Precision Farming Data Management Using Geographic ......farming, prescription farming, or precision farming, the term to be used in the remainder of this paper, apply a combina-tion

PEER.REVIEWED ARI ICTE

Precision Farming Data Management UsingGeographic Information Systems

E. Lynn Usery, Stuart Pocknee, and Broughton Boydell

AbstractAgricultural ecosystems are inherently variable entities. Tomonage spatial variabilitSr, modern farmers are looking foradvanced technological solutions. Management strategies in-corporuting rcmote sensing, the Global Positioning System(cps), and variable rate beatment (var) offer the possibilitvof positioning inputs exactly in order to optimize'farm ,e-turns and minimize chemical inputs and environmental haz-ards. The use of geographic information systems (cts) isessential for such a management paradigm. This paper intro-duces the concept of precision farming and discusses therole of GIS as a centralized data management and analysistool. Results from a survey of the use of cts in precisionfarming are included in order to determine strengths andweaknesses of the technology and to provide impetus for im-provements in GIS to support agricultural applications.

IntroductionInnovative agricultural techniques known as site-specificfarming, prescription farming, or precision farming, the termto be used in the remainder of this paper, apply a combina-tion of new technologies to improve production and reduceenvironmental pollution. Taking advantage of recent devel-opments including the Global Positioning System (ces), re-mote sensing, geographic information systems (cIs) andvariable rate treatment (vRT), precision farming is used tomanage spatial variability in fields through determination ofspatially referenced inputs, such as nutrients which affectsoil fertility and chemical applications which control insectand weed pests (Chancellor and Goronea, 1994). The resultis optimized production with minimum inputs of chemicalsand a corresponding minimum impact on the environment.The precision farming approach has been used in the U.S.Midwest to measure yield variations in wheat and corn(Pringle ef o1., 1993) while the National EnvironmentallySound Production Agriculture Laboratory (Nnsear.) in Geor-gia has developed technology for making similar yield meas-utements for peanuts and cotton in the Southeast. Thepotential to use cIS to integrate the various types of informa-tion needed to manage and control crop production on a site-specific or within-field approach is rapidly being researchedand developed (Schueller, 1992). The potential benefit of theintegration of these technologies to improve agricultural pro-duction while simultaneously reducing environmental degra-

E.L. Usery is with the Department of Geography, Universityof Georgia, Athens, GA 30602-2502.

S. Pocknee and B. Boydell are with the Department of Cropand Soil Science, University of Georgia, Athens, GA 30602-2502.

PE&RS

dation may be one of the greatest contributions of cls tohuman populations.

It is the purpose of this paper to explore cIS as a tool forintegration and analysis of precision farming data and topresent a brief survey of the use of crs in precision-farmingresearch and applications. The next section of this paper pro-vides a brief review of developments in the maior technolo-gies associated with precision farming, i.e.. GpS, continuous-yield sensors, remote sensing, vRt, and GIS. The third sectionprovides an examination of cIs as the hub of orecision-farm-ing data management and analysis as being implemented bythe authors and others at NESPAL. The fourth section presentsresults of a questionnaire survey of precision-farming re-searchers and practitioners concerning their use of GIS, itsstrengths and its weaknesses. The final section draws conclu-sions concerning precision-farming applications of cIs andpresents areas for future work by cIS researchers to betterserve this growing application area.

Precision FamingIdeas of within-field variability, which are the basis of preci-sion farming, surfaced as early as 1929 with approachei tomeasuring the spatial variability of soil acidity (Linsley andBauer, 1929). The modern manifestation of the concept is aresult of environmental awareness and economically viabletechnological innovations which allow global positioning,precision application of variable inputs, and measurement ofvariable yields (Goering, 1993). Precision farming can be rep-resented as incorporating three main areas of management(Figure 1). Data pertaining to yield and potential yield-affect-ing factors are initially collected. These data are then ana-lyzed to determine whether or not yield is affected by thefactors studied. If yield is being affected, a farm manager de-cides the type, distribution, and amount of treatment to ap-ply. Remedill measures can then be undertaken in such away that the correct treatment is applied at the required rateand to the appropriate area within a field. Evaluation oftreatment effects can then begin and this essentially bringsthe manager back to the start of the cycle. Economic analysisof treatment measures and yields is critical to successfulmanagement. Variable application of inputs may not increaseyields but simply hold them constant while reducing inputcosts, Thus, the farmer reaps increased profit through bettermanagement and application of less chemical treatmentwhich also helps preserve the environment. An immensevolume of information is collected and interpreted during

Photogrammetric Engineering & Remote Sensing,Vol . 61, No. 11, November 1995, pp. 1383-1391.

009s-1 1 12l95/61 1 1-1 383$3.00/0O 1995 American Society for Photogrammetry

and Remote Sensing

PEER.REVIETVED ARI ICTE

Figure 1. The conceptual com-ponents of precision farming.

this process. If this management method is to be successful,a suitable data storage, analysis, and decision support systemmust be implemented.

Key TechnologiesTechnologies that have advanced the introduction of preci-sion farming include cPS, continuous data sensors, remotesensing, vRT, and GIS. Each of these technologies is brieflyexamined below.

