fields in central Iowa.
22000
20000
168 kg ha-1
Yield = 3022 + 27.9 Water Use r2= 0.63
18000
116 kg ha-1
232 kg ha-1
)
16000
58 kg ha-1
-1
14000
kg ha
d (
12000
ielY 10000
8000
6000
4000
200
250
300
350
400
450
500
550
Water Use (mm)
Fig. 3. Water use efficiency for corn for multiple years with different nitrogen application rates based on observation of seasonal totals of transpiration and grain yield.
3.2 Field scale response to nitrogen
Corn yield response to N applications varied across fields and years (Fig. 4). These data were separated by field and year to remove the potential confounding effects of weather
variation among years. Analysis of the relationship between N rate and yield revealed
that most fields showed no response to N (Table 3). The Sac Field 1 showed a negative
response to N rate while the other fields showed a positive response to N (Table 3). The
different response to N poses a problem in developing a general set of guidelines for N
management and raises questions about why these fields differ in their yield response to
N rate. However, this observation provides an opportunity for application of precision
agricultural tools if the underlying mechanism for the difference can be identified and is consistent among growing seasons to allow for application of this information for
decision-making.
Spatial Patterns of Water and Nitrogen Response Within Corn Production Fields
83
12000
Corn Strip Trials 2000
11000
) 10000
-1
9000
kg ha
ld (
8000
Calhoun Field 1
ie
Dallas Field 1
Y
Sac Field 1
7000
Shelby Field 1
Story Field 1
6000
Story Field 2
5000
0
50
100
150
200
250
12000
Corn Strip Ni
Trial tr
s ogen Rate (kg h
2001
a-1)
11000
) 10000
-1 a 9000
kg h
d (
8000
ielY 7000
Story Field 1
Story Field 2
6000
Story Field 3
5000
0
50
100
150
200
250
14000
Nitrogen Rate (kg ha-1)
Corn Strip Trials 2002
12000
)-1
ha 10000
(kg dielY 8000
Calhoun Field 1
6000
Dallas Field 1
Coon Rapids Field 1
0
50
100
150
200
250
Nitrogen Rate (kg ha-1)
Fig. 4. Corn yield response to applied nitrogen in different fields in central Iowa from 2000
to 2002.
84
Agricultural Science
Year/Field Significance
Slope
2000/Carroll 1
ns
2000/Carroll 2
ns
2000/Sac *
-
2000/Shelby ns
2000/Story 1
ns
2000/Story 2
**
+
2001/Story 1
ns
2001/Story 2
***
+2000/Story 3
**
+
2002/Calhoun East
*
+
2002/Dallas South
ns
2002/Coon Rapids
**
+
ns- Not significant, *-p<0.1, **-p<0.05, ***-p<0.001
Table 3. Analysis of the effect of nitrogen rates on corn yields from the fields in central Iowa.
There was a range of soils within these fields and across the study sites. Variation among strips within a field for a given N rate was not significant for all fields when evaluated with a simple analysis of variance (ANOVA) using like strips as replicates. Understanding that N
response is not consistent among the different fields creates questions about reasons for N
responses observed among fields. Development of processes for the application of N within a field that takes advantage of information about corn response to N would greatly enhance the efficiency of N use. To explore this question and the lack of a consistent response,
further evaluations were conducted on the data set. An additional data set on N response
across a single field obtained by Hatfield and Prueger (2001) showed that N response was
related to soil water holding capacity (Fig. 5).
20000
Corn Nitrogen Study 1998-2001
18000
16000
Clarion
)
Webster
-1
Canisteo
14000
Nicollett
Okoboji
kg ha 12000
d (iel 10000
Y
8000
6000
4000
40
60
80
100
120
140
160
180
200
220
Nitrogen Rate (kg ha-1)
Fig. 5. Corn yield response to applied nitrogen across five different soils for a field in central Iowa for 1998 to 2001.
