FINAL REPORT ABSTRACT

Size: px
Start display at page:

Download "FINAL REPORT ABSTRACT"

Transcription

1 FINAL REPORT PROJECT TITLE: Corn Response to Nitrogen Using a Modified Strip Trial Design PROJECT NUMBER: SP PRINCIPAL INVESTIGATOR AND CO-INVESTIGATOR(S): Daniel Kaiser, Jeffrey Vetsch, and John Lamb ABSTRACT Commercial variable rate equipment can be adapted for use in establishing large scale small plot research trials and field scale experiments using commercial equipment. Replicated nitrogen trials were established at six locations in Minnesota utilizing eight nitrogen rates replicated sixteen times within each location. Urea was applied using a prescription map input into a Raven Viper Pro which controlled the rates of two Gandy fertilizer deliver boxes. A tractor equipment with auto-steer and RTK was utilized for swath width control. Corn was planted and nitrogen sufficiency was evaluated at three times (V5, V1, and R2) during the growing season with ground based active sensors and multi-spectral imagery collected with a UAV. The ground based sensors uses were a Greenseeker, Crop Circle, and SPAD chlorphyll meter. Aerial imagery was collected with a Tetracam Mini-MCA 6. Corn grain yield was measured from the middle four rows of the eight row plots. Summary of findings The optimum nitrogen rate required to grow corn did not vary by elevation within field, by soil organic matter content in the top six inches, or by soil series. The only soil property that may explain variability in corn response to nitrogen is the two foot soil test collected in the spring prior to fertilizer application. Overall, the data indicate that more information I needed before variable rate nitrogen can accurately apply the nitrogen required on every acre within a field. Ground and aerial sensors did not provide any prediction of corn grain yield or the optimum rate of nitrogen applied to corn. Sensor data collected at V5 exhibited the poorest correlation with the economic optimum nitrogen rate. Sensors could not differentiate among rates of N applied near the economic optimum nitrogen rate early in the growing season. The data indicate that sensors provide very limited utility for Minnesota corn farmers to manage nitrogen in rain-fed corn production. Collection of aerial imagery by UAV provides a rapid method to collect information. Aerial imagery requires more processing and interpretation that ground based sensors. In this study we only utilized digital values to develop different band indices. These values are inadequate to fully assess N sufficiency. With further processing it is still doubtful that the aerial images will provide better information that could be collected with ground based sensors. Deficiencies still need to be visible for the sensor to detect them. The use of Rx prescription maps allows for the establishment of complex strip trials within fields. Minnesota corn farmers could use this technology to establish trials in their own fields to determine their optimum nitrogen rates. 1

2 INTRODUCTION Strip trials are increasingly being used in research to study within site variability and to study differences in response to treatments across a landscape. When combined with grid sampling researchers have focused on studying treatment effects and soil test values. To do this, strips are usually divided into segments with soil samples being collected along the length of the strips. While convenient, this type of design presents problems in the statistical analysis since it violates the assumption of randomization and can make the data be difficult to analyze. Small replicated plots have been the staple of many research programs for studying response to nutrients. Many times the number of replications for small plot studies is limited to 3, 4, or maybe 5 replicates within a field. With increased adoption of precision agriculture growers are looking to researchers for data that focuses on how fertilization should be changed based on soil type or chemical properties. With a large number of locations it is possible to make assumptions on this type of data. However, differences in management inherent between locations can sometimes mask effects such as response differences between soil types. Some of the same technology used by farmers could aid researchers in better designing studies to focus on site specific application of nutrients. As technology has advanced the accuracy of many GIS sources has increased. In the case of RTK systems many producers can have sub inch accuracy of control. This type of equipment can be used in research work to better apply treatments within fields and auto steer guidance can help better control planting equipment. Research work with this type of equipment is needed in order to determine to efficacy for use in fertilizer research. Better management of nitrogen has been a goal of many researchers as well as farmers and commodity groups due to the potential for applied N to contaminate surface waters. Nitrogen rate studies have been very beneficial to determine optimum application rates for corn. However, many growers are interested in variable rate application of N to better apply the nutrient to limit loss but also in order to maximize yields. Current research plots seldom encompass large field areas so it is difficult to determine if optimum N rate changes across a landscape. While others have studied this using a single application rate of N as a field length strip we propose to modify that design and variable rate apply the N along the treatment strip to eliminate the lack of randomization in the strip trial. Farmers are also interested in technology that can help assess in-season N stress in order to better manage inputs putting the optimum rate where it is needed, when it is needed. Much research has focused on using tools such as the Greenseeker to aid in in-season management. While there has been research showing this technology to be effective in other states, the level of N mineralization from the soil can significantly limit the usefulness of this technology. The Greenseeker measures total biomass and in most cases N stress does not show up until it is too late to effectively use. Newer technology that actually measures the greenness of a plant using hyper spectral imagery are available but have not been widely tested. This technology includes a new satellite that can measure different wavelengths of light that may better correlate to the chlorophyll within the plant. If this technology works as promised it should be better at predicting where N should be applied than the current Greenseeker. OBJECTIVE AND GOAL STATEMENTS The purpose of this study is to use some of our new technology, funded in part by the Minnesota Corn Research and Promotion Council and the Agriculture Fertilizer Research and Education council to determine the efficacy of a modified strip trial design at determining the influence of differing soils or landscapes on the response to nitrogen. 2

