An Integrated Data Platform for Agricultural Data Analyses based on

Agricultural ISOBUS and ISOXML

Franz Kraatz, Heiko Tapken, Frank Nordemann, Thorben Iggena, Maik Fruhner and Ralf T

onjes

Faculty of Engineering and Computer Science, Osnabr

uck University of Applied Siences,

Albrechtstr. 30, Osnabr

uck, Germany

Keywords:

Agricultural Data Platform, Data Integration, Data Warehouse, Data Lake.

Abstract:

Over the last years many different agricultural online management portals got to market. The focus of these

portals is on documentation, accounting and task planning. Data analyses and process planning are often not

considerd. For this reason, the existing data in the data platforms of present portals is often badly integrated

and consequently not designed for data analyses. This paper introduces a new architecture concept for an

integrated agricultural data platform. With this new data platform agricultural data analyses for precision

farming become possible. Furthermore, the integration of the agricultural devices and external sources into

one platform changes task planning for one machine into a process planning for cooperated machines. Several

challenges for the integration of agricultural data and data types for agricultural data analyses are discussed.

1 INTRODUCTION

The agricultural industry is still in an extreme state

of change. Since electronic equipment on the agri-

cultural machinery is established, the digital develop-

ment of equipment progresses. One of the essential

development steps was the launch of ISOBUS, stan-

dardised in ISO11783 (ISO, 2007).

Parallel to ISOBUS a huge variety of sensor sys-

tems got to market. Two well-known systems are

the ISARIA sensor for mineral fertilising (Fritzmeier

Umwelttechnik, 2018) and the NIRS sensor for dry

substance measurement used for maize harvesting

(Maschinenfabrik Bernard Krone, 2015).

To handle this digital data, the farmer can organ-

ise inventory and build tasks in farm management in-

formation systems (FMIS). For task planning service

providers or web services, like weather forecast or

precision farming providers, are connected to FMIS.

Machinery logged process data is archived and can be

displayed for evaluation and documentation.

Most available FMISs are not designed for anal-

yses, process planning and process automation. The

main issue is the separate handling of heterogeneous

data. FMISs are often not based on an integrated

database, storing information, like soil and yield in-

formation, in a common way for data analyses. For

that reason the potential of collected process data is

not fully used to optimise agricultural processes.

New electronic equipment and partial area based

documentation enable precision farming. With this

method the ﬁeld is not considered as a homogeneous

area. The ﬁeld is divided into subareas and can be

farmed individually. As a result the harvest rises with

the same resource input or less resources are needed

for the same output. In both cases the efﬁciency rises

and the environmental damage is decreased. Precision

farming will be the future direction for agricultural

development because there is an increased demand for

food to feed the growing population and to handle the

reduction of available farmland. In Germany agricul-

tural used areas shrank from 53.5% in 2000 to 51.6%

in 2015 (Umweltbundesamt, 2018). At the other hand

the population stays nearly constant at 82 million peo-

ple (Federal Agency for Civic Education, 2015). In

comparison the worlds population raised from 6.1 bil-

lion to 7.3 billion in this period (Federal Agency for

Civic Education, 2017).

The research project OPeRAte - ”Orchestration of

Process Chains for data-driven Resource Optimisa-

tion in Agricultural Business and Engineering” (OP-

eRAte, 2018) addresses these topics for the next step

in farming 4.0. To realise data analyses and process

optimisation in agriculture one project target is an in-

tegrated data platform. The existing inventory and

log data are combined with heterogeneous informa-

tion from web services, service providers and other

external data portals.

422

Kraatz, F., Tapken, H., Nordemann, F., Iggena, T., Fruhner, M. and Tönjes, R.

An Integrated Data Platform for Agricultural Data Analyses based on Agricultural ISOBUS and ISOXML.

