Why Models Need Standards
Why Models Need Standards
Joe Russo, Jean Batzer, Mark Gleason and Roger Magarey Midwest Weather Working Group AgGateway Mid-Year Meeting Johnston, Iowa June 19, 2013
Information Technology (IT) Paradigm Hierarchy of Information Flow
Copyright © 2013 ZedX Inc.
Models/Data/Risk in Management Decision Making In Situ Sensors
Data Viewing
Equip Sensors
Record Keeping
April
May
Planting
Application
Field Communication
Data Management
Guidance
As-Applied
June
Irrigation
Alerts
July
Scouting
Yield Monitor
Remote Sensing
August
Tillage
September
Harvest
Field Communication
Irrigation Scheduling Fertilizer/Chemical Applications
Soil/Field Sampling
Degree Day Tracking Scouting
Web Access
April
Data
Copyright © 2013 ZedX Inc.
Models
May
Models
June
Models
July
Models
August
Risk Analysis September
Scale of Weather Data Sources Relative to Farm, Field, Canopy
CFSv2 GFS, NOGAPS NAM, RUC
CFSv2
WRF, RTMA, GFSMOS
+ NWS station network + NWS graphical forecast (grid) + 1000 m SkyBit grid + 800 m PRISM grid
Agricultural decision-making scales in blue box Precision Agriculture
Copyright © 2013 ZedX Inc.
Grid Versus Station Weather Data
.
. .
.
.
. .
.
Weather Stations
Source: http://www.research.noaa.gov/climate/images/modeling_grid.png Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales.
Uncertainty Associated With Spatial Scales
Copyright © 2013 ZedX Inc.
Uncertainty Associated With Temporal Scales and Forecasts
Value
Observation
1- hour
6-hour
1-day
1-week
1-month 6-month
1-year
Value
Forecast
6-day Copyright © 2013 ZedX Inc.
5-day
4-day
3-day
2-day
1-day Observation
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy.
Standard Glossary of Variables and Units
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
Copyright © 2013 ZedX Inc.
Standard Variable Nomenclature (1) Give variable name.
Examples,
temperature maximum temperature
(2) Give original sampling interval. Examples,
hourly temperature daily maximum temperature
(3) State any statistics used to create a new variable from observations. Examples,
daily average temperature (1/2maximum + 1/2minimum)
(4) State any statistics used to derive a mean of the original observations for longer periods. You must state period of mean as days, months, years, etc. Examples,
monthly mean of daily average temperature for 1961 monthly means of daily average temperature for 1961 thru 1995 annual mean of hourly temperature for 1961 annual means of hourly temperature for 1961 thru 1995
(5) State any statistics used to derive a normal for a long-term climatology. You must state period of normal as days, months, years, etc. Examples,
30-year normal of monthly means of daily average temperature for 1961-1990 10-year normal of annual means of hourly temperature for 1961 thru 1970
Note: It is implicitly understood that the monthly mean is for the same period as the normal.
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature.
Model Output Precision and Accuracy Input
Output
Model
~
(Black Box)
Model output compared (~) to truth set Observation/ = Forecast Input Accuracy
Copyright © 2013 ZedX Inc.
=
Model Output Precision
=
Model Output Accuracy
Model Output Precision and Accuracy
How do we measure dew? Truth vs. Electric • • • • • • •
resistance threshold establishment droplet size sensor surface sensor angle compass direction canopy location
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy.
Copyright © 2013 ZedX Inc.
Terms for Plant Pathology
• Presence or absence of a disease in a canopy. • Disease incidence is the proportion of a plant community that is infected.
• Disease severity is the proportion of a plant that is infected. • Pathogen (urediniospore) intensity is the number of urediniospores per unit host area (cm2). • Disease counts are the number of uredinia or pustules per unit area . • Primary disease counts are the number of uredinia resulting from primary infection due to the first deposition of urediniospores.
Terms for crop loss assessment depends on the disease
Disease severity: percentage or proportion of plant area or fruit volume destroyed by a pathogen
Disease incidence: proportion of a plant community that is diseased
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output.
Observational Protocol for Model Verification Can you use a 5 km area model to predict disease in a 100 m orchard?