Spatial Patterns of Water and Nitrogen Response Within Corn Production Fields
85
All soils showed an increase in yield above the 56 kg ha-1 rate but only the Webster soil showed a positive relationship to increasing N above this rate (Fig. 5). These results show that soils with a higher water holding capacity (Webster) had a different response to N
compared to soils with lower water holding capacity (Clarion, Canisteo). Even though the
studies were conducted at a different location, the soils were similar to those in these N
studies so these results help explain the results shown in Fig. 4. The lack of response to N
across the majority of the fields was related to the distribution of soils within the field. The Sac Field, in 2000, had a decrease in yield with increasing N was caused by the large amount of highly eroded Clarion soils within the field resulting in a limitation on the water holding capacity in the soil profile causing yields to be severely limited by water. These results are similar to those reported by Massey et al. (2008) in which the eroded soils had the lower yields and in this study lower yields were associated with soils having lower water holding capacity. Similar findings were reported by Sadler et al. (2000b) in which they suggested that understanding of site-specific yield maps would be enhanced by observations of water stress within the field. Their observations and those from this study suggest that spatial yield patterns in response to N management are dictated by soil types within the field and the interaction with soil water availability.
In order to understand the spatial patterns of yield within a strip, a stepwise approach was taken to evaluate these patterns. The first step was to evaluate the frequency
distribution of yields for the different N rates as shown for the Coon Rapids field (Fig. 6).
These frequency distributions are similar to other fields in this study. In all fields, the lower N rate had a lower mean yield but a similar range of minimum and maximum
yields compared to the other N rates; however, the distribution showed a wider
dispersion and more variation. As the N rate increased the variation pattern showed
reduced dispersion with a higher frequency of values near the mean value (134 kg ha-1
compared to 200 kg ha-1). In all fields we observed that the higher the N rate the less
variation in frequency distribution with a similar shape of the yield distribution and the range of maximum and minimum values. The distribution of the yields based on the
percentiles showed the 134 and 200 kg ha-1 rates were the same. All three rates showed a
skewed distribution toward the lower yields.
To further evaluate the spatial patterns within fields, yields were summarized by each soil type within each field for N rates. Spatial variation of yields within fields was significant in their relationship to soil variation within the field. Across all of the N rates there was a similar pattern with the higher yields in the Webster soils and the lower yields in the
Clarion soils (Fig. 7). Yields in the Webster soils were larger than the yields in the Clarion soils at all N rates and showed less variation than those in the Clarion soils (Fig. 7). This type of analysis was completed for all of the different fields evaluated in this study and the results shown in Fig. 7 were consistent among all of the fields with the soils having a higher water holding capacity producing the larger yields compared to those soils with lower
water holding capacity.
When the yields are aggregated to create a N response curve across fields, yield differences were significant using a simple T-test between these two soils. These two soils were chosen because they were the most dominant in all of the different fields measured in this study.
Yield differences between the Clarion and Webster soil at the 134 and 190 kg ha-1 rates were over 1000 kg ha-1 and there was a decrease in the corn yield with the N rate of 190 kg ha-1 in 86
Agricultural Science
40
Coon Rapids 2002 (67 kg ha-1)
30
ync
20
requeF
10
04000
6000
8000
10000
12000
14000
16000
18000
50
Yield (kg ha-1)
Coon Rapids 2002 (134 kg ha-1)
40
y
30
uenc
eqFr 20
10
04000
6000
8000
10000
12000
14000
16000
60
Yield (kg ha-1)
Coon Rapids 2002 (200 kg ha-1)
50
y
40
quenc
30
eFr 20
10
04000
6000
8000
10000
12000
14000
16000
Yield (kg ha-1)
Fig. 6. Frequency distribution of corn yields at the 67, 134, and 200 kg N ha-1 rate for the Coon Rapids field in 2002.
Spatial Patterns of Water and Nitrogen Response Within Corn Production Fields
87
14
N Rate 78 kg ha-1
12
10
Okoboji
Nicollett
cy
Clarion
8
Webster
quene
6
Fr
4
2
06000
7000
8000
9000
10000
11000
12000
7
Yield (kg ha-1)
N Rate 134 kg ha-1
6
Okoboji
5
Nicollett
Clarion
y
Webster
4
uenc
eq
3
Fr
2
1
05000
6000
7000
8000
9000
10000
11000
12000
8
Yield (kg ha-1)
N Rate 190 kg ha-1
7
6
Okoboji
Nicollett
y
Clarion
5
Webster
enc
4
equFr 3
2
1
02000
4000
6000
8000
10000
12000
Yield (kg ha-1)
Fig. 7. Frequency distribution of corn yields for the 78, 134, and 190 kg N ha-1 rate for the four soil types within the field in the Calhoun East field in 2002.
