Recent advances in deep learning, including the development of AlexNet, Residual Network (ResNet), and transfer learning, offer unprecedented classification accuracy in the field of machine vision. A developing application of deep learning is the automated identification and management of aquatic invasive plants. Classification of submersed aquatic vegetation (SAV) presents a unique challenge, namely, the lack of a single source of sensor data that can produce robust, interpretable images across a variable range of depth, turbidity, and lighting conditions. This paper focuses on the development of a multi-sensor (RGB and hydroacoustic) classification system for SAV that is robust to environmental conditions and combines the strengths of each sensing modality. The detection of invasive Hydrilla verticillata (hydrilla) is the primary goal. Over 5000 aerial RGB and hydroacoustic images were generated from two Florida lakes via an unmanned aerial vehicle and boat-mounted sonar unit, and tagged for neural network training and evaluation. Classes included “HYDR”, containing hydrilla; “NONE”, lacking SAV, and “OTHER”, containing SAV other than hydrilla. Using a transfer learning approach, deep neural networks with the ResNet architecture were individually trained on the RGB and hydroacoustic datasets. Multiple data fusion methodologies were evaluated to ensemble the outputs of these neural networks for optimal classification accuracy. A method incorporating logic and a Monte Carlo dropout approach yielded the best overall classification accuracy (84%), with recall and precision of 84.5% and 77.5%, respectively, for the hydrilla class. The training and ensembling approaches were repeated for a DenseNet model with identical training and testing datasets. The overall classification accuracy was similar between the ResNet and DenseNet models when averaged across all approaches (1.9% higher accuracy for the ResNet vs. the DenseNet).

Unoccupied aerial application systems (UAAS) are gaining popularity for weed management to increase applicator safety and to deliver herbicide treatments where treatment sites limit ground-based spray equipment. Several studies have documented UAAS application strategies and procedures for weed control in terrestrial settings, yet literature describing remote spray technology for use in aquatics remains limited. Currently, applicators seek guidance for UAAS deployment for aquatic weed management to overcome site access restrictions, deal with environmental limitations, and improve ground-based applicator safety in hazardous treatment scenarios. In the present case studies, we evaluate a consumer-available UAAS to deliver the herbicide, florpyrauxifen-benzyl, as both foliar and directed in-water spray applications. The first case study showed that the invasive floating-leaved plant, yellow floating heart, was controlled 80% to 99% by 6 wk after treatment (WAT) following UAAS foliar herbicide treatments. The second case study demonstrated that UAAS directed in-water herbicide application reduced variable-leaf watermilfoil visible plant material by 94% at 5 WAT. Likewise, directed in-water applications from UAAS eliminated the need to deploy watercraft, which improved overall operational efficiency. Data from both case studies indicate that UAAS can provide an effective and efficient treatment strategy for floating-leaved and submersed plant control among common herbicide treatment scenarios. Future integration of UAAS in aquatic weed control programs is encouraged, especially among smaller treatment sites (≤4 ha) or where access limits traditional spray operations.

Invasive aquatic plants constantly threaten freshwaters and associated environs globally. Water resource managers frequently seek new control tactics to combat invasive macrophytes, especially when the availability of herbicides registered for submersed plant control is limited. The synthetic auxin herbicide, florpyrauxifen-benzyl, recently registered (2018) for aquatic site applications in the United States, has shown success in controlling several invasive aquatic weeds. Studies were conducted to evaluate responses of native and invasive submersed plants to florpyrauxifen-benzyl under growth chamber conditions to provide insight on the selectivity of varying herbicide concentrations in New Zealand. Florpyrauxifen-benzyl concentrations evaluated ranged from 0.01 to 107.86 µg ai L⁻¹, encompassing the maximum use concentration (48 µg L⁻¹) for submersed plant applications. Dose–response metrics indicated the New Zealand native species watermilfoil [Myriophyllum triphyllum Orchard] was highly sensitive to florpyrauxifen-benzyl following a 21-d static exposure, having a dry weight 50% effective concentration (EC₅₀) value of 1.2 µg L⁻¹. The invasive species oxygen-weed [Lagarosiphon major (Ridley) Moss] and Canadian waterweed (Elodea canadensis Michx.) were less sensitive, with dry weight EC₅₀ values of 35.4 and >107.86 µg L⁻¹, respectively. Brazilian waterweed (Egeria densa Planch.) was most tolerant to the tested concentrations, as EC₅₀ values were not achieved. Overall, results indicate florpyrauxifen-benzyl demonstrates potential for controlling L. major, with further large-scale screening required to confirm control among field site applications. As the native species (M. triphyllum) was most sensitive to florpyrauxifen-benzyl compared with the invasive plant evaluated (I/N ratio indicated >31.3 times more sensitive), any targeted concentration used for invasive plant control for field applications would likely injure the native M. triphyllum plants. Future studies should investigate additional native and invasive species for management guidance and consider how exposure times influence plant response using similar florpyrauxifen-benzyl concentrations tested in the present study.