May 5, 2009
Biglerville, PA
Copyright © 2011 ZedX Inc.
Source: pa-pipe.zedxinc.com
Observational Protocol for Model Verification Biglerville, PA Experimental Site
1000 m
100 m
Copyright © 2011 ZedX Inc.
Observational Protocol for Model Verification Biglerville, PA Experimental Site
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
Observational Protocol for Model Verification Biglerville, PA Experimental Site
10 m
Copyright © 2013 ZedX Inc.
Observational Protocol for Model Verification 2009 Scab Incidence on Shoot Leaves, Red Delicious University Drive Orchard, PSU-FREC, Biglerville, PA
North Top of Block (R1) Center Canopy
Center Canopy
Top Canopy
60 40 20 0
80 60 40 20 0
5-5
5-18 5-26 Date
6-1
5-5
Center Canopy
Center Canopy
Top Canopy
80
80
% Disease
100
60 40 20 0
5-18 Date
5-26
5-26
6-1
South Block (R4
100
5-5
5-18
Date
Center Block (R3)
% Disease
Top Canopy
100
80
% Disease
% Disease
100
Slope Side of Block (R2
6-1
Top Canopy
60 40 20 0
5-5
5-18 5-26 Date
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
6-1
Observational Protocol for Model Verification 2009 Scab Incidence on Shoot Leaves, Red Delicious University Drive Orchard, PSU-FREC, Biglerville, PA
North Top of Block (R1) Center Canopy
Center Canopy
Top Canopy
PA-PIPE 5 km Model
60 40 20 0
80 60 40 20 0
5-5
5-18 5-26 Date
6-1
5-5
Center Canopy
Center Canopy
Top Canopy
80
80
% Disease
100
60 40 20 0
5-18 Date
5-26
5-26
6-1
South Block (R4
100
5-5
5-18
Date
Center Block (R3)
% Disease
Top Canopy
100
80
% Disease
% Disease
100
Slope Side of Block (R2
6-1
Top Canopy
60 40 20 0
5-5
5-18 5-26 Date
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
6-1
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output. • Specify experimental protocols for model verification.
Standard Model Configuration and Output Files Model Configuration File File Name: corn_country_state_xxx (county fips)_xxx (user number).conf Group= Corn Models Country= US State= PA (042) County= 027 (Centre) Site Name= ZedX Site ID= US_PA_027_001 Latitude, LAT= 40.8880 Longitude, LON= -77.7759 Elevation, ELV= 329 (meters) Standard Time Offset= -5 Daylight Start Date= 20040404 Daylight End Date= 20041031 Data Type= Obs Data Type= Fct
Product= Standard Corn Phenology Sim Start Date= 20040301 Sim End Date= 20030930 User Para= crop= crn User Para= var= pioneer 3394 User Para= cmr= 110 Std Para= oset Std Para= phen User Para= seed= 2 Std Para= crdx User Para= pldate= 20040310 User Para= stex= sicl User Para= sdep= 20 Std Data= gcrnut Std Data= gcrnuc User Data= scrn Std Data= crdcrn
Standard Model Application Programming Interface (API) Manufacturer Seed Company 1
API
Models (3rd Party)
Wireless
Management Template (USDA)
Media Farmer FMIS 2
API Seed Company 2
API Product XML
Equipment Company 1
Telematics
Output XML
Controller
API FMIS Standard XML
Equip XML
AgGateway Consultant FMIS 1 Core Data
ISO Standard XML
FMIS XML
ISO XML
Service Provider FMIS 3
AgFleet
Local Data
Implement
XML
AgGateway Standard
XML
FMIS XML Standard
XML
ISO 11783-10 Standard
Tractor
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output. • Specify experimental protocols for model verification.
• Define a standard application programming interface (API) for exchanging model configuration and output files in a standard format.
Thank You! Questions?
Joe Russo, Jean Batzer, Mark Gleason and Roger Magarey Midwest Weather Working Group AgGateway Mid-Year Meeting Johnston, Iowa June 19, 2013
Information Technology (IT) Paradigm Hierarchy of Information Flow
Copyright © 2013 ZedX Inc.
