at a rate (approximately 1000 measurements second-1)
and averages measurements between outputs. It displays NDVI at a rate of 10
readings second-1. The two middle sorghum rows were sensed by
passing the sensor over each row with the sensor held at a height of
approximately 0.7 m above the crop canopy and oriented so that the 0.6 m
sensed width was perpendicular to the row (nadir) and centered over the row.
Sensor readings in each year were taken at growth stage 2 (collar of 5th
leaf is visible which occurs approximately 20 days after emergence), 3
(growing point differentiation which occur 35 days after emergence), 5 (boot
stage), 6 (half bloom), and 7 (soft dough).
Final plot grain yield was
obtained from two middle rows which were harvested using a Massey Fergusson
8XP plot combine equipped with automatic scale and moisture meter. Total N
in the grain was determined using a Carlo Erba (Milan, Italy) NA-1500 dry
combustion analyzer (Schepers et al., 1989) after grain samples were dried
(70 oC for three days) and ground to pass a 0.125 mm (120-mesh)
sieve.
In-season estimated yield
(INSEY) was calculated as a ratio of average plot NDVI and the number days
from planting to sensing. Relationship among NDVI, INSEY and grain yield
were determined using Proc Reg procedure in SAS (SAS, 1999).
Results and Discussion
Crop Year 2004
Results of simple linear
regression analysis showed that there was a significant relationship between
NDVI and final sorghum grain yield from sensor readings collected at growth
stage 3. When evaluating this relationship over stages of growth, the best
correlation between grain yield and red and green NDVI was obtained at
growth stage 3 with r2 (coefficient of simple determination) of
0.68 and 0.71, respectively (Figures 1 and 2). Similarly, simple regression
analysis of the combined location data revealed that INSEY was significantly
correlated with both red (r2=0.68) and green (r2=0.71)
NDVI (Figures 3 and 4). It should be noted that the use of INSEY did not
significantly improve this relationship over that of NDVI alone for either
red or green. At the same growth stage, total N in the grain was also
significantly correlated with red (r2= 0.49) and green (r2=
0.52) NDVI (Figures 5 and 6).
Crop Year 2005
Results over
sites (Efaw and Hennessey) from simple regression analysis of grain yield on
red NDVI showed only limited correlation for growth stages 2 and 3 (Figures
7 and 8). Similar results were noted for green NDVI at growth stage 3 (data
not presented). Relationship between grain yield and NDVI was weak at growth
stages beyond growth point differentiation. In-season estimated yield where
NDVI values were divided by the number of days from planting to sensing did
not improve this relationship over sites.
Combined years
In both 2004 and 2005 crop
years, it was observed that the relationship of NDVI and grain yield was
better at growth stage 3. This stage is a period of rapid growth and
nutrient uptake by the sorghum plant (Vanderlip, 1993). The relationship
with final grain yield was improved at growth stage 3 when compared to other
stages evaluated. This indicates the importance of collecting early season
sensor measurements for projecting grain yield subsequently enable
adjustment of N needs.
Combined data
over all locations and years appeared to show a trend somewhat similar to
what was found from data on individual location on grain yield and NDVI
relationships at growth stage 3. The scatter plot of grain yield on both
green and red NDVI illustrated a detectible trend for most sites, excluding
the 2005 Lake Carl Blackwell site (Figures 9 and 10). Since this site was
irrigated throughout the growing period, higher grain yields were obtained
compared to the other locations and the response was also noticeably
different. However, it should be noted that when using INSEY, the combined
data were normalized and the outer boundary for detecting yield potential
was quite clear using the green NDVI sensor (Figure 12). Scatter below the
outer boundary is expected since post sensing conditions can lead to the
underestimation of yield potential (drought stress, disease, insect and bird
damage, etc.). But, what is important to consider is that both rainfed and
irrigated sites could be combined on one graph when using INSEY (green NDVI
sensor), further suggesting that early season detection of growth rate
(biomass produced per day, estimated using NDVI divided by the number of
days from planting to sensing) is in fact related to final grain yield.
