In recent years, transient expression systems in plants have emerged as an attractive platform for recombinant protein production. Green fluorescent protein (GFP) is widely used as a reporter; however, its expression changes over time, highlighting the need for a practical, non-destructive method to monitor expression dynamics. Here, we developed an image-analysis method to evaluate GFP expression in cucumber (Cucumis sativus L.) cotyledons infected with a GFPexpressing cucumber mosaic virus (CMV) vector. Cotyledons were illuminated with blue light-emitting diodes (peak wavelength, 480 nm) to excite GFP, and fluorescence images were acquired using a digital camera equipped with a bandpass filter (500-800 nm). Images were binarized to segment regions with detectable fluorescence, and the fluorescent area was quantified. The fluorescent area estimated by image analysis showed a moderate positive correlation with GFP levels measured by ELISA (p < 0.01, R = 0.71). During the first three days after infection, fluorescence remained localized near the inoculation site, suggesting cell-to-cell or short-distance movement. In contrast, the fluorescent area increased markedly by day 6 under high photosynthetic photon flux densities (500 and 1000 mu mol m(-2 )s(-1)), whereas this increase was delayed until day 9 under low PPFD (100 mu mol m(-2 )s(-1)). These increases may reflect long-distance movement through the phloem. Although further improvement is needed to reduce noise (e.g., fluorescence originating from necrotic tissue), this method provides a simple, non-destructive means to monitor GFP-expressing areas in plants. It may facilitate optimization of recombinant protein production.
Pollination services play an important role in global food production. This study advances an agent-based simulation that reproduces the flower-visitation process in which pollinators search for flowers using floral volatile compounds as cues. In particular, we develop theory for (i) the representativeness of an agent with respect to the number of pollinators it denotes in the simulation and (ii) the conversion between simulation and real-world spatiotemporal scales. We show that these scales can be converted from the representativeness of agent and flower counts within the simulation. As an example, when an agent represents one pollinator and an attractant source represents one flower, the results convert to 24% of total attractant consumed in 84 s. This study is expected to help realize the quantitative evaluation, via simulation, of the impacts of environmental change on pollination services.
This study, using semantic segmentation, compared the performance of forest damage detection between aerial, Sentinel-2 and World View-2 (WV2) images, which differ in terms of ease of image acquisition and spatial resolution. In the case of high-resolution aerial images, forest damage can be detected more accurately by semantic segmentation than by conventional maximum likelihood classification. The study conducted segmentation in two stages: segmentation of normal and damaged areas and segmentation of landslides and fallen trees from the damaged area. Two-stage segmentation demonstrated better performance than three-class segmentation. Through the segmentation processes, class weights were applied to mitigate differences in the number of pixels in each category. Although the performance of the segmentation in the Sentinel-2 images was not comparable to that of aerial images, the segmentation appeared to be useful for a rapid and rough estimation of the damaged area. WV2 images can be used as alternatives to aerial images. In the segmentation with aerial and Sentinel-2 images, the network trained at a certain site could be used for another site without any additional training for another site, or with training using a small number of other site images.
This study evaluated a low-cost, closed-loop freshwater recovery system employing hydroponically grown lettuce (Lactuca sativa L.) irrigated with either a standard nutrient solution or a recycled nutrient solution derived from biogas digestate aerobically nitrified (hereafter, nitrified biogas digestate). The system operated under natural sunlight, enabling examination of the effects of solar radiation fluctuations, air temperature, vapor pressure deficit (VPD), CO2 concentration, and irrigation water quality on water transfer, recovery efficiency, and plant performance. Diurnal monitoring and regression analyses demonstrated strong positive correlations of radiation, temperature, and VPD with water transfer and harvest rates (R-2 > 0.82), whereas elevated CO2 concentrations correlated negatively with transpiration-driven fluxes. Despite the higher salinity and osmotic load of the nitrified biogas digestate (EC = 6.8 dS m(-1); psi(s) = 0.277 MPa, where psi(s) denotes the solute potential), water fluxes, recovery efficiency, and biomass production remained comparable to those with the standard nutrient solution. Lettuce sustained stable physiological activity under moderate salinity stress, with condensation recovery efficiencies exceeding 80% in planted systems versus 63.8% in the non-vegetated control. Condensation reduced the electrical conductivity of the recovered water to 0.01 dS m(-1) and restored a near-neutral pH, confirming the effectiveness of vapor-phase purification. Overall, the results demonstrate the potential of plant-based, sunlight-driven hydroponic systems combined with nitrified biogas digestate for sustainable freshwater recovery and reuse in closed, resource-limited environments.
This study examines labor-saving methods for creating training data in the construction of a deep learning model for fruit detection. The study examined the relationship between the number of training images required for normal training and detection accuracy for peaches and grapes. The results of the normal training were used as a reference to examine the other methods, a method based on training between different fruit species, ones using fruit model images, and the effect on accuracy of the presence or absence of background was also compared. In the normal training, accuracy was achieved with around 200 images, and the effect of further increases was limited. Furthermore, in the learning of interspecific fruits, it was demonstrated that even learning model trained on different species can achieve high detection accuracy if it can utilize image data of a small number of target fruit species. The combination of learning with fruit replica images and fine tuning with a small amount of real images was shown to achieve high accuracy while reducing the dependence on the real environment. These results contribute to labor-saving fruit detection using deep learning.