88
Agricultural Science
the Clarion soil (Fig. 8). This was a similar response as to that observed in the Sac field in 2000 (Fig. 4). The N response curve shown in Fig. 5 suggests that the improved water
holding capacity of the Webster soil allows for enhanced yield compared to the other soils because of the increased available soil water during grain-filling. Seasonal water use
patterns between a Clarion and Webster soil were significant during the reproductive
stage of growth because at this time of the year, crop water use was dependent upon
stored soil water in the soil profile (Fig. 2). These water use patterns lead to differences in crop water stress which affects yield patterns as suggested by Sadler et al. (2000b). This was a consistent finding across all of the fields examined in this study in which the higher water holding capacity soils had a higher yield regardless of N rate. Observations from
the 78 kg
11000
Calhoun East 2002
10000
)-1a 9000
kg h (
ield
8000
Y
7000
Clarion
Webster
6000
78
134
190
N rate (kg ha-1)
Fig. 8. Corn yield response to applied nitrogen for Clarion and Webster soils using three N
rates in 2002.
N ha-1 rate showed that yields in the Clarion soil were distributed across the range of yields observed in the field. There were changes in the statistical moments for the yields in the different soils within N rates as shown in Table 3. Yields in the 134 and 190 kg N ha-1 rate were not significantly different for any of soils (Table 4). The yield distribution within the soil types reveals the effects of the soil water availability as a major factor in determining yield response to N rates (Fig. 7). Evaluation of the N response across fields will have to account for the water holding capacity of the soil and the precipitation during the growing season in order to interpret the results. Analysis of the yield distributions within fields segregated by soil type demonstrates the impact that available soil water has on
determining the spatial pattern of corn yield.
Spatial Patterns of Water and Nitrogen Response Within Corn Production Fields
89
Soil Type
N Rate
Mean
Std. Dev.
Skewness
Kurtosis
78 129.9 14.8 -0.09 -1.46
Clarion
134 140.6 21.6 -0.59 -0.65
190 140.8 20.7 -2.29 7.63
78 124.5 22.5 0.65 -1.72
Nicollett
134 143.2 15.9 -0.37 -0.49
190 114.7 10.6 0.41 -1.21
78 147.0 12.3 -0.50 1.62
Webster
134 158.8 11.5 -0.21 -0.48
190 149.2 16.9 -2.55 8.80
78 162.2 7.4 -7.7 -0.42
Okoboji
134 165.4 6.9 0.08 0.35
190 160.5 5.9 0.42 -0.03
Table 4. Mean, standard deviation, skewness, and kurtosis for corn yields within each soil type for different N rates within the Calhoun East field in 2002.
3.3 Seasonal patterns in fields
Harvested yield represents one point in the season which is the result of all of the interacting factors during the season. One question is whether the factors that affect yield patterns at harvest persist throughout the growing season or are there changes which occur and are
detectable only in grain yield. Application of techniques related to improved management
decisions require that observations within a field be able to detect a plant response that is ultimately related to crop yield as part of the decision making process. Sadler et al. (2000b) suggested that yield patterns could be explained by following the patterns of crop stress during the season. These data sets contain a sequence of measurements during the growing
season that may be related to crop yield. The hyperspectral remote sensing data allowed for several indices to be calculated; however, one of the relationships we examined was the red (0.681µm) /green (0.561µm) ratio. This ratio was selected because of the strong relationship to crop biomass and crop yield. There was an inverse relationship between yield and the
August red/green index for the fields (Fig. 9). Although there was a large variation about the regression line, this index showed a significant relationship with yield compared to
other vegetative indices. Seasonal patterns of different vegetative indices provide insights into the spatial patterns of vegetative response during the course of the growing season. In this study we evaluated the NIR (0.819 µm) /red (0.681 µm) ratio and found that this
vegetative index was not consistently related to yield across all of the fields while the red/green relationship showed a more consistent relationship across all of the fields. There are large varieties of vegetative indices that can be computed from the wavebands shown in Table 2; however, the consistency of this index across the fields in this study was one of the primary reasons for its use in these analyses. Hatfield et al. (2008) reviewed the different indices derived from remote sensing signals and their relationship to various agronomic
variables and there are a variety of different indices which can be applied to these fields and in this study the red/green index provided a useful method of assessing response across
fields.