3 1. Evaluate corn response to nitrogen fertilizer across a variable landscape 2. Compare the use of multi spectral imagery collected with a UAV versus ground based sensors for predicting corn N response 3. Study how landscape position may effect nitrogen use efficiency of corn 4. Assess fertilizer rate control of our new application controllers in larger trials within farmers fields MATERIALS AND METHODS Six locations were studied over the span of three years (2 locations per year) Eight nitrogen rates were applied. Rates were changed based on previous crop o Corn following soybean or small grains:, 3, 6, 9, 12, 1, 18, and 21 lbs N per acre o Corn following corn or sweetcorn:, 4, 8, 12, 16,, 24, and 28 lbs N per acre 16 replications of each nitrogen rate were included at each location Spring applied urea was applied and incorporated prior to planting (P, K and S fertilizer was applied with urea at non-limiting rates based on soil test levels) 8 row wide plots (3 rows) Corn Hybrid: Pioneer 9917 AM1 seeded at 35, plants per acre was used at all locations Aerial imaging taken at 3 times (V6, V1, and R2: 213 and 214 sites only) Crop Circle and Greenseeker used to collect data over the middle two rows at V5 and V1 at all sites. Greenseeker was only used to collect data at V6 in 212 SPAD chlorophyll meters were used to collect data from 2 plants at V1 (uppermost fully developed leaf) and R2 (ear leaf) Other Measured variables o Composite soil samples (-6, 6-12, and ) from an area 4 wide by long (two plots wide by two plots long) o Whole plant, cob, and grain N uptake at PM (average of 6 plants sampled per plot) were used to determine nitrogen use efficiency (NUE) o Plant population for each plot at V5 o Grain Yield adjusted to 15.5% moisture o Grain Harvest Moisture To complete the nitrogen study a small air flow equipped with two fertilizer bins was used to apply nitrogen to the research plot area. The fertilizer applicator is connected to a Raven Viper Pro and has the capability to variable rate 2 products on the go based on a prescription map. The controller is similar to that used in commercially available equipment. A tractor equipped with auto steer was used to plant the trial areas to ensure the corn rows follow the treated area. This type of equipment can simplify time setting up the trials and reduce operator error by controlling application rates with the Viper Pro unit eliminating stopping within the trial areas. We have used the planter unit for the past two years without RTK and have used the variable rate application with success at a number of trials. This study will be both beneficial for us to learn the capabilities of the new application equipment and the study designs will also provide useful data to MN corn farmers. 3

4 RESULTS AND DISCUSSION Sites conducted in 212 were located in Sibley County, one South of Stewart and one West of Gaylord. In 213, one site was established in Kandiyohi County near Willmar and a second location was near Janesville in Waseca County. In 214, sites were located in Winona County north of Saint Charles and in Waseca County northwest of New Richland. Soybean was the previous crop at Stewart New Richland, and Saint Charles, spring wheat at Gaylord, corn at Willmar, and sweet corn at Janesville. Field areas were selected that had some variation in topography and soil types. The goal was to have 2-3 distinct soil series or regions of elevation in the field. Major soil series that were included in the study areas are given in Table 1. Elevation data will be considered either using data collected during planting or fertilizer application with the tractor mounted GPS or using LIDAR data when identifying zonal areas within fields. Table 1. Summary of predominant soil series within the study areas for each location. Predominant Soil Series Year Location Gaylord Webster Canisteo -- Stewart Clarion-Swanlake Canisteo Nicollet 213 Janesville Le Sueur Reedslake Cordova Willmar Arctander Wadenil-Sunburg Grovecity 214 New Richland Canisteo-Glencoe Webster Nicollet Saint Charles Seaton Eitzen -- Table 2. Soil test summary for samples collected from the nitrogen studies prior to treatment application in 212. Nitrate-N Location Depth Olsen-P K ph SOM Min Avg. Max ppm % lb N/ac Gaylord Stewart Janesville Willmar New Richland Saint Charles Minimum (Min), Average (Avg.), and Maximum (Max) nitrate nitrogen values for the 32 soil samples collected at each location. 4

