interpreting gis data and developing site-specific ... › presentations › 2008 ›...

2/8/20082/8/2008 Louisiana Agricultural Consultants Association Louisiana Agricultural Consultants Association 2008 Conference2008 Conference

11

Interpreting GIS Data and Interpreting GIS Data and Developing SiteDeveloping Site--Specific Specific Treatment Prescriptions Treatment Prescriptions for Precision Agriculture for Precision Agriculture

Applications Applications Kevin S. McCarterKevin S. McCarter

Department of Experimental StatisticsDepartment of Experimental StatisticsLouisiana State UniversityLouisiana State University

Eugene BurrisEugene BurrisNortheast Research Station

Louisiana State University AgCenter


22

Steps in Developing a Steps in Developing a VariableVariable--RateRate

Treatment PrescriptionTreatment Prescription

2/8/2008 Louisiana Agricultural Consultants Association 2008 Conference

3

Developing a VariableDeveloping a Variable--RateRateTreatment PrescriptionTreatment Prescription

1.1. Design a field trial for the purpose of Design a field trial for the purpose of obtaining data that will allow the obtaining data that will allow the comparison of treatments within the comparison of treatments within the various management zones present in the various management zones present in the field.field.

2.2. Perform the field trial according to the Perform the field trial according to the design to the extent possible. Note all design to the extent possible. Note all problems and deviations from the design problems and deviations from the design in performing the field trial.in performing the field trial.


4


3.3. Gather and process dataGather and process dataa)a) Obtain data from various sourcesObtain data from various sourcesb)b) Consolidate using GIS software.Consolidate using GIS software.c)c) Clean and possibly smooth data.Clean and possibly smooth data.d)d) Add additional variables that may be Add additional variables that may be

necessary for subsequent statistical analyses.necessary for subsequent statistical analyses.e)e) Dataset format Dataset format –– one observation per yield one observation per yield

point.point.f)f) Transfer data to statistician.Transfer data to statistician.


5


4.4. Statistical analysis of dataStatistical analysis of dataa)a) Develop appropriate model based on design Develop appropriate model based on design

and performance of the experiment.and performance of the experiment.b)b) Use model to compare treatments within the Use model to compare treatments within the

various management zones.various management zones.

5.5. Extract results of the treatment Extract results of the treatment comparisons necessary for building data comparisons necessary for building data structures for subsequent processing.structures for subsequent processing.


6


6.6. Build one or more treatment prescriptions. Build one or more treatment prescriptions. a)a) Identify actual or potential producer preferences.Identify actual or potential producer preferences.b)b) For each such preference, develop a preference For each such preference, develop a preference

specification and build a data structure that specification and build a data structure that implements that preference specification.implements that preference specification.

c)c) Combine the preference specification with the Combine the preference specification with the treatment difference information.treatment difference information.

d)d) Assign treatments to the field management zones.Assign treatments to the field management zones.


7


7.7. Supply prescriptions to producer / Supply prescriptions to producer / researcherresearcher

a)a) Graphs of the various treatment prescription Graphs of the various treatment prescription options.options.

b)b) Export each treatment prescription to a Export each treatment prescription to a csvcsvfile or other appropriate format.file or other appropriate format.


8


►► The statistical analysis upon which the treatment The statistical analysis upon which the treatment prescription development process is based is prescription development process is based is dynamic in nature and requires flexibility in its dynamic in nature and requires flexibility in its implementation:implementation:

depends on the experimental designdepends on the experimental designdepends on the data (e.g. nature of the covariates)depends on the data (e.g. nature of the covariates)requires statistical expertiserequires statistical expertisechanges as methodology improves and new software changes as methodology improves and new software becomes availablebecomes availableis therefore is therefore notnot a good candidate for automationa good candidate for automation


9


►►The steps involved in developing a The steps involved in developing a treatment prescription following the treatment prescription following the statistical analysis appears to be astatistical analysis appears to be a

systematic and relatively static processsystematic and relatively static processgood candidate for automationgood candidate for automation


1010

Software for Developing Software for Developing Treatment PrescriptionsTreatment Prescriptions


11

Software for Developing Treatment Software for Developing Treatment PrescriptionsPrescriptions

►►We have developed specifications for a We have developed specifications for a software system for creating treatment software system for creating treatment prescriptions.prescriptions.

►►We have implemented the specifications in We have implemented the specifications in SASSAS

►►Scripts have been developed to generate Scripts have been developed to generate SAS code that meets these specifications.SAS code that meets these specifications.

►►A variety of software could be used.A variety of software could be used.


