This session will be held from 1:30 pm to 5:00 pm under the following Agenda.Objectives :Data processing using big data and AI is considered promising in the agricultural field, and many researchers are conducting research in this field. The APAN Agricultural Working Group has focused on this area for several years, and has held a session on "Big Data and AI in Agriculture" at the APAN Meetings. Unfortunately, international cooperation in this area among APAN members is not enough. Therefore, APAN Agricultural Working Group will continue this session in APAN 48, promoting information exchange and international cooperation.
In this session, to report on sensor network technology, management and sharing of open data, use of AI in agriculture, and discuss collaborative research. etc. are expected. Looking forward to your contributions.
Session Chairs :Takuji Kiura Royboon Rassameethes , HAII, ThailandAgenda1. [[Session Keynote]] Technology thought leadership and how it can support building resilient communities -
Alan Deciantis, Forest Technology Systems ,CanadaAs the world’s population is projected to reach 9 billion people by 2050, global food production will need to increase by 60%. With agricultural land size remaining fixed for the last 30 years, sustainable methods and new technologies will be required. Discover how the North American public sector agri-geomatics and private sector investment in precision-ag and imagery is working help meet this need. As IoT connectivity and AI analytics technologies advance, find out what one of the next big challenges is for the agriculture sector.
2. Treating Field Server Data files with ZFS -
Takuji Kiura There are over 100 million raw data files from field data published at http://fsds.dc.affrc.go.jp/data[1-5]/, and more new files are being added. Since observation data can only be acquired at that time, it is necessary to protect these files from data loss due to ransomware or disasters. Backups are an important way, but the rsync command, we used before, took considerable time just to find files that were modified or added. Because of this, We used Btrfs, a copy on write file system on linux. A year before we started using ZFS, also available on linux, because it has a reputation on other OSs. Here, we introduce a new system using ZFS, comparing it with our two old systems.
3. Multi-Scale Time-Series Analysis of HTPP data collection frequency -
Soumyashree KarHigh Throughput Plant Phenotyping (HTPP) not only generates a large amount of data but also a large amount of information. However, it is often noticed that appropriate information extraction is limited,
either due to redundancy in the observed variables and/or noise, which could either be phenomenon-based or latent (i.e. as part of the data collection procedure itself). In this work, 15-minute frequency plant
evapotranspiration (ET) observations are used to enable proper identification of genotypic differences, by: 1- identifying the appropriate level of de-noised high-frequency time-series (TS) data using Discrete Wavelet Transform (DWT) and entropy analysis at each level of decomposition, 2- determining the optimum sampling frequency/intervals by multi-scaling the ET TS data. For the second objective, ARIMA based models are used for each scaling frequency, and model efficiency is assessed based on the respective AIC and RMSE estimates. Subsequently, clustering is performed and the resultant cluster indices are used to determine classification accuracy at each scale of the TS. The results suggest an acceptable level of
classification accuracy could be obtained till a maximum of 90-minute frequency, beyond which the genotypic differences seem to greatly dissolve.
4. Development of the high-resolution historical gridded daily meteorological data set over Japan using the JRA-55 reanalysis data. -
Yasushi Ishigooka, NIAESWe had developed a high spatial resolution (approximately 1km × 1km) gridded daily meteorological data set by combining the high-resolution monthly climate data with the JRA-55 reanalysis data, which covers for a long-term period of 55 years (1958-2013) and contains most of major meteorological elements that enable to implement many crop and hydro-logical models. The data set can be used for estimating detailed spatiotemporal information on the effects of observed past climate change and changes in the cultivation conditions on the potential productivity of Japanese crops.
