A database of vegetation treatments in New Mexico, USA and surrounding regions.

Active management practices to reduce or promote particular vegetation, known as vegetation treatments, are a common part of environmental management and they are conducted for a variety of purposes including wildfire risk mitigation, invasive species management, and ecological restoration. Vegetation treatment for wildfire mitigation in particular have increased dramatically in the Western United States in the past several decades. While vegetation treatments are common, data regarding the timing, location, and type of treatments conducted are often only maintained by the organization that conducted the work, hampering the ability of managers and researchers to understand the distribution and timing of vegetation treatments across a landscape. This dataset is a collection of spatially referenced records of vegetation treatments such as mechanical thinning, prescribed burning, and herbicide applications that were conducted in the state of New Mexico, USA and adjacent parts of Colorado, Oklahoma, and Texas. Spatial data were collected through requests to the regional or state offices for the relevant agencies (e.g., The Bureau of Land Management, the U.S. Forest Service, New Mexico State Forestry Division). The accuracy of this data collection approach was assessed by conducting more intensive data collection in five randomly selected focal watersheds across New Mexico. In these watersheds local offices of the larger agencies were contacted, as well as any smaller groups (e.g., soil and water conservation districts, municipalities, and environmental non-profits), and in person visits were made to gather any information on vegetation treatments possible. The overall dataset includes records of treatments spanning a century and includes records of 9.9 million acres of treatments conducted by more than a dozen different organizations. In the five focal watershed that we surveyed the database contained 7.4 % fewer acres of treated land than the more intensive interview approach. This spatially extensive dataset on vegetation treatments will be useful for researchers quantifying or modelling the effect of vegetation management on fire risk and behaviour. Additionally, this data will be useful to ecologists studying the distribution, movement, and habitat associations of a variety of plant an animal species. Finally, this data will be useful for research on landscape conservation and management.

An intelligent spraying system for weeds in wheat fields based on a dynamic model of droplets impacting wheat leaves.

Introduction: Targeted herbicide application refers to precise application of herbicides in weed-infested areas according to the location and density of farmland weeds. At present, targeted herbicide application in wheat fields generally faces problems including the low herbicide adhesion rate, leading to omission and excessive loss of herbicides. Methods: To solve these problems, changes in the impact force of herbicide and the weed leaves in the operation process of a spraying system were studied from the interaction between weeds and herbicides applied. A dynamic model of weed leaves was established. On this basis, the research indicated that the herbicide adhesion rate is highest under spraying pressure of 0.4 MPa and flow rate of 0.011 kg/s when the spray height is 300 mm. To study the dynamic deformation of weed leaves and the distribution of liquid herbicides in the external flow field under weed-herbicide interaction, a dynamic simulation model of herbicide application was built using the finite element method. Results and Discussion: The results show that when the spray height is 300 mm, the maximum weed leaf deformation index (LDI) is 0.43 and the velocity in the external flow field is 0 m/s under spraying pressure of 0.4 MPa and flow rate of 0.011 kg/s. This finding indicates that the herbicide is not splashed elsewhere and the turbulence intensity in the weed area is 2%, implying steady flow of the herbicide, most of which can be retained on weed leaves. Field test results of application quality of the herbicide show that the maximum LDI is 0.41 and the coverage of the herbicide in the sheltered area below the leaves is 19.02% when the spraying pressure is 0.4 MPa, flow rate is 0.011 kg/s, and spray height is 300 mm. This solves the problem of a low rate of utilization of herbicides because the herbicide passes through weed plants, and achieves the precision herbicide application in wheat fields.

Detection and coverage estimation of purple nutsedge in turf with image classification neural networks.

BACKGROUND: Accurate detection of weeds and estimation of their coverage is crucial for implementing precision herbicide applications. Deep learning (DL) techniques are typically used for weed detection and coverage estimation by analyzing information at the pixel or individual plant level, which requires a substantial amount of annotated data for training. This study aims to evaluate the effectiveness of using image-classification neural networks (NNs) for detecting and estimating weed coverage in bermudagrass turf. RESULTS: Weed-detection NNs, including DenseNet, GoogLeNet and ResNet, exhibited high overall accuracy and F1 scores (≥0.971) throughout the k-fold cross-validation. DenseNet outperformed GoogLeNet and ResNet with the highest overall accuracy and F1 scores (0.977). Among the evaluated NNs, DenseNet showed the highest overall accuracy and F1 scores (0.996) in the validation and testing data sets for estimating weed coverage. The inference speed of ResNet was similar to that of GoogLeNet but noticeably faster than DenseNet. ResNet was the most efficient and accurate deep convolution neural network for weed detection and coverage estimation. CONCLUSION: These results demonstrated that the developed NNs could effectively detect weeds and estimate their coverage in bermudagrass turf, allowing calculation of the herbicide requirements for variable-rate herbicide applications. The proposed method can be employed in a machine vision-based autonomous site-specific spraying system of smart sprayers. © 2024 Society of Chemical Industry.