5 Soil samples were collected by sampling the center of 4 treatment plots 2 wide by 2 long. Due to the small plot size we did not feel it was necessary to sample every sub plot within the study. The method used gave a good estimate of the variability of soil chemical properties across the field. For the nitrogen study samples were collected at -6, 6-12 and depth increments. The -6 samples were tested for a number of chemical properties while only nitrate-n was determined on the 6-12, and depths. Field average values are given for phosphorus, potassium, soil ph, and soil organic matter values while the average, minimum, and maximum values are given for nitrogen at each location. Two foot residual nitrate in the spring averaged 4 to 6 lbs of N. This is generally average to slightly high for some areas. We were curious to see if the dry weather conditions in 212 would lead to higher carryover of N. In this case there did not appear to be anything significant following soybean at Stewart and Wheat at Gaylord. The wet spring in 213 resulted in low levels of N at the Willmar site. Soils were somewhat sandy at Willmar which increased the potential for less N carried over (total N at Willmar averaged 22 lbs of N per acre in the top two feet). Average N carryover was higher at Janesville which was expected based on the previous crop (sweet corn). The lowest carryover of nitrate occurred for the 214 locations. Average residual nitrate at both 214 locations was around 2 lb of N per acre at both locations. The carryover of nitrate appeared to be slightly higher at Saint Charles in 214 compared to New Richland which is surprising as the soil at Saint Charles is a silt loam which is better drained and would have a higher potential for nitrate leaching. Remote sensing data was periodically collected from the nitrogen trial throughout the summer. Normalized difference vegetative index (NDVI) data was collected at the V5 growth stage using a Greenseeker model 72 and Crop Circle model 43. The NDVI measurements were also collected at V1 but using the Crop Circle only in 212 while both the crop circle and Greenseeker were used in 213 and 214. Some data are missing for the Greenseeker in 214 as we had problems with the electronics on the system that resultrf in some of the data being lost. In total we had issues with both the crop circle and the Greenseeker. The crop circle was prone to shorting out during use. However, the system allowed for us to know when this occurred so data was collected from all plots. We did not know when data collection issues occurred for the Greenseeker until it was too late to return to collect the data. Since both units measure NDVI we do have ground based sensor data collected from all plots by the crop circle. The NDVI calculation reflects the total amount of vegetative biomass and are calculated based on light reflectance in the near infrared (NIR) and visible red (red) wavelengths (Eq. 1) which are not sensitive to greenness differences within the plant. The Crop Circle used also measures a third band in the near rededge (red-edge) part of the spectrum which is supposed to be more sensitive to differences in chlorophyll. This extra band allows for the calculation of a second index, the NDRE (Eq. 2). SPAD chlorophyll meters were used at V1 on the upper most fully developed leaf of the plant and on the ear leaf (leaf opposite and below the ear) at approximately R2 to determine overall greenness of the plant. The SPAD chlorophyll meter has been used extensively in nitrogen research and has been found to correlate well to differences in plant color due to N stress. However, their use is more labor intensive and does not allow for on-the-go measurement. A Tetracam Mini MCA (multiple camera array) was purchased for use with an unmanned aerial vehicle for 213. The Tetracam measures reflectance in ambient light for six wavelengths (Table 3). Additional wavelengths allow for the calculation of additional indexes such as the GNDVI (Eq. 3) which uses reflectance in the green spectrum rather than the red. We are interested in knowing whether measuring reflectance outside of red would offer better stress detection and would correlate better to differences in chlorophyll in the plant. A greater level of detection is needed in order to detect stress earlier when it is more feasible to apply fertilizer. The major drawback of the Tetracam is the use of ambient lighting. Eq [1] NDVI = (Refl NIR Refl red)/ (Refl NIR + Refl red) 5

6 Eq [2] NDRE = (Refl NIR Refl red-edge)/ (Refl NIR + Refl red-edge) Eq [3] GNDVI = (Refl NIR Refl green)/ (Refl NIR + Refl green) Table 3. Examples of sensors used in the studies and specific wavelengths of reflected light measured Sensor Blue Green Red Red-Edge NIR nm SPAD na na 6 na 94 Greenseeker na na 656 na 774 Crop Circle Na na Tetracam & 9 na, not available Site Climatic Data Summary Rain gauges were installed at both locations to measure precipitation but the data. Total daily precipitation is summarized in Appendix Figure A for both locations. Since the rain gauges were installed about one week after planting, rainfall that fell prior to planting and since fertilizer application was not accounted for in the total collected beginning on May 7. Total rainfall at Gaylord was 1.91 inches with the heaviest amounts falling late May to early June. The total precipitation that fell at Stewart was nearly double that of Gaylord at inches which is interesting since both of the locations were within the same county. Similar periods of heavy rainfall were seen at Stewart with one exception, a period of 3+ inches of rain over several days in mid-june. There were several periods of rainfall totaling more than 1 inch at Janesville and Willmar during 213 and New Richland and Saint Charles in 214. The 214 growing season was particularly wet in late May and June. We were able to plant the field earlier in 214 as rainfall in May was less than in 213. Rainfall for the 213 and 214 growing season made for easier comparisons of the sensing technologies due to increased likelihood of N loss. Nitrogen Uptake and Grain Yield Summary The uptake of nitrogen measured at R6 is summarized in Figure 1. Nitrogen rate significantly (P<.5) affected the uptake of N in the plant (stover) and grain at all sites, and cob uptake of N at all site except for Gaylord (Appendix Table A). When significant, the effect of N rate on the uptake of N was curvilinear with a plateau. The exception was for plant N uptake at Gaylord and Janesville, and plant, cob, and grain uptake at Saint Charles which increased linearly and did not plateau. It appears that nitrogen supply may have been low at the previously mentioned sites. 6

7 Nitrogen Uptake (lb N/ac) Nitrogen Uptake (lb N/ac) Gaylord 212 Stewart Grain Grain Stover Cob Stover Cob Nitrogen Rate (lbs N applied + lbs in 2' soil sample) Nitrogen Rate (lbs N applied + lbs in 2' soil sample) Janesville 213 Willmar Nitrogen Uptake (lb N/ac) 1 Grain Stover Cob Nitrogen Uptake (lb N/ac) 1 Grain Stover Cob Nitrogen Rate (lbs N applied + lbs in 2' soil sample) Nitrogen Rate (lbs N applied + lbs in 2' soil sample) New Richland 214 Saint Charles Nitrogen Uptake (lb N/ac) 1 Grain Stover Cob Nitrogen Uptake (lb N/ac) 1 Grain Stover Cob Nitrogen Rate (lbs N applied + lbs in 2' soil sample) Nitrogen Rate (lbs N applied + lbs in 2' soil sample) Figure 1. Summary of Nitrogen uptake and Partitioning among stover, cobs, and grain based on nitrogen rate applied plus nitrate-n in a 2 soil sample taken in the spring at two field locations in Minnesota. Overall, total N uptake was around 3 lbs of N per acre less at Gaylord and New Richland than the next lowest total N uptake at each site. The lower uptake of N at both sites is likely a result of loss of N or plant material prior to sample collection at either site. The Gaylord site was sampled later than Stewart and dry, hot, and windy days late in the season caused rapid drying of the plants within the fields. An early frost affected plants at New Richland. The percentage of total N taken up in the plant, cob, or stover was calculated but the data are not shown. The percentage of N taken up in the plant as a portion of the total remained the same across N rates. The 7

