Using Maps, Models and Open Science in Plant Pathology

Decisions in agriculture are made every day that require information to maximise benefit, reduce risk, and increase production or profitability. This demand for information is a driving factor behind the creation of many plant disease models. Models are used to make tactical on-farm decisions, including how and when plant diseases could or should be controlled while also informing high-level policymaking decisions to determine where to focus breeding efforts both geographically and for future investments. Because of the broad range of modelling applications, many different approaches have been used in botanical epidemiology from non-parametric statistical models such as generalised additive models (GAM) to empirical crop growth models with functions for yield losses due to pests and disease. While these approaches are effective in communicating with end-users, often much more impact can be realised by openly sharing the data and code behind the models when it is possible. Recognising this, along with Emerson Del Ponte from the University of Viçosa, Brazil, in 2018 I co-founded Open Plant Pathology, a community of plant pathology, plant disease epidemiology, pathogen population biology, microbial ecology and genomics researchers that supports the several forms of open science practices in plant pathology. I will present a variety of modelling approaches covering a range of crops and diseases that support decision making at scales from in-season sprays to long-term breeding program prioritisations, which need to account for climate change and the possible related changes in plant disease and how and why we should be sharing our data and code.

Transcript

1. Using Maps, Models and
   Open Science in Plant
   Pathology
   

   View Slide

2. –G. E. P. Box
   “Essentially, all models are wrong, but some are useful.”
   

   View Slide

3. Guiding policy
   How does climate change affect rice diseases in Tanzania?
   Mitigating The Impact Of Climate Change On
   Rice Disease Resistance In East Africa
   [MICCORDEA Project]
   

   View Slide

4. =
   Losses due to
   pest injuries
   Actual Yield
   -
   Production-Situation Driven
   Crop growth model
   (Attainable Yield)
   RICEPEST
   

   View Slide

5. • Simulates yield losses due to pests
   • Data requirements
   o Radiation
   o Temperature
   • Output: Attainable and Actual yield
   o Sheath blight
   o Sheath rot
   o Weeds
   o Bacterial leaf blight
   o Brown spot
   o Leaf blast
   o Neck blast
   o Brown plant hopper
   o Defoliators
   o White heads
   o Dead heart
   o Stem borer
   o Disease severity
   o Production Situation
   RICEPEST
   

   View Slide

6. Bacterial Leaf Blight
   Average AUDPC
   1983-1998
   EPIRICE
   Data: NASA/POWER
   

   View Slide

7. • Simulates unmanaged epidemics
   • Data requirements
   o Precipitation
   o Temperature
   o Relative humidity
   o Planting date
   • Output: Unmanaged disease severity (AUDPC)
   o Sheath blight
   o Tungro
   o Bacterial leaf blight
   o Brown spot
   o Leaf blast
   EPIRICE
   

   View Slide

8. Duku, C., Sparks, A. H. and Zwart, S. 2015. Spatial modelling of rice yield losses in Tanzania due to bacterial leaf blight and leaf blast in a changing climate. Climatic Change 135(3).
   EPIRICE
   RICEPEST
   Data: General Circulation Model (GCM), Commonwealth Scientific and Industrial Research Organization mark 3 (CSIRO-MK3)
   

   View Slide

9. 5 20
   3 0
   18 10
   20 25
   27 24
   25 23
   RICEPEST Model
   { = 3.5 3.2
   3.6 3.6
   Satellite derived,
   georeferenced data
   x
   y
   x
   y
   x
   y
   Georeferenced
   Yield Loss Estimates
   x
   y
   Precip
   Temp
   Rad
   Tons/HA
   Fertilizer
   Irrigation,
   etc.
   Production
   Situation
   EPIRICE
   Model
   (% Disease Severity)
   25 20
   30 19
   27 24
   25 23
   + x
   y
   x
   y
   Blast
   BLB
   +
   

   View Slide

10. Duku, C., Sparks, A. H. and Zwart, S. 2015. Spatial modelling of rice yield losses in Tanzania due to bacterial leaf blight and leaf blast in a changing climate. Climatic Change 135(3).
    Change in yield loss due to BLB
    

    View Slide

11. Guiding breeding efforts
    Disease and submergence or drought
    Stress-Tolerant Rice for Africa and South
    Asia (STRASA)
    

    View Slide

12. International Rice
    Research Institute
    

    View Slide

13. FAO’s Global Administrative Unit
    Layers (GAUL)
    NCEI’s Global Surface Summary of
    the Day – GSOD
    2001-2008*, interpolated
    NASA/JAXA Tropical Rainfall
    Measuring Mission (TRMM) 2001-
    2008*
    Data sources
    

    View Slide

14. EPIRICE BB Output with GAUL overlaid
    AUDPC
    

    View Slide

15. The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by IRRI.
    

    View Slide

16. The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by IRRI.
    

    View Slide

17. May 2015, New Delhi
    Met with partners in person
    

    View Slide

18. The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by IRRI.
    

    View Slide

19. The boundaries and names shown and the designations used on this map do not imply official endorsement or acceptance by IRRI.
    

