Some livestock producers in Hokkaido, Japan, divide their pastureland into small areas and move their cattle herds from one pasture area to another. This rotational grazing method is effective for maximizing pasture utilization, but it is time consuming for producers to measure the grass height in each pasture area regularly. We developed a system in which a trail camera and rubber poles are installed in a pasture and the producer receives the information obtained from the resulting images. The trail cameras were connected to a cellular phone network, and the photos were made available online. The rubber poles were used as height indicators to determine the grass height. Specifically, the pole color (orange) in the pasture image was extracted by Python code, and the grass height was estimated by quantifying the degree to which the pole was hidden by the grass. In addition, Google Apps Script was used to construct a system that automatically collects email attachments (photos from the trail camera) online (in Google Drive) and supplies producers with the grass height data at a fixed time every day by using a messenger app, LINE (https://www.line.me/en/). This system can help with pasture management without taking up the producers’ time, thus leading to more efficient use of grass resources.

An annual supply system for major vegetables in Japan, increasing the demand for processing and food services. With the increased demand, using a crop growth model to predict growth is valuable for understanding and predicting the growth conditions for cabbage production in open fields. Here, this study improved a previous cabbage growth model that focuses solely on growth from the heading stage. We re-examined the distribution rate and added new model variables that enabled the model to predict the start of heading and the amount of growth from the transplant date as the projected area of leaves. We targeted the spring and autumn cropping periods, which were not considered in the previous model, and we estimated the parameters individually for the state variables required to account for weather changes. As a result of the evaluation, the projected area of leaves, the dry matter weight, and the fresh weight of the leaf head were predicted with RMSE values of 123.1 (cm²), 16.7 (g), and 240.8 (g), and MAPE values of 39.2%, 14.9%, and 8.1%. The prediction error for the fresh weight of the cabbage head was within the acceptable range of practical use (10–20%), demonstrating sufficient accuracy. In addition, by determining the dry matter distribution rate throughout the entire cultivation period and the radiation use efficiency according to the growth stage, we found that we could make detailed predictions about heading initiation and growth in the early growth stage.

Along with the introduction of new sugarcane cultivars with superior ratooning ability, mechanization advancements since 2000 have reduced labor on Tanegashima, leading to the emergence of large-scale operations. To evaluate the management of these operations, we developed a mathematical programming model based on mechanized sugarcane farming. The model is characterized by its annually based process design and the consideration of increased ratoon crop cycles. Analysis of the model revealed that the use of high-yield cultivars contributes to income improvement. In addition, we found that, for the introduction of billet planters, which enable labor-saving planting, a farm needs to be ≥30 ha, the operation must have a sufficient labor force, and there should be substantial benefits in avoiding competition with spring operations.

We conducted a smart agriculture demonstration project in Japan to clarify the effects of introducing smart agriculture technology. Agricultural management indicators were developed on the basis of the project’s results. In the process, we considered that visualized agricultural information on the causal relationship among agricultural technology, farm work, and the effects of technology adoption (“technology–work–effect table”) would facilitate the development of the indicators. However, such information is based mostly on documents, the manual deciphering of which is considerably labor intensive, time consuming, and limited. Large language models (LLMs) may be effective in solving this problem. Therefore, we investigated the possibility of using an LLM to generate technology–work–effect tables from agricultural documents. Specifically, experiments used GPT-3.5 Turbo, developed by OpenAI, to generate tables from document data in the “smart agriculture demonstration project results portal”, provided by NARO, and we evaluated and analyzed the accuracy of the tables. Analysis of 4400 technology–work–effect tables generated from agricultural documents by the LLM revealed that they were not always correct, and that the content of the documents significantly influenced the output variability. Furthermore, the appropriate experimental conditions are to use plain-text documents in a more deterministic LLM setting, such as setting the temperature parameter to 0, under which conditions the frequency of the average output is high. These findings indicate that the accuracy of LLM output tables must be improved if these tables are to be put to practical use.

We performed a numerical analysis to obtain a new design guide for a sap flow sensor using the stem heat balance method from the perspective of thermo-fluid dynamics, and evaluated the findings experimentally. The non-destructive stem heat balance method measures sap flow rate by solving the stem heat balance on the basis of the temperature of the heated stem surface. The object of the numerical analysis is a model tube simulating a plant stem. It solves 3-dimensional steady Navier-Stokes equations and advection-diffusion equations. The input flow rate corresponding to the plant sap flow is set to 0.06–1.23 g/min. The numerical results showed that the flow rate calculated by the conventional stem heat balance method was >20% larger than the true value under almost all conditions. The thermo-fluid analysis showed that the error was caused by not considering the quantity of heat transferred from other than the back of the heater. In addition, the wall surface temperature was then used instead of the mixed mean temperature, as a representative temperature of the fluid, to calculate the heat transport. The flow-rate calculation error was reduced to ~10% with the above correction. These results suggest that the flow-rate calculation error can be reduced by installing thermocouples to measure the heat radiated from areas other than the back of the heater, and avoiding temperature measurement positions close to the heater, where the difference between the wall surface temperature and the mixed mean temperature is large.