8 portion taken up by the cob was greater at lower N rates and the portion in grain increased with increasing N rate at a few of the locations. The increase in N content with higher rates of applied N indicates that more protein was likely being produced in the grain as the availability of N increased. The greatest total uptake of N occurred at the Janesville location (Figure 1). Cob N uptake was similar among the six locations but there was substantially greater uptake of N in both the plant and grain at Janesville than the other locations. Residual nitrate supply was the greatest at Janesville which could account for the increased uptake of N. Grain N uptake was high relative to plant and cob at Willmar, but the increased potential for N loss coupled with the previous crop of corn at this site resulted in less total N uptake than the other four locations. In addition, the uptake of N was continuing to increase beyond the highest N rates in the study. This indicates that N uptake will continue beyond the point of which grain yield is maximized. Much of this increase was due to increased uptake in stover and not in the cob or grain. Table 4. Summary of the economic optimum nitrogen rates at each location based on price ratios (price of N/bushel value of corn) from to.15 and the estimation of the percentage of the maximum yield produced at that nitrogen level. Data were calculated based on data considering only the amount of nitrogen fertilizer applied (soil N test was not used). EONR % Max. Yield Year Site lbs N acre % Gaylord Stewart Janesville Willmar New Richland St. Charles Yield data summarized across all field areas is given in Figure 2a through 2c and the economic optimum nitrogen rates are given in Table 4. There was a significant (P<.5) response of corn yield to nitrogen fertilizer rate at all locations (Appendix Table A). The relationship was analyzed both by studying the applied nitrogen rates alone and with the total amount of N in the two foot soil test taken prior to treatment application in the spring (Figure 3). Yield was increased up 222 lbs applied + soil N at Gaylord, 25 lbs at Stewart, 21 lbs at Janesville, 23 lbs at Willmar, 171 lbs at New Richland, and 38 lbs at Saint Charles. The predicted economic optimum N rate (not including N in a two foot soil sample) was lower than expected at Gaylord and Janesville and higher than expected at Stewart, Willmar, New Richland, and Saint Charles (Table 4). Overall conditions were conducive to N loss at both locations due to heavy rainfall events early in the growing season. However, there did not appear to be major loss at Gaylord of Janesville since the predicted N need based on the EONR was lower than expected. There appeared to be greater N loss at the remaining site where the EONR was much higher than what is suggested for use on corn following soybean or corn following corn. 8

9 Zone 1 Zone 1 Zone 4 Zone 2 Zone 2 Zone 3 Zone 3 Gaylord, MN 212 Stewart, MN Corn Grain Yield (bu/ac) 1 Zone 1 Zone 2 Zone 3 Corn Grain Yield (bu/ac) 1 Zone 1 Zone 2 Zone 3 Zone Nitrogen Rate (lbs N/acre) Nitrogen Rate (lbs N/acre) Figure 2a. Response to applied nitrogen fertilizer summarized for the management zones delineated by elevation at the Gaylord and Stewart locations. Individual points on the graphs represent an average of the nitrogen rate within each management zone. Field areas were delineated into separate management zones using elevation to study whether the optimum N rate was affected based on elevation. Both fields were separated into three main zones. Data for the 213 sites is summarized in Figure 2a. A fourth zone was included at Stewart which encompassed the areas where significant drought stress occurred. Areas not under drought stress responded similarly to N fertilizer. Yields increased to about 12 lbs of N at Gaylord and 14 lbs of N at Stewart at which response slowed and the return per lb of N was at a point where it was not economical to apply. Zone 4 at Stewart showed a different response where yield continued to increase even beyond the highest application rate (21 lbs N/ac). While it may have been possible to increase yields further the cost to do so would outweigh any potential increase in profit from N application. In fact, the net return per lb of N was very low that if the ratio of the price of N to the value per bu of corn was above.5 it would not be economically feasible to apply N. This alone was the only direct evidence of a variation in N requirement by corn across the areas studied. Since the area of the field affected was small it likely would not have been economically feasible to vary the amount of N applied. 9

10 Zone 3 Zone 3 Zone 2 Zone 1 Zone 2 Zone 1 Janesville, MN 213 Willmar, MN Corn Grain YIeld (bu/ac) Zone 1 Zone 2 Zone 3 Corn Grain YIeld (bu/ac) Zone 1 Zone 2 Zone Nitrogen Rate (lbs N/acre) Nitrogen Rate (lbs N/acre) Figure 2b. Response to applied nitrogen fertilizer summarized for the management zones delineated by elevation at the Janesville and Willmar locations. Individual points on the graphs represent an average of the nitrogen rate within each management zone. Zone 1 was lost at Janesville due to June flooding. The field sites in 213 were broken into three zones. June flooding resulted in no data being collected from Zone 1 at Janesville. Similar to the 212 data, there was no clear evidence of differences in yield and N response based on the zones at either Janesville or Willmar (Figure 2b). Elevation and soil series were both considered after fields were zoned. 1

