ABSTRACT
Fuzzy logic inference systems (FISs) can help provide within-eld nitrogen (N) fertilization recommendations by combining critical plant-and soil-based spatial information. This chapter describes how, based on spatially distributed information, FIS can be used to develop in-season N recommendations. A sample problem is provided. Soil and plant information considered in this analysis included apparent soil electrical conductivity (ECa), elevation (ELE), and the remote sensing-based N sufciency index
6.1 Executive Summary ...................................................................................... 101 6.2 Introduction .................................................................................................. 102 6.3 Materials and Methods ................................................................................. 103
6.3.1 Extracting the Field Parameters ....................................................... 103 6.3.2 NDVI and Calculating the NSI ......................................................... 104 6.3.3 Background Knowledge about Soil and Plant Status Needed
to Determine N Needs ...................................................................... 105 6.3.4 FIS for Estimating Spatial N Needs ................................................. 106
6.3.4.1 Design of the FIS ............................................................... 108 6.3.5 Step-by-Step Exercises Using ArcGIS 9.2 ........................................ 109
6.4 Results ........................................................................................................... 118 6.5 Conclusions ................................................................................................... 119 Acknowledgment ................................................................................................... 119 References .............................................................................................................. 119
(NSI = NDVIsample/NDVIwell-fertilized reference). Expert knowledge for formulating fuzzy rules was developed from corn growth data following an in-season N application. The best mid-season growth response to in-season N occurred in areas of low ECa and high ELE. Under these favorable soil conditions, maximum mid-season growth was obtained without in-season N irrespective of the NSI values. Where soil conditions were less favorable (i.e., high ECa and low ELE), mid-season growth beneted from high in-season N rate only when NSI was low. These relationships were modeled using a simple FIS having three inputs (ECa, ELE, and NSI) fuzzied with only two sets (low and high), an output (optimum N rate) with three fuzzy sets (low, medium, and high) and a set of eight simple rules. The FIS appeared to be a useful and handy tool for incorporating expert knowledge into spatially variable N recommendations. An example describing a basic implementation of the FIS in ArcGIS is included.