12

Software for Developing Treatment Software for Developing Treatment PrescriptionsPrescriptions

►►Statistical analysis Statistical analysis -- SAS, SSAS, S--Plus / R, etc.Plus / R, etc.►►Subsequent steps in the processSubsequent steps in the process

Any database software, for exampleAny database software, for example►►MySQLMySQL►►AccessAccess►►OracleOracle

SASSAS►►Data step programmingData step programming►►Proc SQLProc SQL


1313

Developing a Treatment Developing a Treatment Prescription for a Prescription for a

Commercial Cotton FarmCommercial Cotton Farm


14

Turner Farm Field TrialTurner Farm Field Trial

►►Commercial cotton farming operationCommercial cotton farming operation

►►Embedded field trial conducted in 2006 to Embedded field trial conducted in 2006 to gather data for developing a variablegather data for developing a variable--rate rate nitrogen treatment.nitrogen treatment.


15


►►Field characteristic data Field characteristic data ElevationElevationSoil electroSoil electro--conductivity (EC)conductivity (EC)


16


►►ElevationElevationSpatially referencedSpatially referencedRanges from 39.66 to 41.03 ftRanges from 39.66 to 41.03 ft


17


18


►►ElectroElectro--conductivity (EC) measurementsconductivity (EC) measurementsUsed as a proxy for clay contentUsed as a proxy for clay contentEC_12 : Shallow EC (down to 12 in.)EC_12 : Shallow EC (down to 12 in.)EC_36 : Deep EC (down to 36 in.)EC_36 : Deep EC (down to 36 in.)Spatially referencedSpatially referenced


19


20


►►ElectroElectro--conductivity (EC) measurementsconductivity (EC) measurementsEC_ZoneEC_Zone: {(EC_12, EC_36)} : {(EC_12, EC_36)} →→ {1,2,3}{1,2,3}►►Defined by researchersDefined by researchers►►Classification by clay content:Classification by clay content:

EC_ZoneEC_Zone 1 : Lowest amount of clay1 : Lowest amount of clayEC_ZoneEC_Zone 2 : Medium amount of clay2 : Medium amount of clayEC_ZoneEC_Zone 3 : Highest amount of clay3 : Highest amount of clay


21


22

Turner Farm Field Trial LayoutTurner Farm Field Trial Layout

►►3 Reps 3 Reps ►►Nitrogen treatments randomly assigned to Nitrogen treatments randomly assigned to

plots within repsplots within repsStrips running entire length of field, as well asStrips running entire length of field, as well asEmbedded plots within such stripsEmbedded plots within such strips

►►Plots were 24 rows widePlots were 24 rows wide


23

Turner Farm Nitrogen TreatmentsTurner Farm Nitrogen Treatments

7090

120

150170

020406080

100120140160180

N1 N2 N3 N4 N5

Nitrogen Rate (pounds per acre)


24

►►Application pass 12 rows wideApplication pass 12 rows wide►►Nested within plotNested within plot►►2 application passes per treatment 2 application passes per treatment

plotplot

Nitrogen ApplicationNitrogen Application


25

►► 6 rows wide6 rows wide►► Harvest pass nested within application Harvest pass nested within application

passpass►► 2 harvest passes per application pass2 harvest passes per application pass

Harvest PassesHarvest Passes


26


►►Response variable Response variable Cotton lint yield Cotton lint yield Measured every 2 seconds Measured every 2 seconds Spatially referenced Spatially referenced Pounds per acrePounds per acreGIS software was used to estimate GIS software was used to estimate elevation and EC values at each yield elevation and EC values at each yield point locationpoint location


27

Statistical ModelingStatistical Modeling

►►Topological experimental design (TED)Topological experimental design (TED)Willers, Milliken, O’Hara, et. al. (2004)Willers, Milliken, O’Hara, et. al. (2004)

►►Statistically analyze using a linear mixed Statistically analyze using a linear mixed model analysis of covariance incorporating model analysis of covariance incorporating spatial componentsspatial components

Willers, Milliken, O’Hara, et. al. (2004)Willers, Milliken, O’Hara, et. al. (2004)

►►Implement using SAS Proc MixedImplement using SAS Proc MixedLittellLittell, Milliken, Stroup, et. al. (2006), Milliken, Stroup, et. al. (2006)


28

Preferences SpecificationsPreferences Specifications

►► Preference Specification 1Preference Specification 1Top ProducerTop ProducerAt each level of At each level of EC_ZoneEC_Zone and Elevation, and Elevation, choose the N rate that has the highest choose the N rate that has the highest lsmeanlsmean..

►► Preference Specification 2Preference Specification 2Willing to drop up to 1 rate levelWilling to drop up to 1 rate level


29

Preferences SpecificationsPreferences Specifications

►► Preference Specification 3Preference Specification 3Willing to drop up to 2 rate levelsWilling to drop up to 2 rate levels

►► Preference Specification 4Preference Specification 4At each level of At each level of EC_ZoneEC_Zone and Elevation, and Elevation, choose the lowest N rate from the class of N choose the lowest N rate from the class of N rates that are not significantly different than rates that are not significantly different than the top producer.the top producer.Can also be called the “willing to drop up to 4 Can also be called the “willing to drop up to 4 N Rate levels” preference.N Rate levels” preference.