GPSLocation expressed in geographic coordinates - latitude andIongitude - can be determined with cPS to accuracies betterthan + 1o0 m. By using a known base source position andsignal, differential positioning (ocps) can attain accuracies inreal time of -r 1. m and with sufficient time - severalminutes to several hours at a collection point - accuraciesof a few centimetres are attainable. Precision farming hasmany operations that rely on immediate and precise knowl-edge of field location and ncps is an ideal approach to ac-quiring that knowledge. For example, to link vRr of fertilityto a desired geo-referenced fertilizer application map, the op-erator must know field location and map location simultane-ously. From the map location the operator can identify thecorrect application rate for the current field position (Del-court and Baerdemaker, 1994). Similarly, spatially variabletreatments have been tested for control of pests from pestmaps (Schueller and Wang, 1994a). Yield mapping combinesthe accurate location information of nGpS with the results ofa variable-flow rate (yield) sensor. The resulting yield-varia-bility map can then be used to spatially locate high- andIow-yielding areas of managerial interest for furtherinvestigation (Aurenhammer ef o/,, 1994).

Data CollectionPrecision farming relies on the collection of geo-referencedenvironmental data to provide relevant information for usein management planning. Direct field attribute measurementand remote sensing are the two most common forms of datacollection. Although direct measurement techniques are veryaccurate, their cost and intense labor requirements may re-strict the precision of sampling. For example, the manualgrid soil sampling of a field to determine spatial variabilityof fertility is both costly and time consuming and the sampleanalysis results may take days to return. In an effort to re-

1384

duce the cost of obtaining spatially referenced data and atthe same time increase data resolution, farmers are turning tocontinuous measurement techniques and remote sensing formore accurate within-field variability information.

Continuous-Yield SensorsThe development of continuous-yield sensors and their sub-sequent linking to Dcps-provided location information hasbeen perhaps the most important and influential develop-ment in precision farming data collection. Yield rates whichvary spatially require different sensing techniques dependingon the type of crop being monitored. The greatest progresshas been achieved with grain flow measurements for cornand wheat (Schueller and Bae, 1987; Searcy et al., 1.g9gPang and 2orcb,1.990; Schueller and Wang, 1994a; Guitiens,1992). Continuous sensors for cotton yields have been testedby Hunsaker (1992) and Wilkerson ef 01. (1994). Recently,personnel ftom NESPAL developed and tested sensors formeasuring yield variation in peanuts (Figure 2) and testedcommercial grain flow monitors in low-yielding conditionsin Australia (Figure 3). Maps produced from these systemsare hardcopy evidence of the degree of within-field variabil-ity. The magnitude of this variability is a good indication ofthe suitability of implementing a spatially variable manage-ment plan.

Remote SensingRemote sensing of environmental factors important to cropgrowth, both from long range, such as aerial photographyand satellite images, and short range, such as ground-pene-trating radar (cpR) and electromagnetic induction, provideaccurate information of field variability with geopositioning.The ability to collect a large amount of data rapidly andcheaply fits with precision farming requirements. Examplesof long-range sensing include determination of soil type vari-ability from aerial photographs to estimate spatial relationsof soil fertility (Gerberman ef o1., 19BB), detecting soil surfaceconditions with multispectral video (Everitt ef a1., 19Bg), useof aerial video images fo identify vegetal condition arrd dis-criminate between crop and weed species (Nixon ef o1.,1985; Richardson ef o1., tgB5; Everitt et al.,1gg2; Brown etal,, 1gg4), and detection of plant stress and insect infestationfrom video images (Everitt and Nixon, 1986; Everitt ef a1.,1994). Short-range sensing with cpR has been used to meas-ure soil characteristics such as location and attributes ofhardpans in clay soils (Raper et aI., 1.ggo) and depth to bed-rock (Schellentrager and Doolittle, 1991). Electromagnetic in-duction is a short-range sensor used to determine soilconductivity, and to estimate salt content, soil texture, watercontent, and yield potential across the field (Kachanoski efo1., lgBB; Suddeth et al., 1.994).

VRTVariable rate treatment provides the precision farmer withthe ability to accept information pertaining to within-ffeldvariability and plan management operations to best releasethe potential of a field. Examples where VRTs have been im-plemented include the use of multiple flow-rate fertilizeripreaders that vary application across the field to match thelocal requirements and the management of weeds with flow-rate controlled sprayers to match previously collected weedincidence maps. VRT herbicide applicators have been devel-oped by Shearer and Jones (1991) and sprayer design exam-ined by Schueller and Wang (1994b).

PEER.REVIEWED ARI ICTE

Peanut Yield MapTifton GA, USA. f 1i14l1994

,()00 lbJrcrc

,500 lbe/rcrc

3000 lbs/.cre

Figure 2. Variable peanut yield map constructed from data gatheyield sensors and cPS.

red with

GISThe use of cts in farm related matters is not new. cIS tech-niques have been used for agricultural management in manyareas but primarily on a whole-field approach, that is, man-agement by application of fertility and pest-control inputs byan average value for an entire field. For example, Wylie et o/.(fgg+) used a GIS combined with a nitrate-leaching modelingpackage to describe the distribution of leached nitrates ingroundwater. Everitt et al. (1,994) used cIS to develop a mapof blackfly infestations of citrus crops in southern Texas. Astreet map from U.S. Census Bureau Topologically IntegratedGeographic Encoding and Referencing (ucnn) files was com-bined in the Atlas*cls software with cps locations of the cit-rus black{ly locations. In attempts at weed eradication, Lassand Callihan (1993) and Prather and Callihan (1993) re-corded weed-infested areas in a cIS with data from U.S. Geo-logical Survey topographic maps to development manage-ment areas by treatment methods and environmental andhealth requirements for controlled spraying. Wilson ef o1.(19931 coupled cts with models of weed control to determinegroundwater pollution by herbicides. Swindell (199s) usedGIS along with cPS to study data quality and generate yield-surface maos.