11 Zone 3 Zone 1 Zone 2 Zone 3 Zone 2 Zone 1 New Richland, MN 214 Saint Charles, MN Corn Grain YIeld (bu/ac) Zone 1 Zone 2 Zone 3 Series5 Corn Grain YIeld (bu/ac) Zone 1 Zone 2 Zone Nitrogen Rate (lbs N/acre) Nitrogen Rate (lbs N/acre) Figure 2c. Response to applied nitrogen fertilizer summarized for the management zones delineated by elevation at the New Richland and Stewart locations. Individual points on the graphs represent an average corn grain yield of the nitrogen rate within each management zone. Fields were again separated into three management zones for the 214 sites. At Saint Charles there was no evidence that economic optimum N rate variety by management as yield was similar even though corn grain yield appeared to be elevate for Zone 1 over Zone 2 and 3. Zones were separated for New Richland. Zone 2 represented a lower area on the landscape that where yield was affected by saturated soils due to major rainfall events in June. What is interesting is that yield was affected but the amount of N needed to maximize grain yield was the same for all zones. This indicates that the water issues affected yield potential but did not affect the responsiveness of the crop to N. It would be likely that a farmer would side-dress areas like Zone 2 at New Richland as the area was showing visible N deficiency. Since the optimum N rate did not differ it is doubtful that additional nitrogen would benefit the crop. Sensors also would not have aided in the determination of rate to apply in Zone 2 as there was no detectable differences that could be discerned among the N rates with any of the sensors used at the time the data was collected. Yellowing due to lack of oxygen was reduced the effectiveness of the sensor in detecting differences among N rates so using a sensor would have resulted in a significant application of N across the area when corn grain yield was maximized resulting in wasted fertilizer for some of the N rates. 11

12 22 Gaylord Stewart Yield (bu/ac) Yield (bu/ac) When N < 222 lb YieldI = (Ntot)-.12(Ntot) 2 R 2 =.46 P< ' soil test N + N Fertilizer Applied (lb N/ac) 8 6 When N < 25 lb YieldI = (Ntot)-.3(Ntot) 2 R 2 =.74 P< ' soil test N + N Fertilizer Applied (lb N/ac) 22 Janesville, Willmar, 213 Yield (bu/ac) Yield (bu/ac) 1 12 When N < 21 lb Yield = (Ntot) -.14 (Ntot) 2 R 2 =.4 P< ' soil test N + N Fertilizer Applied (lb N/ac) When N < 23 lb Yield = (Ntot) -.22 (Ntot) 2 R 2 =.68 P< ' soil test N + N Fertilizer Applied (lb N/ac) Yield (bu/ac) New Richland, 214 When N < 174 lb Yield = (Ntot) -.29 (Ntot) 2 R 2 =.85 P<.1 When N < 172 lb Yield =.1 +.6(Ntot) -.16 (Ntot) 2 R 2 =.66 P< ' soil test N + N Fertilizer Applied (lb N/ac) Figure 3. Yield response data for the Gaylord and Stewart locations summarized for the total amount of N applied in fertilizer and the total amount of N based on a 2 foot soil N test taken in the spring prior to treatment application. Yield (bu/ac) Saint Charles, 214 When N < 38 lb Yield = (Ntot) -.11 (Ntot) 2 R 2 =.77 P< ' soil test N + N Fertilizer Applied (lb N/ac) 12

13 14 Minnesota Data Minnesota Data Relative Yield (% of Maximum) Applied N (lb N/acre) When Applied Urea N <19 Yield= (N) -.1(N) 2 P<.1 R 2 = Applied N + 2' Soil N test (lb N/acre) Figure 4. Summary of data collected among six Minnesota field trial locations conducted from The graph on the left compares relative yield (percent of maximum yield) to the amount of N applied as spring urea. The graph on the right compares relative yield versus N applied as urea plus the amount of nitrate-n calculated from a two-foot soil sampled collected in the spring. The two foot N test was evaluated in order to determine if the optimum N rate could be better predicted across locations by factoring in the amount of nitrate-n in the soil prior to fertilizer application. The left graph in Figure 4 represents the data only considering the amount of N applied in fertilizer. There is a significant amount of variation among replications across sites for the same N rate. Factoring in the soil N test did slightly improve the correlation (R 2.59 versus.48). While the nitrate-n test may improve the prediction the improvement looks to be marginal and may not be better than using a standard recommendation system for N. One thing we did not do was separate out the sites by previous crop which may further improve the relationship among locations. If the soil N test would work it would offer a better choice than sensors in predicting the optimum N rates with and across locations. The advantage to the N test would be a proactive approach to N management. Relative Yield (% of Maximum) When Total N <28 Yield= (N) -.1(N) 2 P<.1 R 2 =.59 13

