From Smart Farming towards Agriculture 5.0: A Review on Crop Data Management

: The information that crops offer is turned into profitable decisions only when efficiently managed. Current advances in data management are making Smart Farming grow exponentially as data have become the key element in modern agriculture to help producers with critical decision-making. Valuable advantages appear with objective information acquired through sensors withtheaimofmaximizingproductivityandsustainability.Thiskindofdata-basedmanagedfarms rely on data that can increase efficiency by avoiding the misuse of resources and the pollution of the environment. Data-driven agriculture, with the help of robotic solutions incorporating artificial intelligent techniques, sets the grounds for the sustainable agriculture of the future. This paper reviews the current status of advanced farm management systems by revisiting each crucial step, fromdataacquisitionincropfieldstovariablerateapplications,sothatgrowerscanmakeoptimized decisions to save money while protecting the environment and transforming how food will be produced to sustainably match the forthcoming populationgrowth.


INTRODUCTION
Theagriculturesectorisundergoingatransfor mationdrivenbynewtechnologies,whichseems very promising as it will enable this primary sector to move to the next level of farm productivity and profitability [1]. Precision Agriculture, which consist of applying inputs (what is needed) when and where is needed, has become the third wave of the modern agriculture revolution (the first was mechanizationandthesecondthegreenrevolutionwith itsgeneticmodification [2]),andnowadays, it is being enhanced with an increase of farm knowledge systems due to the availability of larger amounts of data. The United States Department of Agriculture (USDA) already reported in October 2016 that Precision Agriculture technologies increased net returns and operating profits [3]. Also, when considering the environment, new technologies are increasingly being applied in the farms to maintain the sustainability of farm production. However, the adoption of these technologies involves uncertainty and trade-offs. According to a market analysis, the factors that would facilitate the adoption of sustainable farming technologies include better education and training of farmers, sharing of information, easy availability of financial resources, and increasing consumer demand for organic food [4]. When applying these new technologies, the challenge for retrieving data from crops is to come out with something coherent and valuable, because data themselves are not useful, just numbers or images. Farms that decide to be technology-driven in some way, show valuable advantages, such us saving money and work, having an increased production or a reduction of costs with minimal effort, and producing quality food with more environmentally friendly practices [5]. However, taking these advantages to the farm will depend, not only on the willingness ofproducers foradoptingnewtechnologiesintheirfields,butalsoon eachspecificfarmpotentialintermsofscale economies, as profit margin increases with farm size. The USDA reported that, on average, corn farm operating profit of Precision Agriculture adopters was 163 dollars per hectare higher than for non-adopters, taking into account that the highest adoption rates for three technologies (computer mapping,guidance,andvariablerateequipment)wereonfarmsover1500hectares [3].S uchmargins canevengoupto272dollarsdependingonthecrop.Agre ateruseofSmartFarmingservicesisvital tonotonlyimprovingafarm'sfinancialperformance,b utalsotomeetthefoodneedsofanexpanding population [6].
The final purpose of this paper is to demonstrate how making decisions with the modern data-based agriculture available today can lead to sustainable and profitable actuation to nourish peoplewhilereducingharmtotheenvironment.Inorder the agrifood chain, because with advanced technologies and new thinking, young people can transform the agricultural sector [8].
Internet of Things: CollectingInformation Internetofthings(IoT)inanagriculturalcontextreferst otheuseofsensorsandotherdevicesto turn every element and action involved in farming into data. It has been reported that an estimation of a 10% to 15% of US farmers are using IoT solutions on the farm across 1200 million hectares and 250,000 farms [11]. IoT drives Agriculture 4.0 [12]; in fact, IoT technologies is one of the reasons why agriculture can generate such a big amount of valuable information, and the agriculture sector is expected to be highly influenced by the advances in these technologies [13]. It is estimated that, with newtechniques,theIoThasthepotentialtoincreaseagri culturalproductivityby70%by2050 [14], which is positive, because according to Myklevy et al., the world needs to increase global food production by 60% by 2050 due to a population growth over nine thousand million [15]. The main advantagesoftheuseofIoTareachievinghighercropyie ldsandlesscost.Forexample,studiesfrom OnFarm found that for an average farm using IoT, yield rises by 1.75% and energy costs drop 17 to 32 dollars per hectare, while water use for irrigation falls by 8% [12].