DOI: 10.5220/0007760304220429

In Proceedings of the 4th International Conference on Internet of Things, Big Data and Security (IoTBDS 2019), pages 422-429

ISBN: 978-989-758-369-8

This paper introduces an architectural concept for

an integrated agricultural data platform designed for

data analyses and distributed mobile data. Chap-

ter 2 discusses the FMIS, the agricultural data for-

mat ISOXML and different data sources for precision

farming. This is followed by the state of the art on

data lake and Internet of Things (IoT). Afterwards it

is shown how IoT functionality can be provided on

the agricultural machine. The challenges of data inte-

gration into the data platform are discussed in chapter

5. Before concluding the paper with the summary,

chapter 6 presents the new architecture concept of the

integrated agricultural data platform in detail.

2 INTRODUCTION OF

AGRICULTURAL DOMAIN

2.1 Farm Management Information

Systems

Farm management information systems have been es-

tablished to organise a farm with machinery, ﬁelds,

resources, and task planning as a desktop software.

During the last years many FMIS online portals got to

market (NEXT Farming, 2019), (365FramNet, 2019).

Thereby the farmer can access the FMIS with a com-

puter, mobile device and other electronic equipment.

In Germany the number of agricultural companies

decreases about 20 - 30 times faster (Federal Institute

for Population Research, 2018) than the agricultural

surface (Umweltbundesamt, 2018). Consequently the

remaining farmers need the assistance of an FMIS to

handle the additional ground and work economically.

In order to achieve a suitable workload of the large

and expensive machines, FMISs allow the farmer to

plan tasks digitally. Afterwards the tasks can be trans-

ferred to and integrated in to the machinery. Finished

tasks combined with logged process data are the re-

sult of these jobs and will be transferred to the FMIS

for documentation and interpretation (Figure 1).

The information exchange between FMISs and

machinery is based on the ISOXML format and is not

prepared for a data exchange during task runtime. A

summary of this format follows in section 2.2. Most

of the time the physical transport is realised by mobile

storage mediums. The transfer via mobile communi-

cation and internet is not provided by every system.

Proprietary telematic systems (Claas E-Systems,

2019) are able to collect and store parameters of the

machinery during task runtime for the manufacturers.

In a portal the farmer has access to this information

for live monitoring and parameter evaluation.

Figure 1: Data exchange between FMIS and agricultural

machine.

2.2 Agricultural Data Format

Based on the ISO11783 standard, called ISOBUS,

the ISOXML structure is the agricultural data ex-

change format between machinery and FMISs. Each

XML element has several mandatory and optional at-

tributes. Representative for each attribute name a

character is deﬁned. Similar to this, the element name

in the xml schema is represented by a shortcut of three

characters. Some of the attributes hold references to

other elements in the structure.

In the following ISOXML example a farm called

Hof Herrmann and a ﬁeld called Exx Platz is shown.

1:<FRM A="FRM1" B="Hof Herrmann"/>

2:<PFD A="PFD1" C="Exx_Platz" F="FRM1">

3: <PLN A="1">

4: <LSG A="1">

5: <PNT A="2" C="52.28.." D="8.02.."/>

6: <PNT A="2" C="52.28.." D="8.02.."/>

7: <PNT A="2" C="52.28.." D="8.02.."/>

8: <PNT A="2" C="52.28.." D="8.02.."/>

9: <PNT A="2" C="52.28.." D="8.02.."/>

10: </LSG>

11: </PLN>

12:</PFD>

The ﬁeld (line 2) has a reference to the farm (line

1) at the attribute F. Part of the ﬁeld element is a poly-

gon (line 3) with a line element (line 4) and ﬁve points

(line 5-9) for the ﬁeld border. All elements inside the

ﬁeld do not have an additional identiﬁcation as they

are assigned inside the ﬁeld element.

Main problem is the interpretation of the stan-

dard by the manufacturers. For example, optional at-

tributes are deﬁned as mandatory attributes. A sec-

ond problem is the data exchange between different

FMISs without user interaction because of the unique

identiﬁcation only inside the transfer ﬁle. For this rea-

son only information of executed tasks from machin-

ery can be automatically imported, if the planed task

was generated by this FMIS before. For an automatic

integration in every situation a comparison of selected

An Integrated Data Platform for Agricultural Data Analyses based on Agricultural ISOBUS and ISOXML

423

descriptive string attributes has to be made. This does

not preclude duplicate records in the database and

rises the amount of redundant linked data from the

external sources discussed in the next section.