Schueiler (t992) presents a case for using a GIS as thehub of an integrated system for precision-farming data man-agement. Such a model is a logical choice because of thedata management and integrating capabilities of crs. In-creased implementation of continuous-data collection tech-niques and remote sensing coupled with new applications ofthese technologies could rapidly lead to information over-load. Several companies have recognized the unique abilityof crs to organize these types of information and presentthem in a form useful for management decisions and havebegun to develop cIS software to specifically support the pre-cision farming application. Examples of these software pack-ages include SOILECTIoNTM from Ag Chem (1SS+) ofMinnetonka, Minnesota, for VRT planning and control; CRop-SIGHT'" from Applications Mapping (1994) of Frankfort, Illi-

PE&RS

nois, which has GIS layering capabilities and can associatefarming records such as seed type, equipment use, weatherconditions, and fertilizer; and the VISION SYSTEMTM fromRockwell (rssq) which is a whole-farm multiseason cIS capa-ble of data collection, presentation for management assis-tance, and map generation to control vRt equipment.

GIS Data Management ApptoachAs suggested by Schueller (1992), Grs is the logical hub forprecision-farming data management. At NESPAL, cIS formsthe core of three activities: data collection, data analysis anddecision-making, and variable application treatment (Figurea). While the process is cyclic with the output of one cropseason used as the input to the next season, some variablesare monitored and controlled within a single season. For ex-ample, several photographic missions may be used to meas-ure weed infestation and results of herbicide applicationswithin a single crop season, whereas the final yield will bemeasured at harvest to provide data for analysis and predic-tion to help determine nutrient and pest control inputs forthe next season.

Data CollectionFor precision farming, the cIS database must consist of manyIayers of spatial data, each of which has precise control ofground position in the field. Among the layers are physicallymeasured inputs such as field boundaries, slope and aspectof the terrain, water content, particle size distribution, root-ing volume, and drainage. While some of these inputs can beinterpreted from soil maps, conventional soil survey mapsfrom the Soil Conservation Service are not accurate enoughfor precision farming, thus requiring intensive soil samplingand soil map generation for each field to be managed. Layerswith chemical inputs such as nutrient levels, cation exchangecapacity, pH, salinity, pollution potential, and plant tissueelement levels are also collected but all with a cost. Meas-ured biological data may include layers of yield quantity,

PEER.REVIEWED ARI IC I .E

Wheat Yield Map for "Marinya"N.S.W. Australia. 12127 194

36 Ha. (88 ac.)

25OOl<e/Ha-

lOOOKg,/Ha-

sOOl<e/Ha-

3OOKg,/Ha.

l5Ot<E/Ha-

Ol<E/H.a-

--'-);

2 O O d

Figure 3. Variable yield of wheat from a field in NewSouth Wales, Australia.

yield quality, disease distribution, insect distribution, weeddistribution, and organic matter content.

Data Analysis and Decision-MakingData manipulation includes rectification functions to correctthe geometry of the digital image data such as scanned aerialphotography and digital video images. To aid image interpre-tation, both spatial and spectral enhancement are performed.Classification of the digital images, using statistical classifi-ers, in both supervised and unsupervised modes yields infor-mation on weed or insect infestations, soil types, vegetationtypes, and plant stress.

Visualization of images, including multiple image dis-play and linking, allows NESPAL scientists, including ecolo-gists, agro-economists, consultants, state extension agents,and others, to use their discipline-specific knowledge andunique human abilities to spot important data characteristics.It also provides a well-defined and appropriate tool to linkthese sciences for a better understanding of the complex bio-logical systems. Using GIs analytical functions such as Bool-ean overlay, cluster analysis, clumping functions, reclassifi-cation, indexing, and spatial searching, analysts can createnew data layers which reflect particular characteristicsdeemed to be important.

Modeling capability, including expert systems, linear

1386

programming, and statistical and economic tools, allows sci-entists and consultants to develop decisions concerning ap-plication of inputs and effects on yields. Development ofspatial models allows eventual production of maps of fertil-ity and pests which are used to drive vnr applications. Themodels include raster, vector, and tabular data, as well asscalars. matrices. and a varietv of analvtical functions fiomsimple arithmetic to eigenvalues and principal components.Developing landscape assessments requires examining areasbroader than the field itself, The cls modelins functions areused to develop pollution and habitat models"and to inte-grate these with the GIS data layers.

The precise positioning of the data layers allows an ana-lyst using GIS to determine locational coincidence among theyield rates and the various fertility and pest-control inputs.The locational coincidence can then be used to create mapswhich guide VRT applicators to enhance production and re-duce investment, For some inputs and yield measurementsresulting in VRT, the process is cyclic, using the yield mapsfrom one crop season as input with the fertilizer, pesticide,herbicide, and soil fertility from the GIS to predict requiredinputs to increase yields for the next crop season.

Data ManagementAs shown in Figure 4, GIS provides the capability to integratediverse datasets, but, because precision farming requires dataof many different types from different sources, a process isneeded to insure compatibility of the various data sources.One approach being used at NESPAL concentrates on fourdata aspects: control, sampling, resolution, and generaliza-uon.

