Site-Specific Weather Data for Disease Forecasting: Reality or Pipe Dream? Bob Seem Cornell University New York State Agricultural Experiment Station Geneva,

Site-Specific Weather DataSite-Specific Weather Datafor Disease Forecasting:for Disease Forecasting:Reality or Pipe Dream?Reality or Pipe Dream?

Bob SeemCornell University

New York State Agricultural Experiment StationGeneva, NY 14456

Midwest Weather Group Meeting 25 July 2008

What is site-specific weather?

Weather data and associated information determined for a specific location (lon-lat) or grid-based information at a resolution of ~1km2

Why is site-specific weather important?

• Emerging alternative to regional-scale data

• Incorporates local physiographic features

• Economical alternative to automatic weather stations

• Natural link to disease forecasts and graphical representation of disease risk

Examples of site-specific weather implementation

• Application of mesoscale weather models

• Application of NWS numerical forecast models

• Application of statistical/interpolation schemes

Mesoscale Models

• MM5• North Amercian Meso (NAM,

formerly Eta)• MASS (LAWSS)• WRF

Input PreprocessorsSurface Data

Climatic Data

Raw Data

f

ModelsLAWSS

SWEB

DMCast

Elevation, Landuse …

NDVI, SST …

Reanalysis, AWS …

Temp, RH, Wind, Rad …

Leaf wetness …

Downy Mildew Risk …

Output Processing

Data Transformation

Data Analysis

Merging, Concatenation

Mapping, Plotting …

Sample output - weather variables

Variable Name UnitSurface Air Temperature CAir Temperature at 2m CSkin Temperature CDew Point Temperature CU Wind m/sV Wind m/sAltimeter Setting mbSensible Heat Flux W/m**2Latent Heat Flux W/m**2Total Evaporation mmTotal Shortwave Radiation MJ/m**2Outgoing IR Radiation W/m**2Cloud Cover fractionMean RH SFC - 500MB percentTotal Precipitation mmShallow Soil Layer Water Content vol. fractionCover Layer Water Content m

Grid Horizontal Resolution

# Grid points

Domain Size (km2)

Hours / Domain

# Domains for Finger Lakes

A 9km x 9km 90 x 90 656100 2 1

B 3km x 3km 100 x 100 90000 7 1

C 1km x 1km 100 x 100 10000 12 1

D 333m x 333m 100 x 100 1109 25 9 (parallel)

E 150m x 150m 100 x 100 225 50 36 (parallel)

Domain size and running hours

Grid A (9km resolution)Grid A (9km resolution)

B1B1 B2B2 B3B3

B4B4 B5B5 B6B6

B7B7 B8B8 B9B9

NestedDomains

C1C1 C2C2 C3C3

C4C4 C5C5 C6C6

C7C7 C8C8 C9C9

Grid A

B5B5

Grid B5 (3km resolution)Grid B5 (3km resolution)NestedDomains

D1D1 D2D2 D3D3

D4D4 D5D5 D6D6

D7D7 D8D8 D9D9

Grid B5

C5C5

Grid C5 (1km resolution)Grid C5 (1km resolution)NestedDomains

Grid C5

D2D2 D3D3

D5D5 D6D6

D1D1

D4D4

D7D7 D8D8 D9D9

NestedDomains

Grid D1-D9 (333m resolution)Grid D1-D9 (333m resolution)

14

16

18

20

22

24

26

28

Tem

per a

t ure

( C)

Jun 14 Jun 29 Jul 14 Jul 29 Aug 13

Date (2003)

Simulated by LAWSS

Observed by AWS

Geneva

40

60

80

100

Re l

a ti v

e H

umi d

ity ( %

)

Jun 14 Jun 29 Jul 14 Jul 29 Aug 13

Date (2003)

Simulated by LAWSS

Observed by AWS

Geneva

12

16

20

24

28

Tem

per a

t ure

( C)

Jun 14 Jun 29 Jul 14 Jul 29 Aug 13

Date (2003)