2.3 External Data Sources for Precision

Farming

To realise precision farming different data sources

have to be used to represent the inﬂuencing factors

for crop production. An example is a combination of

yield information of the last years to get a yield poten-

tial map. To improve quality an additional combina-

tion with the soil type map (State Ofﬁce for Mining,

Energy and Geology, 2019) is used in practice.

A large part of external data sources are geo-

graphic data, like historical weather data (Deutscher

Wetterdienst, 2018), weather forecasts, soil maps,

topological maps and process data. Furthermore

satellite images (German Aerospace Center, 2019)

provide information of crop growth and the condition

outside of the ﬁeld, e.g. water bodies. To get match-

able geographic data for analyses, the granularity of

the data must be dissolved.

Important external data sources are legal re-

quirements of the government for nitrate (European

Comission, 2019) and pesticide (European Comis-

sion, 2018) application. Failure to comply will result

in penalties and compromise the environment. Only

in combination with a geographic map and the apply-

ing product it is usable for precision farming.

Furthermore data sheets of the agricultural ma-

chinery are external data sources used for precision

farming. A precision farming application can also be

optimised for the properties of the machinery or for

required absolute values for application rates adapted

to the product.

3 STATE OF THE ART

3.1 Data Lake

A data warehouse (Inmon et al., 2008) combines data

from different sources in an integrated schema based

on requirements of deﬁned analyses, evaluation, and

reporting tasks. Distributed data is combined for anal-

yses, evaluation and reporting tasks in a data lake.

This approach is deﬁned as a non ﬁxed coupling of

structured, semi structured and unstructured data in a

central data management system (Maccioni and Tor-

lone, 2017), (Tomcy and Pankaj, 2017) and is able

to apply the data warehousing concept. The main

components of a data lake are the connectors with

different quantity and complexity levels. This set-

ting is mostly chosen at semantic-aware big data sys-

tems (Nadal et al., 2017) and context-aware data lakes

(Ahmed et al., 2017) for the data management.

For data management at data lakes, extract, trans-

form and load (ETL) tools are used (Zhu, 2017). If

there are already different kinds of data silos avail-

able, methods of machine learning ﬁt for the data in-

tegration (Wibowo et al., 2017). Additional concepts

load all data in the data lake and do the integration

and organisation of the data lake afterwards (Terriz-

zano et al., 2015).

Figure 2: Architecture of mediator-wrapper technology.

With the mediator-wrapper technology (Figure 2)

external, heterogeneous data can be connected to a

data lake (Wiederhold, 1992). The wrapper adapts

the data and communication from external sources to

the mediator and provides a uniﬁed access to a data

source. The mediator combines the adapted data from

wrappers at the same domain as a single data source to

higher applications. To have an efﬁcient implemented

mediator it is fundamental to get the right balance at

service level and domain level. An example for wrap-

per design is web wrapper.

After connecting the external sources to the data

lake, the distribution of data can be handled by a

publish/subscribe system in combination with a mes-

sage passing system. Published data will be automati-

cally forwarded to the subscribed users. For handling

the data ﬂow from a publisher to the subscribers a

message passing system like kafka (Apache Software

Foundation, 2019) can be used.

3.2 Internet of Things

The Internet of Things (IoT) combines several tech-

nologies to connect cyber physical systems with each

other. The systems provide their information in a

open format to the network. A lot of information

and functions combined with a interface description

are available as commercial or open source solutions.

IoTBDS 2019 - 4th International Conference on Internet of Things, Big Data and Security

424

By using ontologies (Bermudez-Edo et al., 2016) it is

possible to automatically process and integrate these

interface descriptions.

