Deep Learning Enables Instant and Versatile Estimation of Rice Yield Using Ground-Based RGB Images.

Rice (Oryza sativa L.) is one of the most important cereals, which provides 20% of the world's food energy. However, its productivity is poorly assessed especially in the global South. Here, we provide a first study to perform a deep-learning-based approach for instantaneously estimating rice yield using red-green-blue images. During ripening stage and at harvest, over 22,000 digital images were captured vertically downward over the rice canopy from a distance of 0.8 to 0.9 m at 4,820 harvesting plots having the yield of 0.1 to 16.1 t·ha-1 across 6 countries in Africa and Japan. A convolutional neural network applied to these data at harvest predicted 68% variation in yield with a relative root mean square error of 0.22. The developed model successfully detected genotypic difference and impact of agronomic interventions on yield in the independent dataset. The model also demonstrated robustness against the images acquired at different shooting angles up to 30° from right angle, diverse light environments, and shooting date during late ripening stage. Even when the resolution of images was reduced (from 0.2 to 3.2 cm·pixel-1 of ground sampling distance), the model could predict 57% variation in yield, implying that this approach can be scaled by the use of unmanned aerial vehicles. Our work offers low-cost, hands-on, and rapid approach for high-throughput phenotyping and can lead to impact assessment of productivity-enhancing interventions, detection of fields where these are needed to sustainably increase crop production, and yield forecast at several weeks before harvesting.

Application of consecutive polyethylene glycol treatments for modeling the seminal root growth of rice under water stress.

The response of root growth to various osmotic potentials is quite important in assessing the drought resistance of rice. This study modeled seminal root growth by applying consecutive polyethylene glycol (PEG) treatments (from 0 to 25%, 1% step), mathematical equations and noncontact image analysis to quantitatively evaluate the root response. Treatment began after seeds were germinated, and root growth was recorded by a digital camera every day from 7 to 20 days after seeding (DAS). Although the seminal root length (SRL) measured by image analysis slightly varied with DAS, the equations explained the differences in SRL increases under each PEG concentration relatively well (R2 = 0.774). The equations also suggested that the maximum seminal root length was observed at 5.9% PEG. This numerical characterization of root growth is an effective means of evaluating drought resistance.

Impacts of the continuous maize cultivation on soil properties in Sainyabuli province, Laos.

In tropical mountainous areas, soil degradation and yield decrease have been anticipated due to conversion from shifting to continuous cultivation and the introduction of cash crops. In our previous report, we quantified the decrease in maize yield under continuous cultivation in farmers' fields in Laos. In this report, we focused on soil nutritional conditions under continuous cultivation in the farmers' fields. For the purpose, twelve soil properties were investigated over two years from three sample sites in each of the 40 farmers' fields with the duration of continuous cultivation varying from 1 to 30 years. Total carbon (TC), total nitrogen (TN), available phosphorus, exchangeable potassium, and exchangeable calcium in the soil decreased with increasing duration of continuous cultivation in the sloped fields. These soil nutrients decreased to around half of the initial content in these 30 years. However, the decreasing rates of TC and TN were negligible in the flat fields. Other soil properties such as clay and exchangeable magnesium were not related to the duration of continuous cultivation in both sloped and flat fields. The reduction in maize yield was mainly explained by TC, but the determination coefficient was only 0.24. Although further analysis is required to quantify the effect of soil nutrients on maize production, the development of integrated soil management would be necessary in the sloped fields for sustainable crop production in the study site.

Parameterization of the vertical distribution of leaf area index (LAI) in rice (Oryza sativa L.) using a plant canopy analyzer.

