Can rainfall be a useful predictor of epidemic risk across temporal and spatial scales?

Plant disease epidemics are largely driven by within-season weather variables when inoculum is not limiting. Commonly, predictors in risk assessment models are based on the interaction of temperature and wetness-related variables, relationships which are determined experimentally. There is an increasing interest in providing within-season or inter-seasonal risk information at the region or continent scale, which commonly use models developed for a smaller scale. Hence, the scale matching dilemma that challenges epidemiologists and meteorologists: upscale models or downscale weather data? Successful applications may be found in both cases, which should be supported by validation datasets whenever possible, to prove the usefulness of the approach. For some diseases, rainfall is key for inoculum dispersal and, in warmer regions (e.g., tropics) where temperature is less limiting for epidemics, rainfall extends wetness periods. The drawbacks of using rainfall at small scales relate to its discrete nature and high spatial variability. However, for pre- or early-season predictions at large spatial scales sources of reasonably accurate rainfall summaries are available and may prove useful. The availability of disease datasets at various scales allows the development and evaluation of new models to be applied at the correct scale. We will showcase examples and discuss the usefulness of rainfall as key variable to predict soybean rust and wheat scab from field to region.

Transcript

1. Can rainfall be a useful predictor of epidemic risk across temporal and spatial scales? Emerson Del Ponte Discussions with: Adam Sparks, Nik Cunniffe, Larry Madden Collaboration: Kaique Alves, Gustavo Beruski
2. Epidemic risk context for this talk Fungal disease (inoculum not limited) Field crops ("large" scale) Weather is key predictor Risk → Fungicide use or planning
3. Agrios (2005) "Classical" disease (infection) risk model Relative humidity and T Dew Rainfall Leaf / Hour Field / Day LWD measured or estimated LWD
4. Which wetness-variable? What is the model-building approach? Data-driven or Concept-driven? Biological or technical reasons Are Data available/accurate for the Scale?
5. Several rainfall-based variables have been selected/used in disease risk models Together with other variables Rainfall duration in hours (continuous) Presence/absence (binary) Number of rain days (count) Daily total (continuous) Weekly total (continuous) Weekly rain events (count) Rain gauges Satellite Gridded data Model estimations
6. Dispersal Infection Cloud cover wet deposition Survival (build-up) moisture splash moisture splash Effects of rainfall on epidemics?
7. Soybean rust (SBR) in Brazil Data-driven model - natural field epidemics Only rainfall (30-day after detection) explained variation in final disease severity Del et (2006)
8. SBR epidemics and rainfall Frequency and amount of rainfall during the epidemics
9. Three Uses of the SBR Rainfall Model 1) Risk analysis 2) Warning system 3) Early-season risk SEV = -2.14 + 0.18CR30 + 1.28RD30 Historical rain (day) Maximum Risk Region Real-time rain (day) More likely Risk Field Real time/forecast rain Likely Risk Large region Cumulative rain n. of Rain Days
10. 1) Risk analysis Del Ponte et al (2011) How frequent severe epidemics? When during the season? Are they influenced by ENSO?
11. 1) Risk analysis Risk of SEV > 70% 1 in 3-4 years 30 years daily rainfall (1979 to 2009) 30 disease onset dates > Jan 1st 24 locations
12. ENSO effects Warm Warm = El Niño Cold = La Niña
13. Time and space ENSO effects
14. Was the model/result too wrong? Warm Warm Cold Cold Warm Cold
15. Warning system Site-specific risk advisory system Daily severity values and action thresholds SEV15 = -2.14 + 0.18CR15 + 1.28RD15 15A 15B DSV SEV15A x 0.3 + SEV15B x 0.7 3-day mean daily severity value
16. Rainfall warning system Site-specific risk advisory system DSV50 action threshold - more conservative DSV80 action threshold - less conservative
17. Number of sprays (9 trials - 3 locations, Gustavo Beruski PhD Dissertation at Esalq) 4.6 4.0 2.6 4.8 5.3
18. Yield gains from using the systems (9 trials - 3 locations, Gustavo Beruski PhD Dissertation at Esalq)
19. Is it worth not follow calendar? Fungicide efficacy (%)
20. 3) Early Season Risk (regional) PR RS Database: first report of soybean rust in commercial fields of a municipality Source: Consórcio Antiferrugem
21. 2014/15 2015/16
22. 2014/15 2015/16
23. Nov 2015 Rainfall These Rainfall data were obtained from the NASA Langley Research Center POWER Project funded through the NASA Earth Science Directorate Applied Science Program Nov 2014 Rainfall
24. These Rainfall data were obtained from the NASA Langley Research Center POWER Project funded through the NASA Earth Science Directorate Applied Science Program Nov 2015 Rainfall Nov 2014 Rainfall
25. These Rainfall data were obtained from the NASA Langley Research Center POWER Project funded through the NASA Earth Science Directorate Applied Science Program Nov 2016 Rainfall Nov 2017 Rainfall
26. http://origin.cpc.ncep.noaa.gov/products/analysis_monitoring/ensostuff/ONI_v5.php What happened in 2015? El Niño! ENSO predictions for OND 2018
27. Past 30-day observed rainfall 7-day forecast rainfall Favorability risk maps
28. So, can rain be useful? Yes! for some diseases and crop production situations, is a very important weather variable driving epidemics from field to regional scales Tha y ! @edelponte