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!