Simulated by LAWSS

Observed by AWS

Fredonia

40

60

80

100

Re l

a ti v

e H

umi d

ity ( %

)

Jun 14 Jun 29 Jul 14 Jul 29 Aug 13

Date (2003)

Simulated by LAWSS

Observed by AWSFredonia

Disease model – Downy mildew

AA BB CC

DD EE FFLake Erie

Seneca Lake

Cayuga Lake

LocationCriteria of daily comparison

Agreement ratio between observed and simulated data during:

the entire simulation non-rainy days

Geneva(42.9N, 77.0W)

Wetness > 0.1 33/62 (53%) 28/32 (88%)

DMrisk > 0 41/62 (66%) 28/32 (88%)

Fredonia(42.7N, 78.9W)

Wetness > 0.1 39/62 (63%) 28/37 (76%)

DMrisk > 0 36/62 (58%) 27/37 (73%)

Weather Research and Forecast (WRF) Model

A next-generation mesocale numerical weather prediction system designed to serve both operational forecasting and atmospheric research needs. It features multiple dynamical cores, a 3-dimensional variational data assimilation system, and a software architecture allowing for computational parallelism and system extensibility. WRF is suitable for a broad spectrum of applications across scales ranging from meters to thousands of kilometers.

www.wrf-model.org

USDA/ZedXSoybean Rust PIPE

24-Hour Precip and RH

USDA/ZedXSoybean Rust PIPE

Spore Wet Deposition

USDA/ZedXSoybean Rust PIPE

Soybean Development Stages

USDA/ZedXSoybean Rust PIPE

Rust Development Stages

NWS numerical forecast models

• Resolution not as high as mesoscale models

• Convergence is occurring• Products greatly increased and

not focused just on aviation weather

NWS Graphical Weather Forecasts

Copyright(c) 1995 SkyBit, Inc. Phone: (800) 454-2266 E-Weather Forecast for: EXPERIMENT STATION-GENEVA, NY Forecast beginning : TUE May 16, 1995 <----------------- 0-48 HOUR FORECAST----------------> May 16 May 17 HOUR (EDT) 8a 11a 2p 5p 8p 11p 2a 5a 8a 11a 2p 5p 8p 11p ----------------------------------------------------------------------------- TEMP (F) 50 65 70 70 65 58 57 56 59 65 69 70 65 60 2"- SOIL TEMP (F) 56 61 67 70 68 63 57 56 57 62 67 69 68 63 REL HUM (%) 75 48 39 40 49 62 68 75 76 67 65 66 73 84 6HR PRECIP(in) .00/ .00/ .00/ .05/ .26/ .62/ .11/ 6HR PRECIP PROB(%) 1/ 0/ 14/ 48/ 84/ 64/ 46/ 3HR WETNESS (hrs) 0 0 0 0 0 2 3 3 3 3 3 3 3 3 WIND DIR (pt) SW WSW W W SE SSE S S S SW SW W W WSW WIND SPEED (mph) 4 3 8 9 3 6 8 8 10 12 13 14 11 9 CLOUD COVER SCT SCT BKN BKN OVC OVC OVC OVC OVC OVC OVC OVC BKN OVC 3HR RADIATION (ly) 27 150 198 149 56 0 0 0 7 37 54 47 40 1 DRYING (key) 3 7 8 8 7 5 4 4 4 6 6 6 5 3 SPRAYING (key) 8 9 6 6 10 7 5 5 4 3 2 2 3 4

Daily products - E-Weather

• Considerable advancements made over recent years (e.g., WRF)

• Computing power is greatly improved

• Flexible links for disease models (and precision ag in general)

• Need better input data (soils, soil moisture, vegetation cover)

• Some variables not reliable…yet

Conclusions I

• Need better biology• How do we validate site-specific

information–Validation sites vs total sites–Do management decisions change

• Don’t let infatuation overtake reality

Conclusions II