We present a novel approach that uses multiple linear regression to combine the CPU temperature from nearby SBCs and remote weather stations, to estimate the temperature at outdoor locations that do not have temperature sensors. We use sensor data to train and test multiple regres-sion models. We investigate the efficacy of using different smoothing techniques and we account for the computational load of SBCs at the time of measurement and data collection. We find that our approach enables a prediction error that is less than 1.5 degrees Fahrenheit, while past work results in errors of 1–14 degrees Fahrenheit for similar datasets. We integrate sensor synthesis into Hypatia and use it to facilitate automatic and scalable model selection, as well as visualization for different data sets and recommendations. Finally, we developed a new approach to distributed scheduling for analytics applications in IoT settings: sensor-edge-cloud deployments. Our scheduler takes advantage of remote resources when available, while fully utilizing local edge systems, as it optimizes for time to completion for applications and workloads. The scheduler uses remote resources only if doing so reduces the latency of providing actionable insights locally. The scheduler uses histories of both computation and communication time the applications, which it uses to construct a job placement schedule that minimizes application response latency . Hypatia then uses this schedule to automatically deploy workloads across edge systems and cloud computing systems. We empirically evaluate Hypatia using both clustering and regression services and show that it is able to achieve better end-to-end performance than using the edge or cloud alone. The result is the first end-to-end system that fully utilizes edge computing resources as it serves the needs of precision agriculture.
It does so by accounting for resource constraints at the edge, the lack of or intermittent connectivity to the public cloud,pots with drainage holes and the expense of transmitting the data to/from remote cloud systems. Moreover, the system is open-source and integrates a wide range of analytics, scoring methods, and visualization tools, which can be easily extended with new and emerging techniques. By doing so, we enable others to easily build upon, extend, reproduce, and compare it to our work in the future. Moving forward, we hope to encourage adoption of Hypatia by growers, farm consultants, and data analysts interested in taking advantage of the locality of edge systems to provide low latency analytics. Given the existing infrastructure, we plan to add new sensors, develop more synthesized, sensors, and to integrate additional analytics and scoring methods. Specifically, we plan to extend Hypatia with support for image classification and to use analytics accelerators at both the edge and in the cloud when available. Other future work includes investigating new data sources and machine learning algorithms that inform a more refined scheduling algorithm that can take advantage of even more granular resources. In addition, Hypatia error analysis can benefit from additional abstractions that account error propagation, which has the potential for making the results and recommendation more informative and trustworthy.Fragmentation of natural habitats during conversion of wild lands to agriculture and the subsequent increase in agrochemicals has resulted in a loss of biodiversity and a deterioration of ecosystem function, including natural pest control. Non-crop habitats harbor natural enemies to crop pests . Such habitats also harbor beneficial songbirds that consume insect pests , and provide perching sites for raptors that deter avian and rodent pests . Balancing the role of agricultural lands in providing habitat for biodiversity while simultaneously avoiding bird damage and reducing food safety risks is the primary goal behind the concept of co-management, which is recommended by the Food Safety Modernization Act .
Wild and domestic animals destroy crops by eating and trampling them, and can pose food safety risks due to the deposition of potentially contaminated feces on or near the crops . Birds are one of the most challenging animals to keep out of agricultural fields, and they may harbor food borne pathogens. For example, European starlings are a source of Salmonella enterica at concentrated animal feeding operations , posing a greater risk of pathogen transfer than other variables like cattle density, facility management operations, and environmental variables . They also may be a significant source of other Salmonella spp, Escherichia coli O157, and other shiga toxin–producing E. coli . During a study at a CAFO in southern Arizona, 103 birds were tested for food borne pathogens. Two tested positive for Salmonella, and five tested positive for non-O157 Shiga toxinproducing E. coli. Other studies have shown similar results as documented in a review by Langholz and Jay-Russell where they listed 23 studies on food borne pathogen prevalence in birds, including positive results for ducks, gulls, starlings, and pigeons. A more recent review listed food borne pathogens specifically transmitted by wild birds . All reviews discuss a 2008 outbreak of Campylobacter related to pea consumption because it was one of few outbreaks directly linking the pathogen to a wildlife source, in this case, sandhill cranes . This highlights the potential risks to food safety associated with migratory birds. Damage and food safety risks from wildlife activities remain significant economic problems despite the use of a variety of methods to control bird and rodent pests . Yield loss and economic impacts vary by crop and region, but can be a substantial burden on growers . Growers of fresh produce try countless methods to deter birds. These deterrents fall into nine general categories . This paper is not intended to be an exhaustive review of bird deterrents, but instead we present an overview of the ones most used in the field, as well as methods that utilize multiple techniques in an effort to develop integrated pest management for nuisance bird control.
