Neural Network Forecasting of Storm Surges along the Gulf of Mexico

1. Neural Network Forecasting of Storm Surges along the Gulf of Mexico Philippe Tissot*, Daniel Cox**, Patrick Michaud* * Conrad Blucher Institute, Texas A&M University-Corpus Christi * * Civil Engineering Department, Texas A&M University

2. Presentation Outline • Frontal Passages and the Texas Gulf Coast • Need for better Water Level Forecasting Models • Application of NN Modeling to Water Level Forecasting • Model Performance • Conclusions

3. Frontal Systems and the Texas Coast • Regular frontal passages from late September to mid May every 7 to 10 days • Wind gusts regularly up to 40-45 mph along the coast • Effects of the frontal passages last up to 4-5 days • Changes in temperature and barometric pressure • The resulting changes in water levels exceed the tidal range

4. Study Location: Galveston, Texas Galveston

6. Computed Harmonic Tidal Sea Level at Pleasure Pier, Galveston (Spring 97)

7. Comparison of Actual and Computed Sea Level (Spring 97)

8. Importance of the Problem • Gulf Coast Ports account for 52.3% of total US tonnage (1995) • 1240 ship groundings from 1986 to 1991 in Galveston Bay • Large number of barge groundings along the Texas Intra Coastal Waterways (ICW) • Worldwide increases in vessel draft

9. Water Level Forecasting in non Tidally Driven Coastal Water Bodies • Water level forecasting is important for a number of coastal users (ports, emergency management, recreational users, …) • Forecasting models need to account for other factors then tidal forces and therefore will necessarily be “near real time” models

10. TCOON Stations

11. Typical TCOON station page Primary Water Level Water Temperature Wind Speed Wind Gust Wind Direction

12. Streaming Data Modeling • Real time data availability is rapidly increasing • Cost of weather sensors and telecommunication equipment is steadily decreasing while performance is improving • How to use these new streams of data / can new modeling techniques be developed

13. Streaming Data Modeling • Classic models (large computer codes - finite elements based) need boundary conditions and forcing functions which are difficult to provide during storm events • Neural Network modeling can take advantage of high data density and does not require the explicit input of boundary conditions and forcing functions • The modeling is focused on forecasting water levels at specific locations

14. Neural Network Modeling • Started in the 60’s • Key innovation in the late 80’s: Backpropagation learning algorithms • Number of applications has grown rapidly in the 90’s especially financial applications • Growing number of publications presenting environmental applications

15. Neural Network Features • Non linear modeling capability • Generic modeling capability • Robustness to noisy data • Ability for dynamic learning • Requires availability of high density of data

16. Neural Network Forecasting of Water Levels • Use historical time series of previous water levels, winds, barometric pressure as input • Train neural network to associate changes in inputs and future water level changes • Make water level forecasts using a Static Neural Network Model

17. Wind Stress Factor in Water Level Changes / Forcing Functions

18. Neural Network Forecasting of Water Levels Water Level History  (X1+b1)  (a1,ixi) Wind Stress History  (X3+b3) b1  (a3,ixi) H (t+i) Wind Stress Forecast b3 Water Level Forecast  (a2,ixi)  (X2+b2) Barometric Pressure History b2 Input Layer Hidden Layer Output Layer Philippe Tissot - 2000

19. Harmonic Analysis NN Model, 24 Hr prediction

20. Comparison between measured water levels (black), tidal chart forecasts (blue), and 24 hour neural network forecasts (red) for Galveston Pleasure Pier during the spring of 1999 (Cox, Tissot, Michaud). The neural network model was trained for a period of 90 days during the spring of 1997 and is applied here to a frontal passage during the spring of 1999. The accuracy of the 24 hour neural network forecast shows the ability to predict the timing and the intensity of frontal passages.

21. Performance of the Model Performance index E Hi are the water levels observations and Xi the water level forecasts

22. Performance Analysis of the Model • Spring ‘97, ‘98, ‘99 data sets covering 90 days with hourly water levels and weather data • Train the NN model using one data set e.g. ‘97 for each forecast target, e.g. 12 hours • Apply the NN model to the other two data sets, e.g. ‘98, ‘99 • Repeat the performance analysis for each training year and forecast target and compute the error index

23. Model Comparisons for Varying Forecasting Time

24. Conclusions • Neural network modeling shows excellent promises for local forecasting of water levels during frontal passages (6 to 30 hour forecasts) • Computationally and financially inexpensive method • The quality of the wind forecasts will likely be the limiting factor for the accuracy of the water level forecasts • Expanding the application of the model to other locations along the coast of Texas

25. Neural Network Forecasting of Storm Surges along the Gulf of Mexico Presentation End

26. Simulated Wind Forecast using Gaussian filter Observed Simulated

27. NWS Predictions and TCOON Observations (Actual Forecast) Galveston Pleasure Pier, 1999 12 Hr Predictions

28. Training of a Neural Network Philippe Tissot - 2000

29. Water Level Changes and Tides • There is a large non tidal related component for water level changes on the Texas coast • Other factors influencing water level changes:

30. Forecasted Water Levels vs. Observed Water Levels Neural Network Forecasts Tidal Forecasts RMS Error: 1ft RMS Error: 3ft

31. Comparison During Frontal Passages Neural Network Forecasts Tidal Forecasts RMS Error: 1ft RMS Error: 3ft