Modeling spatiotemporal dynamics and time to regional outbreaks of soybean rust in southern Brazil

Transcript

1. Modeling spatiotemporal dynamics
   and time to regional outbreaks of
   soybean rust in southern Brazil
   Kaique Alves | Adam Sparks | Emerson Del Ponte
   

   View Slide

2. Why Soybean rust? More models?
   

   View Slide

3. Fungicides are key, but..
   AZOX + CYPR TEBU
   

   View Slide

4. Background … from my talk last year
   

   View Slide

5. Follow up: now using 15 years - SBR prevalence
   All years (2005 - 2019)
   2 States (south) PR and RS
   Commercial soybean fields
   First (date) report in a county
   2,027 records
   ~15,000 records
   PR
   RS
   

   View Slide

6. Epidemic time
   MaxPrev
   AUDPC
   r
   Time_10
   10%
   Prev90
   Prev120
   Temporal progress description / analysis
   Time to outbreak
   (Survival analysis)
   

   View Slide

7. NND
   Nearest neighbour distances
   Epidemic area
   Monthly maps
   Spatial description and analysis
   Initial
   Final area
   

   View Slide

8. Results: reports over time
   Peak in
   January
   Peak in
   February
   PR State
   RS State
   

   View Slide

9. Early onset
   Late onset
   Results: reports space and time
   

   View Slide

10. Results: NND over time
    

    View Slide

11. Results: Epidemic area over time
    Weak correlation between
    Initial and final epidemic area
    

    View Slide

12. Results: Correlations initial and final area
    Week after Sep 15
    Jan 15
    

    View Slide

13. Results: Correlations initial and final area
    Jan 1st
    

    View Slide

14. Results: Time to outbreak
    ~ 70 days
    

    View Slide

15. Results: Principal Components
    

    View Slide

16. Results: Principal Components
    

    View Slide

17. From my talk last year, using 4 seasons
    

    View Slide

18. Euclidean distances
    Clustering method "ward.D2"
    Is the evidence consistent with other years?
    

    View Slide

19. Survival analysis
    Cox modeling
    Is time to outbreak affected by ENSO?
    

    View Slide

20. From my talk last year, using 4 seasons
    

    View Slide

21. What about the next season?
    

    View Slide

22. Conclusions
    ● Large scale spatiotemporal spread varies among seasons and states
    ● Multiple inoculum sources affect initial epidemic area
    ● Early- and mid- season weather plays a major role
    ● ENSO conditions useful as early warning
    Tactical
    Strategical
    Pre-season Growing season
    Risk prediction
    Outlook Forecasting
    Warning
    What's next?
    Develop models for predicting and
    mapping disease risk at the
    regional level
    Thank you!
    @edelponte
    

    View Slide