The array of visual bird deterrents is expansive, and includes lights that are flashing or rotating, searchlights, mirrors and reflectors, reflective tape, flags, rags, streamers, lasers, dogs, humans, scarecrows, raptor models, corpses, balloons with eye spots, kites, kite hawks, mobile predator models, and water dyes or colorants . All of these methods work to some degree for a short period of time until habituation. Lasers that were used to disperse crows, for example, resulted in an initial dispersion, but crows reoccupied their roosts the same night that the lasers were used, and none of the roosts were abandoned . Kite balloons were shown to be effective in the short term, but birds quickly become habituated, reducing the effectiveness over time . Similarly, balloons with eye spots have been used in an attempt to reduce damage to vineyard grapes in New Zealand, but growers reported no economically significant effect . Generally, balloons, scarecrows, hawk kites, and reflective tape work best with sound cannons or netting, described below . Noise deterrents are generally effective, but much like visual deterrents, birds easily become habituated to them,drainage planter pot decreasing their efficacy over time. They have the added issue that growers who use them are subject to complaints of nuisance noise from neighbors . Propane sound cannons are the most commonly used noise deterrent, but they need to be repositioned weekly and set to go off randomly every 7-20 minutes during daylight hours for the greatest effect. Since sound cannons usually make a hissing noise before sounding off, they give birds a warning to leave the area, and then they return after the explosive noise. Some of the other common noise deterrents include bangers, screamers, squawkers, whistlers, scare cartridges, and noise bombs . Even human presence can be used as a noise deterrent if they rattle cans, crack whips, yell, honk horns, or shoot guns . Human activity can be very effective at keeping nuisance birds out of fields when fields are small enough to drive or walk around, but it can be expensive to maintain a human guard on duty. Instead, some growers use synthetic sounds that offer unambiguous messages that elicit inter specific responses, like distress calls . They prevent habituation by varying the rhythm and number of signals emitted . In a study of alarm calls from crimson rosellas in orchards,researchers found that these birds were effectively deterred in the short- to medium-term . However, distress calls offer another challenge since they may be an invitation to nearby predators indicating that their next meal is ready. Broadcast units are a less expensive, more technologically advanced noise deterrent that reproduce accurate and effective birds calls that significantly reduce damage in vineyards .
Another moderately effective noise deterrent is the sonic net, which overlaps with the frequency range of bird vocalizations, making communication among a flock ineffective. When used at an airfield, researchers demonstrated an 82% reduction in birds in the sonic net area, and it remained effective after four weeks of exposure . Fencing is an effective non-lethal, long-term method used as a standard technique to minimize wildlife intrusion into agricultural lands . While fencing cannot be used to deter birds, netting can be. While noise deterrents used against juvenile starlings in a cherry orchard were shown to be ineffective, research suggests that the netting-in of an orchard would be more effective . However, while netting is the most effective method, it is also has some drawbacks. It is one of the most expensive methods for deterring birds due to the massive areas of crops that need to be covered . It can also be easily damaged, and it can be a hazard to wildlife. Other exclusions that are used with birds are electric fencing, overhead wires, and anti- perching devices, such as spikes, some of which are also considered tactile deterrents and forms of habitat modification described below. The concept of habitat modification to deter nuisance birds includes a wide array of activities, from providing better quality forage or shelter in alternate locations through lure crops or sacrificial crops to simply removing roost structures, food, and shelter, forcing birds to go elsewhere. In many cases, deterring nuisance birds from one field causes them to negatively impact neighboring farms. For that reason, Ainsley and Kosoy propose collective action on the part of neighboring farmers in which communal feeding plots are constructed to protect the fields of all farmers in a single area, thereby evenly distributing crop losses and maintaining stable bird populations in the ecosystem . The USDA’s Wildlife Services attempted this method when they began to cost share eight hectare Wildlife Conservation Sunflower Plots with sunflower growers to lure migrating blackbirds away from commercial sunflower fields. The targeted blackbirds ended up removing 10 times more sunflower seeds from the WCSP than from commercial fields, making this strategy an important part of an integrated pest management plan for commercial sunflower growers . Monk parakeets tend to damage corn and sunflower fields that are closest to man-made structures and adjacent trees, areas with tree patches around the crop fields, and areas with high availability of pasture and weedy and fallow fields . The removal of these landscape features that attract birds, like areas with structures for perching, breeding, and shelter, can cause birds to move out of an area . A recent study indicated that hedgerows harbor higher biodiversity of rodents, but that biodiversity does not spill over into wildlife intrusion into fields . While rodents differ from birds, the concept of wildlife utilizing adjacent habitat without affecting agricultural crops or impacting food safety is similar. Physiological methods of bird control include such things as chemo-sterilants, contraceptives, and immune- contraceptive vaccines . These are rarely, if ever, used by growers in agricultural areas because they require extensive permitting and veterinary oversight, often times making their use unfeasible. Linz, Bucher, et. al. identified four limiting factors hindering the use of contraceptive methods and lethal control of birds in agriculture, including: 1) the high cost of implementation combined with challenges related to maintaining long-term control of birds, 2) determining the population level in an area that would be considered acceptable and therefore serves as a level of success, 3) ensuring that the treatment would be directed only at the birds actually causing crop damage, and 4) managing immigration of non-treated birds. Chemical bird deterrents, such as taste and behavioral repellants, are expensive, difficult to apply, not as effective in the field as they are in the lab, need to be licensed, and some overlap with lethal deterrents.