Convolutional neural network based on the fusion of image classification and segmentation module for weed detection in alfalfa.

BACKGROUND: Accurate and reliable weed detection in real time is essential for realizing autonomous precision herbicide application. The objective of this research was to propose a novel neural network architecture to improve the detection accuracy for broadleaf weeds growing in alfalfa. RESULTS: A novel neural network, ResNet-101-segmentation, was developed by fusing an image classification and segmentation module with the backbone selected from ResNet-101. Compared with existing neural networks (AlexNet, GoogLeNet, VGG16, and ResNet-101), ResNet-101-segmentation improved the detection of Carolina geranium, catchweed bedstraw, mugwort and speedwell from 78.27% to 98.17%, from 79.49% to 98.28%, from 67.03% to 96.23%, and from 75.95% to 98.06%, respectively. The novel network exhibited high values of confusion matrices (>90%) when trained with sufficient data sets. CONCLUSION: ResNet-101-segmentation demonstrated excellent performance compared with existing models (AlexNet, GoogLeNet, VGG16, and ResNet-101) for detecting broadleaf weeds growing in alfalfa. This approach offers a promising solution to increase the accuracy of weed detection, especially in cases where weeds and crops have similar plant morphology. © 2024 Society of Chemical Industry.

Semi-supervised learning methods for weed detection in turf.

BACKGROUND: Accurate weed detection is a prerequisite for precise automatic precision herbicide application. Previous research has adopted the laborious and time-consuming approach of manually labeling and processing large image data sets to develop deep neural networks for weed detection. This research introduces a novel semi-supervised learning (SSL) approach for detecting weeds in turf. The performance of SSL was compared with that of ResNet50, a fully supervised learning (FSL) method, in detecting and differentiating sub-images containing weeds from those containing only turfgrass. RESULTS: Compared with ResNet50, the evaluated SSL methods, Π-model, Mean Teacher, and FixMatch, increased the classification accuracy by 2.8%, 0.7%, and 3.9%, respectively, when only 100 labeled images per class were utilized. FixMatch was the most efficient and reliable model, as it exhibited higher accuracy (≥0.9530) and F1 scores (≥0.951) with fewer labeled data (50 per class) in the validation and testing data sets than the other neural networks evaluated. CONCLUSION: These results reveal that the SSL deep neural networks are capable of being highly accurate while requiring fewer labeled training images, thus being more time- and labor-efficient than the FSL method. © 2024 Society of Chemical Industry.

Evaluation of convolutional neural networks for herbicide susceptibility-based weed detection in turf.

Deep learning methods for weed detection typically focus on distinguishing weed species, but a variety of weed species with comparable plant morphological characteristics may be found in turfgrass. Thus, it is difficult for deep learning models to detect and distinguish every weed species with high accuracy. Training convolutional neural networks for detecting weeds susceptible to herbicides can offer a new strategy for implementing site-specific weed detection in turf. DenseNet, EfficientNet-v2, and ResNet showed high F1 scores (≥0.986) and MCC values (≥0.984) to detect and distinguish the sub-images containing dollarweed, goosegrass, old world diamond-flower, purple nutsedge, or Virginia buttonweed growing in bermudagrass turf. However, they failed to reliably detect crabgrass and tropical signalgrass due to the similarity in plant morphology. When training the convolutional neural networks for detecting and distinguishing the sub-images containing weeds susceptible to ACCase-inhibitors, weeds susceptible to ALS-inhibitors, or weeds susceptible to synthetic auxin herbicides, all neural networks evaluated in this study achieved excellent F1 scores (≥0.995) and MCC values (≥0.994) in the validation and testing datasets. ResNet demonstrated the fastest inference rate and outperformed the other convolutional neural networks on detection efficiency, while the slow inference of EfficientNet-v2 may limit its potential applications. Grouping different weed species growing in turf according to their susceptibility to herbicides and detecting and distinguishing weeds by herbicide categories enables the implementation of herbicide susceptibility-based precision herbicide application. We conclude that the proposed method is an effective strategy for site-specific weed detection in turf, which can be employed in a smart sprayer to achieve precision herbicide spraying.