Models/Data/Risk in Management Decision Making In Situ Sensors
Data Viewing
Equip Sensors
Record Keeping
April
May
Planting
Application
Field Communication
Data Management
Guidance
As-Applied
June
Irrigation
Alerts
July
Scouting
Yield Monitor
Remote Sensing
August
Tillage
September
Harvest
Field Communication
Irrigation Scheduling Fertilizer/Chemical Applications
Soil/Field Sampling
Degree Day Tracking Scouting
Web Access
April
Data
Copyright © 2013 ZedX Inc.
Models
May
Models
June
Models
July
Models
August
Risk Analysis September
Scale of Weather Data Sources Relative to Farm, Field, Canopy
CFSv2 GFS, NOGAPS NAM, RUC
CFSv2
WRF, RTMA, GFSMOS
+ NWS station network + NWS graphical forecast (grid) + 1000 m SkyBit grid + 800 m PRISM grid
Agricultural decision-making scales in blue box Precision Agriculture
Copyright © 2013 ZedX Inc.
Grid Versus Station Weather Data
.
. .
.
.
. .
.
Weather Stations
Source: http://www.research.noaa.gov/climate/images/modeling_grid.png Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Converting Weather Data Sources to Standard Scales
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales.
Uncertainty Associated With Spatial Scales
Copyright © 2013 ZedX Inc.
Uncertainty Associated With Temporal Scales and Forecasts
Value
Observation
1- hour
6-hour
1-day
1-week
1-month 6-month
1-year
Value
Forecast
6-day Copyright © 2013 ZedX Inc.
5-day
4-day
3-day
2-day
1-day Observation
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy.
Standard Glossary of Variables and Units
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
Copyright © 2013 ZedX Inc.
Standard Variable Nomenclature (1) Give variable name.
Examples,
temperature maximum temperature
(2) Give original sampling interval. Examples,
hourly temperature daily maximum temperature
(3) State any statistics used to create a new variable from observations. Examples,
daily average temperature (1/2maximum + 1/2minimum)
(4) State any statistics used to derive a mean of the original observations for longer periods. You must state period of mean as days, months, years, etc. Examples,
monthly mean of daily average temperature for 1961 monthly means of daily average temperature for 1961 thru 1995 annual mean of hourly temperature for 1961 annual means of hourly temperature for 1961 thru 1995
(5) State any statistics used to derive a normal for a long-term climatology. You must state period of normal as days, months, years, etc. Examples,
30-year normal of monthly means of daily average temperature for 1961-1990 10-year normal of annual means of hourly temperature for 1961 thru 1970
Note: It is implicitly understood that the monthly mean is for the same period as the normal.
Copyright © 2013 ZedX Inc.
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature.
Model Output Precision and Accuracy Input
Output
Model
~
(Black Box)
Model output compared (~) to truth set Observation/ = Forecast Input Accuracy
Copyright © 2013 ZedX Inc.
=
Model Output Precision
=
Model Output Accuracy
Model Output Precision and Accuracy
How do we measure dew? Truth vs. Electric • • • • • • •
resistance threshold establishment droplet size sensor surface sensor angle compass direction canopy location
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy.
Copyright © 2013 ZedX Inc.
Terms for Plant Pathology
• Presence or absence of a disease in a canopy. • Disease incidence is the proportion of a plant community that is infected.
• Disease severity is the proportion of a plant that is infected. • Pathogen (urediniospore) intensity is the number of urediniospores per unit host area (cm2). • Disease counts are the number of uredinia or pustules per unit area . • Primary disease counts are the number of uredinia resulting from primary infection due to the first deposition of urediniospores.
Terms for crop loss assessment depends on the disease
Disease severity: percentage or proportion of plant area or fruit volume destroyed by a pathogen
Disease incidence: proportion of a plant community that is diseased
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output.
Observational Protocol for Model Verification Can you use a 5 km area model to predict disease in a 100 m orchard?
May 5, 2009
Biglerville, PA
Copyright © 2011 ZedX Inc.
Source: pa-pipe.zedxinc.com
Observational Protocol for Model Verification Biglerville, PA Experimental Site
1000 m
100 m
Copyright © 2011 ZedX Inc.