ControlBecause location in the field is the key to precision farmingdata management, each collected data set must be preciselyregistered to a standard set of control. The approach is to es-tablish ground control points (ccps) within each field and touse these cCPs to establish the locational coordinates foreach data layer gathered. The cCPs are established by precisencrs techniqueJto a horizontal and vertical accur""y i 0.1m. A minimum number of ccps are determined based onfield size and shape with an absolute minimum of fourpoints. The GCPs are targeted to appear in all aerial photo-graphs and video images. The GCPs form the starting pointfor units using on-the-go DGPS determination.

SamplingOnce established, the cCPs are used by all Nespat scientiststo collect precise data which are introduced into the com-mon cIS database. While the cCPs account for locational cor-respondence among data layers, other factors such as preci-sion of the collected data and variable types and rates of datacan introduce inaccuracy. To facilitate management, standardsampling methods are being developed to insure that pointdata collected for one laver can be effectivelv combined withpoint data from anotheriayer. Interpolation is critical butonly produces results adequate for overlay in a cIS if theoriginal samples are representative. For data requiring ran-dom samples over a field. a procedure called svstematicstratified unaligned random sampling is used (berry andBaker, 1968; Congalton, 1988; Wollenhaupt ef al., tsgs).This procedure has been demonstrated to maintain system-atic coverage of a target area while providing randomness insubareas, allowing statistical testing to be used.

PE&RS

PEER.REVIEWED ARTICTE

i::" 6?- - - / . txpertsvstems -" -

.4,;"{;;ffid;$lW

Figure 4. The components of precision farming organized with a Gls as acentral hub and cPS as the correlating base for geographic reference'

ResolutionThe resolution of data collection for different data types isestablished to facilitate data management and analysis. Ulti-mately, the resolution is a product of the crop being man-aged, but various types require different scales of analysis.For example, insect pests may cause crop damage with smallinfestations while, to cause equivalent damage, weed pests orsoil infertility may need to cover much larger areas. Collect-ing the data at the appropriate resolution facilitates use inthe GIS in which all data lavers are eventuallv reconciled to acommon resolut ion.

GeneralizationMany of the data layers are collected on a grid or raster pat-tern. These datasets are directly used in raster-based cls soft-ware requiring only resampling to a common raster cell size.The layers are maintained in their initial form for manipula-tion and analysis. Other data, such as point samples, are en-tered directly in vector-based cIS software, but are interpo-lated to provide correspondence with the raster data model.While the interpolated data are used in some analysis andpresentation procedures, the original points are maintainedin the GIS databases. Data collected as lines of attribute or re-sponse values, such as yield rates, also require interpolationto yield raster cells matching the other datasets. Data col-lected by interpretation of aerial photographs require digitiz-ing and processing to convert to a raster format. In essence,all datasets undergo transformation to support analysis in a

PE&RS

cIS in raster or vector formats, and conversion between theforms is essential.

Integrating Precision Farming Data Usin$ Commercial SoftwareAt Nispai data integration is managed through application of

standard functions in commercially available GIS software.The cIS software must be suited to a wide range of users,permitting the precision-farming specialists to interact with

ihe system andanalyze their own data. To assess the appli-

cability of generic cIS software packages to precision-farming

needs, a study was conducted'This study consisted of a survey of precision-farming

workers. One hundred and eighty questionnaires were

mailed to individuals across the U.S. based on a mailing list

compiled from past precision-farming conferences. All sur-veyJwere sent with a cover letter, instructions, and astamped return envelope. The survey was written in two sec-tions, one section for those who use a GIS and one for thosewho don't. The survey comprised a variety of short answerand multiple choice questions. A copy of the survey is avail-

able from the authors.Of the 180 questionnaires mailed, 82 were returned. Of

these, eight were unread because of relocation of the in-

tended recipients. Of the remaining 74, a further 13 were de-

termined to have been received from individuals outside ofthe target field and were disregarded. Thus, a total of 61

valid zurvey responses from 1B different states was received.Of the respondents, 82 percent are researchers, defined as

PEER.REVIEWED ARTIC IE

TneLe 1. SoFMARE Pncxlcrs UsED FoR PRecrsror.r FARN4TNG

GIS SoftwareSoftware Package Percentage of Respondents

Arc/InfoGRASSIdrisiMaplnfo

SurferSigma-PlotSASOther

Non-GIS Software

people primarily engaged in investigating improvements inagricultural methods and production, 3 percent are farmers,10 percent are suppliers of equipment or software, and 16percent are involved in some other way with precision farm-ing, These percentages do not total to 100 percent becauseseveral researchers also consider themselves to be involvedin precision-farming activities other than research. Becausethe majority of responses came from researchers, it was de-cided that this group would be made the focus of the analy-sis. The percentage time spent by the respondent researchersin precision farming-related research ranges from 2 percentto 100 percent with an average of 32 percent. Precision farm-ing is still a relatively small field of research despite rapidgrowth over the past few years. It is probably reasonable toassume that 52 researchers from 18 states comprises a repre-sentative sample of all researchers in the field. Tables 1through 4 summarize responses to the questionnaire.