The heterogeneity and variety of IoT also includes

the individual system properties, e.g. the currency or

reliability of information. In the end, you have to

trust the provided description or merge several sys-

tems with identical information to derive information

about the individual system properties.

For information exchange, message bus systems

are used in the IoT. The information is published on a

mesage bus, like Message Queuing Telemetry Trans-

port protocol (MQTT) (Stanford-Clark and Nipper,

2019), to reach the subscribed system. The exchange

of sensitive information in IoT is realised by access

control to the message bus.

4 IoT ENABLED SMART

FARMING

For an independent integration of the different sensor

systems of a modern agricultural machine in an FMIS,

a self description of the machinery is needed. With

the description additional standalone sensor systems

can be used as Internet of Things devices to support

the agricultural application.

In the research project OPeRAte such a universal

valid description to handle a machine like an Internet

of Things device has been developed. To get the avail-

able process data from the machine during a running

agricultural process, an essential function is missing.

The data exchange between machinery and FMIS

is currently only possible before and after an agricul-

tural task is processed. An extension of the ISOBUS

standard with a streaming protocol for data stream-

ing during application is still an ongoing task (Harris,

2018). For process regulation or adaptation there is

no controlling function available.

To solve this problem, the OPeRAte project is de-

veloping a separate ISOBUS module. An individ-

ual adaptation of the task processing unit (Task Con-

troller) on the machine is too complex. With a Task

Controller (TC) client the module can have a direct

online connection and a peer-control connection via

ISOBUS to the implement. Similarly, the client can

be commissioned by the TC for the recording of the

log data. This client is much easier to adapt to new re-

quirements and can still take over almost all functions

from the TC. Only a simpliﬁed part of a task needs to

be integrated into the TC.

5 CHALLENGES FOR THE DATA

ENRICHMENT IN

AGRICULTURAL AREA

5.1 Different Level of Granularity

Data analyses at the ﬁnest level of granularity often

do not show all relationships. A ﬁrst abstract view

at higher granularity level opens up new insights for

detailed analyses at ﬁner resolution. An example is

a summary of the different agricultural resources to

categories. At the ﬁrst step winter and spring barley

together make up the category barley. Barley together

with wheat and rye forms the next category cereals.

Similar a temporal summary of information can be

useful for analyses. A good example is the water sup-

ply for plants over the full growth phase and the im-

pact on work planning and the resulting yield.

For realising data analyses at these different levels

of granularity, information must be available in the

smallest granularity level, if possible. The summary

in a higher level of granularity takes place via sepa-

rately stored ﬁlters, categorization or linkage.

For the representation of such information in dif-

ferent levels of granularity so-called data cubes are

used (Totok, 2000). Here the data is arranged in

a multi-dimensional cube and the user can analyse

within the different planes. Most of the time the data

itself is already stored as a data cube in order to be

quickly available during analyses.

5.2 Data Identiﬁcation at ISOXML

For ISOXML ﬁles, a simpliﬁed example shown in

section 2.2, no unique identiﬁcation is deﬁned. Only

a manufacturer given machine identiﬁcation ISON-

AME is included. Alternatively, the coordinates of

the ﬁeld can be used for identiﬁcation. For the other

data only an identiﬁcation with the available links to

the machines and ﬁelds or a string comparison of de-

scription is possible.

Figure 3: Two different ﬁeld borders for the same ﬁeld.

An Integrated Data Platform for Agricultural Data Analyses based on Agricultural ISOBUS and ISOXML

425

A problem of identiﬁcation by location is the dis-

crepancy in locations at the different FMISs shown

in Figure 3 with green and red colour. To solve that

problem, a variation rate has to be deﬁned. For ex-

ample, a difference for the ﬁeld area matching of ﬁve

percent.

5.3 Process Data Cleansing

The sensor value recording takes place during task

processing without observing the working state for a

deﬁned interval or by threshold values. Yield record-

ing is carried out, for example, position according to

a distance interval. If a machine turns at the end of

the ﬁeld during harvest, a yield of 0 is recorded. The

same recording problem occurs for start-up, run out

and back ride. These errors must be removed, other-

wise the average yield, for example, is falsiﬁed.