Evaluationof compact air-induction flat fan nozzles for herbicide applications: Spray drift and biological efficacy.

Nozzles are the most critical component of a sprayer for pesticide applications. Recently, air-induction nozzles and twin flat-fan air-induction nozzles have started to be used for herbicide applications. In order to evaluate the potential of compact air-induction nozzles for herbicide spraying, this paper compares the effects of air-induction nozzles and standard flat-fan nozzles on spray atomization, deposition, drift, and weed control efficacy in maize and wheat. Droplet spectra were measured by a laser particle size analyzer, and drift potential values were determined using a drift test bench (ISO 22401). A field study was conducted to compare the spray drift and biological efficacy between Lechler standard flat-fan nozzles and compact air-induction nozzles including different nozzle sizes. In the range from 0.2 to 0.4 MPa, the droplet size classes of the LU and ST nozzles were very similar and ranged from fine to very fine, while the droplets of the air-induction nozzles IDK and IDKT were medium or coarse depending on the spray pressure and nozzle size. The drift potential trials showed that the droplet size characteristics, mainly V 100, are strongly linked with the drift reduction potential. Both drift potential and field results showed that the compact air-induction nozzles had a good performance in drift reduction. In terms of weed control biological efficacy, there were no significant differences between standard flat-fan nozzles and air-induction nozzles. In all cases, the efficacy values were above 80% both in maize and in wheat. In conclusion, air-induction nozzles are recommended for herbicide applications as they provide good biological efficacy while significantly reducing the amount of spray drift, which is of great significance for the protection of the environment and the surrounding sensitive crops.

Deep learning for detecting herbicide weed control spectrum in turfgrass.

BACKGROUND: Precision spraying of postemergence herbicides according to the herbicide weed control spectrum can substantially reduce herbicide input. The objective of this research was to evaluate the effectiveness of using deep convolutional neural networks (DCNNs) for detecting and discriminating weeds growing in turfgrass based on their susceptibility to ACCase-inhibiting and synthetic auxin herbicides. RESULTS: GoogLeNet, MobileNet-v3, ShuffleNet-v2, and VGGNet were trained to discriminate the vegetation into three categories based on the herbicide weed control spectrum: weeds susceptible to ACCase-inhibiting herbicides, weeds susceptible to synthetic auxin herbicides, and turfgrass without weed infestation (no herbicide). ShuffleNet-v2 and VGGNet showed high overall accuracy (≥ 0.999) and F1 scores (≥ 0.998) in the validation and testing datasets to detect and discriminate weeds susceptible to ACCase-inhibiting and synthetic auxin herbicides. The inference time of ShuffleNet-v2 was similar to MobileNet-v3, but noticeably faster than GoogLeNet and VGGNet. ShuffleNet-v2 was the most efficient and reliable model among the neural networks evaluated. CONCLUSION: These results demonstrated that the DCNNs trained based on the herbicide weed control spectrum could detect and discriminate weeds based on their susceptibility to selective herbicides, allowing the precision spraying of particular herbicides to susceptible weeds and thereby saving more herbicides. The proposed method can be used in a machine vision-based autonomous spot-spraying system of smart sprayers.

A deep learning-based method for classification, detection, and localization of weeds in turfgrass.

BACKGROUND: Precision spraying of synthetic herbicides can reduce herbicide input. Previous research demonstrated the effectiveness of using image classification neural networks for detecting weeds growing in turfgrass, but did not attempt to discriminate weed species and locate the weeds on the input images. The objectives of this research were to: (i) investigate the feasibility of training deep learning models using grid cells (subimages) to detect the location of weeds on the image by identifying whether or not the grid cells contain weeds; and (ii) evaluate DenseNet, EfficientNetV2, ResNet, RegNet and VGGNet to detect and discriminate multiple weed species growing in turfgrass (multi-classifier) and detect and discriminate weeds (regardless of weed species) and turfgrass (two-classifier). RESULTS: The VGGNet multi-classifier exhibited an F1 score of 0.950 when used to detect common dandelion and achieved high F1 scores of ≥0.983 to detect and discriminate the subimages containing dallisgrass, purple nutsedge and white clover growing in bermudagrass turf. DenseNet, EfficientNetV2 and RegNet multi-classifiers exhibited high F1 scores of ≥0.984 for detecting dallisgrass and purple nutsedge. Among the evaluated neural networks, EfficientNetV2 two-classifier exhibited the highest F1 scores (≥0.981) for exclusively detecting and discriminating subimages containing weeds and turfgrass. CONCLUSION: The proposed method can accurately identify the grid cells containing weeds and thus precisely locate the weeds on the input images. Overall, we conclude that the proposed method can be used in the machine vision subsystem of smart sprayers to locate weeds and make the decision for precision spraying herbicides onto the individual map cells. © 2022 Society of Chemical Industry.