14 Gaylord 212 Stewart 212 Nitrogen Use Efficiency (%) y = -.19x R² =.85 Nitrogen Use Efficiency (%) y = -.x x R² = Fertilizer Nitrogen Rate (Lbs N/acre) Fertilizer Nitrogen Rate (Lbs N/acre) Janesville 213 Willmar 213 Nitrogen Use Efficiency (%) y =.x x R² =.89 Nitrogen Use Efficiency (%) y = -.6x R² = Fertilizer Nitrogen Rate (Lbs N/acre) Fertilizer Nitrogen Rate (Lbs N/acre) New Richland 214 Saint Charles 214 Nitrogen Use Efficiency (%) y = -.x 2 +.1x R² =.93 Nitrogen Use Efficiency (%) y = -.2x R² = Fertilizer Nitrogen Rate (Lbs N/acre) Fertilizer Nitrogen Rate (Lbs N/acre) Figure 5. Nitrogen use efficiency calculated for four field locations in central Minnesota using total uptake of N in the plant, cob, and grain from each N rate minus the control then divided by the amount of N applied in fertilizer. Nitrogen use efficiency (NUE) was calculated based on the total amount of nitrogen taken up divided by the total amount applied in fertilizer. At both locations the maximum NUE was around % and there were differences as to which treatments produced the highest NUE. The relationship at Gaylord, Janesville, Willmar, New Richland, and Saint Charles is what was expected where the highest NUE occurred at the lowest rate of N. In contrast, NUE increased then decreased at Stewart with the highest NUE occurring near the EONR. There was no evidence of a difference in NUE by zone (Appendix Table A) which indicates that one rate of fertilizer could have been applied and not reduce yield or NUE. Nitrogen use efficiency decreased with increasing rates of N at Janesville, New Richland, and Saint Charles but did not differ among N rates at Willmar. In fact, NUE was relatively high across the rates of N at Willmar averaging near % while the NUE was similar between the to 9 lb N application rates at New Richland. The high NUE at Willmar could be due to the lack of residual soil N. More applied N 14

15 would be required to maximize crop yield at Willmar if the soil could not supply a smaller portion of the amount of N required. Remote Sensing Data and Grain Yield Table 5. Relationship between sensor values (relative to the site maximum for each sensor) taken from each plot and the difference (Δ) of N applied and the economic optimum nitrogen rate (at the.1 price ratio) across sites. ΔEONR values are summarized for the plateau (% relative yield) and 95% of maximum yield. ΔEONR (.1) Timing Sensor Index n R % 95% R.Y. ---lb N ac V5 Greenseeker NDVI Crop Circle NDVI NDRE Tetracam NDVI NDRE GNDVI V1 Greenseeker NDVI 4. na Na Crop Circle NDVI NDRE Tetracam NDVI NDRE GNDVI SPAD R2 Tetracam NDVI NDRE GNDVI SPAD n, number of sites included in the particular dataset Four indices were calculated from the various sensor platforms. At each site, the data was summarized by relating values collected from each plot to the maximum value from each site in order to smooth out variation in data related to when and how the samples were taken. A summary of sensor data, across locations, using the individual plot data is given in Table 5. Relationships between relative sensor value from individual plots was related to the difference from the economic optimum N rate of the plot versus the study location are presented. For the EONR calculation the amount of N in the two foot soil test was considered with the amount of N applied in the spring. The amount of nitrate in the soil was not weighted differently for the economic calculation versus the amount in the applied fertilizer. Four sites were sampled but data are not available for all sensors, thus the number of sites included in the dataset is given in Table 5. Data that includes only 1 or 2 sites would be considered weaker than that which includes 4 sites and would have the tendency to be greatly affected by the addition of further data which is planned following the 214 cropping season. There were significant relationships between all sensing values. However, the R 2 values indicate large differences in the overall predictability of the models. The poorest relationship was found for sensing values taken at the early growth stage (V5) and was poor for NDVI calculations throughout the sampling timings. The exception for the NDVI measurements was the Tetracam sensing at V5. However, this dataset only includes plots from 1 site, the Willmar location, which was highly N deficient early in the growing season (site was corn after corn). For the other sites with higher residual N there was very little 15