Big Data: Analysis of MassiveData Inthecurrenttechnologybasedera,theconceptofbigdataispresentinmanyecon omicsectors, but is it already available to agriculture? The ever-growing amount of data available for field management makes necessary the implementation of some type of automatic process to extract operational information from bulk data. However, the volume of data currently retrieved from most commercialfieldsis,arguably,notyetatthelevelconsid eredtobeclassifiedasbigdata.Accordingto Manyica et al. [16], big data has three dimensions: Volume, velocity, and variety. Kunisch [17] added a fourthVforveracity.Finally,afifthVwasaddedbyChiet al.fortheextradimensionvalorization [18]. Overall, the five V (dimensions) of big data standfor: • Volume refers to datasets whose size is beyond the ability of typical database software tools to capture,store,manage,andanalyzeinformation.Thisd efinitionincludesanestimateofhowbig adatasetneedstobeinordertobeconsideredbig,anditca nvarybystudysector,dependingon softwaretoolsthatarecommonlyavailableandcommo nsizesofdatasets,typicallystartinginthe terabyte range [16].
• Velocity refers to the capability to  [24][25][26][27], pushing agricultural systems to the new concept of Agriculture 5.0. According to Reddy et al. [28], the advent of robots in agriculture drastically increased the productivity in several countries and reduced the farm operating costs. As said before, robotic applications for agriculture are growing exponentially [27],whichofferspromisingsolutionsfo rSmartFarminginhandlinglaborshortageand a longtime declining profitability; however, like most innovations, there exist important limitations to cope with at the current early stages. These technologies are still too expensive for most farmers, especially those with small farms [29], because scale economics make small individual farms less profitable [30].Nevertheless,thecostoftechnologydec reaseswithtime,andagriculturalrobotswillbe surelyimplementedinthefutureasthealternativetobrin gabouthigherproduction [4,31].Theworld agricultural production and crop yields slowed down in 2015. The concept of agricultural robotics was introduced to overcome these problems and satisfy the rising demand for high yields. Robotic innovations are giving a boost to the global agriculture and crop production market, as according to the Verified Market Intelligence report, agricultural robots will be capable of completing field tasks with greater efficiency as compared to the farmers [32].
Agricultural tech startups have raised over 800 million dollars in the last five years [31]. Startups using robotics and machine learning to solve problems in agriculture started gaining momentum in 2014,inlinewitharisinginterestinAI [33].Infact,ventu recapitalfundinginAIhasincreasedby450% inthelast5years [34].Thiskindofnewagricultureprete ndstodomorewithless,becausenourishing people while increasing production sustainably and taking care of the environment will be crucial in the coming years, as the Food and Agriculture Organization of the United Nations (FAO) estimates that, in 2050, there will be a world population of 9.6 billion [35]. Advanced sensing technologies in agriculture can help to meet the challenge; they provide detailed information on soil, crop status, and environmental conditions to allow precise applications of phytosanitary products, resulting in a reducedusedofherbicidesandpesticides,improvedwa teruseefficiencyandincreasedcropyieldand quality [2].

