An approach for de-identification of point locations of livestock premises for further use in disease spread modeling

Big Data Privacy in Smart Farming: A Review

Smart farming aims to improve farming using modern technologies and smart devices. Smart devices help farmers to collect and analyze data regarding different aspects of their business. These data are utilized by various stakeholders, including farmers, technology providers, supply chain investigators, and agricultural service providers. These data sources can be considered big data due to their volume, velocity, and variety. The wide use of data collection and communication technologies has increased concerns about the privacy of farmers and their data. Although some previous studies have reviewed the security aspects of smart farming, the privacy challenges and solutions are not sufficiently explored in the literature. In this paper, we present a holistic review of big data privacy in smart farming. The paper utilizes a data lifecycle schema and describes privacy concerns and requirements in smart farming in each of the phases of this data lifecycle. Moreover, it provides a comprehensive review of the existing solutions and the state-of-the-art technologies that can enhance data privacy in smart farming.

Modelling the domestic poultry population in the United States: A novel approach leveraging remote sensing and synthetic data methods

Comprehensive and spatially accurate poultry population demographic data do not currently exist in the United States; however, these data are critically needed to adequately prepare for, and efficiently respond to and manage disease outbreaks. In response to absence of these data, this study developed a national-level poultry population dataset by using a novel combination of remote sensing and probabilistic modelling methodologies. The Farm Location and Agricultural Production Simulator (FLAPS) (Burdett et al., 2015) was used to provide baseline national-scale data depicting the simulated locations and populations of individual poultry operations. Remote sensing methods (identification using aerial imagery) were used to identify actual locations of buildings having the characteristic size and shape of commercial poultry barns. This approach was applied to 594 U.S. counties with > 100,000 birds in 34 states based on the 2012 U.S. Department of Agriculture (USDA), National Agricultural Statistics Service (NASS), Census of Agriculture (CoA). The two methods were integrated in a hybrid approach to develop an automated machine learning process to locate commercial poultry operations and predict the number and type of poultry for each operation across the coterminous United States. Validation illustrated that the hybrid model had higher locational accuracy and more realistic distribution and density patterns when compared to purely simulated data. The resulting national poultry population dataset has significant potential for application in animal disease spread modelling, surveillance, emergency planning and response, economics, and other fields, providing a versatile asset for further agricultural research.

Veterinary informatics: forging the future between veterinary medicine, human medicine, and One Health initiatives-a joint paper by the Association for Veterinary Informatics (AVI) and the CTSA One Health Alliance (COHA)

Objectives

This manuscript reviews the current state of veterinary medical electronic health records and the ability to aggregate and analyze large datasets from multiple organizations and clinics. We also review analytical techniques as well as research efforts into veterinary informatics with a focus on applications relevant to human and animal medicine. Our goal is to provide references and context for these resources so that researchers can identify resources of interest and translational opportunities to advance the field.

Methods and Results

This review covers various methods of veterinary informatics including natural language processing and machine learning techniques in brief and various ongoing and future projects. After detailing techniques and sources of data, we describe some of the challenges and opportunities within veterinary informatics as well as providing reviews of common One Health techniques and specific applications that affect both humans and animals.

Discussion

Current limitations in the field of veterinary informatics include limited sources of training data for developing machine learning and artificial intelligence algorithms, siloed data between academic institutions, corporate institutions, and many small private practices, and inconsistent data formats that make many integration problems difficult. Despite those limitations, there have been significant advancements in the field in the last few years and continued development of a few, key, large data resources that are available for interested clinicians and researchers. These real-world use cases and applications show current and significant future potential as veterinary informatics grows in importance. Veterinary informatics can forge new possibilities within veterinary medicine and between veterinary medicine, human medicine, and One Health initiatives.