GIS UsersThree-fifths of the respondents utilize GtS in their research.These researchers spend roughly double the amount of time(40 percent) than non-cls users spend on precision farmingresearch (18 percent). The GIS software packages in use varygreatly. By far the most prevalent is Arc/Info with 37 percentof respondents listing this as one of the packages used. Nextmost prevalent is GRASS (20 percent) followed by IDRISI (17percent) and Maplnfo (13 percent) (Table 1). Other productsused include PC-MAPS (Texas A&M), EPPLT (University ofMinnesota), MapsII, Arcview2, rIs (University of Illinois),Grouper (Soil Teq), scIS (AG Chem), and ERDAS Imagine. Byfar the most widely used operating system is MS-DOS (ZZ per-cent) followed by Unix (33 percent), osl2 (2o percentJ, andMacintosh operating systems (10 percent).

cIS are currently used mostly for soil parameter mappingand yield mapping (Table 2). It is rare that a GIS is utilizedfor only one purpose. In 89 percent of cases, researchers in-dicated that they use cIS to map more than one factor withina field. This is perhaps to be expected because researchersoften seek correlation and coincidence among observed fac-tors. Only 59 percent of respondents, however, indicated thatthey performed correlation analysis between maps of differ-ent factors. This large discrepancy may indicate that the GISused does not have the ability to correlate different factors. Itmay also indicate that, if the cIs is capable of performingthis operation, the researcher has not yet been able to fullyutilize this option.

Forty-six percent of respondent researchers use their GISto perform advanced statistical analysis on their data. For aresearch community this is a low percentage and may indi-

1388

cate a shortcoming in many of the cIS packages used. Alter-natively, it may be indicative of the newness of this researchor a tendency for researchers to perform statistical analysisoutside the GIS environment. Many researchers may simplynot be cognizant of the features available to them in their GISbeyond the simpler functions. Nearly half of the respondentssuggested forms of statistics that they would like to see in-corporated into their cIS. The desire for kriging-related geo-statistics was highly prevalent. Statistical tests to performmultivariate analysis, linking yield response to other meas-ured factors, was also an obvious priority. Other require-ments include trend analysis, clustering, principal compo-nents analysis, and fuzzy-logic statistics as well as suchsimple statistical tests as t-tests and F-tests. Obviously, thereis concern among the research community that the statisticalfurrctions currently available to them in crs packages are in-adequate.

A cIS which incorporates and manipulates data in realtime would be usefuI for instantaneous viewing of results be-cause data are gathered in the field. It would also be requiredfor instantaneous application of amendments in respons-e togathered site data. Only 17 percent of the respondents areusing a GIS capable of performing real-time manipulations.For most research applications it is probable that post-proc-essed data are quite adequate. For some of the spatial-rela-tionship research, linking factors which cannot be measuredsimultaneously requires post-processing. However, 48 per-cent of researchers do consider that incorporation and ma-nipulation of data in real time is necessary (Table 2). Thereis probably some conflict between responses dealing withfundamental agronomic research and responses related moredirectly to applied precision-farming research in this answer.The lack of sensors reliable enough to be used in real-timetreatment is probably another reason for the lack of apprecia-tion for this feature; however, the future in pest managementappe€us to be in real-time sensing and treatment.

Incorporation of some form of economic analysis was re-garded, almost unanimously, as essential for precision-farm-ing purposes (Table 2). Again, however, practice is laggingbehind the theory, probably a result of the early stage of vnr,and only 36 percent of the respondents have actually at-tempted incorporation of economics into their cIS applica-tions.

Despite apparent deficiencies in current cIS, more than75 percent of the respondents scored their GrS a 3 or aboveon a suitability scale of 0 to 5 (Table 3). There was muchless satisfaction in terms of user-friendliness of cls software.

Trele 2. SuvvnRv or Quesrronnntne Resporuses

Percentage of RespondentsSoftware Use GIS-Users Non-GIS Users

Yield MappingSoiI Parameter MappingInsect/Weed MappingAspect/Elevation MappingCorrelation of Data LayersReal-Time AnalysisAdvanced StatisticsEconomic Analysis

GIS Software Needs

Real-Time AnalysisEconomic Analysis

J /

201.71 3

362950a /

92/ J

t 7J J

42

46

/ J

8939+ J

747 74636

PEER.REVIEWED ARI IC IE

TaeLr 3. Surrlarury lruo Usen-FnrEnouress or GIS

Low High

GIS SuitabilityGIS User-Friendliness

710

although there are obviously systems available which arevery friendly. Among the respondents, Arc/Info was viewedas the system considered most unfriendly. No single packagewas viewed as being particularly user-friendly, althoughMacintosh and Windows-based systems fared best.

There appears to be quite some variability in the expec-tations of useful lifetimes of current cIS (Table 4). Nearly 50percent of researchers think that their systems are alreadyobsolete or will be within two years. Thirty-two percent ofresearchers said their cIS will not become obsolete in the im-mediate future (within two years) and 19 percent were un-able to say. This variability is probably to be expected, giventhe different aims and cIS of the respondents. It is strikingthat nearly a third of researchers believe that their cIS is ;1-ready inadequate for their needs.

A maior concern communicated about GIS is the com-plexity of most of the cIS software. There are many functions6n the large GIS packages which are superfluous for preci-sion-farming work and hence only contribute to complexityand result in user confusion. Ease of use is also a major con-cern. Some of the smaller customized packages in use hadmuch better reviews with respect to ease of use. One respon-dent stated another major concern: "[it is] difficult for end-users to perform complex analysis with real understanding.Knowledge of theory and principles of cts is required forreal agility with the software to be acquired."