Data-Driven Management for Advanced Farming: PrincipalStages
The raw measurements of key parameters  The following paragraphs and Figure1explain the cycle that embodies a general data-driven management system for advanced agriculture, including representative examples for each stage. Table1classifies the scientific works referenced in this study into the di fferent categories of Figure1. StageI:TheCropastheBeginningandEndoft heAgriculturalManagementCycle-AnalyzingVariability Regardless how the crop will be managed, some degree of spatial variability is assumed for all fieldsbynature.AccordingtoSearcy [37],naturalvaria bilityisinfluencedbyweatherwithinagrowing season and from year to year; then, data from several years may be needed to determine trends in the parameters of interest, and hence, data becomes a regular input to the farm management system. Therefore,thenecessityofmonitoringcropscomesfro mtheexistenceofvariability,butthereisaneed fortheproducertomanagethatvariabilityinafeasiblew ay,andthewidelyacceptedwaytodoitisby settingwithinfieldmanagementzones.Managementzonesaresubfie ldareasthathavehomogeneous features,sofieldpracticescanbecustommadetoeachofsuchareas,resultinginapracticaland cost-effectiveapproachtoPrecisionAgriculture [41].Thead optionofmanagementzoneswouldreduce thecostoffertilizing,improvecropyields,reducetheus ageofpesticides,providebetterfarmrecords that are essential for sale, and provide better information for management decisions [4]. According to Zhang et al. [38], the number of management zones is a function of the natural variability within the field, the size of the field and certain management factors. If the variability is high, the minimum size of a zone is limited by the possibility of each farmer to differentially manage regions within a fieldineconomicandlogisticsterms.Inadditiontodeci detheareaofworkingzones,theselectionof the specific parameters to be tracked within those zones must be carefully made early in theprocess. Rovira-Más and Saiz-Rubio [65] classified crop biometric traits in a tri-level division of crop features dependingonthefocusofinterestbeingatsoillevel,plan tlevel,orproducelevel. Thisdivision allowed the superimposition of various layers in a standardized map with the aim of determining a data-basedwinequalityindexdefinedastheQualityPotentia lIndex(QPI)foreachsubfieldareaina vineyard. Nevertheless,theremaybespecificcaseswherethespat ialvariabilityofafieldissolowthat a single mapping event can be sufficient, as reported by Klassen et al. [42] when characterizing soil variability in ricefields.
Stage II: Platforms SupportingSensors Sensorsaretheuniversaldevicestomonitorcr opsandtoobtainobjectiveinformationfromthem. They are usually integrated in a platform, which is the general term used in Figure1to name the structureswheresensorsareplacedandcarried.Thesepl atformsmaybeattachedtooff-roadvehicles orfixedtothegroundwithinfieldssuchaslocalweathers tations.Oneofthemosturgentchallengesto copewithinthenextfewyearswillbegettingawiderrang eofnon-invasivesensorsabletomeasure on-the-go. This approach would be closer to Agriculture 5.0, as these sensors could be attached to autonomous platforms and robots. Nowadays, not all the parameters of interest can be measured noninvasivelyandatadistancefromthetarget;however,so metechnologiessuchasmultispectralor hyperspectral imaging are making significantimprovements.

Remote Sensing Platforms:Satellites
Remote sensing has played a key role in the progress of Smart Farming when field data became Importan tsatellitesprovidingagriculturalinformationare the American Landsat satellites (eight satellites take spectral data from the Earth each 16 to 18 days), the European Sentinel 2 satellite system (it provides multispectral data at 10 m pixel resolution for NDVI-NormalizedDifferenceVegetationIndeximagery,soil,andwatercovereverytendays),the RapidEyeconstellation(fivesatellitesprovidemultisp ectralRGBimagery,aswellasred-edgeandNIR bandsat5mresolution),theGeoEye-1system(capturesmultispectralRGBdataandNIRdata ata 1.84 m resolution), and the WorldView-3 (collects multispectral data from the RGB bands including the red-edge, two NIR bands, and 8 SWIR bands with a resolution of 1.24 m at nadir). IKONOS andQuickBird have been already decommissioned. There exist several reviews on satellite sensing applications,havingrecentstudiesfocusedonthepoten tialapplicationsofthermaltechnologiesusing remote sensing [44] and nutritional status in commodity crops [45].

AircraftSystems
The distance between crops and satellites is considerable, typically around 700 km, and deeper insights are reachable when sensors remain closer to the targets. For aircraft systems, the distance to land can be around 100 m. For example, there is a legal limit of 120 m above the ground in Spain for unmanned flying vehicles. Unmanned aerial vehicles (UAV) and remotelypiloted aircrafts(RPA) can basically be of two kinds: Fixed-wing aircrafts and multirotor aircrafts.

