ForecastPresentation.ppt
Download
Report
Transcript ForecastPresentation.ppt
Update Of The DNR
Real-Time Predictive
Infrastructure
April 6, 2007
James Davis
Niall Durham
Information Touchpoint
Leveraging Outside-The-Box
Thinking To Develop
Cyberinfrastructure For
Providing Mission-Critical
Forecasts For The Now
Economy
Goals
Integrate NWS predictions with NBDC and
TCOON/DNR collected data
Provide ANN models on DNR website
Water Level
Water temperature
Salinity
Improve ANN display and presentation
Bigscreen TV display
Overview of Predictive System
Useful Terms
Source
Bob Hall Pier
Global Forecast Systems model (NCEP)
Location
Station
Offset (hours)
Predicted data have positive offset
Series
Collection (a.k.a. griddled)
“Six Minute” Database
Designed in late 1980s or early 1990s
Imported data pushed to nearest six minutes
(:00, :06, :12, :18, ... , :48, :54)
One value per time point
Values stored as 16-bit integers [-32768, 32767]
in fixed-point format
Still in use today!
“Six Minute” Database Limitations
Forces data to six minute intervals
What if I want fifteen minute intervals?
Assumes one location per source (station)
The Problem With Predictions
18 Z
O
12 Z
O
6Z
O
0Z O
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Time
(hours) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Each data set contains 15 predictions, from +6 hours to +48 hours at 3 hour intervals.
O = when we receive the data
X = data point
New Database
Multiple values per time
Unique offsets
Store time data was actually collected
One second resolution
Values stored as 32-bit integers [-2147483648,
2147483647] fixed-point format
Each source may have one or more locations
Six minute Data Retrieval
Specify station(s)
Specify series
Select output and formatting options
“Not Six Minute” Data Retrieval
Specify data source(s)
Specify location
Presently only supports one location
Specify series
Specify an offset (optional)
Explicit offset (single)
Maximum offset (upper bound)
Select output and formatting options
Query Examples
Collected Data
Predicted Data
Compare Predicted and Collected Data
What's Next?
Update ANN water level module to include
forecast data from NWS
Bob Hall Pier
Generalize ANN code to support NN models for
water temperature and salinity
Improve graphical displays
The Big Screen
Kiosk example
Build on existing object-oriented code
Water level prediction graph(s)
Rotate through multiple stations
Target: Hurricane
Season
Provide up-to-date ANN
water level predictions for
2007 hurricane season
Questions?
Suggestions?
Thanks For Attending!