HAL Id: hal-01773571
https://hal.inrae.fr/hal-01773571v2
Submitted on 25 May 2020
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of
sci-entific research documents, whether they are
pub-lished or not. The documents may come from
teaching and research institutions in France or
abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est
destinée au dépôt et à la diffusion de documents
scientifiques de niveau recherche, publiés ou non,
émanant des établissements d’enseignement et de
recherche français ou étrangers, des laboratoires
publics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
Case-based Reasoning for Forecasting the Allocation of
Perennial Biomass Crops
Florence Le Ber, Jean Lieber, Marc Benoit
To cite this version:
Florence Le Ber, Jean Lieber, Marc Benoit. Case-based Reasoning for Forecasting the Allocation of
Perennial Biomass Crops. ERCIM News, ERCIM, 2018, Smart Farming, pp.34-35. �hal-01773571v2�
of GPS points and polygon data togetherwithfieldphotographs.Beside thelocationandextent,localwater regime,habitatquality(biodiversity) andobservationoffloraandfaunais alsoincludedinthedataset.Thedata-baseincludesthesurroundsofthethree largestHungarianlakes:LakeBalaton, Lake Tisza and Lake Neusiedl (Hungarianside).Thedatabasecontains 150testpatches,outofwhich113are wetlands,providingagoodrepresenta- tionofthevarietyofHungarianwet-lands.Besideswetlands,thereare23 grasslands(alsoanimportanthabitat type)and14croplands. Themainfocusoftheprojectistouse thisdatabasetoinvestigatetemporal, spectralandspatialcharacteristicsof wetlandsusingSentinelmultispectral satelliteimagery,andfromthisto developanautomaticdetectionprocess, whichcanprovidefurtherinformation aboutnaturalhabitats.Thedetection processappliesafusionbasedMarkov RandomField(MRF)technique[1]on multitemporalandmultispectralimage series.InourfusionMRFmodelan unsupervisedmethodtrackstemporal changesofwetlandareasbycomparing theclasslabellingofdifferenttime layers[2].Aclassificationmapbasedon existingairbornelaserscanningdata[3] hasbeenusedforwetlandclassification improvingthediscriminationofland- coverclasseswithsimilarspectralchar-acteristics.Theproposedmethodcanbe extendedbymachinelearningandfea- turebasedclassificationformoreaccu-ratemonitoringofwetlands. Links: [L1]http://mplab.sztaki.hu [L2]https://kwz.me/hbv [L3]https://kwz.me/hbw References: [1] T.SzirányiandM.Shadaydeh: “SegmentationOfRemoteSensing ImagesUsingSimilarityMeasure BasedFusion-MRFModel”,IEEE Geosci.RemoteSens.Lett.,vol. 11,no.9,pp.1544-1548,2014 [2] M.Shadaydeh,etal.:“Wetland MappingbyFusionofAirborne LaserScanningandMultitemporal MultispectralSatelliteImagery”, Int.J.RemoteSens.,vol.38,no. 23,pp.7422-7440,2017 [3] A.Zlinszky,etal.:“Categorizing wetlandvegetationbyairbornelaser scanningonLakeBalatonandKis-Balaton,Hungary”,RemoteSens., vol.4,no.6,pp.1617-1650,2012 Please contact: AndreaManno-Kovacs MTASZTAKI,MPLab andrea.manno-kovacs@sztaki.mta.hu
ERCIM NEWS 113 April 2018
34
Special theme: Smart Farming
Figure2:WetlandmappingonLakeBalaton-Thetestpatchesofthewetlanddatabaseareshowninyellow,followedbytheMRF-based multitemporalsegmentation[2]andthetestsiteinGoogleMaps. Perennialbiomasscropsarenewrenew-ableenergyresourcesthatcouldplayan importantroleinreplacingfossilfuels. Theexpansionofthesecropsseems unavoidable,andwillraiseglobalissues suchascompetitionforspacebetween foodandnon-foodcrops.Toforecasttheir introductioninfarmingsystemsandtheir spatialallocationisthusamainchallenge. Whilstseveralland-usechangemodels dealwithbiomasscropallocation,most ofthemsimulatelarge-scaleallocation processes,takingintoaccountnumerous biophysicalvariablesbutonlyafewtrue-to-lifehumanvariables.Incontrast,we aimedtomodelfarmers’allocationdeci-sionsregardingnewperennialbiomass cropsasacomplexagriculturalmanage-mentsystem,couplingsocial,technical andenvironmentalvariables. Asthisobjectiveraisesknowledge acquisitionandknowledgeintegration methodologicalissues,weproposedto modelbiomasscropallocationrelying oncase-basedreasoning(CBR),a problem-solvingparadigm.CBRcon-sists of solving new problems by reusingthesolutionofsimilarproblems thathavealreadybeensolved[1].A casecorrespondstoaproblem-solving
Case-based Reasoning for Forecasting
the Allocation of Perennial biomass Crops
by Florence Le Ber (Université de Strasbourg, ENGEES), Jean Lieber (Université de Lorraine, Inria) and Marc Benoît (INRA)
A farmer’s decision to plant new perennial biomass crops is a complex process that involves numerous parameters and can change the food / non-food balance. The paradigm of case-based reasoning is able to deal with the kind of complex and sparse information that is available in this situation, to model and forecast the allocation of these crops.