Realistic assumptions about spatial locations and clustering of premises matter for models of foot-and-mouth disease spread in the United States

Spatially explicit livestock disease models require demographic data for individual farms or premises. In the U.S., demographic data are only available aggregated at county or coarser scales, so disease models must rely on assumptions about how individual premises are distributed within counties. Here, we addressed the importance of realistic assumptions for this purpose. We compared modeling of foot and mouth disease (FMD) outbreaks using simple randomization of locations to premises configurations predicted by the Farm Location and Agricultural Production Simulator (FLAPS), which infers location based on features such as topography, land-cover, climate, and roads. We focused on three premises-level Susceptible-Exposed-Infectious-Removed models available from the literature, all using the same kernel approach but with different parameterizations and functional forms. By computing the basic reproductive number of the infection (R₀) for both FLAPS and randomized configurations, we investigated how spatial locations and clustering of premises affects outbreak predictions. Further, we performed stochastic simulations to evaluate if identified differences were consistent for later stages of an outbreak. Using Ripley’s K to quantify clustering, we found that FLAPS configurations were substantially more clustered at the scales relevant for the implemented models, leading to a higher frequency of nearby premises compared to randomized configurations. As a result, R₀ was typically higher in FLAPS configurations, and the simulation study corroborated the pattern for later stages of outbreaks. Further, both R₀ and simulations exhibited substantial spatial heterogeneity in terms of differences between configurations. Thus, using realistic assumptions when de-aggregating locations based on available data can have a pronounced effect on epidemiological predictions, affecting if, where, and to what extent FMD may invade the population. We conclude that methods such as FLAPS should be preferred over randomization approaches.

When modeling the spread of infectious livestock diseases such as foot-and-mouth disease (FMD), the distance between premises is an important aspect. In the U.S., locations of premises are not available, forcing modelers to make assumptions about their coordinates. Such assumptions can be more or less crude and will impact the conclusions drawn from the model. To investigate the impact of such assumptions, we modeled outbreaks of FMD within the cattle population of the U.S. under two assumptions about premises locations. Their position was either randomly distributed within counties or informed by a state-of-the-art method developed specifically to simulate realistic locations of agricultural operations. We found that the higher degree of spatial clustering of premises associated with more realistic assumptions about locations leads to a substantially higher risk of outbreaks. Our results also show that the amount with which the risk is under-estimated by randomizing locations is unevenly distributed across the landscape. Together, these findings show a clear support for using informed methods to determine the spatial locations of premises and highlight the importance of spatial clustering when modeling FMD-like diseases.

Simulating the Distribution of Individual Livestock Farms and Their Populations in the United States: An Example Using Domestic Swine (Sus scrofa domesticus) Farms

Livestock distribution in the United States (U.S.) can only be mapped at a county-level or worse resolution. We developed a spatial microsimulation model called the Farm Location and Agricultural Production Simulator (FLAPS) that simulated the distribution and populations of individual livestock farms throughout the conterminous U.S. Using domestic pigs (Sus scrofa domesticus) as an example species, we customized iterative proportional-fitting algorithms for the hierarchical structure of the U.S. Census of Agriculture and imputed unpublished state- or county-level livestock population totals that were redacted to ensure confidentiality. We used a weighted sampling design to collect data on the presence and absence of farms and used them to develop a national-scale distribution model that predicted the distribution of individual farms at a 100 m resolution. We implemented microsimulation algorithms that simulated the populations and locations of individual farms using output from our imputed Census of Agriculture dataset and distribution model. Approximately 19% of county-level pig population totals were unpublished in the 2012 Census of Agriculture and needed to be imputed. Using aerial photography, we confirmed the presence or absence of livestock farms at 10,238 locations and found livestock farms were correlated with open areas, cropland, and roads, and also areas with cooler temperatures and gentler topography. The distribution of swine farms was highly variable, but cross-validation of our distribution model produced an area under the receiver-operating characteristics curve value of 0.78, which indicated good predictive performance. Verification analyses showed FLAPS accurately imputed and simulated Census of Agriculture data based on absolute percent difference values of < 0.01% at the state-to-national scale, 3.26% for the county-to-state scale, and 0.03% for the individual farm-to-county scale. Our output data have many applications for risk management of agricultural systems including epidemiological studies, food safety, biosecurity issues, emergency-response planning, and conflicts between livestock and other natural resources.