Reduce Crop Damage Research Articles

Individual responses of GPS‐tagged geese scared off crops by drones or walking humans

Abstract Scaring of wildlife is commonly used in attempts to reduce crop damage in agricultural landscapes, but relatively few studies exist on its actual effect. We tracked GPS‐tagged greylag geese (Anser anser) in south‐central Sweden, before and after scaring by approaching them either by walking or by drone. On the field level, we studied the scaring effect by estimating return rate to the field where scared. On the landscape level, we tested if geese were less prone to use agricultural fields after being scared. Geese immediately left the field when scared and 5 min later they were on average 990 m (±56 SE) from the scaring position. The proportion of GPS positions near the scaring position decreased significantly for at least 4 h after scaring. Geese showed a significant shift from agricultural fields to wetland habitats the first 4 h after scaring. However, the effect of scaring soon levelled off; after 24 h the field where scaring had occurred was used to the same extent as any other field in the landscape, and agricultural fields were used to the same extent by scared and undisturbed geese. We did not find any differences in response depending on scaring technique. The probability to return and use agricultural fields after scaring was higher for geese scared in the morning compared to in the afternoon. Moreover, the probability to return and use of agricultural fields were higher in spring than in other seasons. Practical implication. We found that scared geese tend to visit agricultural fields soon after scaring and that scaring alone tends ‘to move the problem around’. This suggests that scaring needs to be repeated across the landscape, but also that accommodation fields where geese do not cause damage may be needed to keep geese off conventional fields. However, our study presents a glimpse of promise as the rather simple drone used covers large areas quickly and minimizes walking in growing crops. With technological advancement and possible autonomous techniques, drones may be capable of providing repeated scaring over large areas in the future.

Can repellent crops reduce the abundance of the common vole (Microtus arvalis) as a way to reduce crop damage?

The common vole is one of the most damaging rodents in agriculture. A number of methods are used to suppress its numbers and limit damage. One option is growing crops that are not suitable for voles; this might limit their numbers and distribution. Through long-term monitoring of common voles throughout the Czech Republic, their abundance in eleven crop types (annual, biennial and perennial crops) was evaluated through active burrow counts. Reference crops were selected from perennials and biennials and annual crops were selected as potential repellents. The perennial crops were clover with alfalfa, which serve as the primary habitat for voles. These crops are where voles are found in their highest densities, and from which they spread to the surrounding crops during periods of high abundance. The biennial winter rape was selected because it provides favourable conditions for voles to overwinter and multiply rapidly in the spring. Compared to perennial crops and winter rape, significantly lower numbers of voles were found in onions, poppy seeds and maize. Additionally, fewer voles were also found in mustard compared to perennial crops. Onion and poppy were the only crops to show a significantly lower abundance of voles in comparison to wheat. The annual crops tested are unattractive to voles and due to the cultivation practices used, they do not even have enough weeds as food. As profitable crops they can be cultivated over sufficiently large areas to potentially prevent the spread of voles to surrounding crops, especially during outbreak years.

Suppression of Thrips palmi population by spray-on application of dsRNA targeting V-ATPase-B

The RNA interference (RNAi)-based gene silencing technique has enormous potential as a non-chemical and eco-friendly alternative to hazardous pesticides. This study reports a spray-induced gene silencing (SIGS) approach for managing Thrips palmi by lowering survival and offspring development. Vacuolar ATP synthases (V-ATPases) are responsible for survival, egg-laying, and viability of eggs in insects. In the current study, T. palmi V-ATPase-B was targeted to suppress the pest population by spray-on application of double-stranded RNA (dsRNA). Silencing of V-ATPase-B was first validated by oral administration of dsV-ATPase-B. The expression of V-ATPase-B was reduced by 5.40-fold post-dsRNA feeding leading to increased mortality (57.03 %) and reduced reproductive fitness (67.73 %). Spray-on application of naked dsV-ATPase-B at concentrations of 3.0 μg/mL and 5.0 μg/mL effectively suppressed the population by 30.00 % and 43.33 %, respectively. The expression of the target gene was downregulated by up to 4.24-fold. Two consecutive sprays at a concentration of 5.0 μg/mL provided substantial protection against the fresh release of T. palmi for up to 10 days. The spray-on application of dsV-ATPase-B would be an eco-friendly alternative for managing T. palmi populations thereby reducing crop damage and limiting the spread of orthotospoviruses.

