6.1   Explaining Spatial Variability in Grain Yield

Principal Investigators

Tom Mueller, Assistant Professor, Agronomy
Scott A. Shearer, Associate Professor, Biosystems and Agricultural Engineering
Tasos Karathanasis, Professor, Agronomy
Ed Perfect, Assistant Professor, Agronomy
Ken Wells, Extension Professor, Agronomy
Lloyd Murdock, Extension Professor, Agronomy

Don Nielsen, Professor, Land, Air and Water Resources, University of California, Davis
Richard Barnhisel, Professor, Agronomy
Haluk Cetin, Mid-America Remote Sensing Center, Murray State University
Steve Higgins, Research Specialist, Biosystems and Agricultural Engineering
Carl Dillon, Associate Professor, Agricultural Economics
Than Hartsock, Graduate Research Associate, Agronomy

Cooperators

Mike Ellis, Farmer, Worth and Dee Ellis Farms, Shelby County, Kentucky
Charles Stuecker, Farmer, Stuecker Farms
Rick Murdock, Farmer, Ponderosa Farms

Introduction

The number of yield monitoring systems in Kentucky is growing yearly because producers recognize the great value of yield maps.  Nevertheless, growers have become frustrated because they lack the tools and knowledge base to fully utilize multiple years of archived yield map data.  The potential is great potential for farmers to utilize yield maps to make better agronomic decisions which can improve the economy and efficiency of agriculture in Kentucky.

There are three steps for utilizing yield maps.  The first is describing variability.  Yield maps allow producers to visualize and quantify variability within agricultural fields.  Profit maps can be calculated from yield maps in order to describe the spatial variability of economic gains and losses within agricultural fields.  The second step is understanding the causes of yield variability.  Spatial variability in grain yield is due in part to soil physical, chemical, biological, and morphological properties.  In addition, topography, pests (e.g. weeds, insects, wildlife), micro variations in weather, and genetic differences in seed cause yield variability within agricultural fields.  Perceived yield variability may be due to errors introduced by yield monitoring equipment.  The third step is managing yield variability.  This may include variable rate application of fertilizers, pesticides, irrigation, and seeds.  In addition, site-specific tillage may be used in areas where compaction is a problem.  And corn hybrids and soybean varieties may be matched to specific soils within a given field.

The greatest challenge for utilizing yield maps is the second step, understanding the causes of the variability.  Because of the costs involved with sampling, producers can only measure those factors that are expected to have the greatest impact on yield and the factors that are manageable.   The problem is with the factors that limit yield within any particular field, farm, or physiographic region of Kentucky vary from year to year and are not known prior to sampling.

Objectives

This portion of the overall investigation has two objectives:

1)      To develop an information base for use in yield map interpretation based on the relationship between grain yield, topography, soil physical, chemical, morphological characteristics, precipitation, and pest patterns in three Kentucky agricultural fields for two and a half years; and

2)      To develop an economically feasible approach for producers to create management opportunity maps.  This map will identify the yield limiting factors for each map unit.  The limitations may or may not be manageable.  If they are manageable, the producer must decide if the management changes would be cost effective. 

Background

A number of spatial studies have shown little correlation between grain yield and soil chemical properties (e.g. Everett and Pierce, 1996; Blackmore et al. 1999) although not without exception (e.g. Johnson et al., 1999).  Crops often do not respond to soil fertility for two reasons: 1) soil nutrients may be above sufficiency levels where a crop response is expected, or 2) soil water or other factors may be even more limiting than nutrients, thereby driving yield variability.  Yield variability often relates to various other factors that relate to plant available water i.e., landscape position (e.g. Solohub et al., 1996; Karen et al., 1999; Timlin et al, 1999), soil aggregate size (Cambardella et al. 1996), and soil water storage precipitation (Runge and Hons, 1999).  Often, researchers have found that crop yield is not temporally stable (Runge and Hons, 1999) and this is likely due to annual weather patterns and to differences in crops.  Some researchers have used simple regression or correlations to explain yield variability (Khakural et al., 1996; Lamb, et al., 1996; Braum, et al., 1999; Khakural, et al., 1999) while others have used multiple regression approaches (Sudduth, et al., 1996; Cambardella et al., 1996; Blackmore, et al., 1999; Khakural, et al., 1999; Heiniger, 1996).  A survey of the factors that control yield variability in Kentucky will provide an information base that will be useful for yield map interpretation.

To create management opportunity maps that are economical, a map unit should be defined as a homogeneous productivity zone rather than a homogeneous soil unit (e.g. soil map unit).  Historical yield maps are used to define these productivity zones. 

The next step is to determine probable causes of this variation.  There are already some existing spatial data bases available on the Internet that will be useful in this regard for Kentucky agricultural fields (e.g. NRCS digital soil surveys, USGS DEM’s, digital orthophotos, digitized topographic maps; http://www.state.ky.us/agencies/finance/depts/ogis/gisdept.htm).  Additional information (e.g. soil electrical conductivity sensor data, remote sensed imagery, high resolution DEM’s) may be used if they are found to be related to factors that influence yield variability (e.g. soil physical properties, pest damage) and if they are not cost prohibitive.  Producers and equipment operators usually have a great deal of insight into the causes of yield variability (e.g. planter or herbicide applicator failure, compaction near field entrances, etc.) and should be consulted by a crop/soil specialists in generating the management opportunity maps. 

The next step is field measurements and sampling.  This includes composite soil fertility sampling within each productivity zone.  Depending on the results from Objective 1, soil survey evaluations may also be required (soil structure, testing for compaction, texture by the feel method, soil drainage, depths to limiting layers: hardpans, bedrock, fragipans).  These measurements will require training in soil survey techniques.