Rotary-wing
UAVs are morestablefliersastheyarecapableofaverticaltakeoffandlanding;however,theyareslowerand cannotcoverasmuchareaduringtheirbatterylife.Fixed -wingplatforms,ontheotherhand,cancover more area per flight and carry larger payloads, but tend to be more expensive and break moreeasily aftermultiplelandings [45].Whencomparedtoremotes ensing,theadvantagesofUAVsforPrecision Agriculture are their flexibility in frequency (revisit time of satellites) and better spatial resolutions. When compared to ground vehicles, UAVs can get data from inaccessible places where conventional equipmentcannotstand;however,theyrequireaprofes sionalplanningoftheflightroutebeforehand, and certain machine vision applications may require flying at midday to avoid vegetation shadows on the ground causing errors with imagery data. Furthermore, post processing the data and image mosaicking is often quite challenging. An important disadvantage of UAVs is the limited payload theycancarry,whichoftenlimitsthesuiteofsensorsonb oard,aswellastheincapacityofflyingwith strongwind.
Proximal Sensing: Ground Autonomous Systems-the Great Push for Agriculture5.0 Whenmonitoringplatformsoperatefromtheground,th edistancefromthesensorstothetarget cropdiminishestolessthan2m.Duetotheproximityoft hesensortotheplant,whendataisacquired fromgroundbasedplatforms,itiscalledproximalsensing. Groundvehiclesarepolyvalentinrelation tothepayloadofsensors.Asthesevehiclesmovenearth ecrop,thedataacquiredincreaseinaccuracy, andresolutionsofoneormoresamplespermeterarefeas ible,beingonlylimitedbythespecifications of the particular sensors implemented. When active sensors are used, weather conditions such as strong sunlight or poor illumination are not a serious problem anymore, and, in case of on-the-fly processing, real-time applications are possible, as spraying weeds with the previous detection of the pest [47].Therehasbeenasignificantimpulseinthelastf iveyearsfortheparticularcasewheredatais retrievedfromanautonomousplatform(unmannedgro undvehicleorUGV) [48][49][50][51][52].Aravindetal. [48] reviewed ground robots for tilling, soil analysis, seeding, transplanting, crop scouting, pest control, weed removal and harvesting, where crop scouting has been defined as the process of continuously monitoring the field to acquire information on the plant status, disease incidence, and infestations affectingcropgrowth.Shamshirietal. [27]describedre centachievementsofUGVsforweedcontrol, field scouting, and harvesting, highlighting that, if successfully integrated and implemented, field scouting robots can play a key role in reducing production cost, increasing productivity and quality, and enabling customized plant and crop treatments.  [56]. RowBot Systems LLC (Minneapolis, MN, USA) patented a robotic platform whose structure was configured to perform several field tasks, as selectively applying fertilizer, mapping growth zones, or seeding cover crop [57]. Over the 20th century, farm productivity has been increasing by augmenting the size of machines, which has led to heavy and oversized equipment. In order to invert this trend, researchers and growers have started to think about alternatives to tractors to avoid soilcompaction.
Shamshiri et al. [27] suggested using various machines instead of one heavy machine. In the same line,Hameed [58]proposedatechnologythatenabledas inglefarmertocontrolateamofautomated vehicles, and Ball et al. [59] used cooperative robots as a measure to control weeds. In fact, there have been several projects implementing more than one machine operating in collaborative work, as the Flourish European project that combines UAVs and UGVs to retrieve information for decision support [46], or the RHEA project where a fleet of autonomous robot units performed treatments in crops [82].