Observational Protocol for Model Verification Biglerville, PA Experimental Site
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
Observational Protocol for Model Verification Biglerville, PA Experimental Site
10 m
Copyright © 2013 ZedX Inc.
Observational Protocol for Model Verification 2009 Scab Incidence on Shoot Leaves, Red Delicious University Drive Orchard, PSU-FREC, Biglerville, PA
North Top of Block (R1) Center Canopy
Center Canopy
Top Canopy
60 40 20 0
80 60 40 20 0
5-5
5-18 5-26 Date
6-1
5-5
Center Canopy
Center Canopy
Top Canopy
80
80
% Disease
100
60 40 20 0
5-18 Date
5-26
5-26
6-1
South Block (R4
100
5-5
5-18
Date
Center Block (R3)
% Disease
Top Canopy
100
80
% Disease
% Disease
100
Slope Side of Block (R2
6-1
Top Canopy
60 40 20 0
5-5
5-18 5-26 Date
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
6-1
Observational Protocol for Model Verification 2009 Scab Incidence on Shoot Leaves, Red Delicious University Drive Orchard, PSU-FREC, Biglerville, PA
North Top of Block (R1) Center Canopy
Center Canopy
Top Canopy
PA-PIPE 5 km Model
60 40 20 0
80 60 40 20 0
5-5
5-18 5-26 Date
6-1
5-5
Center Canopy
Center Canopy
Top Canopy
80
80
% Disease
100
60 40 20 0
5-18 Date
5-26
5-26
6-1
South Block (R4
100
5-5
5-18
Date
Center Block (R3)
% Disease
Top Canopy
100
80
% Disease
% Disease
100
Slope Side of Block (R2
6-1
Top Canopy
60 40 20 0
5-5
5-18 5-26 Date
Credit: Jim Travis, Noemi Halbrendt, Penn State Fruit Research & Extension Station, Biglerville, PA
6-1
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output. • Specify experimental protocols for model verification.
Standard Model Configuration and Output Files Model Configuration File File Name: corn_country_state_xxx (county fips)_xxx (user number).conf Group= Corn Models Country= US State= PA (042) County= 027 (Centre) Site Name= ZedX Site ID= US_PA_027_001 Latitude, LAT= 40.8880 Longitude, LON= -77.7759 Elevation, ELV= 329 (meters) Standard Time Offset= -5 Daylight Start Date= 20040404 Daylight End Date= 20041031 Data Type= Obs Data Type= Fct
Product= Standard Corn Phenology Sim Start Date= 20040301 Sim End Date= 20030930 User Para= crop= crn User Para= var= pioneer 3394 User Para= cmr= 110 Std Para= oset Std Para= phen User Para= seed= 2 Std Para= crdx User Para= pldate= 20040310 User Para= stex= sicl User Para= sdep= 20 Std Data= gcrnut Std Data= gcrnuc User Data= scrn Std Data= crdcrn
Standard Model Application Programming Interface (API) Manufacturer Seed Company 1
API
Models (3rd Party)
Wireless
Management Template (USDA)
Media Farmer FMIS 2
API Seed Company 2
API Product XML
Equipment Company 1
Telematics
Output XML
Controller
API FMIS Standard XML
Equip XML
AgGateway Consultant FMIS 1 Core Data
ISO Standard XML
FMIS XML
ISO XML
Service Provider FMIS 3
AgFleet
Local Data
Implement
XML
AgGateway Standard
XML
FMIS XML Standard
XML
ISO 11783-10 Standard
Tractor
Why Models Need Standards • Convert input weather and other data to standard spatial and temporal scales. • Quantify spatial and temporal uncertainties associated with standard input (weather) data scales. The combined uncertainties can define input accuracy. • Create a standard glossary of variables and units.
• Create a standard variable nomenclature. • Establish standard truth sets to evaluate model output precision and accuracy as a function of input accuracy. • Create standard terms for model output. • Specify experimental protocols for model verification.
• Define a standard application programming interface (API) for exchanging model configuration and output files in a standard format.
Thank You! Questions?
