String Theories Fuzzy Logic and Forecasting Inconsistencies Applying

String Theories, Fuzzy Logic and Forecasting: Inconsistencies Applying Empirical Plant Disease Models Daniel Cooley and Jon Clements University of Massachusetts Amherst dcooley@microbio. umass. edu UMass. Amherst

Case study: sooty blotch flyspeck complex of apple • Models target period following end of primary apple scab to ~ 1. 5 in. apples • Can eliminate several early summer fungicide ‘cover’ applications Scab fungicides: green tip to fruit set 1 to 4 applics. 3 to 6 applics. SBFS fungicides: fruit set to harvest UMass. Amherst

Basic concept • Most or all SBFS inoculum develops on reservoir hosts in borders and moves into orchards • Development in borders and on fruit is driven by surface moisture or very high humidity • FS is focus UMass. Amherst

Original model • Brown & Sutton 1995 – empirical model based on first signs • Biofix – 10 days after petal fall • 273 accumulated leaf wetness hrs. for periods ≥ 4 hrs. UMass. Amherst

Original model – LW sensing • De. Wit monitor – “string” based • Wet if ≥ 50% deflection • Placed inside dripline of tree • 1. 5 meter above ground UMass. Amherst

Original model action threshold • First appearance of signs: 209 to 310 ALWH • Benzimidazole trt. at 200 to 225 ALWH • “… the threshold that we have established with the de. Wit sensor may have to be modified if other sensors are used. ” UMass. Amherst

Hartman revision • Electronic sensor rather than de. Wit • Used a 175 hr. treatment threshold • Counted all wet hrs. – no 4 hr. minimum • Biofix of the first postpetal fall fungicide treatment UMass. Amherst

Illinois / Iowa / Wisconsin • Babadoost et al. 2004 used Hartman modification • Compared electronic on-site with mesoscale interpolated (Skybit) data • Skybit LW accumulated more rapidly than on-site • SBFS Incidence higher in modeldirected plots in 12 of 28 site yrs. UMass. Amherst

LW vs. RH • Duttweiler et al. 2008 • Accumulated hrs. of RH ≥ 97% better predictor in IA, but ALWH better in NC • Regional differences in climate expected with empirical model UMass. Amherst

What does a user ask? • When should I spray? • Commercial model software and monitoring software bundle – Spectrum • Commercial remote monitoring and model delivery – Sky. Bit • Public web-based weather and model delivery – NEWA and Orchard Radar • On-site monitoring and published Extension recommendations UMass. Amherst

Accumulated Leaf Wetness or High RH Hours First SBFS recs. , 5 models 400 350 Orchard Radar using Sky. Bit LW, temp. adj. 270 threshold Jun 6 300 250 Sky. Bit model, 350 threshold Jun 16 Spectrum model, Ext. rec. , 300 threshold on-site Hobo Jul 17 270 threshold Jun 12 NEWA, on-site Hobo 170 threshold Jun 2 200 150 100 50 5 wks. 2 – 3 fungicide applications PF 0 5/3/06 5/10/06 5/17/06 5/24/06 5/31/06 6/7/06 Date 6/14/06 6/21/06 6/28/06 7/5/06 7/12/06 UMass. Amherst

Key differences • Biofix • Petal fall (cultivar? ) • 10 days after petal fall • Last fungicide targeting scab • Accumulated leaf wetness hrs action threshold • Count all hours or exclude short periods • Choose one: 170, 200, 259, 270, 300, 350 … UMass. Amherst

Key differences • Method of data collection • On-site • Remote site-specific • On-site • Placement of grid • 45º facing north 1. 5 meters • Canopy or open? • Is anyone still using strings? • Grid sensitivity: 40% of range? UMass. Amherst

Key differences • Remote site-specific • Sources of data – leaf wetness not common • Algorithms for location • Algorithms to interpolate from common data such as RH and wind speed to LW • Should we just use RH? UMass. Amherst

Accumulated Leaf Wetness or High RH Hours Comparing four data sources Date UMass. Amherst

Issues to resolve • Biofix – last primary scab fungicide vs. phenology • If fungicide, depletion should be used • Differentiate action threshold from damage threshold, i. e. first SBFS fungicide trigger from first visible signs on non-sprayed fruit UMass. Amherst

Issues to resolve • Standardize sensor placement for on-site equipment considering ease of use and maintenance (not in canopy) • Standardize sensor wetness threshold • Determine accurate LW estimators for off-site sensing UMass. Amherst