Stage III:Data
One of the fundamental differences between traditional and modern farming is, apart from the mechanization level, the data collected directly from the crops. In traditional farms where growers judge by visual assessment, decisions are relative and subjective. Modern farming offers assessment by quantitative data producing objective decisions. Sensors allow data acquisition in the field, but the special case of non-invasive technologies in combination with on-the-fly sensing from moving platforms has opened the window of massive data collection, a forerunner of big data in agriculture. However, the excess of data is also a serious challenge to cope with, as vital

Maps Containing Relevant FieldFeatures
Displayingdatainacoherentformatiskeyforf inaluserstounderstandwhatishappeninginthe field.Themostcommonwaytodisplayagriculturaldata hasbeenintheformatofmaps,asmappingis usefultodefinespatialtrendsandhomogeneouszones. However,displayingagronomicalinformation inbeautifulmapsshouldnotbethegoalofmapgeneratio n.Mapsneedtobeusefulformakingdecisions, they need to be a help to answer a question, providing an interpretation of spatial information [39]. Thegoalofbuildingmapsisobtainingafewmanagemen tzoneswiththeparametersofinterestsothata treatmentcanbeefficientlyapplied.Togetplausiblema nagementzones,krigingisoneofthemostused interpolationtechniquestodelimitareasofmanageable sizes [43].Takingintoaccounttheconsiderable amount of data that Smart Farming generates, there are many software applications to cope with interpolation,ingeneral,orkriginginparticular[66].Al so,whenbuildingamap,acoordinatesystem needs to be supplied along with the map. One ideal alternative for agricultural maps is brought by theLocalTangentPlane(LTP)coordinatesystem,whic hfeaturesEuclideangeometry,allowsuser-set origins, and employs the intuitive coordinate frame eastnorth. Regarding the coding and display of datainthemaps,gridsallowthesystematicquantization oftheLTPcoordinatesystemtomanagecrop production information more efficiently, facilitating the exchange of information among successive seasons and the comparison of multiple parameters on the same field [67]. A practical example of grid-based maps using LTP coordinates is shown inFigure3. Taking into account the key role of positioning systems, a map-based approach is the method in which a Global Positioning System (GPS)-or any other Global Navigation Satellite System (GNSS)-receiver and a data logger (e.g., an onboard computer) are used to record the position of a particularmeasurement(georeferenceddata),sosever almapscanbegeneratedandprocessedalong with other layers of spatially variable information [68]. In general, GNSS receivers are the universal position devices used to build maps; however, in some cases, for example in greenhouses or dense fieldsoftalltrees,GNSSisnotthebestoptiontousedueto thedifficultyofgettingsignalswithreliable accuracy;so,insomecases,alternativesolutionssuchas machinevisionmustbeimplemented [69].
Data Management Software to Ease the Process of DecisionMaking A popular way to manage field data displayed on maps and culminate with a practical solution is through the use of Geographic Information Systems (GIS). This set of computer-based tools (or dataplatforms)allowstostore,analyze,manipulateand mapanytypeofgeoreferencedinformation. A specific GIS system called the Field-level geographic Information System (FIS) was developed for Precision Agriculture applications [70], but it was set for old computer operativesystems such as Windows 3.1×, 95, 98, or NT [71]. The updated version of FIS is the farm management comply with agricultural standards, and maintain high product quality and safety, guiding growers to make the best decisions possible [95]. Farm management software solutions support the automationofdataacquisitionandprocessing,monitor ing,planning,decisionmaking,documenting, and managing the farm operations [64], and include basic functions for record keeping like crop production rates (harvests and yields), profits and losses, farm tasks scheduling, weather prediction, soil nutrients tracking, and field mapping, up to more complex functionalities for automating field management accounting for farms and agribusinesses (accounting, inventory management, or labor contracts).Inmanycases,growersdonotneedtobefluid ondatamanagementbecausethesoftwarecan build