|Scharf, P - UNIVERSITY OF MISSOURI|
|Shannon, D - UNIVERSITY OF MISSOURI|
|Palm, H - UNIVERSITY OF MISSOURI|
Submitted to: European Conference on Precision Agriculture Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: December 2, 2008
Publication Date: N/A
Technical Abstract: The nitrogen (N) supplying capacity of the soil available to support corn (Zea mays L.) production can be highly variable both among and within fields. Thus, the amount of N fertilizer applied should be site-specific and also climate-sensitive. In recent years, canopy reflectance sensing has been investigated for in-season assessment of crop N health. Multiple groups have conducted research leading to variable-rate N recommendation algorithms based on this information. To date, it has proven difficult to define a single algorithm that works well under a wide variety of conditions. This may be due to inherent variability in soils, climatic conditions, or other factors that the existing algorithms do not sufficiently represent. The objective of this research was to determine whether incorporating auxiliary information (e.g., soil texture, landscape position, weather variables) could improve the accuracy of canopy reflectance-based N recommendations. A total of 16 field-scale experiments were conducted over four growing seasons (2004-2007) in three major soil areas of Missouri, USA: river alluvium, deep loess, and claypan. Multiple blocks of randomized N rate response plots traversed the length of the field. Each block consisted of 8 N treatments from 0 to 235 kg N/ha on 34 kg N/ha increments, top-dressed between vegetative growth stages V7 and V11. Adjacent to the response blocks, N-rich (235 kg N/ha) reference strips, along with strips of the usual producer practice and of candidate N algorithms were established. These ran the full length (400 to 800 m) of the field, and were replicated from 3 to 6 times. Crop canopy reflectance sensor measurements were obtained from the N response blocks and adjacent treatment strips at the time of top-dress N application. Grain yield was measured with a yield-monitor equipped combine. Other available data included soil electrical conductivity (EC), topographic attributes calculated from RTK-DGPS data, growing-season weather, and aerial imagery. In a previous study, we fit quadratic-plateau functions to estimate yield response to N rate for the response blocks and then determined economically optimum N recommendation algorithms for various corn and nitrogen prices. In this paper, we develop modified algorithms incorporating soil EC, topography, weather, and/or remote sensing data and compare them to those baseline algorithms. Analyses consider these auxiliary data as (1) modifiers in calculating a baseline N-rich reflectance reading for the field, (2) classifiers in the development of sub-field management zones within which separate algorithms may apply, and (3) potential covariates in the algorithm itself. Preliminary results have shown an advantage to using auxiliary information such as soil EC in recommendations on the more highly variable river alluvium fields. This research is expected to provide an increased understanding of why canopy reflectance-based N recommendations are of varying quality among fields and insight into how such recommendations can be improved.