NETWORK ANALYSIS FOR FORECASTING THE SPATIAL PROGRESS OF SOYBEAN RUST EPIDEMICS AND OPTIMIZING SENTINEL PLOT STRATEGIES

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: SPECIAL GRANT
• Reporting Frequency: Annual
• Accession No.: 0221691
• Grant No.: 2010-34103-20964
• Proposal No.: 2010-02462
• Project Start Date: Jun 1, 2010
• Project End Date: May 31, 2013
• Grant Year: 2010
• Program Code: [QQ.NC]- Integrated Pest Management - North Central Region

Recipient Organization
KANSAS STATE UNIV
(N/A)
MANHATTAN,KS 66506

Performing Department
Plant Pathology

Non Technical Summary
Soybean rust is an important threat to US soybean production. The pathogen overwinters in the US south. For the North Central states, a primary question for management is whether the disease will reach soybean fields early enough in the season that use of a fungicide is economically supported. The sentinel plot network established to support soybean rust management has saved farmers huge amounts of money by supplying information about where soybean rust was present in time for farmers to anticipate whether fungicides were needed or not. As funding for the sentinel plot network has been reduced, it is important to optimize the usefulness of disease monitoring. We have developed a dynamic network model for soybean rust in the US using the sentinel plot data from previous years. This model uses information from the sentinel plots in a very direct way, such that we can estimate how useful the information from any given county has been for predicting movement of soybean rust to the north. This allows us to estimate how important particular sampling regions will be as an epidemic is ongoing. We will provide this information to ipmPIPE and other collaborators so that sampling to inform soybean growers will be as efficient as possible. We will also provide our estimates of where soybean rust will occur directly to farmers through a project website linked with ipmPIPE.

Animal Health Component

50%

Research Effort Categories

Basic

50%

Applied

50%

Developmental

(N/A)

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
216	1820	1060	50%
216	1820	1170	50%

Knowledge Area
216 - Integrated Pest Management Systems;

Subject Of Investigation
1820 - Soybean;

Field Of Science
1060 - Biology (whole systems); 1170 - Epidemiology;

Keywords

efficient sampling, epidemiology, integrated pest management, network

models, sampling, soybean rust

Goals / Objectives
Goals: Apply network models to determine which locations of sentinel or mobile sites lead to the best predictions of soybean rust risk for US north central counties under limited resources. Objectives: Characterize US soybean rust epidemics using network models for previous sentinel site locations, and provide epidemic predictions to stakeholders through a project website linked with ipmPIPE. Select subsets of sentinel sites and mobile sites that optimize epidemic prediction given financial and practical constraints identified by stakeholders. Provide recommendations for optimal short-term and long-term sampling sites and times throughout the project, incorporating new soybean rust data in models as it becomes available. Expected Outcomes: The recommendations from this project will support more efficient sampling for soybean rust, reducing the risk of unneeded fungicide applications.

Project Methods
(1) Stakeholder involvement in problem identification, planning, implementation and evaluation. Soybean rust has been identified as a key threat to soybean production. Its importance has motivated the USDA and collaborating groups to develop one of the most impressive sampling and monitoring program every implemented to address the epidemiology of a plant pathogen. In times of economic challenge such as the world currently faces, the important problem now is how to optimize such systems under financial constraints. (2) Description of the proposed project activities in the sequence in which it is planned to carry them out. Model development. We have developed our general soybean rust epidemiological model using data from 2005-2008 epidemics. We will continue to improve our model throughout the course of the project, as more soybean rust data becomes available from the sentinel plot network and other sources, and we receive feedback from model users. Providing recommendations for sentinel plot and mobile sampling locations. We will provide recommendations for optimal sampling based on our current epidemiological model at the outset of the NC RIPM project in 2010. We will then update those recommendations on at least a monthly basis, and provide predictions about disease movement. Collaborative meetings with stakeholders. On at least a quarterly basis we will have phone conferences with ipmPIPE representatives and stakeholder groups to explain the development of the model and our most recent predictions and recommendations. We will also seek their recommendations to improve the utility of our modeling efforts for their needs. Year 1. At the beginning, we will make recommendations for sampling and epidemiological predictions based on the preliminary model we had developed before the NC RIPM project. During the course of the year we will improve on the model and recommendations as new data become available. Year 2. We will continue to improve the model and provide recommendations and predictions, as well as publishing our soybean rust project in peer-reviewed journals. Year 3. We will continue to improve the model and provide recommendations and predictions, and we will publish general recommendations for network modeling and sampling in plant disease epidemiology. (3) Techniques to be employed, including their feasibility and rationale for their use in this project. Network models offer an exciting new technology for epidemiology. We have already implemented a preliminary version of the network model for predicting soybean rust epidemic progress in US counties, demonstrating its feasibility (Figures 1 and 2 at end of Project Narrative). The current model, based on soybean density per county, distance between counties, and wind patterns between counties, provides a quite low prediction error, as discussed in more detail below. The rationale for use of such network models includes their natural adaptation for analysis of the effects of the presence or absence as well as timing of data collection at each node (county in our case) on model prediction.

