Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
On the Horizon: Smart Agriculture and Big Data
Dr. Rozita Dara Assistant Professor
School of computer Science University of Guelph
Data Collec9on Data Storage
Data Prepara9on Data Analysis
& Usage
Data Management and Privacy Governance Lab
“an all encompassing term for any collec>on of data sets so large and complex that it becomes
difficult to process using on hand data management tools or tradi>onal data
processing applica>ons”
Source: hEps://ipp.cifs.cornell.edu/sites/ipp.cifs.cornell.edu/files/shared/Wiedmann%20IPP%20summit.pdf
Four Vs of Big Data
hEp://www.ibmbigdatahub.com/sites/default/files/infographic_file/4-‐Vs-‐of-‐big-‐data.jpg
hEps://csironewsblog.files.wordpress.com/2013/06/smart-‐farming-‐infographic_final-‐png.jpg
Agriculture Data Growth
Supply Chain Complexity
• A cup of Starbucks coffee can depend on 19 countries: coffee, milk, sugar, paper cup, and other factors.
extensions to their drug product portfolio to layer relevant data sets together to provide transparency into the status of an individual animal at any point in time along its lifecycle. Regulatory bodies and policy makers will also soon realise the bene!ts of improved data collection, as some European countries now require farm-level data and documentation on antibiotic use (Maron et al., 2013). In turn, these digital technologies are ushering in a new era of best practices by enabling farmers and veterinarians to increase clinical treatment performance, and improve farm productivity and overall animal welfare at the same time. The IoAHT will create a new level of transparency and become a necessity in the study of translational medicine and for global research initiatives such as ‘One Health’. Digital technology, and the transformational societal bene!ts that it promises, have an entirely di"erent research and development process to drug technology and present a relatively new ‘space’ for pharmaceutical companies. This emerging ‘space’ presents new challenges, and not just for the pharmaceutical animal health company but rather for all
stakeholders dependent upon our societal food production capabilities.We carry a societal responsibility to use data in a positive way to maximise value to the production chain, while protecting the rights of individuals. With the development of the IoAHT and the intensi!cation of farming driving the rise of PLF, we now have a potential mechanism to support the collection of redacted data from individual animals to utilise big data for societal bene!t. The use of data in isolation does not ful!l its potential bene!ts: greater transparency of the food chain, improved traceability, as well as further improvements to animal health and welfare. This big data is essential when de!ning governmental policies, identifying new population trends and cultural shifts and allocating resources e#ciently; think of the value of human census data, which is personal data used for a societal bene!t. We have a moral responsibility to use big data e"ectively to oversee animal wellbeing, attempt to mitigate the forecast GAP index shortfall and to enhance food security as a result of a rapidly growing world population.
GLOBAL DEMAND FOR MEAT2005 vs 2050(in tonnes)
2005
2050
64M
106M
13M25M
100M
143M
82M
181M
62M
102M
BEEF MUTTON PORK POULTRY EGGS
Figure 2: Global demand for meat in 2050 (adapted from FAO, 2012; Gates Notes 2013)
Global broiler produc>on market stands at approximately 82 million tones of meat.
hEp://image.slidesharecdn.com/kpmgbigdatainhealthcare-‐13454169797539-‐phpapp01-‐120819175826-‐phpapp01/95/big-‐data-‐in-‐healthcare-‐6-‐728.jpg?cb=1345399180
Supply Chain Complexity
• A cup of Starbucks coffee can depend on 19 countries: coffee, milk, sugar, paper cup, and other factors.
hEp://www.fao.org/3/a-‐ae930e/ae930e09.htm
Data Driven Business Model
Case Study | Internet of Animal Health Things (IoAHT): Opportunities and Challenges
5
Data-Driven Business Model of Precision Livestock Farming (PLF)
PLF technologies can be incorporated into the Data-Driven Business Model (DDBM) framework, described by Brownlow et al. (2015), see Figure 3. In this section, we answer the six fundamental questions for PLF-DDBM innovation:
1. What do we want to achieve by using big data?2. What is the PLF-DDBM desired o!ering?3. What are the key data sources for PLF-DDBM?4. What are the key activities?5. What are the potential revenue streams?6. What are the challenges to us accomplishing our goal?
Figure 3: PLF-DDBM Innovation Blueprint
1 Target OutcomeUsing big data to improve the PLF process management and for targeted delivery of drugs to individual animals.
2 O!eringData: Continuous sensing of outputs (process responses) at appropriate scale and frequency, with data fed back to the process controller.
Information: A target value and trajectory for each process output such as growth rates, behaviour patterns.
Knowledge: Actuators and a predictive controller for the process inputs.
Data ErasureIssuesThird Party
Involvement Issues
Consent Quality Issues
User Access and Control Issues
Data Collection Issues
3 Data SourceInternal: Batch data collected by sensor devices such as herd/flock camera systems, automatic weighing devices, vocalisation monitors, cough monitors, electronic identification ear tags (EID) and pedometers.
External: Data obtained from collaboration with related parties, for example, feed manufacturers working with weight-monitoring PLF companies.
4 Key O!eringData acquisition: Capturing and recording multiple attributes of each animal such as age, pedigree, growth rates, etc.
Aggregation: Integrate data from di!erent devices.
Descriptive analytics: Temporal trend analysis, for example, monitor animals’ size and weight gain.
Predictive analytics: Predict the estimated real-time process output.
Prescriptive analytics: Enable interventions to ensure target trajectory is met.
5 Revenue ModelPotential revenue streams include usage fees, purchase of sensor devices, subscription fees. Data use may support other business core products providing market insight.
hEp://cambridgeservicealliance.eng.cam.ac.uk/Resources/Monthly%20Papers/2015JulyCaseStudyIoAHT_HQP.pdf
Enabling Technology: Decision Support System
Decision
Ac>on
Report
Weather
Lab data
Farm Data
Maps
Extract Transform Load (ETL)
Big Data Challenges 1. Poor integra>on into prac>>oners workflow
2. Low level of uptake by the stakeholders
3. Cost and >me required to develop the data management sysem
4. Inadequate infrastructure (e.g. IT)
5. Interoperability: technical and opera>onal
6. Poor evalua>on of stakeholders’ needs
7. Lack or limited commitment of the subject maEer experts
8. Data availability and quality
9. Analy>cal challenges
10. Scalability
Data Prepara9on Fa9gue
Solu9ons: Building the Team
• Subject maEer expert: biologist, engineer, health specialist, Vet, someone who knows the problem domain
• Data scien>sts: sta>s>cian, data mining expert
• Privacy specialist: engineer, requirement engineer, policy analyst, lawyer
• Stakeholders: end user (e.g. farmer), policy maker, consumer, general public
• Story teller: end user, subject maEer expert, marketer
Thank you!