Sensors and Monitoring for Production and Distribution of a Tropical Fruit

Abstract

able development goals (SDGs) of the United Nations (2016). These comprehensive goals cover all possible topics toward sustainable development. Agriculture is a human activity which heavily relies on natural resources for food production, and thus directly and indirectly impacts natural systems. Considering the balance between environment and natural resources management including agriculture, the water-energy-food (WEF) nexus (Flammini et al., 2014) has emerged as a key approach towards SDGs. Self-evidently, human beings depend on WEF, and conserving them under the pressure of population growth and climate change is an important challenge. Worldwide, agricultural water use accounts for 70% of total water use, and more than 90% in the least developed countries (WWAP, 2012). It is expected to increase because of population growth and changing food demands. Water use for agriculture is in competition with domestic and industrial sectors. Climate change may further foster such conflicts, resulting in overexploitation of both surface water and groundwater resources. Therefore, agricultural water use and distribution need to be efficiently managed considering local conditions. Energy is another important factor for agriculture, covering the entire chain from production, harvesting, processing, distribution and consumption. Pressurized irrigation systems (e.g. drip and sprinkler irrigation) are important tools for water saving and increasing production but consumes significant amounts of energy. Fuels are used for processing, storage and transportation of fresh agricultural products to achieve a long shelf life during a supply chain. Advanced storage and transportation allow for supplying quality products to distant markets. A comprehensive research framework, adopting the WEF nexus approach, is needed for research and developed for an improved food system (e.g., from field to fork) from a viewpoint of sustainable development. In 2013, the research consortium established a project aiming at the development of intensive production system with improved distribution systems of fresh mango fruit (Mangifera indica L.) between Thailand and Japan. To this end, various kinds of data have been collected continuously from farm to table using Information and Communications Technologies (ICTs) such as advanced sensors and semi-real-time field monitoring devices. In addition, laboratory experiments were conducted to characterize ecophysiological traits (e.g., respiration rates and climacteric ripening process) of fresh mango fruit. We have applied machine learning methods in order to evaluate the effects of production environments including irrigation regimes on the yield and quality of fresh mango fruit (Fukuda et al., 2013) as well as the effects of distribution environments such as storage temperature on the dynamics of the fruit quality during transportation (Fukuda et al., 2014). Process-based fruit quality prediction models are being tested for predicting the quality changes of mango fruits according to their physiological activities such as postharvest ripening and respiration. By integrating these models, a comprehensive fruit quality prediction system will be established, which can contribute to improved production systems for high quality fresh mango fruit and advanced distribution systems allowing for a long supply chain from Thailand to Japan with minimum quality loss. This special issue was organized after the special session “Sensors and Monitoring in Environmental and Agricultural Sciences” at the 9th International Conference on Intelligent Robotics and Applications (ICIRA2016). The opening paper, “Random forests as a tool for analyzing partial drought stress based on CO2 concentrations in the rootzone of longan trees” (Fukuda et al., 2018) reports how an advanced machine learning, namely random forests (Breiman, 2001), can be applied for estimating CO2 concentrations in the rootzone based on ambient climate data and soil moisture content under different irrigation regimes (i.e., full irrigation and partial root-zone drying). The model performance was found to be moderate. Response curves and variable importance computed from random forests were used as a tool of knowledge extraction. Variable importance indicated the significance of soil moisture content for the CO2 estimation in the rootzone, while response curves indicated ecophysiological responses of longan trees under a given condition and irrigation treatment. Such an approach can be used as a first examination to find an input-output relationship in observed data and the interpretation