16 detectable N deficiency among treatments that received N (control plots were noticeable). It is likely that the predictability of the Tetracam data at V5 will be less when the data is added from the two sites in 214. The fact that NDVI only reflects differences in vegetative growth is primarily why there was such a low sensitivity of this index to the nitrogen need of the crop. Reflectance in the green and the red-edge (zone between red and near-infrared) have been shown to be more sensitive to differences in chlorophyll in the plant. The NDRE index uses the red-edge and the GNDVI uses green reflectance instead of red reflectance used to calculate NDVI. The Crop Circle 47 used in this study measures both in the red and red-edge reflectance spectrums while the tetracam also measures in green reflectance. The NDRE and GNDVI index values were poorly correlated to the difference from the economic optimum N rate at V5 for all of the sensors, while the NDRE measurement did show improved correlation when using the Crop Circle at V1. The tetracam data collected at V1 did not show a good relationship to the economic optimum N rate. Some of the variability in the tetracam data at V1 was likely due to lighting when the pictures were taken which highlights one of the problems using technology that relies on ambient lighting. The data from R2 indicates a better predictability of the model generated for GNDVI and NDRE and the difference from the economic optimum N rate compared to NDVI alone. Whether the sensors could accurately predict EONR will be discussed shortly. The sensor that provided the model that could explain most of the variation in the difference form the economic optimum N rate was the SPAD chlorophyll meter. While sample taking is labor intensive, a direct measurement on the leaf appears to still have the best overall correlation with corn N response. Overall, the R2 data collected still showed the best overall correlation to nitrogen rate response than at V1. Late season data collection should have the best correlation with N response as a majority of the nitrogen should be taken up thus the relative sufficiency/deficiency of N should be more readily detectable. The difference from the economic optimum N rate at % relative sensor value is summarized in Table 5. The difference from the EONR was based on a.1 price ratio which is common for the price of N relative to value of corn. An acceptable rang plus or minus 15 lbs of N from would be acceptable considering variability in data from field research studies. In this case, there were very few of the sensors that were within this range at their maximum values. In fact, sensing measurements taken at V5 tended to under-predict the EONR, both under- and over-predicted EONR at V1, and most indexes tended to overpredict at R2. The worst under-prediction tended to occur with the NDIV measurements (except for V5 Greenseeker which has a deonr = but also has a relatively low R 2 ), but was not surprising since NDVI is not greatly correlated to plant chlorophyll. More data would help better fine-tune the models but also would likely result in a general lowering of the R 2 value which would indicate the model would explain less of the overall variability within the data set. Correlation among sensor indices is summarized in Appendix Table B. Although most values showed at least some weak correlation, we were mainly interested in what correlated to the SPAD data collected at V1 and R2 since the data from these collections best explained differences in N availability. The best degree of correlation occurred between the SPAD data collected at V1, SPAD data collected at R2, and the NDRE measurements with the Crop Circle at V1. The fact that the Tetracam data did not seem to correlate well is likely due to the quality of images at V1. We have found the Tetracam data to be more sensitive to smaller differences due to N than the crop circle or greenseeker which tend to reach canopy saturation earlier in the growing season. However, the Tetracam data does not appear to be clearly superior to some of the ground based sensors considering the fields where the data were collected. It is common to use a 95% value relative to the maximum for determining where a yield response would occur when using sensors such as a SPAD chlorophyll meter. This may help considering an overprediction of N requirement but would result in a further under-prediction of need for indices with a 16

17 predicted deonr of less than. While not calculated, the % relative to maximum for each index will be calculated for a deonr of for the final report after all of the data are collected. Sensor Relationship to Total Nitrogen Uptake at R6 Four of the sensor indices (Greenseeker NDVI at V5, Crop Circle NDRE at V1, and SPAD chlorophyll readings at V1 and R2) were compared to the relative total N uptake across sites (relative values were calculated based on total N uptake at each site to give a weighted difference for comparison). Data are given in Appendix Figure B. There was a better correlation between all of the sensor readings and total N uptake. However, correlations were poorest with NDVI measured at V5, somewhat better with NDRE measured at V1, and the best correlations were to V1 and R2 SPAD readings. While all relationships were curvilinear, there was no clear plateau in the season readings at the measured Total N uptake. We did not measure N uptake at each sampling time but the end goal was to measure total uptake which all sensors provided some measure. The fact that yield did plateau and N uptake did not was not surprising since corn does have the potential for luxury uptake. The fact that better correlations existed with N uptake than yield is interesting but not particularly useful in helping make N rate decisions. Variable rate equipment performance One goal of this work was to assess whether complex N trials could be laid out with commercially available equipment. The Raven Viper Pro saves as applied maps that can be utilized to determine as applied N rates. The dual bin until would allow for the application of urea along with another product at the same time. The primary limitation of the new set up is the design of the Gandy delivery system. In order to ensure that we could apply up to 3 with the applicator the Gandy bins used on the applicator contain medium size wheels. Larger wheels recommended for many fertilizers are available but would not allow for the width of application desired. In 212 we used two separate applications methods. At Stewart the Urea was applied using a single bin. The primary issue we encountered was plugging of the individual fertilizer delivery wheels at the high N rates. At Gaylord, both bins were used and which applied ½ of the needed fertilizer rate for each plot. Table 6. Summary of as applied nitrogen rate data collected from the Raven Viper Pro during application at Stewart and Gaylord in 212 and the standard deviation amongst the data points collected for each N rate. Target Nitrogen Rate Applied Standard Deviation Within Plots N Rate Stewart Gaylord Stewart Gaylord lb N/ac as Urea More difficulties were encountered with the single bin application. The target N rates could not be achieved at the highest rate, 21 lbs of N per acre and there was substantially more variation in the rate at the highest 3 application rates. The double bin arrangement achieved a closer as applied N rate to the target rates as indicated by the Gaylord data. There was also less variation within the plots based on the standard deviation data at Gaylord. Overall, the use of this technology to lay out complex N rate studies 17