maps or decision-making models with basic information introduced by growers. Furthermore, a critical feature of these applications is that they even help in the early warning of weather-related hazards that enables farmers, policy makers, and aid agencies to mitigate their exposure to risk [83]. However,itmustbetakenintoconsiderationthattheeffi ciencyofarecommendationforaparticular agent will depend on the factors included in the algorithms of the software (technical, economic, safety-wise. . . ). In this sense, a DSSAT (Decision Support System for Agrotechnology Transfer) provides outputs with experimental data for evaluation of crop models, allowing users to compare simulatedoutcomeswithobservedresults,whichiscriti califreal-worlddecisionsorrecommendations arebasedonmodeledresults [84].Table2gathersarepre sentativesetofcommerciallyavailableFMIS programs specifically configured to deal with the usual data generated in the farm. It includes the name of each application program, the company commercializing it with its headquarters location, and the main features of the program. The table is focused on programs managing crop data as the primary tool, and its purpose is not the compilation of all available FMIS software, which would be futile given the rate new applications are constantly released, but bringing a proof of the global effort realized in the last decade to deploy Smart Farming in actual farms, accelerating the move from academics to agribusiness. The examples show that some smartphone and tablet applications alreadyincludecomplexfeaturessothatgrowerscanins ertdatadirectlyinthefield;othercompanies, on the contrary, prefer having a basic application for mobile devices to increase complexity in the cloudbased desktop version. In the majority of cases, it is not necessary to have wireless connection while the grower is entering data in the field, because as soon as the mobile device finds a wireless connectiontotheinternet,itsynchronizesthedataprevi ouslyintroducedbythegrowerinthemobile device with the data safely stored in the cloud. Many of the programs listed below offer the option of upgrading the software depending on specific grower needs, increasing the price accordingly. The most advanced tools include features for financial and machinery management, help in the decisionmakingprocess,releasewarnings,orevenproposeman agementadvice.Inmanycases,these softwareapplicationsarenotonlyaddressedtothegrow erorproducer,butalsotootherstakeholdersin agriculturesuchasinputssuppliers,servicesuppliers,a ndfooddistributors,whichmakesadifference for Smart Farming, where multiple agriculture agents are connected. Regarding exploitation rights, variousagriculturalmanagementsystemshavebeenpa tented,asthesoftwarefromTheClimateCorp. togenerateagricultureprescriptions [85],whichentere dintopartnershipwithAGCOCorporationin 2017 [4]. Decisive Farming Corp. [73,74], AgVerdictInc. [75] or Trimble [86] have also patented their commercialsolutions.   The use of commercial data management systems, as the ones listed in Table2, often implies that producers need to share their crop data with a software platform owned and run by private companies. This fact creates some controversy regarding the ownership of the data. In the Software Services Agreement (SSA), it is stated that the person or entity providing the data to the farm management software company shall own and retain all rights, title and interest in and to their data, so that the data belongs to the provider [76]. However, when data are aggregated with other growers'data,thecombineddatatypicallybecomeprop ertyofthesoftwarecompany [96].Thelistof applications included in Table2proves that there is a global interest in developing software for farm datamanagement,andmostofthefeaturesrequestedby end-usersaresimilareverywhere.Thistable alsogivesanideaoftheinterestraisedinindustrybysoft ware-basedmanagementsystems.However, manyapplicationsusetheirownproprietaryformats,w hichcomplicatestheshareofdataamongdata acquisition and processing systems. A standardization effort is needed among software developers and providers. The ADAPT toolkit of Table2 [77] is an example of how to face this challenge, as it providesanopensourceapplicationthateliminatesabarriertothebroadu seofPrecisionAgriculture data by enabling interoperability between different hardware and softwareapplications.