Lack of a soils database at a scale appropriate for preci-sion farming is another problem. This particular concernprobably holds for most of the yield-determining factorswithin a field. Other failings recorded repeatedly includeddifhculties in interfacing particular GIS data with other sys-tems, slow speed, lack of statistics, high cost, and long train-ing time.

Positive features of cIs for use with precision-farming re-search varied with the particular cls used. There were, how-ever, a number of recurring positive general features. Themapping abilities of most cIS are highly valued, as are layer-ing and overlay features. A good user interface and compati-bility with other software and databases is highly prized as isthe ability to perform statistical and economic analysis.

Non.GlS UsersOver a third of the respondent researchers do not use a cISin their work. A summary of the statistics from these respon-dents is also listed in Table 2. These researchers spend, onaverage, only half the time that the cls-using researchersspend on precision-farming work. The majority of these re-spondents use some form of computer-generated maps toview or analyze their spatially referenced data. The two mostoften used pieces of software for this purpose are SAS (usedby 56 percent of researchers) and Sigma-plot (33 percent)(Table 1). A variety of other software is also used, includingSystat, various spreadsheets, Agrilogic, GS*, neural-net soft-ware, and custom software. All respondents use MS-DOS asan operating system although in two cases other systems areused as well (oslz and vtvts).

The work of this group of researchers is concentrated in

PE&RS

yield mapping and soil parameter mapping (Table 2). Thereis less tendency to perform correlations between mappedparameters. This is partly due to the more specialized natureof the work of certain of these respondents who were lookingat relationships within map layers rather than between lay-ers. In general, however, this group of researchers is less ad-vanced in precision-farming research relative to their GIS-using counterparts. If they do expand their research in thearea, it would be expected that they will need to use a cIS.Virtually all of the respondents indicated that they wouldneed to use a cIS in the future.

About 70 percent of the non-clS software used scored athree or above in terms of suitability to researchers aims (Ta-ble 3). This contrasts with the higher suitability ranking ofcIS by cIS users. This was not expected given that the non-GIS software has generally been in use for a much greater pe-riod of time with a much less specialized market and muchmore development. This probably bears witness to the greatpotential for GIS in this field. There were not great differ-ences in the perceptions of potential obsolescence of soft-ware used between the cIS and non-cls groups.

Only one researcher used real time manipulation of data,although a similar proportion to that of the cls-user grouPthought that it was necessary for precision farming. Almostall of the researchers who do not use GIS have considered us-ing it at one time or another. The most common reasons fornot using cIS are high start-up costs and the steep learningcurve.

ConclusionsPrecision farming, with its potential to simultaneously in-crease crop yields and reduce environmental pollution, israpidly expanding in both research and production agricul-ture. As this expansion occurs, cIS will become more inte-grated as a data management and analysis tool for agricul-tural applications, Placing the cls at the center of the processprovides the advantage of a single database and data m€Inage-ment system, To succeed, a precision-farming GIS applicationmust include precise geographic positioning for all data lay-ers and ground control for all image sources and on-the-gocoordinate measurements. A logical approach is to targetground control points in a field and use them as references forall data collected. DGPS provides the needed accuracies andthe capabilities for bothlhtic and dynamic measurement ofcoordiirates associated with precision-farming variables. Inte-grating the DcPs-collected information with crs allows thenecessary manipulation and analysis to support generation offarm-management decisions and digital maps which can beused to drive VRT sprayers and other equipment.

From the questionnaire survey, one may conclude thatcIS are cunently an integral and indispensable part of preci-sion farming. cIS are widely used through much of theU.S.A. for precision-farming research. Almost all respondentseither use a cls or believe that thev will in the future. It is

Tnerr 4. Sorrune OssoLesceNce tru rHE NeaR Furune

GIS-Users Non-GIS Users

010

15 30 30 1827 21 28 10

NoAlreadyWithin one yearWithin two yearsAfter two yearsUnable to say

l o

291 01 0l o

1 9

23J I

8I

I J

I D

PEER.REVIEWED ARI ICTE

also readily apparent that significant development must oc-cur and experience must accrue before precision-farming re-search and data management using cIS are successful appli-cations. Suggestions for improvement in crs were not lackingin responses. Better user interfaces, more statistical and eco-nomic functions, less complexity, incorporation of models,better database resources, real-time linkage with positioningsystems, and faster operation speeds are seen as priority im-provements in GIS. The ability for a cIS to incorporate eco-nomic data is considered essential, while advanced statisticalfunctions and real-time data manipulation abilities are con-sidered highly desirable. It is probably fair to say that, as aresult of problems associated with cIS, few researchers havefully utilized the potential. A more complete utilization ofcIS appears to be essential for a successful precision-farmingoperation and is likely to occur as a result of the use of cpsto link the cIS data layers.

AcknowledgmentsWe acknowledge the support of Craig Kvien and Debbie Wa-ters, coordinating researchers for the precision-farming effort,and the other NnspaL scientists, particularly the engineeringteam led by Mike Tucker, Calvin Perry, George Vellidis, andDon Thomas, who are cooperating to make this long-termproject a successful endeavor. We also acknowledge our in-dustry support from Albany Scales, Applications Mapping,/ohn Deere, John E. Chance and Associates - Omnistar Divi-sion, ERDAS, Inc., Kelly Manufacturing Co., and Microtrack.Thanks must go to all the participants in the survey whotook the time to respond and who imparted some very help-ful insights into the use of GIS in precision farming.