18 will work. However, a higher capacity flow controller is needed to consistently apply high rates of N to lay out a N rate study. CONCLUSIONS Utilization of the Raven Viper Pro system allows for an easy and relatively quick way to establish large on-farm nitrogen rate trials. This technology can be adapted to commercial applicators to establish large plot replicated strips within farmer fields. Within the acre areas studied the optimum rate of N did not vary based on soil type or landscape position unless there was a major change in soil chemical properties. At one location a sand inclusion are yield less than the rest of the field, took more nitrogen to increase yield, but the addition of nitrogen was not economical. There may be differences in the optimum nitrogen rate required for corn in fields, but there is little to no impact on a small scale. Sensors can detect plant nitrogen stress. The ability of sensors to predict the final yield and nitrogen requirement is poor at early vegetative stages. All sensors under predict required nitrogen at V5. Prediction of nitrogen requirement increases at V1 but the sensors would still under apply nitrogen. Overall, the data indicates limited utility of sensors for the management of nitrogen in Minnesota. Aerial images collected with a UAV provide a quick method to sense corn. While there is correlation among indices generated with aerial images and ground based sensors, the data collected with the UAV did not provide a better method for sensing compared to ground based sensors. Nitrogen use efficiency varies by location. There is no indication that sensors can be used to improve nitrogen use efficiency of corn. Nitrogen use efficiency can be increased through increasing corn grain yield without a marked increase in the required N. EDUCATION, OUTREACH, AND PUBLICATIONS Oral extension presentations (# of Farmers) 1. Kaiser, D.E In-season N predictions using canopy sensors. Nitrogen Conference. Rochester, MN. 23 Feb (4) 2. Fernandez, F.G., and D.E. Kaiser In-season nitrogen prediction: The sensing approach. Nutrient Management Conference. Morton, MN. 9 Feb (25) 3. Kaiser, D.E Establishing on-farm strip trials. North America Farm and Power Show. Owatonna, MN. 19 Mar (5) 4. Kaiser, D.E Update on remote sensing for crop nutrients. Soybean Symposium. Chanhassen, MN. 19 Mar (25) 5. Kaiser, D.E Current technology for predicting in-season nitrogen needs. Nitrogen Management Conferences. Saint Cloud, MN 6 Mar () 6. Kaiser, D.E Future for UAV use in agriculture. Freeborn County Corn Growers Field Day. Clarks Grove, MN. 4 Sept (3) 7. Kaiser, D.E Can remote sensing be used to accurately manage nutrient inputs. Benton County SWCD irrigation clinic. Sauk Rapids, MN. 4 Feb (15) 8. Kaiser, D.E Potential and limitations for nutrient management with remote sensing. Institute for Ag. Professionals Research Updates. Morris, MN. 15 Jan (25) 9. Kaiser, D.E Potential and limitations for nutrient management with remote sensing. Institute for Ag. Professionals Research Updates. Willmar, MN. 14 Jan (4) 18

19 1. Kaiser, D.E Potential and limitations for nutrient management with remote sensing. Institute for Ag. Professionals Research Updates. Lamberton, MN. 9 Jan (25) 11. Kaiser, D.E Potential and limitations for nutrient management with remote sensing. Institute for Ag. Professionals Research Updates. Kasson, MN. 8 Jan () 12. Kaiser, D.E Potential and limitations for nutrient management with remote sensing. Institute for Ag. Professionals Research Updates. Waseca, MN. 7 Jan (4) 13. Kaiser, D.E Corn response to nitrogen and nitrogen assessment tools. Crop Input Seminar, Hutchinson, MN. 4 Dec (65) 14. Kaiser, D.E Nitrogen management research studies. Sibley County Corn Growers Summer Plot Tour, Gaylord, MN. 6 Sept (2) 19

20 3 APPENDIX Gaylord Daily Rainfall (inches) /7/12 6/4/12 7/2/12 7/3/12 8/27/12 9/24/ Stewart Daily Rainfall (inches) /7/12 6/4/12 7/2/12 7/3/12 8/27/12 9/24/12 Figure A. Summary of daily total precipitation data from Gaylord and Stewart beginning May 7 and ending at harvest. Total precipitation was 1.91 inches at Gaylord and inches at Stewart. Data does not include all precipitation following fertilizer application. 2

21 Total Daily Precipitation (inches) Total Daily Precipitation (inches) 3 Janesville /5/213 7/5/213 8/5/213 9/5/ Willmar /5/213 7/5/213 8/5/213 9/5/213 Figure A (continued). Summary of daily total precipitation data from Janesville and Willmar beginning June 6 and ending at harvest in 213. Total precipitation was 17.6 inches at Janesville and inches at Willmar. Data does not include all precipitation following fertilizer application. 21

22 Total Daily Precipitation (inches) 3.5 New Richland 214 Total Daily Precipitation (inches) /5/213 7/5/213 8/5/213 9/5/213 3 Saint Charles /5/213 7/5/213 8/5/213 9/5/213 Figure A (continued). Summary of daily total precipitation data from New Richland and Saint Charles beginning on June 3 and ending at corn harvest in 214. Total precipitation was 16.3 inches at New Richland and 11.3 inches at Saint Charles. Data may not include all precipitation following fertilizer application. 22

23 Table A. Summary of significance of main treatments effects of nitrogen zone and rate and their interaction for field two field locations in Minnesota. Variable Location N Zone N rate Zone x Rate P>F Plant N Uptake Gaylord Stewart <.1.51 Janesville <.1.14 Willmar <.1.4 New Richland <.1 <.1 Saint Charles <.1.15 Cob N Uptake Gaylord Stewart Janesville <.1.2 Willmar <.1.23 New Richland <.1 <.1 Saint Charles <.1.2 Grain N Removal Gaylord <.1.38 Stewart <.1.1 Janesville <.1.13 Willmar <.1.12 New Richland <.1.3 Saint Charles <.1.41 Yield Gaylord <.1.99 Stewart <.1.4 Janesville <.1.14 Willmar <.1.19 New Richland <.1 <.1 Saint Charles <.1.9 NUE Gaylord Stewart Janesville <.1.88 Willmar New Richland Saint Charles