Anthropic pressure drives resource selection of an adaptable generalist in human‐dominated landscapes

AbstractFew landscapes on earth remain free of human‐modification, which may influence resource selection in wildlife. To investigate the effects of anthropic pressure on wild boar (Sus scrofa) and explore management implications, we studied how diel resource selection of the species' main life stages changed with spatial variations of human access (e.g., for recreation), temporal changes in hunting pressure, and habitat type. Using 206,461 hourly GPS‐locations of 15 males, 11 females with dependent young, and 17 other females from south‐western Germany, we found anthropic pressure influenced resource selection more than ecological factors. All boars were more likely to select for low human‐access areas than high human‐access areas, regardless of habitat. Hunting pressure was most avoided by females with dependent piglets, followed by males and other females. Since both hunting activity and general human access affected resource selection, they should be considered simultaneously in wildlife management and conservation. We suggest the further establishment of wildlife reserves that are inaccessible to people where boar may remain more localized, thereby reducing the risk of disease transmission, and boar hunting to focus on open lands and refuge boundaries to reduce crop damage. This may also benefit overall human‐wildlife coexistence, animal welfare, and biodiversity conservation in anthropized environments.

Combining biological control approaches for managing insect crop pests in the field can generate interactive effects

Abstract With the loss of effective chemical controls of crop pests, a move towards biological controls is a way to future‐proof our agricultural system. Floral field margins have shown successes in reducing crop pests, though the effect may not be precise enough to control infestations in commercial crops. Entomopathogenic nematodes (EPNs) may provide a more species‐ and time‐specific control and combining these methods may have synergy. Three in‐field vegetation margins, divided into four plots each (two floral and two grass), were established in March 2022. Three strips of sprouting broccoli were planted parallel in April, May and June. Each strip had a split‐plot design, with eight replicates of four nematode treatments placed alongside each floral/grass plot: Steinernema carpocapsae, S. feltiae and Heterorhabditis bacteriophora and a nematode‐free control. Pest pressure was assessed using adult counts. Assessments of crop yield and crop damage associated specifically with the swede smidge (Contarinia nasturtii) were also made. It was found that utilising a combination of control measures was successful in controlling a wider range of pests, though the success of each control method was highly dependent on the time of crop plantings. Field margins had a significant influence on crop damage associated with the swede midge. The EPN H. bacteriophora was an effective control method for the cabbage stem flea beetle and the swede midge, as well as reducing crop damage and yield loss. Field margins and EPNs combined have an interactive effect on crop yield, highlighting the need for control methods to be tested in combination in future integrated pest management research to accurately understand their effects within an integrated system.

Unwelcomed guests: Impact of deer harvest on corn and soybean wildlife damage

While wildlife damages are a relatively small part of crop insurance claims in the United States, reported damages from wildlife are growing and producers are concerned this issue is getting worse. Wildlife agencies advocate for hunting as a control to wildlife damages to crops but to our knowledge, no study has econometrically explored if hunting could reduce crop damage. Therefore, we analyze if county-level white-tailed deer (Odocoileus virginianus) harvest and other factors like habitat and field fragmentation are associated with corn and soybean acres being indemnified due to wildlife damages in Tennessee. We estimate a fractional probit model using crop insurance data to measure crop damages and a county-level white-tailed deer harvest dataset. We find a higher coverage level of a policy increases the likelihood of both corn and soybean acres being indemnified due to wildlife damage. We also find white-tailed deer harvest during the period post-harvest in the previous crop year (November through January) slightly decreases the likelihood of a claim being made due to wildlife damage for corn but was not found to mitigate losses to soybeans. The whitetail hunting season appears to align better with corn production to reduce crop losses than with soybeans. This study will be useful for wildlife and crop agencies in state and federal governments to develop policies to more effectively utilize hunting to reduce crop insurance claims from wildlife damages.