Manually collected data, that has subsequently

been digitised, is still used in agricultural. To use

these data, overlaps, gaps or self-intersects of poly-

gons have to be removed.

The ﬁve coordinated steps of data cleansing take

place with saving all raw data. Thereafter, the require-

ments of the data are derived from the properties of

the data. If the requirements for the data are present,

an analysis shows which data meets the requirements.

Before the data is cleaned up, the data goes through

standardisation to resolve remediable errors.

5.4 Runtime Data Handling

Agricultural machinery record large numbers of sen-

sor values. To provide this captured information di-

rectly to the system during processing, the previously

presented treatment steps have to be integrated di-

rectly into the provisioning process.

Thus the records can be included directly in new

analyses. In order to reduce the resulting queries of

the external data sources, the data sources should be

subdivided into the following categories.

• local available with online connection

• only local available

• access only with online connection

To avoid expensive queries from external sources,

a local part for the existing datasets or the entire ex-

ternal source is the better solution. The information

of the source is available in the prepared state during

task processing and is updated at a ﬁxed interval or at

a new dataset. Sources without an online connected

interface have to read locally in the system.

The different levels of granularity can be used for

an additional optimisation of source queries. Assign-

ing the information of an external source to a higher

granularity level also requires queries only for records

at that level. The data in the underlying levels can be

added to the dataset without a source query.

6 AN AGRICULTURAL DATA

PLATFORM FOR DATA

ANALYSES

6.1 Provided Characteristics of the

Agricultural Data Platform

The data platform has to support different character-

istics focused on FMIS functionality and data anal-

yses. For task management and process planning a

data pool with user, resource, yard, and ﬁeld man-

agement is needed. The logged process data has to

be integrated during task execution and prepared for

analyses, process optimisation and the required docu-

mentation.

Second, the platform has to connect the distributed

external sources and mobile devices to a common ex-

change layer. The exchange layer enables the avail-

ability of the external sources at the agricultural de-

vice without data lake interaction and a data access

from process management for process controlling.

6.2 Concept of the Agricultural Data

Lake

The main component of the data platform, the data

lake, works as a single point of truth and supports dis-

tributed data subsets on mobile devices like agricul-

tural machinery. In the opposite direction the device

transfers the process data back. The data will be en-

riched with additional information from external data

sources during the integration process. The runtime

integration process is necessary to provide the data

for reactive process optimisation across machines per-

formed by a connected process management.

The core of the data lake is build on the ISOXML

format described in section 2.2. Therefore, the archi-

tecture is following the data warehouse approach with

a database, ETL processes and a data warehouse. The

database is realised with relational database technol-

ogy and an extension for spatial functions.

To enrich the core data, different additional data

sources are integrated, e.g. historical weather infor-

mation of Germany (Deutscher Wetterdienst, 2018),

IoTBDS 2019 - 4th International Conference on Internet of Things, Big Data and Security

426

raw data

MD1

MD3

MD4

MD5

S1D1

S2D1

S3D1

S1D2

S2D2

publish/subscribe + messaging

MD6

FMIS

analysis

data

AgriRouter

FMIS

process management

AgriRouter

transform

load

process

storage

Sources /

Domains

Mobile

Device

Agricultural Data Lake

extract

MD2

Figure 4: Architecture of the data platform with agricultural data lake.

soil types of Germany (Federal Institute for Geo-

sciences and Natural Resources, 2019), ﬁeld borders

of Lower Saxony (Servicezentrum Landentwicklung

und Agrarf

orderung, 2019) and sentinel2 satellite im-

ages (German Aerospace Center, 2019). The different

categories of enriched data are as follow:

• Spatio temporal data: A signiﬁcant proportion of

data in agricultural has a spatial and temporal cor-

relation. For example the recorded yield with the

position and associated time stamp is used in anal-

yses, separated in subareas for each year.