Utility of roller wiper applications of dicamba for Palmer amaranth control in soybean.

BACKGROUND: The commercialization of dicamba-resistant soybean has resulted in increased concern for off-target movement of dicamba onto sensitive vegetation. To mitigate the off-target movement through physical drift, one might consider use of rope wicks and other wiper applicators. Although wiper-type application methods have been efficacious in pasture settings, the utility of dicamba using wiper applicators in agronomic crops is not available in scientific literature. To determine the utility of roller wipers for dicamba applications in dicamba-resistant soybean, two separate experiments were conducted in the summer of 2020 and replicated in both Keiser and Fayetteville, AR, USA. RESULTS: Utilizing opposing application directions and a 2:1:1 ratio of water: formulated glyphosate: formulated dicamba were the most efficacious practices for controlling Palmer amaranth. The high herbicide concentrations and wiping in opposing directions increased dicamba-resistant soybean injury when the wiper contacted the crop, but no yield loss was observed because of this injury. Broadcast applications resulted in greater Palmer amaranth mortality than roller wiper applications, and the most effective roller wiper treatments were when two sequential applications were made inside the crop canopy. CONCLUSIONS: Dicamba applications require adequate coverage for optimum weed control. While efforts can be made to increase roller wiper efficacy by optimizing coverage and timing of applications, broadcast applications are superior to roller wiper applicators for weed control. Roller wiper applications did not reduce soybean yield, thus wiper-type applications may be safely used in dicamba-resistant soybean, albeit the likelihood for off-target damage caused by volatilization of these treatments would need to be investigated. © 2022 The Authors. Pest Management Science published by John Wiley & Sons Ltd on behalf of Society of Chemical Industry.

Household herbicide use as a source of simazine contamination in urban surface waters.

Contamination of urban surface waters by herbicides is an increasing concern; however, sources of contamination are poorly understood, hindering the development of mitigation and regulatory strategies. Impervious surfaces, such as concrete in driveways and paths are considered an important facilitator for herbicide runoff to urban surface waters following applications by residential homeowners. This study assessed the transferability of a herbicide from concrete pavers treated with an off-the-shelf product, containing simazine as the active herbicide, marketed for residential homeowner application to impervious surfaces. Commercially available pavers were treated according to label directions and the effects of exposure time prior to irrigation, repeated irrigations, and dry time between irrigations on transferability of simazine to runoff were assessed. Simazine transferability was greatest when receiving an initial irrigation 1 h after application, with concentrations in runoff reduced by half when exposure times prior to the first irrigation were >2 days. Concentrations remained stable for repeated irrigations up to 320 days and exposures to outdoor conditions of 180 days prior to a first irrigation. Dry time between irrigations significantly influenced simazine transfer to runoff. Dry periods of 140 days resulted in approximately a 4-times increase in simazine transferability to runoff. These results suggest that herbicides used by homeowners, or any other users, on impervious surfaces are available to contaminate runoff for prolonged time periods following application at concentrations that may pose risks to aquatic life and for reuse of harvested runoff on parks and gardens. Regulators should consider the potential of hard surfaces to act as reservoirs for herbicides when developing policies and labelling products.

Shelterbelt Management Practices for Maximized Ecosystem Carbon Stocks on Agricultural Landscapes in Saskatchewan, Canada.

