Optimization of crop canopy sensor placement for measuring nitrogen status in corn

Authors: ROBERTS, DARRIN - UNIVERSITY OF NEBRASKA; ADAMCHUK, V - UNIVERSITY OF NEBRAKSA; Shanahan, John; FERGUSON, RICHARD - UNIVERSITY OF NEBRASKA; SCHEPERS, JAMES - ARS - RETIRED

Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/15/2008
Publication Date: 1/15/2009
Citation: Roberts, D.F., Adamchuk, V.I., Shanahan, J.F., Ferguson, R.B., Schepers, J.S. 2009. Optimization of crop canopy sensor placement for measuring nitrogen status in corn. Agronomy Journal. 101:140-149.

Interpretive Summary: Current nitrogen (N) fertilizer management strategies for corn production systems in the U.S. are characterized by low N use efficiency (NUE) and environmental contamination. Alternative strategies are required to sustain corn based farming. One approach being investigated is the use of active crop canopy sensors to assess corn N status and direct spatially-variable in season N application as a means to improve NUE. We have been working with the Crop Circle model ACS-210 active sensor manufactured by Holland Scientific of Lincoln, NE. The sensor generates its own source of modulated light (pulsed at ~ 40,000 Hertz) in two wavebands (amber color and near infrared) using a single bank of light emitting diodes (LED), and measures the percent of modulated light reflected back from the crop canopy, using another bank of LED. While our previous research has shown potential for using this approach to manage N applications, little research has been conducted to establish the optimal spatial scale for sensing and N application to corn. Because of likely high costs associated with active sensors and control equipment to vary N rates for individual rows, there is need to identify an optimal strategy for sensor placement on application booms. Therefore, the objective of this study was to determine optimal sensor spacing for controlling whole- and/or split-boom N application scenarios for a hypothetical 24 row applicator. Sensor readings were collected from 24 consecutive rows at eight cornfields during vegetative growth in 2007 and 2008, and readings were converted to chlorophyll index (CI) values, a measure of canopy N status. A base map of measured CI values was created using square pixels equivalent to row spacing for each site (0.91 or 0.76 m in size). It was assumed that this base map represented the finest spatial resolution for sensor measurements and provided the greatest spatial detail for prescribing N applications. Every map obtained using a reduced sensor number was evaluated against this “base” map. Sensor placement and boom section scenarios were evaluated using MSE (mean squared error) of calculated CI maps vs. the base CI map. Scenarios ranged from one sensor, one variable-rate to 24 sensors, 24 variable-rates for the hypothetical 24-row applicator. In this study, the ability to minimize the number of canopy sensors used to measure chlorophyll index (CI) in corn depends on the site. An average of 2-3 sensors should be an acceptable approach to obtain a single application rate for the entire boom, assuming the boom is not longer than the width of our study areas (22 m). Three of eight sites indicated a potential benefit of splitting the boom into three sections with 2-3 sensors per section. Relatively high row-to-row variability in some fields signified low predictability of sensor measurements obtained from a neighbor row. In fact, due to management-induced systematic patterns, we found sensor measurements from rows equidistant from the center of an 8-row planter had more similarities than rows next to each other. The ability to model this variability provided some improvement over a single rate approach. Relatively low variability in the direction of travel measured in 5 of the 8 measured fields caused varying N rate, based on the averaged or modeled CI prediction from a few sensors, to be inappropriate. However, significant variability in the other three fields suggested potential benefit to site-specific N management practices based on active crop canopy reflectance sensors. The practical implications for this study suggest that spatial variability of sensor readings can differ significantly both within and between fields and a producer could possibly benefit by being prepared to manage small-scale variation (i.e. multiple boom sections with one to multiple sensors per section).

Technical Abstract: Active canopy sensors can be used to assess corn (Zea mays L.) nitrogen (N) status and direct spatially-variable in season N application. The goal of this study was to determine optimal sensor spacing for controlling whole- and/or split-boom N application scenarios for a hypothetical 24 row applicator. Sensor readings were collected from 24 consecutive rows at eight cornfields during vegetative growth in 2007 and 2008, and readings were converted to chlorophyll index (CI) values. A base map of measured CI values was created using square pixels equivalent to row spacing for each site (0.91 or 0.76 m in size). Sensor placement and boom section scenarios were evaluated using MSE (mean squared error) of calculated CI maps vs. the base CI map. Scenarios ranged from one sensor, one variable-rate to 24 sensors, 24 variable-rates for the hypothetical 24-row applicator. The greatest reduction in MSE from the one variable-rate scenario was obtained with 2-3 sensors estimating average CI for the entire boom width, unless each row was individually sensed. Five of eight fields received more accurate prediction of CI by averaging sensor readings across the entire 24 rows rather than predicting CI for several consecutive rows using only one sensor. Results for a 24-row boom (~22 m) indicate there is no value in using more than 2-3 sensors to estimate average CI. Due to spatial variability in CI, some fields may benefit from an increased number of sensors and/or boom sections.