Progress 06/01/10 to 05/31/13

Outputs
Target Audience: The target audience includes research and extension personnel with the responsibility to provide farmers with information and recommendations for disease management, for soybean rust in particular, and for aerially dispersed pathogens and arthropod pests more broadly. We are adapting the strategies from soybean rust for wheat rusts. Changes/Problems: During the course of this project, soybean rust epidemics were not particularly important and interest has moved away from this pathosystem. We anticipated this potential issue, and made sure to develop general results, particularly such that they can be applied to other rusts such as wheat rusts. In ongoing work we are addressing the same general topics for wheat rusts. What opportunities for training and professional development has the project provided? The project also provided training to a graduate student and postdoctoral scientist, as well as engaging faculty in Electrical and Computer Engineering to think about problems in IPM. Garrett has also incorporated examples from this project in teaching. How have the results been disseminated to communities of interest? We have published our results in peer-reviewed journals, presented them at meetings, and discussed them with colleagues What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Major activities completed Our major activity was to develop and evaluate a network model for soybean rust invasions in the US. We have published one paper about this model and its implications for sampling strategies (Sutrave et al., 2012) and have another paper in internal review (Sanatkar et al.) that addresses how the soybean rust epidemic has evolved over time. In the long run, a major contribution of this project will be its role in the foundations of a long-term project in our group, the development of impact network analyses in support of IPM. Impact network analyses are designed to evaluate the effects of IPM strategies, taking into account the structure of linked networks for communication among people and networks for pests in the lands they manage. Specific objectives met We have developed the models for improved sampling strategies for soybean rust, and made them available to researchers with responsibility for developing sampling strategies, through published papers, discussions, and presentations at meetings. We have also developed the more general framework for these analyses for application to wheat rusts and other emerging diseases. Significant results One significant result is demonstrating the benefits of a strategy, as described in Sutrave et al. (2012), for identifying the most important sampling locations in an epidemic network when resources for sampling are limited. We demonstrated the utility of this strategy for soybean rust, but the general principles can be applied in most epidemic invasions. Key outcomes The key outcome of the project is the identification of strategies for sampling and mitigation that are general for IPM, based on the structure of the invasion network.

Publications

• Type: Journal Articles Status: Accepted Year Published: 2012 Citation: Sutrave, S., C. Scoglio, S. A. Isard, J. M. S. Hutchinson, and K. A. Garrett. 2012. Identifying highly connected counties compensates for resource limitations when evaluating national spread of an invasive pathogen. PLoS ONE 7:e37793. Garrett, K. A. 2012. Information networks for plant disease: Commonalities in human management networks and within-host signaling networks. [Invited] European Journal of Plant Pathology 133:75-88.
• Type: Journal Articles Status: Accepted Year Published: 2013 Citation: Skelsey, P., K. A. With, and K. A. Garrett. 2013. Why dispersal should be maximized at intermediate scales of heterogeneity. Theoretical Ecology 6:203-211. Cox, C. M., W. W. Bockus, R. D. Holt, L. Fang, and K. A. Garrett. 2013. The spatial connectedness of plant species: Potential links for apparent competition via plant diseases. Plant Pathology, in press.
• Type: Book Chapters Status: Awaiting Publication Year Published: 2013 Citation: Garrett, K. A., P. D. Esker, and A. H. Sparks. 2013. An introduction to key distributions and models for epidemiology using R. Exercises in Plant Disease Epidemiology, 2nd Edition. K. Stevenson and M. Jeger, eds. APS Press, Minneapolis, MN. In press.

Progress 06/01/11 to 05/31/12

Outputs
OUTPUTS: We continue collaborating with ipmPIPE aerobiologist Scott Isard and communicating the results of strategy development with ipmPIPE. Our project provides a strategy for making sampling of soybean rust, or other pests and diseases, more efficient by identifying the nodes in the epidemic network that are most important for movement of the disease. The extensive sentinel plot network that was deployed to address the new US soybean rust epidemic was an important strategic investment in understanding risks from soybean rust. This project built on that investment to develop sampling strategies for years when fewer resources are available for sampling. PARTICIPANTS: Karen Garrett, Professor, KSU Scott Isard, Professor, Penn State Bala Natarajan, Associate Professor, KSU Mohammad Sanatkar, graduate student, KSU Caterina Scoglio, Associate Professor, KSU Pete Skelsey, postdoctoral scientist, KSU Sweta Sutrave, graduate student, KSU The project provided training for two graduate students and a postdoctoral scientist, and developed collaboration across Land-Grant universities. TARGET AUDIENCES: Our project is aimed primarily at the teams of researchers and extension agents who study soybean rust, and ultimately wheat rust, epidemics and develop strategies for sampling and evaluating epidemics to provide recommendations about management to farmers. PROJECT MODIFICATIONS: Our project generally proceeded as planned, but we interacted with people sampling in the field less than anticipated because of the slow soybean rust epidemics. This also made us emphasize further the importance of determining how methods for soybean rust could best be adapted to other systems such as wheat rusts.