• Spatial data: If the time factor disappears from

the data, the data only provides information about

a speciﬁc location. Here you can call soil maps as

examples. Once captured, this data is used with-

out any time limit.

• Meta data: In order to get properties of resources,

consumables and plants, additional information

needs to be collected, if it is not supplied by the

product. An interface for machine-speciﬁc prop-

erties for the manufacturers database for business

and service purposes would help. Meta data on

mineral fertilizers and pesticides can often be ob-

tained from the central licensing authority.

In parallel to the ﬁxed linked data, temporal linked

data is possible too, for example weather forecasts.

The forecast is linked to the core to include the in-

formation into decision making. If the process plan-

ning is not ﬁnished or cancelled, the weather forecast

is deleted like a garbage collector. Only permanent

linked data stays in the data lake to keep the growth

of data small. This ensures to keep important correla-

tions for decision reconstruction and evaluation.

The measured data from the agricultural ma-

chineries is transferred to the data lake during applica-

tion execution. This data is enriched by the following

integration with additional data in real time. Missing

additional data for this dataset as well is reloaded in

real time from the external data source.

The technical architecture is built-up with classic

data warehouse technology. There is a multi-layer

mediator/wrapper in combination with a publish sub-

scribe system shown in Figure 4. With the publish

subscribe system other applications can get enriched

data from the data lake or directly from external data

sources. The integrated topic structure enables a se-

lective access to sensitive data.

The functions of wrappers are divided into differ-

ent adaptation aspects, which handle the granularity

levels. One function solves the temporal issues like

time zone and resolution. Another function solves

the different spatial issues, e.g. the adaptation be-

tween different geographical coordinate systems and

formats. Wrappers for the machinery also contain

a function to perform data cleansing of the process

and sensor data. At the next level the mediator com-

bines different external data sources with similar in-

formation. By this procedure a temporarily not reach-

able external data source is compensated. In addi-

An Integrated Data Platform for Agricultural Data Analyses based on Agricultural ISOBUS and ISOXML

427

tion to the combination mechanism the different quo-

rum (Ozsu and Valduriez, 1991) techniques best vote,

weighted vote and majority vote are implemented.

The prepared information is available at the pub-

lish subscribe system by the mediator as a publisher.

All different connected players such as the core data

base, the data lake or third party actors, like mobile

devices on the ﬁeld, can register to the available ser-

vices at the publish subscribe system. This technique

reduces the complexity of the architecture and allows

a ﬂexible mapping to new agricultural applications.

For the integration of new ISOXML data from

agricultural machinery a trigger for validation of ex-

isting enriched data in the data lake is performed.

For this purpose, a message-passing system is used,

which receives the new ISOXML data and sum-

marises the data to data sets on the same reference.

Thereby, the veriﬁcation of enriched data must be per-

formed only for the summary and the identiﬁcation

problem, shown in section 5.2, is dissolved.

7 CONCLUSIONS

The need for an increasing resource efﬁciency in agri-

culture, for farm economics and to feed the world’s

growing population, is shown in the introduction. The

establishment of electronic systems such as sensors

and farm management information systems in agri-

culture contributes to this as well as precision farm-

ing. However, the available FMISs, with their badly

integrated data platforms, have limited scope for data

analyses and cross-machine process planning for col-

laborative machines.

The presented data platform for agriculture ad-

dresses this problem. The data lake architecture

already considers the integration of different data

sources as well as the connection of the agricultural

machine itself. For the integration of an agricultural

machine to the data platform, a new ISOBUS module,

comparable to an IoT device, is introduced. Within

the architecture the integration of the external data

source and the agricultural machines takes place by

mediator-wrapper method.

The data exchange between the connected compo-

nents within the data platform is realised via a mes-

sage bus according to the publish/subscribe principle.

With this method, external data sources are available

to the agricultural machinery for assistance in task

processing. Topics in the message bus system are

used for selected access to sensitive data.