Roads reduce breeding occupancy and productivity in barn owls

AbstractUrbanization and expanding road networks threaten some avian populations through habitat loss and degradation. Barn owls (Tyto alba) have been particularly affected through roadway mortality, loss of grassland, and conversion or destruction of nesting sites. To combat declines and bolster reproduction, and as part of integrated pest management to reduce crop damage, some land managers provide supplemental owl nesting sites via the installation of nest boxes. If nearby habitat and land cover characteristics are not considered when placing boxes, such as major roads or other anthropogenic features, owls could be attracted to locations that could hinder populations further. We investigated the effects of roads and urban areas on barn owl breeding occupancy and productivity to provide information to help guide the placement of nest boxes. We monitored >300 nest boxes over the 2020 and 2021 breeding seasons in southwestern Idaho, USA, where substantial roadway mortality of barn owls occurs. Barn owls occupied >60% of nest boxes, but the likelihood of breeding occupancy decreased with increasing proximity of nest boxes to roads. Boxes 500 m from roads had a predicted probability of occupancy of over 0.9, which was nearly double that of boxes within 100 m. Proximity to roads also was associated with reduced productivity such that boxes within 100 m of roads were predicted to produce approximately 1 fewer fledgling than those 500 m away. There was no evidence that the proportion of urban area surrounding nest boxes substantially influenced breeding occupancy or productivity. Thus, land managers can consider placing nest boxes for barn owls farther from roads when possible as a potential tactic to increase the probability of breeding occupancy and to foster higher productivity.

Ragi Pest Control

The rise of global population has put increasing pressure on the agriculture industry to meet the demand for food. However, the growing use of pesticides and insecticides in conventional farming practices has caused significant harm to the environment and human health. Thus, there is a growing interest in using sustainable agriculture practices that reduce the use of these harmful chemicals. One such practice is pest detection, which enables farmers to detect pests in their crops before they cause significant damage. In this context, this project aims to develop a pest detection system using IoT and Arduino. The system will be designed to detect pests in crops through a combination of sensors and machine learning algorithms. The system will consist of an Arduino microcontroller, soil moisture sensors, temperature and humidity sensors, infrared sensors or camera modules, and a Wi-Fi or Bluetooth module. The sensors will collect data on soil moisture levels, temperature, humidity, and pest activity. The data will be sent to a cloud-based server or database for analysis and visualization. The infrared sensor or camera module will detect the presence of pests in the crops. The system will use machine learning algorithms to distinguish between pests and other objects, such as leaves or debris. When pests are detected, the system will alert the farmer through a buzzer or LED connected to the Arduino board. The farmer can then take appropriate action, such as applying pesticide or removing infested plants. The pest detection system has the potential to reduce the use of harmful pesticides and insecticides in agriculture, as farmers will be able to identify pests before they cause significant damage. The system will also provide farmers with real-time information on pest activity, enabling them to take proactive measures to control pests and reduce crop damage. Additionally, the system can track and store the pest detection data over time, allowing farmers to monitor trends and patterns in pest activity. In conclusion, this project proposes the development of a pest detection system using IoT and Arduino that will enable farmers to monitor pest activity in their crops in real-time. The system has the potential to reduce the use of harmful chemicals in agriculture and improve crop yield while ensuring sustainable and environmentally friendly practices.

Positional Effects of Bottle-Baited Traps in Reducing Infestation Level of Coffee Berry Borer Hypothenemus Hampei Ferrari in Kilimanjaro Region, Tanzania

Coffee berry borer (CBB) is among the key insect pests of coffee worldwide. The use of bottle-baited traps has been in practice in several coffee-growing areas including Tanzania. However, there is limited information about the influence of height and spacing of commonly used bottle-baited traps in managing CBB in coffee-growing areas in the country. Therefore, the objective of this research was to evaluate the effect of height where traps were placed (0.6, 1.2, and 1.6 m) on the reduction of infestation level of coffee berry borers at different developmental stages of coffee fruit (green and red fruit) under field conditions. The experiment followed a completely randomized block design with a factorial arrangement and four replications, three (lower, middle, and upper) levels of height and spacing were placed for 7 months. The number of captured CBB and damaged berries percentage was evaluated. The data were analyzed by R Software (2021) through an analysis of variance and means were separated by Turkey’s (0.05). A significant minimum berries damage (0.26%) as an implication of the lowest CBB infestation level was shown at the height of 0.6 m (for all stages of berries). On the other hand, at the red berries stage, the lowest damage (11.12%) was observed at the height of 1.6 m. Generally, this study deduced that the lower the height from which the traps are placed, the lower the infestation level of CBB hence reducing crop damage by the pest.