ERCIM NEWS 113 April 2018 35
episode usually represented by a problem-solution pair. Cases are recordedinacasebase.Asourcecaseis anelementofthecasebase.TheCBR process consists of solving a new problem,thetargetproblem,usingthe casebase.Acommonwayofdoingthis istoselectasourcecasesimilartothe targetproblem(caseretrieval)andto modifytheretrievedcasesothatitpro-videsasolutionthathypothetically solvesthetargetproblem(caseadapta-tion).Aftertheseinferencesteps,some learningstepsaresometimesimple-mented. Thisapproachwastestedonacase studylocatedinBurgundy(Eastof France),whereanewperennialbiomass cropMiscanthus sp.hasbeenestab-lishedsince2010. Inourmodel[2],acaseisdefinedasa specificexperienceofMiscanthus allo-cation(ornon-allocation)inafarm field.Thecasebaseincludes82farm fieldsthathavebeendescribedby farmersinpastinterviews(2011-2012) ashavingMiscanthus allocationpoten-tial.Theproblem-solutionpairisafarm fieldanditsallocationpotentialfor Miscanthus.Eachcaseisrepresented withavectorofqualitativevalues, dividedintotwoparts:theproblempart isasetofattributes,givingthefarm fieldcharacteristics,asdescribedbythe farmer;thesolutionpartdescribesthe Miscanthus allocationpotentialofthe farmfield(0-noallocation,1-alloca-tion,or2-possibleallocation).The modelalsoreliesonasetofrulesthat formalisetheelementsgivenbyfarmers whenexplainingtheirdecisiontoplant (ornotplant)Miscanthus inafield. Todefineasimilaritymeasurebetween cases,weassumedthatsimilarfarm managementandbiophysicalcon-straintsoffarmlandenableanalogue farmers’decisionsaboutcropalloca-tion.Thecomparisonofthesource casesandthetargetproblemisthus basedonacomparisonoftheirattribute valuesintermsofcroppingplan,farm biophysicalfeaturesandspatialfarm-landfeatures. Intheadaptationstep,weusetherules ofthefarmerassociatedtothesource casetobuildthesolution.Adaptation knowledgeallowstheappropriaterule tobechosenandapplied,accordingtoa givenadaptationcontext:thesystem promotestheruleswithconclusion0 (whenthecontextisnotfavourablefor Miscanthus,e.g.,becauseitspriceis lowcomparedwithtraditionalcrops)or thosewithconclusion1or2,iftheeco-nomic context is favourable for
Miscanthus. Experimentsshowedthecentralroleof theuser,whohastochooseparameters andadaptationcontexts,andtoexamine resultsstepbystep:retrievedcases, availablerules,proposedsolutions. Rulesinparticularcanbeanalysedto highlightthefieldcharacteristicsthat are important with respect to the farmer’sdecision.Futureworkwill deepentheuseofrulesforadaptation andexplanation.Thismodelcouldalso betestedonotherinnovativecrops. Finally,aformalmodelfortheadapta- tionofcropallocationhasbeendevel-opedthatisbasedontheapplicationof beliefrevisioninthequalitativealgebra RCC8setting(Dufouretal.,2012): thismodelstillneedstobeappliedand validatedonrealfarmdata. Farmsurveysandknowledgemodelling werefundedbyFUTUROLproject[L1] andtheFrenchgovernment. Link: [L1]https://www.projetfuturol.com References: [1]C.K.Riesbeck,R.C.Schank: 3InsideCase-Basedreasoning”, LawrenceErlbaumAssociates Publishers,1989. [2]F.LeBer,etal.:“AReasoning ModelBasedonPerennialCrop AllocationCasesandRules”, ICCBR.Springer,LNAI10339, 61-75,2017. [3] V.Dufour-Lussier,etal.: “AdaptingSpatialandTemporal Cases”,ICCBR.Springer,LNAI 7466,77-91,2012. Please contact: FlorenceLeBer,ICube,Universitéde Strasbourg,CNRS,ENGEES,France +33388248230 florence.leber@icube.unistra.fr MarcBenoît,ASTER,INRA,France +33329385501 marc.benoit@inra.fr Figure1:AMiscanthusfieldinBurgundy.