Stage IV: Decision-Making Insituationswheremanyfieldparametersnee dbeingconsidered,peoplefindpracticaldifficulties in managing complex information to make effective decisions. In such cases, artificial intelligence (AI) can help with techniques like deep learning or neural networks, fuzzy logic, genetic algorithms, orexpertsystems.AI,withitsmodellingandreasoningc apabilities,canplayakeyroleinagriculture, helpingtomakesenseofallthedataavailable.Fuzzylogi c,tonameoneexamplewithinAI,resembles human reasoning imitating the way of making decisions that involve several possibilities instead of ‗true' or ‗false' alternatives; this technique uses linguistic variables that fit well with the complexity of the challenges posed by the diversity of agricultural decision making. According to Dengel [20], agriculture offers a vast application area for all kinds of AI core technologies as agents operating in uncontrolledenvironments.GiustiandMarsili-Libellia[81]designedafuzzy-baseddecisionsupport system (DSS) taking as input variables soil moisture and rain forecast for kiwi, corn, and potato. Similarly, the DSS developed by Navarro-Hellín et al. [87] estimated weekly irrigation for citrus orchardstakingintoaccountclimateandsoilvariables;i nthatwork,real-timemeasurementsfromsoil parametersinaclosedloopcontrolschemeweredecisivetoavoidtheaccumul ativeeffectduetoerrors in consecutive weekly estimations, as the DSS was allowed to adapt to local perturbations. In the samefashion,LindsayCorporation(Omaha,Nebraska ,USA)wasawardedforitssolutionFieldNET Advisor™ [91] that provides irrigation management decisions for growers. DSS may be morerobust and reliable when different variables are considered, but some procedures remain controversial as objectivescanleadtodifferentsolutionsatdifferenttim esbasedontheprioritysetbydecisionmakers or other people involved in the procedure [88]. SrivastavaandSingh [80]highlightedtheimportanceof incorporatingthegraphicalpartofGISto DSS, which was demonstrated for water management scenarios in India. The importance of using GIS for agricultural DSS lies on using user-friendly graphical interfaces for growers. The result of a questionnairedistributedbyVineScoutproject [36] some kind of DSS, and the preferred ways of delivery were software (28%), paper-based (22%) tools, and mobile apps (10%). These results show that the use of software to manage decisions is growing, but its percentage is still low and comparable to those who preferred paper-based tools. Choosing softwareandmobileapplicationstomakeagriculturald ecisionsmaybeconsideredbeneficialbecause digitaltoolsincreasemanagementefficiencywhencom paredtopaper-basedtools;however,thereis stillalongwaytomaketechnologybasedtoolsattractiveenougheasytounderstand,intuitiveand nice-for growers to adopt. On the producer side, it is important to have access to proper training until these technologies can be comfortablymanaged. The last step for closing the loop in the complete crop management cycle of Figure1is the physical actuation on the crop. Actuation is understood as executing some action on the crop or related to it, and this can be done by making decisions right after obtaining information (realtime applications) or in another moment deferred in time (off-line). For farmers to execute decisions, they need advanced equipment that can receive orders from a computerized control unit. Variable rate machines can execute a number of farming tasks driven by a smart system [60]. Variable rate technology (VRT) applied on site-specific crop management (SSCM) has the potential to increase profit and decrease environmental impact [61] as only what is needed is actually applied. Colaço and Molin [92] conducted a long-term study for six years with the goal of evaluating the effects of variable rate fertilization on fertilizer consumption, soil fertility, and yield in citrus. The outcomes of comparing variable and uniform rates showed that the former achieved higher yields while using less fertilizer: using nitrogen, fruit yield (kg of oranges) respect to the amount of fertilizer resulted in a 32% yield increase in field 1, and 38% in field 2. When using potassium, the yield increase even  [3] showed that variable rate technologies had positive, but small, rate adoptions of 1% due to theirdifficultyofuse.Apartfromefficiencyandutility,c ostisalsoacriticalparametertoconsiderfor the adoption of this technology. In this sense, the ubiquitous availability of low-cost electronics will favortheintroductionofsuchdigitalapplications.Infac t,advancesinautonomousdrivingtechnology for cars, including object detection capabilities through multi-camera systems, have already reduced the cost of developing automated agricultural machines [22].