30

Treatment Prescription 1Treatment Prescription 1


31



32



33



34

Turner FieldTurner FieldTreatment Prescription OptionsTreatment Prescription Options


3535

Prescription Based on Prescription Based on Helena 2007 Field TrialHelena 2007 Field Trial


36

Thank You!


37

Author InformationAuthor Information

►►Kevin S. McCarterKevin S. McCarterDepartment of Experimental StatisticsDepartment of Experimental StatisticsLouisiana State UniversityLouisiana State UniversityBaton Rouge, Louisiana 70803 Baton Rouge, Louisiana 70803

►►Eugene BurrisEugene BurrisNortheast Research StationNortheast Research StationLouisiana State University AgCenterLouisiana State University AgCenterP.O. Box 438P.O. Box 438St Joseph, Louisiana 71366St Joseph, Louisiana 71366


38

BibliographyBibliography►► LittellLittell, Ramon C., George A. Milliken, Walter W. Stroup, Russell D. , Ramon C., George A. Milliken, Walter W. Stroup, Russell D. WolfingerWolfinger, and Oliver , and Oliver

SchabenbergerSchabenberger. . SASSAS®® for Mixed Models, Second Edition. for Mixed Models, Second Edition. Cary, NC: SAS Institute Inc. Cary, NC: SAS Institute Inc. 2006.2006.

►► McCarter, K.S, Burris, E., Milliken, G.A., Clawson, E.L., Wong,McCarter, K.S, Burris, E., Milliken, G.A., Clawson, E.L., Wong, H.Y., H.Y., WillersWillers, J.L., J.L.►► Specifications Of A Prototype Software System For Developing VarSpecifications Of A Prototype Software System For Developing Variableiable--Rate Treatment Rate Treatment

Prescriptions For Use In Precision AgriculturePrescriptions For Use In Precision Agriculture. Proceedings of the Nineteenth Annual . Proceedings of the Nineteenth Annual Kansas State University Conference on Applied Statistics in AgriKansas State University Conference on Applied Statistics in Agriculture, April, 2007, culture, April, 2007, Manhattan, Kansas. Ed. J. Boyer. 2007. Manhattan, Kansas. Ed. J. Boyer. 2007.

►► Milliken, George A., Dallas E. Johnson. Milliken, George A., Dallas E. Johnson. Analysis of Messy Data, Volume III: Analysis of Analysis of Messy Data, Volume III: Analysis of CovarianceCovariance. Chapman & Hall / CRC. Boca Raton. 2002. . Chapman & Hall / CRC. Boca Raton. 2002.

►► SchabenbergerSchabenberger, Oliver, Francis J. Pierce. , Oliver, Francis J. Pierce. Contemporary Statistical Models for the Plant Contemporary Statistical Models for the Plant and Soil Sciencesand Soil Sciences. Taylor & Francis. Boca Raton. 2002.. Taylor & Francis. Boca Raton. 2002.

►► Willers, J.L, G. A. Milliken, C. G. O’Hara, and J. N. Jenkins. Willers, J.L, G. A. Milliken, C. G. O’Hara, and J. N. Jenkins. Information Technologies and Information Technologies and the Design and Analysis of Sitethe Design and Analysis of Site-- Specific Experiments within Commercial Cotton FieldsSpecific Experiments within Commercial Cotton Fields. . Proceedings of the Sixteenth Annual Kansas State University ConfProceedings of the Sixteenth Annual Kansas State University Conference on Applied erence on Applied Statistics in Agriculture, April 25Statistics in Agriculture, April 25--27, 2004, Manhattan, Kansas. Ed. G. A. Milliken. 2004. 27, 2004, Manhattan, Kansas. Ed. G. A. Milliken. 2004.

►► Wong, Hoi Yee. 2007. Wong, Hoi Yee. 2007. Incorporating GIS (Geographic Information System) to Incorporating GIS (Geographic Information System) to Characterize Spatial Variability to Analyze the Effectiveness ofCharacterize Spatial Variability to Analyze the Effectiveness of SiteSite--Specific Management Specific Management (SSM) and to Generate Prescription Maps. (SSM) and to Generate Prescription Maps. Seminar, Department of Experimental Seminar, Department of Experimental Statistics, Louisiana State University. April 11, 2007.Statistics, Louisiana State University. April 11, 2007.

interpreting gis data and developing site-specific ... › presentations › 2008 ›...

Documents