Sign up for email alerts of new Fluid Journal issues!
Fluid Journal : Spring 2011
11 The Fluid Journal Spring 2011 Dr. Shaver is an Assistant Professor in the Department of Agronomy and Horticulture, University of Nebraska-Lincoln. Dr. Khosla and Dr. Westfall are Professors in the Department of Soil and Crop Sciences, Colorado State University. which maize growth stage each sensor performed best. Our results suggest that both the red and amber sensors performed equally in the determination of N variability in maize and that each performed best at the V12 maize growth stage. Therefore, our objectives for this study were: to develop and test an in-season N recommendation algorithm based on NDVI for the red and amber sensors for use at the V12 maize growth stage, which can then be used and further evaluated in farmers' fields across Colorado. Methodology Study site was the Agricultural Research Development and Education Center (ARDEC) at Colorado State University. The site was furrow irrigated, continuous maize, and classified as a fine-loamy, mixed, super-active, mesic, Aridic Haplustalf. Two different locations within the same field were used for this study, resulting in two site years. Sensors included GreenSeekerTM (red) and Crop CircleTM (amber). Sensor readings were collected across four N application rates at the V12 maize growth stage for site years 1 and 2. Sensors were mounted on a telescoping boom, allowing readings to be collected at the proper height above the maize canopy. The red sensor was connected to a Compaq IpaqTM to record NDVI values. The amber sensor was connected to a GeoSCOUT GLS 400 data logger to record all NDVI values. N application. Nitrogen was applied as 32-0-0 urea-ammonium-nitrate (UAN) at maize emergence (no preplant N), using a 4-row sidedress applicator with variable rate capabilities. This applicator applied liquid N below and to the side of the maize plant. The N was applied as close to a scheduled irrigation event as possible to reduce potential N losses due to volatilization. Four N rates were applied: 0, 50, 100, and 175 lbs/A. Plot design. Subplots of each N application rate were set up at two different locations (site years 1 and 2) at ARDEC and each N rate was replicated four times at each site year in a complete randomized block (CRB) design. This resulted in 16 subplots within each site year. Each plot was 4 maize rows in width (30-inch row spacing) and 50 feet long. Site years 1 and 2 had not received applied N for two years prior to this study. N algorithm. The NDVI-based N estimation algorithms for the amber and red sensors were created by using the maize growth stage V12 NDVI readings from the 0, 50, 100 and 175 lbs/A N plots in site years 1 and 2. The algorithm was created at the V12 growth stage because maximum N variability was recorded by the sensors at the V12 growth stage. Additionally at V12 the maize is still small enough to allow N application implements into the field. The overall idea with this algorithm is that an RI can be based on N application differences. If we know the difference in N application rates and the resulting RI, this information can be plotted and an N prediction equation can be formulated through linear regression. Data analysis. All statistical analysis was performed using the Statistical Analysis System (SAS). All regressions were performed using the REG procedure in SAS. The bootstrapping process was accomplished using a bootstrapping macro in SAS. Proc MEANS was used for all means calculations and the CLM option was used in Proc MEANS for all confidence interval calculations. Analysis of variance was performed using Proc Anova. Results promising Yield. Grain yield was significantly increased by applied N fertilizer in both site years. Yields were highest in site-year 1 relative to site-year 2, and the 175 and 100 lbs/A N application rates produced equal yields, suggesting that the 100-lb/A rate supplied sufficient N for maximum yield. All applied N rates yielded significantly more than the check. Yields in site-year 2 were similar to those in site- year 1, and again, the 175- and 100-lb/A Table 1. Amber and red NDVI algorithm-based N recommedations at maize growth stage V12 across 4 N application rates for two site years. N applied at corn emergence (lbs/A) 0 50 100 175 Site Year 1 N recommended by amber algorithm 130 28 13 6 N recommended by red algorithm 124 28 14 8 Site Year 2 N recommended by amber algorithm 122 29 8 5 N recommended by red algorithm 116 26 7 7 N rates produced the same yield. This again suggests that the N sufficiency level was reached at the 100 lb/A rate. Algorithm premise. Previous studies suggest that both sensors perform equally well. The N recommendation algorithms (NRA) were therefore developed for each sensor using the same methodology. As with other NRA our amber and red sensor NRA were based on an RI. One method for determining RI is presented in Equation 2 (see earlier, above, in introduction). Equation 2 is also the format used to determine RI in the algorithms presented in this article. An RI was calculated over a range of N application differences (175, 100, and 50 lbs/A) and then was regressed over the N application difference that created that particular RI. This regression was then used to calculate an N recommendation quadratic equation that predicts crop N need for the amber and red (Figure 1) sensors. The premise for the algorithm methodology we used was that RI is directly related to N differences in the crop. The RI can, therefore, be used to predict the amount of N it would take to make up this difference, which can be used as the N recommendation. Our data clearly show that the amber and red sensor algorithms are unbiased and are a sound methodology for determining NDVI-based N recommendation algorithms (Table 1). This process represents a good first step for algorithm development in Colorado. "Crop sensors help calculate nutrient recommendations"
Early Spring 2011