Impacts
Our key product from this project is the analysis reported in Sutrave et al. (2012), available as an open access article at http://www.plosone.org/article/info%3Adoi%2F10.1371%2Fjournal.pone.00 37793. This analysis identifies sampling strategies for the US soybean rust epidemic that make use of the extensive sampling of the sentinel plot network so that smaller investments in the future can support sampling that is as efficient as possible. These strategies can also be adapted for other rust epidemics, such as wheat rusts, and Ug99 if it is introduced to the US. Following on Sutrave et al. (2012), we have another analysis about ready for submission in the next month that evaluates the importance of specific location histories in the soybean rust epidemic, such that these locations may need special consideration in sampling. These evaluations take into account how environmental conditions change, and we have supported this analysis with a direct assessment of how climate variability can influence epidemic processes (Garrett, Dobson, et al, in press). We also developed a teaching module that introduces the statistical distributions needed in such modeling projects (Garrett, Esker, and Sparks, 2012).

Publications

• Garrett, K. A. 2012. Information networks for plant disease: Commonalities in human management networks and within-host signaling networks. [Invited] European Journal of Plant Pathology 133:75-88.
• Garrett, K. A., A. Dobson, J. Kroschel, B. Natarajan, S. Orlandini, H. E. Z. Tonnang, and C. Valdivia. 2012. The effects of climate variability and the color of weather time series on agricultural diseases and pests, and decision making for their management. Agricultural and Forest Meteorology, in press.
• Garrett, K. A., P. D. Esker, and A. H. Sparks. 2012. An introduction to key distributions and models for epidemiology using R. Exercises in Plant Disease Epidemiology, 2nd Edition. K. Stevenson and M. Jeger, eds. APS Press, Minneapolis, MN. In press.
• Sutrave, S., C. Scoglio, S. A. Isard, J. M. S. Hutchinson, and K. A. Garrett. 2012. Identifying highly connected counties compensates for resource limitations when evaluating national spread of an invasive pathogen. PLoS ONE 7:e37793.

Progress 06/01/10 to 05/31/11

Outputs
OUTPUTS: This has been an unusually slow year for soybean rust development, so we have not worked as extensively with the soybean rust community as we would have. But the disease has started to move up through Florida so we are preparing to provide recommendations on sampling. In the meantime, we have made improvements to the models for predicting disease movement, including evaluations of the effects of scale on epidemic models. We have also initiated discussions with scientists working with wheat stem rust global epidemics, to develop the more generalized version of the model for application to this and other systems. We are planning to do preliminary work with their data sets, as well, to determine the best approaches for adapting this model system. PARTICIPANTS: This project funds a graduate student, Mohammad Sanatkar, and part of the time of a postdoctoral scientist, Peter Skelsey. Both these team members have expertise in modeling epidemic processes. Faculty advisors at KSU are Karen Garrett, Caterina Scoglio, and Bala Natarajan. We also collaborate with Scott Isard (Penn State) and Shawn Hutchinson (KSU), who provide expertise in soybean rust aerobiology and GIS. For work with wheat stem rust, we have begun collaboration with Dave Hodson (FAO). As the project progresses, we will interact more broadly with the US soybean rust community and wheat stem rust community. TARGET AUDIENCES: Our direct target audiences will be researchers and extension personnel with responsibilities for tracking these diseases (and others similar to them). Our indirect target audiences will be those who benefit from improved estimates of the movement of the epidemics. PROJECT MODIFICATIONS: The only modifications to the project have been (1) delayed project initiation because of delayed funding availability to the project, and (2) less interaction with field scientists to date in 2011 because of an unusually slow soybean rust epidemic.

Impacts
At this intermediate point in the project, our main outcome has been an improved version of the model that we will apply to provide sampling recommendations as the soybean rust epidemic progresses, and shortly to global wheat stem rust data. We have developed new estimation methods for the models that should improve model performance in the future.

Publications

• In internal review. 2011. Sutrave, S., C. Scoglio, S. Isard, S. Hutchinson, and K. A. Garrett. Identifying highly connected counties compensates for resource limitations in the study of an invasive pathogen. For submission to PLoS ONE in July.
• In review. 2011 . Garrett, K. A., A. Dobson, J. Kroschel, B. Natarajan, S. Orlandini, S. Randolph, H. E. Z. Tonnang, and C. Valdivia. Multi-scale modeling of pests and disease under increased climate variability. Agricultural and Forest Meteorology.
• In review. 2011. Garrett, K. A. and P. D. Esker. Epidemiology in the R programming environment. Exercises in Plant Disease Epidemiology.
• In review. 2011. Skelsey, P., K. A. With, and K. A. Garrett. The intermediate scale hypothesis. American Naturalist.
• Garrett, K. A., G. A. Forbes, S. Savary, P. Skelsey, A. H. Sparks, C. Valdivia, A. H. C. van Bruggen, L. Willocquet, A. Djurle, E. Duveiller, H. Eckersten, S. Pande, C. Vera Cruz, and J. Yuen. 2011. Complexity in climate change impacts: A framework for analysis of effects mediated by plant disease. Plant Pathology 60:15-30.
• Sparks, A. H., G. A. Forbes, R. J. Hijmans, and K. A. Garrett. 2011. A metamodeling framework for extending the application domain of process-based ecological models. Ecosphere, accepted.