    View Slide

20. Guiding breeding efforts
    Pathogen population monitoring
    GRDC Project GRI2007-001RTX
    

    View Slide

21. View Slide

22. View Slide

23. View Slide

24. Measuring impact
    Rice variety releases in Philippines and Indonesia
    

    View Slide

25. View Slide

26. N
    

    View Slide

27. Studying effects of climate change
    New methods to address gaps in data availability
    

    View Slide

28. • Forms of data available
    • Policy needs
    • Climate change evaluations
    Yuan-Min Shen, Taichung District
    Agricultural Research and Extension
    Station, Bugwood.org
    Modelling and global mapping of potato late blight
    

    View Slide

29. Metamodels
    Metamodel (GAM*)
    Daily or Monthly Data
    SimCast
    Hourly Data
    Blight Units
    *Generalised additive model
    

    View Slide

30. View Slide

31. BU = s(T, RH, k = 150)
    

    View Slide

32. 1:1
    Fitted
    R-squared: 0.7963
    

    View Slide

33. BU = s(T, RH, k = 150)
    Metamodel (GAM*)
    Daily or Monthly Data
    SimCast
    Hourly Data
    Blight Units
    Data: CRU CL2.0* and World Climate Research Programme’s
    (WCRP) Coupled Model Intercomparison Project phase 3 (CMIP3)
    

    View Slide

34. Tactical Decision Making
    In the paddock, in the moment
    

    View Slide

35. blackleg.sporacle R package output
    Example!
    

    View Slide

36. View Slide

37. Can we spatially enable?
    Field pea blackspot spore showers
    Source: Sentinel 2 and CSIRO Melloy and Sparks (unpublished)
    

    View Slide

38. Where to from here?
    Whole farm models for liming
    Acid Management Strategy Comparison
    All net return forecasts are relative to an “acidification not managed” strategy whereby no
    addition of lime occurs.
    Net present value of future production ($/ha)
    Minimum (1:10) 20 Mean 93 Maximum (1:10) 173
    Lime applied at 1.5 times acidification rate on paddocks below target pH and at 1.0 times acidification rate on paddocks at target pH
    on a target 5 year cycle with additions proportional to previous season’s cash flow
    Strategy 1 Remediation responsive to cash flow
    Minimum cash flow ($/ha)
    Minimum (1:10) -78 Mean -53 Maximum (1:10) -26
    Likely Gain Risk
    

    View Slide

39. –G. E. P. Box
    “Essentially, all models are wrong, but some are useful.”
    

    View Slide

40. Open and Reproducible Research
    

    View Slide

41. Science
    Collect
    Analyse
    Publish
    Write
    Review
    Summarise
    Reproduce
    Re-analyse
    (meta-analysis)
    Share data
    Open repository
    Share code
    open/free
    tools
    Collaborative
    tools
    Citation
    Pre-prints
    Open Access
    Ideal simplified research life-cycle
    

    View Slide

42. • Sponsors/journals require data (standard in
    molecular)
    • Allows reproducibility (data and/or methods)
    • Technology (less cumbersome) is becoming available
    • Enhanced visibility/transparency
    • Multiple citable outcomes: data, code, manuscript,
    etc.
    Why embrace Open Practices?
    

    View Slide

43. • Lack of interest/knowledge (supplemental rarely
    posted)
    • Low incentive/pressure - that may change!
    • Takes (huge) time and effort
    • Document data and code
    • Versioning code and maintaining
    • FOBS - Fear of being scooped?
    Barriers for open practices?
    

    View Slide

44. www.openplantpathology.org
    How can we change that?
    

    View Slide

45. Open Plant Pathology (OPP)
    fosters a diverse community
    culture that values open,
    transparent and reproducible
    research using shared data and
    reusable software
    Vision and mission
    

    View Slide

46. By creating a social network
    with a welcoming and
    sharing scientific
    community
    openplantpathology.slack.com
    Expanding network
    Sharing knowledge
    Brainstorming
    Building capacity
    (Hopefully!) more
    transparent, reproducible,
    efficient and reliable Plant
    Pathology research
    1.
    Social Workspace
    

    View Slide

47. to promote the initiative and publish
    community outcomes
    Community chat
    Website &
    Blog
    Data repository
    Code &
    Data Files
    2.
    Infrastructure
    

    View Slide

48. View Slide

49. Twitter
    

    View Slide

50. GitHub
    

    View Slide

51. Website
    

    View Slide

52. R package: hagis
    

    View Slide

53. ICPP2018 - Workshops
    ● Network Analysis in Plant Pathology
    ● Applications of Information Theory in
    Plant Disease Management: Theory and
    Practice
    ● Introduction to Multivariate Statistics
    Using R
    ● Population Genomics in R
    APPS 2019 - Workshop
    ● R Markdown for Scientists
    Previous activities
    

    View Slide

54. Repositories
    • AfricaRice/IRRI Tanzania work
    • https://github.com/adamhsparks/MICCORDEA
    • STRASA mapping disease in India
    • https://github.com/adamhsparks/STRASA-Biotic-Stress-Maps
    • Metamodelling potato late blight
    • https://github.com/adamhsparks/Global-Late-Blight-MetaModelling
    

    View Slide

55. R Package Resources
    epicrop, the EPIRICE (and hopefully EPIWHEAT) model
    https://github.com/adamhsparks/epicrop
    GSODR/nasapower, global daily weather data
    https://github.com/ropensci/GSODR (station)
    https://github.com/ropensci/nasapower (satellite/model)
    chirps, rainfall from TRMM
    https://github.com/ropensci/chirps (satellite/model)
    getCRUCL2.0, CRU Climate data
    https://github.com/ropensci/getCRUCLdata
    hagis, Plant pathogen pathotype complexities, distributions and analysis
    https://github.com/openplantpathology/hagis
    

    View Slide

56. Thank you
    Visit dpird.wa.gov.au
    Important disclaimer
    The Chief Executive Officer of the Department of Primary Industries and Regional
    Development and the State of Western Australia accept no liability whatsoever by reason of
    negligence or otherwise arising from the use or release of this information or any part of it.
    © State of Western Australia 2018
    

    View Slide