There is a significant knowledge gap in the area of management of the vast shelterbelt network currently existing on agricultural lands in Canada and across the world. Throughout eight decades of shelterbelt planting in Saskatchewan, Canada, there are no available records of shelterbelt management practices used by land managers, such as herbicides (H), fertilizers (F), irrigation (I), or tillage (T) applications, collectively referred to as HFIT management. The main objective of this large-scale study was to quantify the effects of HFIT management on shelterbelt carbon sequestration for six common tree and shrub species. Field data from 303 randomly selected shelterbelts across millions of hectares of agricultural land in three soil zones were combined with existing shelterbelt carbon stock curves for Saskatchewan, produced by a shelterbelt carbon management support tool, Belt-CaT, to estimate site-specific total ecosystem carbon (TEC) stocks. Estimated TEC stocks and annual rates for HFIT sites were compared to the no management sites used as a reference. HFIT management increased carbon stocks for the majority of species, four of six, resulting in higher TEC at any tree spacing, mostly at higher suitability sites. However, HFIT management effects were not consistent across individual species, land suitability, or planting designs. The top three HFIT management combinations for hybrid poplar were IT, HIT, and HI, for white spruce they were FT, IT, and FIT, and only FT benefited caragana shelterbelts. The lack of management practices makes unmanaged shelterbelts more unpredictable and unreliable, in terms of tree growth and carbon stocks sequestration potential.

Recurrent Selection with Sub-Lethal Doses of Mesotrione Reduces Sensitivity in Amaranthus palmeri.

Amaranthus palmeri, ranked as the most prolific and troublesome weed in North America, has evolved resistance to several herbicide sites of action. Repeated use of any one herbicide, especially at lower than recommended doses, can lead to evolution of weed resistance, and, therefore, a better understanding of the process of resistance evolution is essential for the management of A. palmeri and other difficult-to-control weed species. Amaranthus palmeri rapidly developed resistance to 4-hydroxyphenylpyruvate dioxygenase (HPPD) inhibitors such as mesotrione. The objective of this study was to test the potential for low-dose applications of mesotrione to select for reduced susceptibility over multiple generations in an A. palmeri population collected from an agricultural field in 2001. F0 plants from the population were initially treated with sub-lethal mesotrione rates and evaluated for survival three weeks after treatment. All F0 plants were controlled at the 1× rate (x = 105 g ai ha-1). However, 2.5% of the F0 plants survived the 0.5× treatment. The recurrent selection process using plants surviving various mesotrione rates was continued until the F4 generation was reached. Based on the GR50 values, the sensitivity index was determined to be 1.7 for the F4 generation. Compared to F0, HPPD gene expression level in the F3 population increased. Results indicate that after several rounds of recurrent selection, the successive generations of A. palmeri became less responsive to mesotrione, which may explain the reduced sensitivity of this weed to HPPD-inhibiting herbicides. The results have significance in light of the recently released soybean and soon to be released cotton varieties with resistance to HPPD inhibitors.

Influence of floating plant density on potential herbicide spray loss.

BACKGROUND: Mesocosm experiments were conducted to evaluate the effect of floating plant density on over-the-top spray solution loss to the column using a tracer dye. Experiments quantified in-water rhodamine water tracer (RWT) dye concentration after foliar treatment at 935 L ha-1 to waterhyacinth [Eichhornia crassipes (Mart.) Solms], waterlettuce (Pistia stratiotes L.) and giant salvinia (Salvinia molesta D.S. Mitchell) at 0, 25, 50 and 100% area covered (PAC). RESULTS: As expected, spray loss to the water surface decreased with increasing plant density in all species. However, each species exhibited an unique relationship between density and percentage spray loss. The plant material required to result in 50% spray loss (ED50 ) was 32, 62 and 55 PAC for waterhyacinth, waterlettuce and giant salvinia, respectively. Greater ED50 estimates in waterlettuce and giant salvinia were attributed to plant architecture and leaf orientation compared to waterhyacinth, which grows more vertically and has a greater overall surface area to intercept and retain spray solution. However, when treated at 100 PAC, waterhyacinth and waterlettuce resulted in 20-25% spray loss, whereas giant salvinia resulted in only 10% loss. Consequently, giant salvinia exhibited a near 1:1 relationship between spray loss and PAC (slope = -0.93). CONCLUSION: These data suggest that potential herbicide spray loss, as affected by plant density, is largely species-specific and dependent on leaf morphology and plant architecture. Further research will confirm these findings under field conditions as well as to identify other parameters that might affect spray loss when treating floating and emergent plants. © 2021 Society of Chemical Industry. This article has been contributed to by US Government employees and their work is in the public domain in the USA.

Monitoring and water quality impacts of an herbicide treatment on an aquatic invasive plant in a drinking water reservoir.