Once the causes of variation are understood, the variability can be managed.  This may include variable rate fertilization, seeding, or pesticide application.  It may also include site-specific tillage or seeding.  The University of Kentucky can provide data from field trials to help make these management decisions, but some recommendations may need to be made at the site-specific level.  Yield monitors may be useful in this regard.  They can be used to make side-by-side management comparisons and to generate profit maps for cost effective decision-making.

Procedures

Procedures for Objective I - Steve Higgins, a doctoral student in Agronomy who is also a Research Specialist in Biosystems and Agricultural Engineering will be responsible for this objective.  Scott Shearer has initiated this research for a field in Shelby County during the 1999 growing season.  There are 3 phases to this investigation.

Phase 1) Relate yield with exhaustive landscape information. Elevation data is being collected at a high intensity for several fields using Real-Time Kinematic GPS where multiple years of yield map data exists.  Digital elevation models (e.g. Fig 1) will be created for each field as part of a USDA sponsored project.  Topographic attributes will be extracted (e.g. slope, curvature, aspect, wetness indices) and related with grain yield maps.  Multiple regression will be used to determine the percentage of yield variability explained by topographic attributes.

Phase 2) Relate grain yield to soil measurements collected on a grid.  Grid sampling was done on several fields on a 100-ft grid (30.5-m; e.g. Fig. 2) at three locations in Kentucky (Shelby Co., Hardin Co., Calloway Co.), and several other fields will be sampled at lesser intensities.  Soil test analysis will be done by UK’s division of Regulatory Services for pH, BpH, P, K, Ca, Mg, CEC, and organic matter (OM).  In addition, particle size, topsoil depth, and depth to bedrock will be measured at each grid point.  Multiple regression will be used to describe the relationship between grain yield, topographic attributes, and the information described above at each grid point.  Plots of predicted versus measured will be used to visually assess the quality of prediction.

Phase 3) Relate grain yield to information at points of interest.  A transect study along topographic and fertility gradients will be established with between 50 and 75 sample points at each of three locations.  At each sample point, soil samples will be collected for soil fertility analysis.  Additional measurements include, volumetric soil water content (top 40-in; weekly during growing season), soil depth (if less than 60-inches), depth to compacted layers (e.g. plowpan, fragipan), soil structure, and pest damage (i.e. weeds, insects, diseases, wildlife).  This data will be analyzed with multiple stepwise regression, state space analysis (with the assistance of Professor Donald Nielsen, University of California, Davis), boundary layer analysis, principle components analysis, and canonical correlations, and with neural network classifiers.

Procedures for Objective II - The second objective will be conducted by a MS student in Agronomy, co-advised by Tom Mueller and Tasos Karathanasis.  We are requesting funding for this graduate student assistantship for 3 years beginning in August of 2000.  In addition to course work, the student will have the following  responsibilities: 1) devising and testing an approach for creating management opportunity maps, 2) co-authoring extension publications, 3) co-authoring refereed scientific publications, 4) writing a thesis, and 5) assist in seeking additional research funding for utilizing yield maps in Kentucky (specifically managing yield variability).  We believe that responsibilities involved justify a three-year masters program.  The model for creating management opportunity maps will have two components, 1) zone delineation and 2) explanation of limitations to grain yield.

 

Zone Delineation - There are a number of approaches for creating management zones including: 1) crop yield, 2) 2nd order soil surveys available for most counties from the NRCS, 3) intensive first order soil surveys; 4) electrical conductivity zones, 5) remote sensed imagery (Pocknee et al., 1996).  It is our hypothesis that zones based on historical grain yield maps in conjunction with 2nd order soil surveys will be most efficient for creating management opportunity maps.  To test this premise, we define efficient management zones as having more variability (in soil fertility or grain yield) between management zones than within management zones.  This principle will be statistically tested with an F test (Steel and Torrie, 1980).  The five strategies for creating management zones will be tested and compared to our proposed method.  The most efficient method will be used for explaining limitations to grain yield.

Explaining Limitations to Grain Yield - We suggest a three step approach for explaining yield variability.  1) Gathering spatial information: free digital spatial data bases include digital second order survey information, USGS DEM’s, digital ortho quads and more expensive data sets which may or may not be cost effective: IR aerial images of crops, high resolution digital elevation models (e.g. Fig 1), and measures of soil electrical conductivity.  These may or may not be cost effective for farmers.  2) Soil fertility sampling: composite soil samples should be collected from each productivity zone.  3) limited scope 1st order soil survey: A set of soil survey measurements will be taken for each productivity zone.  These measurements may include depth to limiting layers (hardpans, perched water tables, bedrock), soil texture, internal drainage.  The measurements that we will recommend to farmers will also depend on the results from Objective 1.

Expected Benefits

We will provide producers in Kentucky 1) an information base for yield map interpretation, and 2). a set of techniques for creating management opportunity maps.  This information will be disseminated through a number of avenues: 1) cooperative extension publications, 2) UK field days, 3) internet, 4) scientific literature, and 5) classroom.  If UK students graduate with these skills, it would be of tremendous value to producers in the state of Kentucky.  Tom Mueller will teach these techniques to students in his “Soil Use and Management” course.  Scott Shearer, Tom Mueller, and Carl Dillon will teach these techniques to students in a new course: “Spatial Applications in Agriculture”.

Deliverables

I. An information base for yield map interpretation, II. a set of techniques for creating management opportunity maps, III. extension publications presenting this information (available on the internet), IV. course material for two courses at the University of Kentucky (Soil Use and Management and Spatial Applications in Agriculture), V. field days, and VI. refereed journal publications.