Monitoring the vertical distribution of leaf area index (LAI) is an effective method for evaluating canopy photosynthesis and biomass productivity. In this study, we proposed a novel method to characterize LAI vertical distribution non-destructively by utilizing LAI-2200 plant canopy analyzer, followed by the application of statistical moment equations. Field experiments were conducted with 5 rice cultivars under 2 fertilizer treatments in 2013 and with 3 rice cultivars under 3 plant density treatments in 2014. LAI readings obtained by a plant canopy analyzer for non-destructive stratified measurements were relatively consistent with LAI estimations using the stratified clipping method for every cultivar and treatment. The parameters calculated using the statistical moment equations numerically showed the changes in LAI vertical distribution with plant growth up to the heading stage. The differences in the parameters also quantified the effect of cultivar, fertilizer, and plant density treatments. These results suggest that the non-destructive stratified measurements and the statistical moments evaluated in this study provide quantitative, reliable information on the dynamics of LAI vertical distribution. The method is expected to be utilized by researchers in various research fields sharing common interests.

Finlay-Wilkinson's regression coefficient as a pre-screening criterion for yield responsiveness to elevated atmospheric CO2 concentration in crops.

The rising atmospheric CO2 concentration ([CO2 ]) can increase crop productivity, but there are likely to be intraspecific variations in the response. To meet future world food demand, screening for genotypes with high [CO2 ] responsiveness will be a useful option, but there is no criterion for high [CO2 ] responsiveness. We hypothesized that the Finlay-Wilkinson regression coefficient (RC) (for the relationship between a genotype's yield versus the mean yield of all genotypes in a specific environment) could serve as a pre-screening criterion for identifying genotypes that respond strongly to elevated [CO2 ]. We collected datasets on the yield of 6 rice and 10 soybean genotypes along environmental gradients and compared their responsiveness to elevated [CO2 ] based on the regression coefficients (i.e. the increases of yield per 100 µmol mol-1 [CO2 ]) identified in previous reports. We found significant positive correlations between the RCs and the responsiveness of yield to elevated [CO2 ] in both rice and soybean. This result raises the possibility that the coefficient of the Finlay-Wilkinson relationship could be used as a pre-screening criterion for [CO2 ] responsiveness.

Diurnal and developmental changes in energy allocation of absorbed light at PSII in field-grown rice.

The allocation of absorbed light energy in PSII to electron transport and heat dissipation processes in rice grown under waterlogged conditions was estimated with the lake model of energy transfer. With regard to diurnal changes in energy allocation, the peak of the energy flux to electron transport, J(PSII), occurred in the morning and the peak of the energy flux to heat dissipation associated with non-photochemical quenching of Chl fluorescence, J(NPQ), occurred in the afternoon. With regard to seasonal changes in energy allocation, J(PSII) in the rapidly growing phase was greater than that in the ripening phase, even though the leaves of rice receive less light in the growing phase than in the ripening period in Japan. This seasonal decrease in J(PSII) was accompanied by an increase in J(NPQ). One of the reasons for the lower J(PSII) in the ripening phase might be a more sever afternoon suppression of J(PSII). To estimate energy dissipation due to photoinhibition of PSII, J(NPQ) was divided into J(fast), which is associated with fast-recovering NPQ mainly due to qE, and J(slow), which is mainly due to photoinhibition. The integrated daily energy loss by photoinhibiton was calculated to be about 3-8% of light energy absorption in PSII. Strategies for the utilization of light energy adopted by rice are discussed. For example, very efficient photosynthesis under non-saturating light in the rapidly growing phase is proposed.

Evaluation of genotypic variation in leaf photosynthetic rate and its associated factors by using rice diversity research set of germplasm.

In order to evaluate genotypic variation, we measured leaf photosynthetic rate (Pn) and its associated factors for the rice diversity research set of germplasm (RDRS) selected from the Genebank in National Institute of Agrobiological Sciences (NIAS). Pn showed large genotypic variation from 11.9 to 32.1 micromol m(-2 )s(-1). The variation in stomatal conductance to CO2 (Gs) explained about 50% of that in Pn, while that in nitrogen concentration (N) in leaves explained about 35%. The genotype group which mainly consists of aus type indica tended to have higher Gs, and the genotype group which corresponds to japonica had a higher nitrogen concentration (N) in leaves. The relationships of Pn with Gs and N were not significantly different among genotype groups, suggesting photosynthetic efficiencies are similar among genotype groups.