ReferencesAg-Chem, 1.994. Soilection, Ag-Chem Equipment Co., Inc., Minne-

tonka, Minnesota.

Applications Mapping, 1594. FMS Cropsight, Version 2.0, Applica-tions Mapping, Frankfort, Illinois.

Aurenhammer, H., M. Demmel, T Muhr, |. Rottmeier, and K. WiId,1994. GPS for Yield Mapping on Combines, Computers andElectrcnics in Agriculture, 11:53-68.

Berry, B.J,L., and A.M. Baker, 1968. Geographic Sampling, SpatialAnalysis: A Reader in Statistical Geography (B.f.L. Berry andD.F. Marble, editors), Prentice-Hall, Englewood Cliffs, N.f., pp.91-100.

Brown, R.8., G.A. Steckler, and G.W. Anderson, 1994. Remote Sens-ing for Identification of Weeds in No-Till Corn, Transactions ofASAE, 37(1,) :2s7-3o2.

Chancellor, W.J., and M.A, Goronea, 1994. Effects of Spatial Variabil-ity of Nitrogen Moisture and Weeds on the Advantages of Site-Specific Applications for Wheat, Transactions of the AmericanSociety of Agricultural Engineers, 37 (3):7 77 -7 24.

Congalton, R.G., 1988. A Comparison of Sampling Schemes Used inGenerating Error Matrices for Assessing the Accuracy of MapsGenerated from Remotely Sensed Data, Photogrammetric Engi-neering & Remote Sensing, 54(5):593-600.

Delcourt, H., and f.D. Baerdemaeker, 1994. SoiI Nutrient MappingImplications using GPS, Computers and Electronics in Agricul-ture, 77i37-51..

Everitt, J.H., and P.R. Nixon, 1986. Canopy Reflectance of TwoDrought-Stressed Shrubs, Photogrammetric Engineering & Re-mote Sensing, 52[B):1 189-1 192.

Ever i t t , J .H. , D.E. Escobar, M.A. Alaniz, and M.R. Davis, 1989. UsingMultispectral Video Imagery for Detecting Soil Surface Condi-

1390

tions, Photogrammetfic Engineering & Remote Sensrng, 55(4):467471.

Everitt, J.H., D.E. Escobar, M.A. Alaniz, R. Villarreal, and M.R. Davis,1992. Distinguishing Brush and Weeds on Rangelands UsingVideo Remote Sensing, Weed Technology, 6:91,3-921,.

Everitt, J.H., D.E. Escobar, K.R. Sunny, and M.R. Davis, 1994. UsingAirborne Video, Global Positioning System and Geographical In-formation System Technology for Detecting and Mapping CitrusBIackfly Infestations, Southwestern Entomologist, 1.9(2lI2g-1,J8.

Gerbermann, A.H., D.E. Escobar, R.R. Rodriguez, and |.H. Everitt,1988. Color and Multispectral Video Imagery for Detecting SoilMapping Units, Frrst Workshop on Videograph.y, American Soci-ety for Photogrammetry and Remote Sensing, pp. 3741,.

Goering, C.E., 19S3. Recycling a Concept, Agricultural Engineering,November, p. 25.

Guitjens, 1.C., lgg2. Interpreting Spatial Yield Variability of IrrigatedSpring Wheat, Transactio n s of AS AE, 3b ( 1 ) :91-9b.

Hunsaker, D.1., 1,992. Cotton Yield Variability Under Level Basin Irri-gation, ?ronso ctions of ASAE, 35(4):7205-1211.

Kachanoski, R.G., E.G. Gregorick,, and I.J. Van Wesenbeeck, 1988.Estimating Spatial Variations of Soil Water Content Using Non-Contacting Electromagnetic Inductive Methods, Canadian lour-nal of Soi l Science,68:715-722.

Lass, L.W., and R.H. Callihan, 1993. GPS and GIS for Weed Survevsand Managem ent, Weed Technology, 7 :249-254.

Linsley, C.M., and F.C. Bauer, 1.929. Test Your Soil for Aciditv, Cir-cular 346, University of Illinois, College of Agriculture and Rgri-cultural Experiment Station.

Nixon, P.R., D.E. Escobar, and R.M. Menges, 1S85. A MultibandVideo System for Quick Assessment of Vegetal Condition andDiscrimination of Plant Species, Remote Sensing of Environ-ment, 77t2O3-2O8.

Pang, S.N., and G.C. Zoerb, 1990. A Grain FIow Sensor for YieldMapping, presented paper, ASAE, December, Chicago.

Prather, T.S., and R.H. Callihan, 1993. Weed Eradication Using Geo-graphic Information Systems, Wee d Technology, 7 :265-269.

Pringle, J.L., M.D. Schrock, R.T. Hinnen, K.D. Howard, and D.L.Oard, 1993. Yield Variation in Grain Crops, Proceedings ofASAE, Paper No. 93-1505, Chicago.

Raper, R.L., L.E. Asmussen, and J.B. Powell, 1990. Sensing Hard PanDepth with Ground Penetrating Radar, Ironsactions of ASAE,33(1 J:41-46.

Richardson, A.J., R.M. Menges, and P.R. Nixon, 1985. DistinguishingWeed from Crop Plants Using Video Remote Sensing, Pftoto-gtammetfi c Engine e rin g & Re mote S e ns i ng, 5 1 (1 1 ) : 1 7BS-1 790.

