Asia Pacific Advanced Network Meeting

Grand Millennium Hotel Auckland 5th - 9th August 2018

Register Now!

Session Details

Big Data and AI in Agriculture

ObjectivesObjectives of the Session We will exchange information on constriction and usage of big data in agriculture field and application of artificial intelligence in agriculture field. For example, high-spacial or temporal density, high-throughput observation methods of plants and/or agricultural environments, analysis methods combined with genome data and others, and determination of plant diseases and pests etc are expected, but not limed.
Target AudienceMembers of Agriculture Working Group, anyone interested in creating and using big data, application of artificial intelligence technology in agriculture
Session Co-ordinator(s)Takuji Kiura, National Agriculture and Food Research Organization (NARO), Japan
Expected No. of Participants20
Seating ArrangementClassroom

Session 1

Date:Tuesday 2018-08-07
Time:13:30 - 15:00
Session Chair(s):Takuji Kiura, National Agriculture and Food Research Organization (NARO), Japan
No. of Participants:26, out of which 1 have provided feeedback
1.  AGRI-WATCH: Case Studies of IT Applications in Thai Agriculture    Slides (PDF)
Asanee Kawtrakul, Kasetsart University, Thailand
In past works, we proposed a software tool called Soft Wheel to help rice farmers make the right decisions regarding rice farming - especially planting and harvesting. Our current project, called AGRI-WATCH, is a continuation of the RICE-WATCH project, in the sense that we plan to extend Soft Wheel in three significant directions: Testing: Soft Wheel was so far tested with real data provided by farmers of the central region of Thailand. We would like to test it with real data coming from the relevant departments of the Ministry of Agriculture and concerning all rice producing areas of Thailand. We plan to extend Soft Wheel to provide services for the traceability of the Thai rice products sold in the national and international markets. This requires the integration of national and international standards of traceability in the Soft Wheel tool. The Ministry of Agriculture strongly supports such an extension. As mentioned earlier Soft Wheel was designed and developed for improving rice farming in Thailand. However, there are several factors that influence not only rice farming but also other kinds of farming (e.g. weather). Inspired by the basic underlying principles of Soft Wheel we plan to design and implement similar tools for the farming of tilapia fish, durian fruit and dairy products. Typically, the use of Soft Wheel requires digital literacy. Therefore, we plan to train a selected group of farmers in the use of Soft Wheel and let them each train the farmers of their communities.
2.  Plant Genetic Conservation Project   Slides (PDF)
Pakarat Danusatianpong, Hydro and Agro Informatics Institute, Thailand
Since 2003, HAII has participated with Plant Genetic Conservation Project under the Royal Initiatives of Her Royal Highness Princess Maha Chakri Sirindhorn. This aims to develop the plant genetic resources for the maintenance of plant varieties. HAII is one of the participated organizations of the project. We have developed the genetic database system in order to collect the data from the local communities.

In addition, HAII has worked with the local about the Community Water Resource Management (CWRM) which most communities have their indigenous knowledge. Therefore, we have expanded the system to the CWRM’s indigenous knowledge database system since 2007. The input data divided into 2 types; plant genetic data and indigenous knowledge data. Both data were combined with Internet-GIS that show the communities’ location on the map.
Currently, there are 17 communities’ data has shown for our communities surrounding the country could share and learn each other via the website.

3.  Java Agricultural Model Framework and its Web API   Slides (PDF)
Kei Tanaka, National Agriculture and Food Research Organaization, Japan
A software framework to develop an agricultural model as Web application, Java Agricultural Model Framework (JAMF) has implemented a function to acquire meteorological data. Because MetBroker that was a middleware to mediate various meteorological databases ended its service on March. JAMF can access Agro-Meteorological Grid Square Data System (AMGSD, 1km gird in Japan, daily data with forecast data), Automated Meteorological Data Acquisition System (AMeDAS, 1300 stations in Japan, daily and hourly data), and POWER Project Data Sets (NASA, 1-degree grid data all over the world, daily data). These three different types of meteorological databases are enabled to execute an agricultural model for anywhere on the Earth. Especially, AMGSD is suitable for an agricultural model because it provides the observed, the predicted, and the 30-year averaged values seamlessly. JAMF-Servlet is another framework to develop a Web application as Java Servlet based on JAMF. One of its features is to run applications with Web APIs. Easily understandable APIs enable to change various parameters simply by string manipulation. Technologies of data acquisition and model execution through Web API, and formatted data output in XML or JSON can easily build a complex application by combination of several APIs.
4.  The Next Step of AI Application in Agricultural Decision Support   Slides (PDF)
Seishi Ninomiya, The University of Tokyo, Japan
AI such as machine learning has been showing its high potential for agricultural decision support in the last 10 years. For example, crop segmentation and head/panicle detection from drone images under varying natural light conditions became highly accurate with the approach, replacing inaccurate and time-consuming human visual judgement. AI is surely promising in agriculture as well but we also have to realize some existing issues such as (1) laborious work required to acquire quality training data set, (2) possibility of extrapolation and (3) new knowledge extraction from the AI approach. This paper will discuss those issues in the applications of AI in agriculture, consideration of the acceleration of AI use in the field.