This same
trend was noted when using the red NDVI at growth stage 3 over sites and
years versus sorghum grain yield (Figure 9). However, for this data, INSEY
failed to normalize all sites, as was noted for INSEY when using green NDVI
(Figure 11). Excluding the Lake Carl Blackwell site, red NDVI and INSEY did
provide reasonable detection of sorghum grain yield potential (outer left
hand boundary of the data). In general the combined location and year data
showed that the INSEY and grain yield relationship could be explained by red
and green NDVI with green having slightly better performance.
CONCLUSIONS
The results obtained from
this experiment suggest that yield potential prediction in sorghum using
spectral measurements should be carried out at a stage of critical biomass
production and nutrient demand. This was shown by the relationship of INSEY
and final grain yield at sorghum growth stage 3. Red and the green NDVI
sensor data collected at growth stage 3 were highly correlated with final
sorghum grain yield. However, the use of INSEY as has been employed in
wheat and corn trials was not as effective in normalizing sites whereby one
yield prediction equation could be established.
Reference
Feinerman, E., E.K. Choi,
and S.R. Johnson. 1990. Uncertainty and split nitrogen
application in crop
production. Amer. J. Agr. Econ. 72:975-984.
Johnston, A. E. 2000. Efficient use of nutrients in agricultural production systems.
Commun Soil Sci Plant Anal 31:1599-1620.
Gitelson, A.A., Y.J. Kaufman, and M.N.
Merzlyak. 1996.
Use
of a green channel in
remote sensing
of global vegetation from EOS-MODIS. Remote Sens. Environ. 58:289–298.
Hergert, G.W., W.L. Pan, D.R. Huggins, J.H.
Grove, and T.R. Peck. 1997. Adequacies of current fertilizer
recommendations
for
site-specific
management. p. 283–300.
In F.J. Pierce and E.J. Sadler (ed.) The state of
site-specific
management
for
agriculture. ASA, CSSA, and SSSA, Madison, WI.
Kitchen, N.R., D.F. Hughes, K.A. Sudduth, and
S.J. Birrell. 1995. Comparison of variable rate to single rate nitrogen
fertilizer application:
Corn
production and residual soil NO3–N. p. 427–439. In P.C.
Robert et al. (ed.)
Site-specific
management
for
agricultural systems. ASA, CSSA, and SSSA, Madison, WI.
Pierce, F.J., and P. Nowak. 1999. Aspects of precision agriculture. Adv.
Agron.
67:1–85.
Ramsey, R.D., A. Falconer, and J.R. Jensen. 1995. The relationship between
NOAA-AVHRR NDVI and ecoregions in Utah. Remote Sens. Environ. 53:188–198.
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.
Roderick, M., R. Smith,
and G. Lodwick. 1996. Calibrating long-term AVHRR-
derived NDVI
imagery.
Remote Sens. Environ. 58:1–12.
SAS Institute Inc. 1998. SAS/STAT User’s guide. 6.03 ed. SAS Institute Inc.
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.
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.
Smith, E.G., R.D. Knutson, C.R. Taylor, J.B. Penson. 1990. Impact of
chemical use reduction on crop yields and costs. Texas A&M Univ., Dep. of
Agric. Economics, Agric. and Food Policy Center, College Station.
Vanderlip, R.L. 1993. How a sorghum plant develops. Cooperative extension
service Contribution No. 1203, Kansas Agricultural Experiment Station,
Manhattan, Kansas.
Vetch, J.A., G.L.
Malzer, P.C. Robert, and D.R. Huggins. 1995. Nitrogen
specific
management by soil condition: Managing fertilizer nitrogen in
corn.
In P.C. Robert et al. (ed.)
Site-specific
management
for
agricultural systems. ASA, CSSA, and SSSA, Madison, WI.
Wade, G., R. Mueller, P. Cook, and P. Doraiswamy. 1994. AVHRR map products
for crop condition assessment: A geographic information systems approach.
Photogramm. Eng. Rem. Sens. 60:1145–1150.