Comparison of time series temperature prediction with auto-regressive integrated moving average and recurrent neural network

The region of Beni Mellal, Morocco is heavily dependent on the agricultural sector as its primary source of income. Accurate temperature prediction in agriculture has many benefits including improved crop planning, reduced crop damage, optimized irrigation systems and more sustainable agricultural practices. By having a better understanding of the expected temperature patterns, farmers can make informed decisions on planting schedules, protect crops from extreme temperature events, and use resources more efficiently. The lack of data-driven studies in agriculture impedes the digitalization of farming and the advancement of accurate long-term temperature prediction models. This underscores the significance of research to identify the optimal machine learning models for that purpose. A 22-year time series dataset (2000-2022) is used in the study. The machine-learning model auto-regressive integrated moving average (ARIMA) and deep learning models simple recurrent neural network (SimpleRNN), gated recurrent unit (GRU), and long short-term memory (LSTM) were applied to the time series. The results are evaluated based on the mean absolute error (MAE). The findings indicate that the deep learning models outperformed the machine-learning model, with the GRU model achieving the lowest MAE.

Using mung bean as a trap crop and biopesticide for control of soybean pest

Many pests attack soybean plants and can cause high-yield losses. To reduce crop damage, using chemicals in the field can potentially cause harmful effects on the environment. Using traps-crops, and biopesticides will be tested as an effective and safe control method. The research was done in the dry season of the tidal land of Tumih Village South of Kalimantan. The design used was RCBD which consisted of six treatments and was repeated four times. The treatment tested were 1). Check, 2). Biopesticides 6 times, 3). Biopesticide 4 times, 4). Biopesticides 6 times + mung bean as trap crop 6%, 5). Biopesticides 4 times + mung bean as trap crop 6%, 6. Insecticides 6 times. The results indicated that armyworms Spodoptera litura and Lamprosema indicata were important pests. Armyworm attacks in the vegetative phase reached 82%. However, the yields obtained were around 2,069 to 2,463 t/ha. Soybean yields in treatment plots with 6% trap crop were 1,664 to 1,969 t/ha. Anjasmoro varietal is not recommended In endemic areas of the armyworm.

Artificial Intelligence & Nature-Based Solutions in Agriculture: A BT Cotton Pest Management Case Study in India

Artificial intelligence (AI)-based pest management advisory, based on integrated pest management (IPM), provided to cotton farmers on smartphones, resulted in a reduction in pest attacks & up to 22% higher income in the 1 year 2020-21 in Ranebennur, Karnataka, and Wardha, Maharashtra states. However, no significant benefit was seen in a multi-state experiment in 2021-22 due to unusually high rainfall, resulting in lower pest attacks. The artificial intelligence was used in pest detection & counting insect numbers in the pheromone trap to decide if threshold numbers were reached for pesticide spraying decisions. This was 1-2 weeks in advance of mass pest emergence and could control it to reduce crop damage. It required manual trap checking by the farmers on a weekly basis, which many farmers disliked. Artificial intelligence coupled to remote sensing, GIS, and/or farm sensors can benefit the farmers by cutting costs, increasing yield, and enabling cleaner production. Lower environmental pollution and less risk to farmers and consumers are co-benefits of the AI-IPM package. However, mating disruption technology, a competitor, includes putting 4-6 pheromone traps per acre for the mass capture of moths. It is organic-compatible, and another competitor is the mechanical growing degree day (GDD)-based IPM advisory, such as that provided by the startup “Fasal.” These are unintelligent, mechanical, but very effective algorithms. Thus, a cautious, logical, and gradual approach is needed in promoting AI in agriculture, also keeping in mind its impact on labour displacement.