Invasive emergent and floating macrophytes can have detrimental impacts on aquatic ecosystems. Management of these aquatic weeds frequently relies upon foliar application of aquatic herbicides. However, there is inherent variability of overspray (herbicide loss) for foliar applications into waters within and adjacent to the targeted treatment area. The spray retention (tracer dye captured) of four invasive broadleaf emergent species (water hyacinth, alligatorweed, creeping water primrose, and parrotfeather) and two emergent grass-like weeds (cattail and torpedograss) were evaluated. For all species, spray retention was simulated using foliar applications of rhodamine WT (RWT) dye as a herbicide surrogate under controlled mesocosm conditions. Spray retention of the broadleaf species was first evaluated using a CO₂-pressurized spray chamber overtop dense vegetation growth or no plants (positive control) at a greenhouse (GH) scale. Broadleaf species and grass-like species were then evaluated in larger outdoor mesocosms (OM). These applications were made using a CO₂-pressurized backpack sprayer. Evaluation metrics included species-wise canopy cover and height influence on in-water RWT concentration using image analysis and modeling techniques. Results indicated spray retention was greatest for water hyacinth (GH, 64.7 ± 7.4; OM, 76.1 ± 3.8). Spray retention values were similar among the three sprawling marginal species alligatorweed (GH, 37.5 ± 4.5; OM, 42 ± 5.7), creeping water primrose (GH, 54.9 ± 7.2; OM, 52.7 ± 5.7), and parrotfeather (GH, 48.2 ± 2.3; OM, 47.2 ± 3.5). Canopy cover and height were strongly correlated with spray retention for broadleaf species and less strongly correlated for grass-like species. Although torpedograss and cattail were similar in percent foliar coverage, they differed in percent spray retention (OM, 8.5± 2.3 and 28.9 ±4.1, respectively). The upright leaf architecture of the grass-like species likely influenced the lower spray retention values in comparison to the broadleaf species.

Expanding the current aquatic herbicide portfolio, reducing total spray volumes, or remotely delivering herbicide using novel spray technologies could improve management opportunities targeting invasive aquatic plants, where options are more limited. However, research on giant salvinia (Salvinia molesta Mitchell) response to foliar herbicide applications at carrier volumes ≤140 L ha⁻¹ is incomplete. Likewise, no data exist documenting S. molesta control with unoccupied aerial application systems (UAAS). Following the recent >100-ha incursion of S. molesta in Gapway Swamp, NC, a case study was developed to provide guidance for ongoing management efforts. In total, three field trials evaluated registered aquatic and experimental herbicides using a 140 L ha⁻¹ carrier volume. Select foliar applications from UAAS were also evaluated. Results at 8 wk after treatment (WAT) indicated the experimental protoporphyrinogen oxidase inhibitor, PPO-699-01 (424 g ai ha⁻¹), in combination with endothall dipotassium salt (2,370 g ae ha⁻¹) provided 78% visual control, whereas control when PPO-699-01 (212 g ai ha⁻¹) was applied alone was lower at 35%. Evaluations also showed diquat (3,136 g ai ha⁻¹) alone, glyphosate (4,539 g ae ha⁻¹) alone, and metsulfuron-methyl (42 g ai ha⁻¹) alone achieved 86% to 94% visual plant control at 8 WAT. Sequential foliar applications of diquat, flumioxazin (210 g ai ha⁻¹), and carfentrazone (67 g ai ha⁻¹) at 6 wk following exposure to in-water fluridone treatments were no longer efficacious by 6 WAT due to plant regrowth. Carfentrazone applications made from a backpack sprayer displayed greater control than applications made with UAAS deploying identical carrier volumes at 2 WAT; however, neither application method provided effective control at 8 WAT. Additional field validation is needed to further guide management direction of S. molesta control using low carrier volume foliar applications.