Effective treatment options are needed for the management of aquatic invasive species. An herbicide treatment was used to control an invasive aquatic plant, yellow floating heart (Nymphoides peltata) in a 3350-acre drinking water reservoir. The purpose of this research was to document the success of the treatment in an individual cove of the reservoir using in-situ sampling and reservoir-wide using remotely sensed Sentinel-2 satellite imagery. We also determined if the dying vegetation negatively impacted biological oxygen demand and dissolved oxygen concentrations in the cove. The aquatic herbicide ProcellaCOR™ (active ingredient = florpyrauxifen-benzyl) was used to treat a 55-acre infestation of YFH at a rate of 3 Prescription Dose Units (PDU)/ac-ft by a certified applicator in July 2019. Total surface coverage of yellow floating heart in the reservoir was reduced by more than 90% within 15 days after the treatment, and to less than 3.0 acres within 50 days after the treatment. No blooming flowers were observed after treatment and the surface coverage was close to 0% within 17 days after treatment in the cove. The effect of the herbicide treatment also appeared to carry over into the following growing season as the total surface coverage of yellow floating heart in the reservoir was less than 8 acres one year after the treatment in July 2020. The herbicide treatment resulted in short term increases in biological oxygen demand and decreases in dissolved oxygen at some sites in the cove within 3-10 days after the treatment. Dissolved oxygen then increased and concentrations were greater 42 days after treatment than they were before the treatment. Our results show that ProcellaCOR™ has the potential to control yellow floating heart infestations with relatively short-term negative impacts on dissolved oxygen concentrations. We also show that Sentinel-2 satellite imagery can be used to monitor the success of herbicide applications over large spatial and temporal scales that would not be possible from ground based monitoring alone.

Dormancy, a critical trait for weed success in crop production systems.

Agricultural practices exert selective forces on weed populations. As these practices change over time, weed adaptive traits also evolve, allowing weeds to persist in the new environment. However, only weeds having individuals showing the trait with adaptive significance will be able to cope with these changes, thus allowing a sub-population to be selected for persistence. In addition, changes in agricultural practices can select new weed species showing functional traits with characteristics adaptive to the modified system. Seed dormancy has long been recognized as a trait with enormous adaptive value to adjust weed biology to cropping systems. In this paper, we illustrate with examples of success and failure, the value of seed dormancy as a functional trait to cope with long-term changes in crop production systems. We show that successful outcomes are mostly related to the existence of sufficient variability for the functioning of physiological mechanisms that control dormancy characteristics as influenced by the agricultural environment. Presented examples illustrate how knowledge about the relationship that exists between agricultural practices and their selective pressure on seed dormancy can be instrumental in predicting changes in weed biotype dormancy characteristics or foreseeing the appearance of new weed species in future agricultural scenarios. © 2019 Society of Chemical Industry.

Weed Detection in Perennial Ryegrass With Deep Learning Convolutional Neural Network.

Precision herbicide application can substantially reduce herbicide input and weed control cost in turfgrass management systems. Intelligent spot-spraying system predominantly relies on machine vision-based detectors for autonomous weed control. In this work, several deep convolutional neural networks (DCNN) were constructed for detection of dandelion (Taraxacum officinale Web.), ground ivy (Glechoma hederacea L.), and spotted spurge (Euphorbia maculata L.) growing in perennial ryegrass. When the networks were trained using a dataset containing a total of 15,486 negative (images contained perennial ryegrass with no target weeds) and 17,600 positive images (images contained target weeds), VGGNet achieved high F1 scores (≥0.9278), with high recall values (≥0.9952) for detection of E. maculata, G. hederacea, and T. officinale growing in perennial ryegrass. The F1 scores of AlexNet ranged from 0.8437 to 0.9418 and were generally lower than VGGNet at detecting E. maculata, G. hederacea, and T. officinale. GoogleNet is not an effective DCNN at detecting these weed species mainly due to the low precision values. DetectNet is an effective DCNN and achieved high F1 scores (≥0.9843) in the testing datasets for detection of T. officinale growing in perennial ryegrass. Moreover, VGGNet had the highest Matthews correlation coefficient (MCC) values, while GoogleNet had the lowest MCC values. Overall, the approach of training DCNN, particularly VGGNet and DetectNet, presents a clear path toward developing a machine vision-based decision system in smart sprayers for precision weed control in perennial ryegrass.

Detection of broadleaf weeds growing in turfgrass with convolutional neural networks.