II. DISCUSSION
After the Industrial Revolution, mainly since the advent of mechanization, and along the Green Revolution,humansandmachineshavebeenefficientl ycollaboratingforgrowingcropstofeedpeople. However, to face the population growth in the coming years, an extra effort is needed to succeed, not only in feeding people by increasing productivity, but also in doing it in the most efficient and respectfulpossibleway,thatis,producingsustainably. To facethischallenge,remarkableadvancesin technologyhavebeenappearingoverthelastdecades,in particulartheaccesstoreliableagricultural dataandadvancedcomputertechniquestogettheoptim almeaningfromthem,eventuallyobtaining maximumbenefitswhilebeingrespectfulwiththeenvir onment.Thisnewapproachdrivenbydigital technologyimpliesthatgrowersmustactassupervisors oftheircropsratherthanlaborers,inanattemptof avoiding repetitive, physically-demanding, and tedious field tasks. In this modern agronomical framework,DATAisthekey,andtheinformationbasedmanagementcycledescribedaboveprovides the practical approach that unites concept and tasks. The following points summarize some of the specific ideas drawn from thisstudy: • Precision Agriculture, which consists of applying what is needed when and where is needed, has further improved the efficiency of managing farms with the addition of data-based digital systemsthatincreasetheknowledgeofproducersabout theirfields;thisisknownasAgriculture 4.0 orDigital Farming. When these data-driven farms incorporate robotics with AI algorithms to their systems, the overall concept is then referred to as Agriculture 5.0. Some studies report that agricultural robots integrating forms of AI can do certain tasks faster than humans [23].Despite thereareotherstudiesthatcontradictthisstatement[63], roboticsisagrowingeconomyandthere exists a great potential for many applications withinagriculture.
• A greater adoption of Digital Farming by professional growers is vital to not only improving a farm's financial performance, but also to meet the food needs of an expanding population [6]. Smallfarmswillsteadilyincorporatebasictechnology whereaslargefieldswilllikelyinvestwith sophisticated equipment, but data-less intuitiondriven management will no longer represent the modus operandi of professional farms in the future. This should be considered a source of opportunities, especially for a new generation of young farmers used to digital technology, who are the ones with the capacity to balance an aging population in rural areas, mainly those in industrializedcountries.

•
After the rapid growth of UAVs, a steadystate is being reached, mostly induced by the factthat dataanalysisandgroundtruthvalidationhasresultedfarmorecomplexanddelic atethanimage • Maps, as the most common way to represent agricultural data, would need to be standardized. Intensely-interpolated colored maps are output by GIS, FMIS, and other software applications, butatthetimeofcomparingdatawiththeprecisionenou ghtograntstatisticalsignificance,itoften becomes an impossible mission without standardization. Figure3, for example, uses the flat representation provided by the local tangent plane (LTP) and formatted in a regular grid. Other programs use UTM projections, and there are even images only given in geodetic coordinates. At the need of overlapping maps, it takes a big effort to make all data compatible. Not only the waycoordinatesarerepresentedneedsastandard,butal sotheunits,intervals,andevencolorsin which parameters are displayed. The combination of aerial and ground data, for instance, will greatly benefit from such standardization in the way data is visually displayed for the average grower tounderstand.
• Table2provides a representative compilation of software applications for farm management. The list is not exhaustive, and yet includes companies from four continents and 14 countries, which provides evidence of the fact that agricultural digitalization is in fact a globalmove.

•
Regardingvariablerateapplications,adoptio nratesneedtoaugment,andtodoso,farmersmust find by themselves the value in this technology for their crops.

III. CONCLUSIONS
This analysis confirms that consistent knowledge about farms leads to optimal decisions. Agricultural management systems can handle farm data in such a way that results are orchestrated to address customized solutions for each farm. This aid for farmers in the form of digital solutions combines forces with robotics and artificial intelligence to launch the imminent idea of Agriculture 5.0. After thirty years of great expectations-and disappointments-by the application of robotics to agriculture, the timing seems right for the first time. However, in order to take the most advantages from Agriculture 5.0, deep training needs to be delivered to users, ideally young farmers eager to learnandapplymoderntechnologiestoagricultureand grantingagenerationalrenewalstilltocome. It seems to be the right time to move forward towards a modern and sustainable agriculture that is capable of showing the full power of data-driven management to face the challenges posed to food production in the 21st Century. The evolution to Agriculture 5.0 is in the agenda of most major farm equipment makers for the next decade, and therefore off-road equipment manufacturers will playakeyroleinthismoveifagriculturalrobotsareconsi deredasthenext-smarter-generationof farmmachines.