90
Agricultural Science
12
Dallas South 2002
11
10
)-1
g ha
9
M
d (
8
ielY 7
6
Data
Yield = -3.105 + 7.343 R/G r2 = 0.45
5
0.65
0.70
0.75
0.80
0.85
0.90
0.95
1.00
Red/Green Ratio
Fig. 9. Relationship between corn yield and red/green ratio observed in August with an
aircraft scanner for the Dallas South field in 2002.
A more detailed examination of the red/green ratio was conducted with the observations
collected four times during the growing season. Observations throughout the season represent unique characteristics of the growing season, May observations represent the soil background, July represents maximum vegetative cover, August the point of mid-grain fill, and September the time near physiological maturity (Fig. 10). At each of these times the frequency distribution of the red/green ratio was computed for each N rate within the field. There is a seasonal trend in the frequency distributions with a decrease in the variation found in the distribution from the May to July or August observations and then an increase in variation for the September observations (Fig. 10). Variation in the red/green ratios early in the season was related to the soil variation within each N rate. The variation in the July and August observations was small for all three N rates. Observations of the water use patterns among soils within a field showed little difference at this time of the growing season because there was adequate soil water in all soils to meet crop demands. Later in the growing season the crop water demand exceeds the precipitation and crop water use is dependent upon stored soil water and variation among
soils becomes evident and the variation in the red/green ratio is similar to the bare soil distribution (Fig. 10). There was no significant difference among the 67, 134, or 200 kg N ha-1
rates for the frequency distributions of the red/green ratio (Fig. 10). The frequency patterns of the red/green ratios within N rates follow the yield patterns. Spatial patterns of reflectance reveal the seasonal dynamics of the interactions of soil types with N rates. These same patterns of red/green reflectance throughout the season were the same across all of the fields within this study. There is a consistent pattern in terms of a decreasing variation as the crop develops until mid-grain fill and then variation increases during the later grain-fill stages. The only difference among fields was whether the early grain-fill observations began to reveal spatial variation because of the lack of soil water in the profile and limited precipitation to meet the crop water demands. In fields with adequate soil water during grain-fill the variation is less pronounced.
Spatial Patterns of Water and Nitrogen Response Within Corn Production Fields
91
500
Coon Rapids 2002 67 kg/ha
400
y
May
300
July
August
September
quenceFr 200
100
0 0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
600
Red/Green
Coon Rapids 2002 (134 kg ha-1)
500
y
400
May
uenc
300
July
August
eq
September
Fr
200
100
0 0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
700
Red/Green
Coon Rapids 2002 (200 kg/ha)
600
500
cy
400
May
July
equen
300
August
Fr
September
200
100
0 0.5
0.6
0.7
0.8
0.9
1.0
1.1
1.2
Red/Green
Fig. 10. Frequency distribution of red/green ratio for the four observation times during the 2002 growing season for the Coon Rapids field with 67, 134, and 200 kg N ha-1 rate.
92
Agricultural Science
Spatial analysis of the red/green index for the May and August periods during the 2002
growing season showed the effect of the soil differences for the May image with the
differences from west to east that were related to the distribution of soil types within the field (Fig. 11). The presence of waterway was very evident in this kriged map of the field.
The range of the samples was 20 m indicating there were detectable differences over
relatively short distances within the field. In other fields the range was considerably
longer and on the order of 80 to 100m. Spatial analysis was able to reveal the patterns of the soil types within the fields. This is in contrast with August image in which there is little variation across the field in the red/green ratio (Fig. 12). There is one spot with poor plant growth that was detectable in the field. In this analysis there was no stable
range in the data because there were no significant spatial patterns detected within the
field. These interpolated maps confirm the analysis conducted within each strip that
showed the July and August periods have the least variation in vegetative indices across
the field because the crop growth is uniform (Fig. 12). The growth of the crop reduces the variation within the field and there is no detectable variation caused by the N rates
within this field. Spatial analysis of the September red/green ratios showed the variation had reoccurred within the field (Fig. 13). This temporal pattern was common across all of the fields in which the variation in the red/green ratio decreased in the July and August observations and there was no correlation of these ratios with soil types within the field.
The reason for this pattern is that during this phase of crop growth the water use rate is small and with the soil profile completely recharged at the beginning of the growing
season there is more than sufficient soil water along with the precipitation to produce a uniform growth across the field. During the grain-fill period when crop water use rates
are larger and precipitation is more infrequent then soil water availability from the soil profile becomes a critical factor and influences the red/green ratio bec