Transcript 14_Gleason-Portability

Spatial portability of
empirical leaf wetness
duration models
Kwang Soo Kim, Mark Gleason,
Elwynn Taylor, Len Coop, Bill Pfender,
et al.
(Submitted 9/09 to Agric. and Forest Meteorology)
Western Weather
Working Group
Midwest Weather
Working Group
Why model LWD?
 Input to many disease-warning systems
 Problems with measurements:
 No calibration standard
 LWD sensor performance is variable


Not measured at most weather stations


Placement, coating, etc.
Expense, logistics of monitoring
LWD is highly variable in crop canopies

Where to measure?
 ALTERNATIVES
ARE NEEDED.
Modeling LWD
 From contributing environmental
factors
 Aim: Avoid pitfalls of measuring LWD
 Spectrum of models:
Physical to empirical
Physical models
Energy balance at crop surface
 PRO:

 Highly accurate anywhere

CON:
 Radiation inputs are not measured at
most weather stations.
Empirical models
 Use
statistical best-fit approaches
 PRO:

Use widely measured inputs
 RH, wind speed, air temperature
 CON:

Portability may be limited
Portability
Portability
Portability
“Hybrid” LWD models
 Physical principles AND empirical
best-fit methods.
 Most LWD models have both
physical and empirical features.
Empirical LWD models
Three models compared:
 RH>90% (Sentelhas et al., 2008)
 CART/SLD/Wind (Kim et al., 2004)
 Fuzzy logic model (Kim et al., 2006)
43 study sites
Midwest
Pacific NW
Brazil
Costa Rica
Italy
Approach
 Meta-analysis of existing data sets
 RH, wind speed, air temperature
 LWD:
 Painted
vs. non-painted sensors
 How well did each model do in
estimating measured (“true”) LWD?
LWD sensors
Painted sensor
Non-painted sensor
Results
 Painted LWD sensors
 Fuzzy
logic model most accurate
 Highest % correct estimates
 Lowest coefficient of variation
 Highest agreement across sites
Results
 Non-painted LWD sensors
 Less
sensitive that painted sensors
 Correction factor applied to fuzzy
logic model

RH-dependent
 Adjusted
Fuzzy model: highest
accuracy and agreement across
sites.
Summary
 Fuzzy logic model had greater spatial
portability than RH or CART.
 Reason:

Fuzzy model incorporates physical principles
more explicitly than the other empirical models.
 RH model may need a site-specific correction
threshold.
MANUSCRIPT ON MWWG WEBSITE