Mathematical Modelling of Host-Pest Interaction in the Presence of Insecticides and Resistance: A Case of Fall Armyworm

Several pest management programs have been developed to control rising agricultural pest populations. However, the challenge of rapid evolution and pest resistance towards control measures continues to cause high production losses to maize farmers in Africa. Few models have attempted to address the issue of fall armyworm (FAW) but have barely incorporated the effect of insecticide resistance. Models with resistance would help predict the dynamics of the FAW population, thus mitigating losses. The main objectives of this work were to develop, analyze, and numerically simulate a susceptible-infected deterministic mathematical model expressing the FAW-maize interaction and population dynamics under insecticidal sprays and resistance FAW larvae. Three model steady states are established. Their local stability is conducted using either the eigenvalue or the Routh–Hurwitz stability criteria, and their global stability is analyzed using either the Castillo–Chavez, Perron eigenvector, or the Lyapunov methods. An expression for the basic reproduction number R0, together with the sensitivity analysis of its parameter values, is provided. Numerical analysis is conducted on various model parameter values. The results established all the model steady states to be locally and globally asymptotically stable at R0≤1. Also, resistance ω increased the infection rates by increasing the FAW larvae survival rate λ and reducing the insecticidal efficacy δR and δN. This work informs the agriculturists and policymakers on pest control with the best ways to use insecticides to minimize pest resistance and enhance efficacy in production. Pest control measures should be modified to lower the FAW survival rate and all model parameters contributing to resistance formation by FAW larvae to minimize FAW-host interaction, thus reducing crop damage.

Reducing the number of grazing geese on agricultural fields - Effectiveness of different scaring techniques

Scaring is a widely used damage mitigation tool to make agricultural fields less attractive to wildlife and by that reduce crop damage. However, few experimental studies exist where the numerical response of different scaring devices has been compared. We tested experimentally the effect of three different scaring devices (kite, scarecrow, inflatable man) on the number of geese in fields with cereals, ley, rapeseed, potatoes, and carrots in Sweden. Geese were counted by camera traps and two approaches were used; in a first (model 1) only geese within 50–150m of the scaring devices were counted, and in a second (model 2) all geese in the field were included. A total of 42,281 geese were counted: Greylag goose Anser anser was the most common species (87%), followed by bean goose Anser fabalis (6%), greater white-fronted goose Anser albifrons (3%), barnacle goose Branta leucopsis (2%), and Canada goose Branta canadensis (2%). During scaring the number of geese significantly decreased for all three devices in model 2. The inflatable man decreased goose numbers by 90.0 %, scarecrow 64.6%, and kite 60.5%. A similar pattern was found in model 1, but the decrease was not significant. Our study shows that the scaring devices studied can reduce goose grazing pressure for some time and locally. However, since geese continue to graze during scaring, we conclude that scaring alone is not a final solution to mitigate crop damage. Future work to develop more effective control measures should address the efficiency of other management tools and scaring techniques in combination.

Experimental assessment of laser scarecrows for reducing avian damage to sweet corn.

Birds damage crops, costing millions of dollars annually, and growers utilize a variety of lethal and nonlethal deterrents in an attempt to reduce crop damage by birds. We experimentally tested laser scarecrows for their effectiveness at reducing sweet corn (Zea mays) damage. We presented 18 captive flocks of free-flying European starlings (Sturnus vulgaris) with fresh sweet corn ears distributed on two plots where laser and control treatments were alternated each day and allowed each flock to forage over 5 days. In 16 trials, fresh sweet corn ears were mounted on wooden sticks distributed from 0 to 32 m from laser units (Stick Trials), and in two trials birds foraged on ripe corn grown from seed in the flight pen (Natural Trials). We aimed to determine if laser-treated plots had significantly less damage overall and closer to the laser unit, and whether birds became more or less likely to forage in laser-treated plots over time. Lasers reduced damage overall, marginally in Stick Trials and dramatically in Natural Trials. Damage increased during each week in both trial types. Damage increased significantly with distance from lasers, and significant treatment effects occurred up to ~20 m from lasers. Our results concur with recent field trials demonstrating strong reductions in sweet corn damage when lasers are deployed. This study provides a first look at how birds respond to repeated laser exposure and whether damage increases with distance from lasers. Key differences between pen and field trials are discussed. © 2023 Society of Chemical Industry. This article has been contributed to by U.S. Government employees and their work is in the public domain in the USA.

Biological control of Magnaporthe oryzae using natively isolated Bacillus subtilis G5 from Oryza officinalis roots.

