The increasing nitrate-nitrogen concentrations in the groundwater that became apparent in the 1980s amid concerns regarding the effects of nitrogen loading caused by agriculture and livestock industries are the problems faced by the two pyroclastic flow plateaus, the Kasanohara and Kanoyabaru Shirasu plateaus, located at the center of the Osumi Peninsula in Kagoshima Prefecture, Japan. Therefore, this study evaluated the long-term trends in spring, groundwater, and river water quality, as well as nitrogen-emitting loads from agriculture and livestock on both plateaus to assess the impact of agriculture and livestock on the aquatic environment. The nitrate-nitrogen concentrations in spring water on the periphery of the two plateaus significantly increased with the nitrogen load from pig farming, reaching a peak from 2000 to the mid-2010s. Since the latter half of the 2010s, nitrate-nitrogen concentrations in a representative spring of Kanoyabaru plateau rapidly decreased from 12.4 mg L⁻¹ to 5.8 mg L⁻¹. Similarly, a decreasing trend was observed on Kasanohara plateau. In recent years, the nitrate-nitrogen concentrations rapidly decreased due to reduced nitrogen load caused by abolition of the percolation treatment of pig manure since the beginning of the 2000s, which was based on the enactment of the Livestock Waste Law of 1999. This study demonstrated that, even in a large field on a plateau area with severe groundwater nitrate-nitrogen pollution, owing to the high environmental impact of agricultural and livestock systems, the aquatic environment can be improved by the consistent implementation of agro-environmental measures.

Ensuring an ideal pulverized soil rate is an important management practice to stabilize germination and seedling establishment in field crops and to improve the efficacy of soil-treated herbicides. The target for pulverized soil rate is approximately ≥70％, but determining the pulverized soil rate in the field is difficult because of the time and labor required to measure the rate on site. Therefore, a simple method should be developed to estimate the pulverized soil rate based on the number and size of soil clods in the image of pulverized soil surface. This study analyzed the relationship between the degree of surface soil clods calculated from the number of soil clods A (with a long side between 5 and 10 cm) and B (with a long side of >10 cm) in the image and the pulverized soil rate. The results indicated a close relationship between the two factors. Furthermore, an image classification model was developed for estimating the degree of surface soil clods (sixth ranks from 0 to 5) using a convolutional neural network, one of the algorithms of deep learning. This image classification model had an accuracy of approximately 80％ if an error of one rank was allowed. The results indicate that the image classification model in this study can be used to easily diagnose whether the pulverized soil rate satisfies the 70％ target, because one rank error is an acceptable error. In addition, the Grad-CAM method was used to visualize the points of interest in the image classification, and the classification focused on the periphery of soil clods rather than soil clods itself.

The appropriate application rates of nitrogen (N) fertilizer were estimated based on the amount of N mineralization in the topsoil of paddy fields to achieve both good quality and taste and stable yield in “Koshihikari”, the main rice cultivar in the Hida region of Gifu prefecture (Hida Koshihikari). The amount of N uptake (N uptake) at rice maturity was correlated with grain yield and protein levels in brown rice. These relationships indicated that the ideal N uptake of Hida Koshihikari was 80-85 kg ha⁻¹. Furthermore, N uptake at the panicle formation stage was significantly correlated with the amount of N supplied by the basal N fertilizer (B-N) in addition to the amount of N mineralized from paddy soils (S-N). The latter value was calculated from N mineralization when moist topsoil taken from paddy fields were incubated under submerged conditions at 30°C for 10 weeks (M10 w). These results demonstrate an accurate prediction of the N uptake at the panicle formation stage from the N supply by B-N and S-N when M10 w was used. B-N, M10 w, and N supplied from sources other than B-N and S-N were used to develop a linear equation for predicting N uptake at the panicle formation stage. This formula, along with the N uptake value at the panicle formation stage and varying values of M10 w, was used to calculate the appropriate N application rates for basal fertilizer in Hida Koshihikari. However, N uptake after the panicle formation stage was strongly affected by the amount of N supplied by top-dressed N fertilizer. Therefore, these relationships could be used to estimate the appropriate N application rates for top-dressing from the N uptake after the panicle formation stage.