Session 2

Date:Tuesday 2018-08-07
Time:15:30 - 17:00
Session Chair(s):Royboon Rassameethes, Hydro and Agro Informatics Institute (HAII), Thailand
No. of Participants:22, out of which 3 have provided feeedback
1.  Taiwan Campus Nutritional Lunch Food Safety Use of Agricultural Big Data: A complete deployment case   Slides (PDF)
Chun-Tang Lu, Taiwan Agricultural Research Institute, Taiwan
The campus lunches in Taiwan are now putting more emphasis on food certifications to improve the safety of the meals for students. In this presentation, we will share our experience and achievements of using agricultural big data to improve the traceability to campus lunch meal’s food material, the providers, kitchens, etc., which one of the purposes is to provide online checking the children's nutritional lunch for the parents of primary schools across the country. The big-data system through the Internet open data API integrates more than 10 scattered independent databases, including the Taiwan TAP, Organic agriculture products, CAS, Agriculture product pesticide residues, Traceable products, food material providers, schools, etc. This big data system can also promote food material providers to increase the use of local food materials and the self-sufficiency rate of domestic agricultural products, which is a pioneering research in Taiwan.
2.  Agri-Map: Agricultural Big Data Management and Analysis Map of Thailand   Slides (PDF)
Noppadon Khiripet, National Electronics and Computer Technology Center (NECTEC), Thailand
Thailand’s Agricultural Map for Adaptive Management (Agri-Map) is an ongoing collaborative project between the Ministry of Agriculture and Cooperatives and Ministry of Science, and has been in operation since 2016. The collaboration allows national, regional, local policy makers and Thai farmers to access to the comprehensive, integrated and updated agricultural information through publicly available GIS applications. Agri-Map Online is a web-based application containing more than 200 national-wide geospatial information layers including land usage and land suitability of 13 major economics crops, livestock and fisheries. Moreover, there are supportive information layers such as weather forecast, water resources, forest boundary, physical soil group and problems. Furthermore, it supports location data layers such as marketplaces and factories, helping and learning centers and cooperative locations. It also contains economics information i.e., the approximate cost and net profit for each of the crops. Agri-Map Mobile is the latest application that enable users with modern mobile devices with internet connection such as mobile phones and tablets to access useful features of the Agri-Map in a user-friendly and easy to use fashion. This is to support the main function of the related project: Zoning by Agri-Map on how to target lands that have more potential for more profitable agricultural alternatives.
3.  High-Throughput Phenotyping of Big Agricultural Data for Improved Crop Selection   Slides (PDF)
Soumyashree Kar, Centre of Studies in Resources Engineering, Indian Institute of , India
Big agricultural data is generated from large breeding experiments which simultaneously phenotype multiple traits. These large field-based High-Throughput Phenotyping (HTP) experiments are essentially conducted in breeding programmes under non-controlled environments to assess the diversity among genotypes (and subsequent crop selection), in terms of their phenotypes. However, differences in micro-agro-climatic conditions results in different phenotypic values of the same genotype in different replications (located randomly). Hence, it becomes necessary to normalize the differences and statistically segregate the genotypes. A statistical and quantitative spatial analysis methodology is adopted for succinct extraction of genotypic clusters. The developed method both pre-processes the big input dataset and spatially models the phenotypic values for efficient and optimal grouping of statistically-similar genotypes, in a single-click process. Thus, it successfully executes the 3Vs (volume, velocity and variety) of Big Data and aids crop selection in High Performance Breeding programmes.
4.  Big Data in Little New Zealand   Slides (PDF)
Matthew Laurenson, New Zealand Institute of Plant & Food Research, New Zealand
Agriculture is the mainstay of the New Zealand economy. Of New Zealand’s annual exports of NZ$54B for the year ended December 2017, exports of dairy, meat, wood and horticultural products made up $30B. The collection and analysis of large quantities of data has contributed to New Zealand’s strength in agriculture. This paper highlights several New Zealand examples, namely milk testing and the dairy breeding database, kiwifruit new cultivar development, and weather-based disease management decision support tools.
5.  NIAES approaches to FAIR data   Slides (PDF)
Takuji Kiura, National Agriculture and Food Research Organization (NARO), Japan
NIAES has created various agricultural environmental data sets and made them public. Some of them, provided as a Web application, were no longer available because they can not be maintained due to lack of budget. Also, many data sets are still stored on researchers' hard disks, without decisions to be provided to the public. NIAES is trying to improve this situation by following FAIR principles, and this report introduces our plan.