The challenges of data processing in the compo-

nents of the new agricultural data platform are dis-

cussed. For example, the data often has a different

granularity, has to be cleaned up for further process-

ing and must be available for new analyses or process

optimisation at runtime. In order to ensure data pro-

vision at runtime, a different integration of external

data sources according to their characteristics has to

be considered additionally.

The distributed and modular design of the new

agricultural data platform ensures adaptation and ex-

tension capabilities for future topics. For an easier

integration of external sources, a partially to fully au-

tomatic integration process would make sense.

ACKNOWLEDGEMENTS

The project is supported by funds of the Federal Min-

istry of Food and Agriculture (BMEL) based on a

decision of the Parliament of the Federal Republic

of Germany via the Federal Ofﬁce for Agriculture

and Food (BLE) under the innovation support pro-

gramme.

REFERENCES

365FramNet (2019). 365FarmNet - the innovative software

solution for your entire agricultural holding. https://

www.365farmnet.com/en/ accessed 20/01/19.

Ahmed, E., Yaqoob, I., Hashem, I. A. T., Khan, I., Ahmed,

A. I. A., Imran, M., and Vasilakos, A. V. (2017). The

role of big data analytics in Internet of Things. Com-

puter Networks, 129(Part 2):459 – 471. Special Issue

on 5G Wireless Networks for IoT and Body Sensors.

Apache Software Foundation (2019). Apache kafka - adis-

tributed strstream platform. https://kafka.apache.org/

25/01/19.

Bermudez-Edo, M., Elsaleh, T., Barnaghi, P., and Taylor, K.

(2016). Iot-lite: A lightweight semantic model for the

internet of things. In 2016 Intl IEEE Conferences on

Ubiquitous Intelligence & Computing, Advanced and

Trusted Computing, Scalable Computing and Commu-

nications, Cloud and Big Data Computing, Internet of

People, and Smart World Congress.

Claas E-Systems (2019). How TELEMATICS

works. http://www.claas-e-systems.com/en/oem-

products/telematics/ accessed 30/01/19.

Deutscher Wetterdienst (2018). Climate Data Center. https:

//www.dwd.de/EN/ourservices/cdcftp/cdcftp.html ac-

cessed 22/12/18.

European Comission (2018). Agriculture and pesticides.

https://ec.europa.eu/agriculture/envir/pesticides\ en

accessed 29/12/18.

European Comission (2019). Agriculture and nitrates.

https://ec.europa.eu/agriculture/envir/nitrates\ en ac-

cessed 29/01/19.

Federal Agency for Civic Education (2015). Deutsche

Bev

olkerungsentwicklung. http://www.bpb.de/

IoTBDS 2019 - 4th International Conference on Internet of Things, Big Data and Security

428

nachschlagen/zahlen-und-fakten/soziale-situation-in-

deutschland/61541/altersstruktur accessed 15/12/18.

Federal Agency for Civic Education (2017). Globale

Bev

olkerungsentwicklung. http://www.bpb.de/

nachschlagen/zahlen-und-fakten/globalisierung/

52699/bevoelkerungsentwicklung accessed 12/12/18.

Federal Institute for Geosciences and Natural Resources

(2019). Boden

ubersichtskarte. https://www.bgr.

bund.de/DE/Themen/Boden/Projekte/

Informationsgrundlagen-laufend/BUEK200/

BUEK200.html accessed 20/02/19.

Federal Institute for Population Research (2018). Agri-

cultural holdings. https://www.destatis.de/EN/

FactsFigures/EconomicSectors/AgricultureForestry

Fisheries/AgriculturalHoldings/Tables/Agricultural

HoldingsandUtilisedAgriculturalAreabySizeofthe

UtilisedAgriculturalArea.html accessed 11/12/18.

Fritzmeier Umwelttechnik (2018). ISARIA -

Your system for intelligent crop management.

http://fritzmeier-umwelttechnik.com/isaria/?lang=en

accessed 15/12/18.

