Recipient Organization
UNIV OF WISCONSIN
21 N PARK ST STE 6401
MADISON,WI 53715-1218
Performing Department
Plant Pathology
Non Technical Summary
The capacity to assemble, analyze and use large commercial agricultural data resources to drive behavioral change and integrated pest management or IPM adoption remains limited in WI potatoes. Our ability to collect, store and retrieve vast quantities of data, along with advances in computational power and sophisticated algorithms for the analysis of large data sets, is offering new opportunities for understanding and predicting the behavior of complex natural, and agricultural systems. We propose the development of research methods that can resolve patterns from these large datasets, and aid in our ability to limit risk of insect and disease pressure for producers. We will use an existing potato scouting dataset (10+ years of field data) and relate pest and disease occurrence, timing of initial incidence, and severity levels with crop data layers to conduct retrospective analyses to inform current and future management practices. We will research linkages to specific insect and disease "risks" and IPM needs based on temporal, regional, and location-based pest levels related to potato intensities. These proposed "IPM risk aversion" tools will enable growers to analyze future IPM risk based on previous locations, pest incidence, and phenological development; ultimately allowing growers to employ tactics more precisely and when the greatest benefits can be realized. Research will focus on WI potatoes but our results will be applicable to broad geographic areas and cropping systems. This work will utilize valuable multi-site and multi-year grower data and advance the development of informatic approaches to better inform practical IPM in agricultural systems.
Animal Health Component
30%
Research Effort Categories
Basic
50%
Applied
30%
Developmental
20%
Goals / Objectives
With the valuable dataset of past scouting pest levels for insects and diseases, we will be able to determine "IPM risk aversion" tools and strategies for growers to make better use of these historical scouting data and more precisely time and direct application of IPM approaches. The research that we propose, will result in a significant shift in IPM thought based on our ability to use new technology to predict pest occurrence and manage pests with greater precision and less risk. We propose to work initially with established Healthy Grown growers to demonstrate the benefits of this research, encourage rapid adoption of new practices and work with these growers to use evidence-based learning to achieve adoption throughout the industry.Objective 1. Using an informatics approach to define temporal and spatial pest onset.Objective 2. Observing temporal changes and phenology in the timing of arrival and infestation/infection onset for select, critically important potato pests and diseases including, CPB, early blight, and late blight.Objective 3. Characterizing insect and disease abundance towards the development of risk maps combining eco-physiological knowledge, weather, and habitat variability.Objective 4. Steward adoption of advanced IPM techniques produced through risk aversion strategies resulting from our work.
Project Methods
We initially propose to pursue the question of whether there is any indication, based on these scouting records, that CPB is changing its emergence phenology in response to the use of at-plant systemic neonicotinoid insecticides. In-plant concentrations of neonicotinoids are known to drop as plants grow larger and deplete the available soil- or seed-applied insecticide. We hypothesized that beetles could be adapting to this pressure by either delaying emergence, thus encountering larger plants with lower in-plant concentrations of systemic insecticides observable as a later peak first-generation emergence, or extending their emergence over a longer period of time observable as an extended period of first-generation emergence with a lower absolute peak magnitude. We will use a generalized linear mixed modeling (GLMM) approach based on Poisson regression (log-link) with random intercepts was used to examine the relative importance of year, week, farm and field on the abundance of key insect pests (e.g. CPB, PLH), diseases (early blight) and weed populations. A multilevel model will be designed for counts by field and the total counts will be offset by the number of transects walked (or sampling effort) in each field. A fixed effects term will represent the model intercept and will be interpreted as the statewide seasonal average insect, disease or weed abundance in selected fields. Random effects (or intercepts) for year, day, farm, and field, respectively, will also be include in the model, and these will be represented by the variance components associated with the temporal and spatial "blocks" of the proposed model. The variance components of the responses will be quantified and variance components assessed in terms of variances (standard deviations) on the latent, or log scale of the model.Mixed-effects models, in general, are used because they associate random effects to observations sharing the same level of a classification factor. Thus, mixed-effects models are useful because they can accurately represent the covariance structure that exists among samples when repeated measurements are taken at the same location or time; essentially we are assuming all observations from a given source (or subject) are correlated. Often, when research emphasis is placed on estimating fixed regression coefficients, random effects are included in a model to account for the covariance among sample groupings prior to estimating the regression coefficients. However, in our case, the variance-covariance structure itself is of interest and all factors in our analyses will be considered random since the primary goal was to examine the nature of different spatial and temporal levels from which the data are presumed to have come. There are many ways that a GLMM can be defined to examine the interactions among different combinations of covariate groupings and to quantify their associated variances. In these specific instances, we are less interested in the contributions of specific years as much as we are interested in the variation due to year. Additionally, we are more interested in the day-to-day variability of count or abundance than the variability of count or abundance on a specific day. The flexibility of the GLMM allows us to specify random variables for year and day and model that variability. Our biological interpretation of this approach is analogous to that of regression analyses where both intercept and slope are allowed to vary among treatments. Estimates of the variability of each grouping (i.e. year, day, and year-day) are obtained and the relative amount of variability described by each level of grouping is reflective of the importance of each factor. Important terms are those that describe larger amounts of variability. Thus, the inclusion of different "interaction terms" as random effects in the GLMM leads to numerous ways to partition the variances associated with count, providing insight about the underlying biology and spatial or temporal scales at which processes important for pest population dynamics are occurring.In our current data set, the temporal and spatial grouping of covariates occurs at different scales. For example, year and day represent a different spatial grain size. Studying the patterns of insect abundance, for example, and associated variation at different temporal scales can provide insight about the scale of the underlying ecological processes driving prevalence. Large variation of L. decemlineata or E. fabae (or disease pressure) abundance among years, relative to other sources, might suggest that numbers are influenced by climatic or biological factors. In contrast, large variation within years might be better explained by processes such as differences in grower management or synoptic weather events. Similarly, insights can be gained by examining variation occurring among and within geographic location. For example, large variation of disease prevalence among geographic locations might imply that the local habitat surrounding a crop field is important. Alternatively, small variations in incidence among geographic locations might suggest larger scale processes, occurring across all locations, drive insect abundance (i.e. mean annual temperatures). Implementation of this objective will occur in the first year of the project as the current dataset is immediately accessible.