Rice blast, caused by Magnaporthe oryzae, is a major threat to global rice production causing significant crop losses and impacting grain quality. The annual loss of rice production due to this disease ranges from 10% to 30%. The use of biologically controlled strains, instead of chemical pesticides, to control plant diseases has become a research hotspot. In this study, an antagonistic endophytic bacterial strain was isolated from the roots of Oryza officinalis using the traditional isolation and culture methods. A phylogenetic tree based on 16S RNA and whole-genome sequencing identified isolate G5 as a strain of Bacillus subtilis. This isolate displayed strong antagonistic effects against different physiological strains of M. oryzae. After co-culture in LB medium for 7 days, the inhibition rates of the mycelial growth of four strains of M. oryzae, ZB15, WH97, Guy11, and T-39800E were 98.07 ± 0.0034%, 98.59 ± 0.0051%, 99.16 ± 0.0012%, and 98.69 ± 0.0065%, respectively. Isolate G5 significantly inhibited the formation of conidia of M. oryzae, with an inhibition rate of 97% at an OD600 of 2. Isolate G5 was able to provide 66.81% protection against rice blast under potted conditions. Whole-genome sequencing revealed that the genome size of isolate G5 was 4,065,878 bp, including 4,182 coding genes. Using the anti-SMASH software, 14 secondary metabolite synthesis gene clusters were predicted to encode antifungal substances, such as fengycin, surfactin, and bacilysin. The G5 isolate also contained genes related to plant growth promotion. These findings provide a theoretical basis for expounding the biocontrol mechanisms of this strain and suggest further development of biogenic agents that could effectively inhibit rice blast pathogen growth and reduce crop damage, while being environmentally friendly, conducive to ecological development, and a sustainable alternative to chemical pesticides. This study also enriches the relevant research on endophytes of wild rice, which proves that wild rice is a valuable microbial resource bank.

Revolutionizing Agriculture: Machine Learning-Driven Crop Recommendations and Disease Detection in Fertilizer Management

Modern agriculture faces a multitude of challenges, including crop failures, disease outbreaks, and suboptimal yields, primarily stemming from the underutilization of advanced farming technologies and a lack of expert guidance. This research proposes a comprehensive solution consisting of three key components: a Crop Disease Detection System, a Fertilizer Recommendation System, and a Crop Suggestion System. The Crop Disease Detection System employs state-of-the-art technology to evaluate crop health by analyzing the condition of plant leaves, enabling early and accurate identification of agricultural diseases. Simultaneously, the Fertilizer Recommendation System leverages soil quality data and environmental factors to provide personalized fertilizer recommendations, optimizing nutrient application. An essential element of this system is a robust soil testing module, recognizing the critical importance of assessing soil quality. Soil fertility evaluation, guided by soil pH measurements, enables precise crop predictions. The proposed system utilizes Machine Learning classification algorithms to predict suitable crops based on essential soil parameters—Phosphorus, Potassium, and Nitrogen levels. It also offers tailored fertilizer recommendations to enhance soil fertility. By implementing these interconnected solutions, this research aims to significantly improve crop yields while reducing crop damage. This holistic approach empowers farmers with the tools and knowledge needed to enhance agricultural productivity and food security. Anticipated outcomes include higher crop yields and a reduced vulnerability of crops to diseases, contributing to a more sustainable and prosperous agricultural sector.

Federated Transfer Learning for Rice-Leaf Disease Classification across Multiclient Cross-Silo Datasets