German Aerospace Center (2019). Copernicus Data and

Exploitation Platform – Deutschland (CODE-DE).

https://code-de.org/en accessed 17/01/19.

Harris, T. (2018). AEF takes on automation,

machine-to-machine communication. https://

www.farmprogress.com/technology/aef-takes-

automation-machine-machine-communication

accessed 20/01/19.

Inmon, W., Strauss, D., and Neushloss, G. (2008). DW

2.0: The Architecture for the Next Generation of Data

Warehousing. Morgan Kaufmann.

ISO (2007). ISO 11783 Tractors and machinery for agricul-

ture and forestry - Serial control and communications

data network.

Maccioni, A. and Torlone, R. (2017). Crossing the ﬁnish

line faster when paddling the Data Lake with KAYAK.

In Proceedings of the 43rd International Conference

on Very Large Data Base, volume 10, pages 1853 –

1856, Munich, Germany.

Maschinenfabrik Bernard Krone (2015). DLG rates Krone

NIR sensor ’good’. http://landmaschinen.krone.de/

english/news/news-archive/news-archive-2015/dlg-

rates-krone-nir-sensor-good/ accessed 14/12/18.

Nadal, S., Herrero, V., Romero, O., Abell, A., Franch, X.,

Vansummeren, S., and Valerio, D. (2017). A software

reference architecture for semantic-aware Big Data

systems. Information and Software Technology, 90:75

– 92.

NEXT Farming (2019). Farm management at the touch

of a button. https://www.nextfarming.de/en/products/

next-farming-live/ accessed 23/01/19.

OPeRAte (2018). OPeRAte - Orchestration of Pro-

cess Chains for data-driven Resource Optimization in

Agricultural Business and Engineering. http://operate.

edvsz.hs-osnabrueck.de/ accessed 18/12/18.

Ozsu, T. M. and Valduriez, P. (1991). Principles of dis-

tributed database systems (2nd ed.). Prentice-Hall,

Inc.

Servicezentrum Landentwicklung und Agrarf

orderung

(2019). Schlaginfo-Portal. https://sla.

niedersachsen.de/agrarfoerderung/schlaginfo/ ac-

cessed 15/02/19.

Stanford-Clark, A. and Nipper, A. (2019). Message queuing

telemetry transport protocol. http://mqtt.org/ accessed

28/01/19.

State Ofﬁce for Mining, Energy and Geology (2019).

Lower Saxony Soil Information System NIBIS.

https://www.lbeg.niedersachsen.de/kartenserver/

nibis/niedersaechsisches-bodeninformationssystem-

nibis-841.htmlengl\ 0503 accessed 10/01/19.

Terrizzano, I. G., Schwarz, P. M., Roth, M., and Colino,

J. E. (2015). Data Wrangling: The Challenging Your-

ney from the Wild to the Lake. In CIDR.

Tomcy, J. and Pankaj, M. (2017). Data Lake for Enterprises

- Leveraging Lambda Architecture for building Enter-

prise Data Lake. Packt Publishing Ltd.

Totok, A. (2000). Modellierung von OLAP- und Data-

Warehouse-Systemen. Deutscher Universit

atsverlag.

Umweltbundesamt (2018). Struktur der Fl

achennutzung.

https://www.umweltbundesamt.de/daten/ﬂaeche-

boden-land-oekosysteme/ﬂaeche/struktur-der-

ﬂaechennutzung accessed 13/12/18.

Wibowo, M., Sulaiman, S., and Shamsuddin, S. M. (2017).

Machine Learning in Data Lake for Combining Data

Silos, pages 294–306. Springer International Publish-

ing.

Wiederhold, G. (1992). Mediators in the architecture of fu-

ture information systems. Computer, 25(3):38–49.

Zhu, D. (2017). Large Scale ETL Design, Optimization

and Implementation Based On Spark and AWS Plat-

form. Master’s thesis, KTH, School of Information

and Communication Technology (ICT).

An Integrated Data Platform for Agricultural Data Analyses based on Agricultural ISOBUS and ISOXML

429