Rockwell, 1994. The Vision Svstem, Rockwell International. CedarRapids, Iowa.

Schueller, 1.K., 1.552. A Review and Integrating Analysis of Spatially-Variable Control of Crop Production, Fertilizer Research,33 I-34.

Schueller, J.K., and Y.H. Bae, 1987. Spatiallv Attributed AutomaticCombine Data Acquisition, Comp-uters and Electronics in Agri-culture, 2i1.1.9-1.27.

Schueller, f.K., and M.W. Wang, 1994a. Spatially-Variable Fertilizerand Pesticide Application with GPS and DGPS, Computers andElectronics in Agriculture, 11 :69-83.

1994b. Boom Hose Diameter and Feed-Forward Precommandfor Spatially-Variable Sprayers, Report N.g4-D162, presented pa-per, AgEng Conference, Milan, Italy.

Searcy, S.W., J.K. Schueller, Y.H. Bae, S.C. Borghelt, and B.A. Stout,1989. Mapping of Spatially Variable Yield During Grain Com-bining, Transactions of the ASAE, 32(3):826-825.

Shearer, S.A. , and P.T. Jones, 1991. Select ive Appl icat ion of Post-Emergence Herbs Using Photoelectrics, Transactions of ASAE,34:4.

Sudduth, K., N. Kitchen, D. Hughes, and S. Drummond, 1994. Elec-

PEER 'REVIEWED ARI ICTE

tromagnetic Induction Sensing as an Indicator of Productivity onClay, Proceedings of Second Site-Specific Crop ManagementSymp o sium, Minneapolis, forthcoming.

Swindel, J., 1995. Mapping Crop Yield Variations by Use of GPS andGIS, Proceedings of the loint European Conference on Geogtaph-ical Information, The Hague, Netherlands, forthcoming.

Wilkerson, ].8., ].S. Kirby, W.E. Hart, and A.R' Womac, 1994. Real-Time Cotton Flow Sensor, presented paper, ASAE, fune, KansasCity.

Wilson, I.P., W.P. Inskeep, P.R. Rubright, D. Cooksey, J.S. facobsen,and R.D. Snyder, 1993. Coupling Geographic Information Sys-tems and Models for Weed Control and Groundwater Protection,W ee d T e ch no logy, 7 1255-264.

Wollenhaupt, N.C., R.P. Wolkowski, and M.K. Clayton, 1994. Map-ping Soil Test Phosphorus and Potassium for Variable Rate Fer-tilizer Application, lournal of Production Agricultute, 7 (4):44I-447.

Wylie, 8.K., M.J. Shaffer, M.K. Brodahl, D.Dubois, and D'G. Wagner,1994. Predicting Spatial Distributions of Nitrate Leaching inNortheastern Colorado, Iournal of Soil and Watet Conservation,a9(3J:288-293.

ducted research in feature-based approaches to GIS and three-dimensional visualization. He ioined the faculty of theUniversity of Georgia Department of Geography and th-e Na-tional Environmenially Sbund Production Agriculture Labo-ratory (NESPAL) in 1994. His current research includes GISdata hodeling including fuzzy-set implementation, featureextraction, and cts data management techniques for preci-sion-farming applications.

Stuart PockneeStuart Pocknee received his B.Agr.Sc. from theUniversity of Queensland, Australia, in 1992.He is currently completing an M.S. degree in

E. Lynn UseryE, Lynn Usery received a BS degree in geogra-phy from the University of Alabama and MAand PhD degrees in geography from the Univer-sity of Georgia. He worked 10 years with theU.S. Geological Survey developing automated

mapping systems and conducting GIS research. He was amembeiof the faculty of the Department of Geography at theUniversity of Wisconsin-Madison for 5 years where he con-

Broughton BoydellBroughton Boydell, a wheat farmer from Moree,New South Wales, has a B.Sc.Ag from the Uni-versity of Sydney, Australia, in 1993. He is cur-rently enrolled in an M.S. degree program. inAgronomy at the University of Georgia' His .$*rilqwiEtr Agfonomy at tne Unlvefsfty OI (JeU-rBra, nrs -

work with NESFAL at the University of Georgia involves thedevelopment of yield sensors for use in peanuts and cotton,investi-gation of iactors affecting the measured yield, and themanagJment of yield-affecting elements using precision farm-

farming techniques.

ing techniques.

Iffi ff ;:ml."llir",Y,t',"fitl'di3:,:ll-.h,T'3'Georgia involves the investigation of soii .spatial variabilityin th6 Southeast and the management of it using precision-

GIS FOR RESOURCE MANAGEMENT:A COMPENDIIN{

Edited by: William J. RiPPle

This book will introduce the reader to the techniques and functional capabilities on GIS, and bring togetherinformation on the use of GIS tbr a variety of resource management applications. This book is intended tobe used by ecologists, fbresters, geologists, geographers, hydrologists, landscape architects, soil scientists,and urban and relional planners,ls welt as indiiiduals woiking in the mapping sciences. It is also used as areader for introductory university courses on GIS.

Chapters include:

o Introduction to GIS Techniques and FunctionalCapabilities

o Land Suitability Studies. urban studies. Water Resource Management

o Soil Resource Management. Vegetation Resource Managemento Global Studieso Bibliography

1987. 296 pp. (sofcover). Srock # 628E 1. NOW ONLY $10.

For ordering information' see the ASPRS Store.