Paddy leaf diseases encompass a range of ailments affecting rice plants’ leaves, arising from factors like bacteria, fungi, viruses, and environmental stress. Precision agriculture leverages technologies for enhanced crop production, with disease detection being a vital element. Prompt identification of diseases in paddy leaves is critical for curtailing their propagation and reducing crop damage. However, manually diagnosing paddy diseases in regions with vast agricultural areas and limited experts proves immensely difficult. The utilization of machine learning (ML) and deep learning (DL) for diagnosing diseases in agricultural crops appears to be effective and well-suited for widespread application. These ML/DL methods cannot ensure data privacy, as they involve sharing training data with a central server, overlooking competitive and regulatory considerations. As a solution, federated learning (FL) aims to facilitate decentralized training to tackle the identified limitations of centralized training. This paper utilizes the FL approach for the classification of rice-leaf diseases. The manuscript presents an effective approach for rice-leaf disease classification with a federated architecture, ensuring data privacy. We have compiled an unbalanced dataset of rice-leaf disease images, categorized into four diseases with their respective image counts: bacterial blight (1584), brown spot (1440), blast (1600), and tungro (1308). The proposed method, called federated transfer learning (F-TL), maintains privacy for all connected devices using a decentralized client-server setup. Both IID (independent and identically distributed) and non-IID datasets were utilized for testing the F-TL framework after preprocessing. Initially, we conducted an effectiveness analysis of CNN and eight transfer learning models for rice-leaf disease classification. Among them, MobileNetV2 and EfficientNetB3 outperformed the other transfer-learned models. Subsequently, we trained these models using both IID and non-IID datasets in a federated learning environment. The framework’s performance was assessed through diverse scenarios, comparing it with traditional and federated learning models. The evaluation considered metrics like validation accuracy, loss as well as resource utilization such as CPU and RAM. EfficientNetB3 excelled in training, achieving 99% accuracy with 0.1 loss for both IID and non-IID datasets. MobilenetV2 showed slightly lower training accuracy at 98% (IID) and 90% (non-IID) with losses of 0.4 and 0.6, respectively. In evaluation, EfficientNetB3 maintained 99% accuracy with 0.1 loss for both datasets, while MobilenetV2 achieved 90% (IID) and 97% (non-IID) accuracy with losses of 0.6 and 0.2, respectively. Results indicated the F-TL framework’s superiority over traditional distributed deep-learning classifiers, demonstrating its effectiveness in both single and multiclient instances. Notably, the framework’s strengths lie in its cost-effectiveness and data-privacy assurance for resource-constrained edge devices, positioning it as a valuable alternative for rice-leaf disease classification compared to existing tools.

Agricultural diversification promotes sustainable and resilient global rice production.

Rice is a staple food for half of the human population, but the effects of diversification on yields, economy, biodiversity and ecosystem services have not been synthesized. Here we quantify diversification effects on environmental and socio-economic aspects of global rice production. We performed a second-order meta-analysis based on 25 first-order meta-analyses covering four decades of research, showing that diversification can maintain soil fertility, nutrient cycling, carbon sequestration and yield. We used three individual first-order meta-analyses based on 39 articles to close major research gaps on the effects of diversification on economy, biodiversity and pest control, showing that agricultural diversification can increase biodiversity by 40%, improve economy by 26% and reduce crop damage by 31%. Trade-off analysis showed that agricultural diversification in rice production promotes win-win scenarios between yield and other ecosystem services in 81% of all cases. Knowledge gaps remain in understanding the spatial and temporal effects of specific diversification practices and trade-offs.

A conceptual evaluation of a weed control method with post-damage application of herbicides: A composite intelligent intra-row weeding robot

Weed management is an important part of crop production. Efficient weeding can help increase the yield and quality of crops. However, traditional weeding methods often fail to consider weeding quality and environmental protection. To address these shortcomings, a weeding method in which herbicides are applied after injury to weed tissue was developed, and an intelligent intra-row weeding robot was designed to evaluate the performance of this method. In control experiments, the proposed innovative weeding method showed a significantly better weed removal rate than purely mechanical weeding; its greatest contribution was to reduce the number of mechanical weeding operations, thereby reducing energy consumption and damage to the environment. Moreover, compared with chemical weeding approaches, the proposed method achieved a good weed removal rate while minimizing the use of herbicides, thereby reducing crop damage, weed resistance, and environmental pollution. In the final maize (Zea mays L.) test area, the robot traveled at a speed of 2.0 km/h and operated at a transmission efficiency of 60 frames per second. Moreover, the robot achieved a weed removal rate of 90.0% and a crop damage rate of 1.9%. Even better performance was obtained in a test area with Chinese cabbage, with a 94.5% weed removal rate and a 0.8% crop damage rate. Meanwhile, this method used only 15.3% of the herbicide required by traditional chemical weed control methods. Thus, the environmental protection properties of the intelligent intra-row weeding robot are more sustainable than those of conventional weeding methods, and the robot can achieve good performance in small-area fields.