BACKGROUND: Weed infestations reduce turfgrass aesthetics and uniformity. Postemergence (POST) herbicides are applied uniformly on turfgrass, hence areas without weeds are also sprayed. Deep learning, particularly the architecture of convolutional neural network (CNN), is a state-of-art approach to recognition of images and objects. In this paper, we report deep learning CNN (DL-CNN) models that are remarkably accurate at detection of broadleaf weeds in turfgrasses. RESULTS: VGGNet was the best model for detection of various broadleaf weeds growing in dormant bermudagrass [Cynodon dactylon (L.)] and DetectNet was the best model for detection of cutleaf evening-primrose (Oenothera laciniata Hill) in bahiagrass (Paspalum notatum Flugge) when the learning rate policy was exponential decay. These models achieved high F1 scores (>0.99) and overall accuracy (>0.99), with recall values of 1.00 in the testing datasets. CONCLUSION: The results of the present research demonstrate the potential for detection of broadleaf weed using DL-CNN models for detection of broadleaf weeds in turfgrass systems. Further research is required to evaluate weed control in field conditions using these models for in situ video input in conjunction with a smart sprayer. © 2019 Society of Chemical Industry.

Strong indirect herbicide effects on mycorrhizal associations through plant community shifts and secondary invasions.

Million of acres of U.S. wildlands are sprayed with herbicides to control invasive species, but relatively little is known about non-target effects of herbicide use. We combined greenhouse, field, and laboratory experiments involving the invasive forb spotted knapweed (Centaurea stoebe) and native bunchgrasses to assess direct and indirect effects of the forb-specific herbicide picloram on arbuscular mycorrhizal fungi (AMF), which are beneficial soil fungi that colonize most plants. Picloram had no effect on bunchgrass viability and their associated AMF in the greenhouse, but killed spotted knapweed and reduced AMF colonization of a subsequent host grown. Results were similar in the field where AMF abundance in bunchgrass-dominated plots was unaffected by herbicides one year after spraying based on 16:1ω5 phospholipid fatty acid (PLFA) and neutral lipid fatty acid (NLFA) concentrations. In spotted-knapweed-dominated plots, however, picloram application shifted dominance from spotted knapweed, a good AMF host, to bulbous bluegrass (Poa bulbosa), a poor AMF host. This coincided with a 63% reduction in soil 16:1ω5 NLFA concentrations but no reduction of 16:1ω5 PLFA. Because 16:1ω5 NLFA quantifies AMF storage lipids and 16:1ω5 PLFA occurs in AMF membrane lipids, we speculate that the herbicide-mediated reduction in host quality reduced fungal carbon storage, but not necessarily fungal abundance after one year in the field. Overall, in greenhouse and field experiments, AMF were only affected when picloram altered host quantity and quality. This apparent lack of direct effect was supported by our in-vitro trial where picloram applied to AMF mycelia did not reduce fungal biomass and viability. We show that the herbicide picloram can have profound, indirect effects on AMF within one year. Depending on herbicide-mediated shifts in host quality, rapid interventions may be necessary post herbicide applications to prevent loss of AMF abundance. Future research should assess consequences of these potential shifts for the restoration of native plants that differ in mycorrhizal dependency.

Estimating economic thresholds for site-specific weed control using manual weed counts and sensor technology: an example based on three winter wheat trials.

BACKGROUND: Precision experimental design uses the natural heterogeneity of agricultural fields and combines sensor technology with linear mixed models to estimate the effect of weeds, soil properties and herbicide on yield. These estimates can be used to derive economic thresholds. Three field trials are presented using the precision experimental design in winter wheat. Weed densities were determined by manual sampling and bi-spectral cameras, yield and soil properties were mapped. RESULTS: Galium aparine, other broad-leaved weeds and Alopecurus myosuroides reduced yield by 17.5, 1.2 and 12.4 kg ha(-1) plant(-1) m(2) in one trial. The determined thresholds for site-specific weed control with independently applied herbicides were 4, 48 and 12 plants m(-2), respectively. Spring drought reduced yield effects of weeds considerably in one trial, since water became yield limiting. A negative herbicide effect on the crop was negligible, except in one trial, in which the herbicide mixture tended to reduce yield by 0.6 t ha(-1). Bi-spectral cameras for weed counting were of limited use and still need improvement. Nevertheless, large weed patches were correctly identified. CONCLUSION: The current paper presents a new approach to conducting field trials and deriving decision rules for weed control in farmers' fields.