AdcircLite-NC Project Updates Jesse Bikman Brian Blanton
AdcircLite-NC Project Updates
Jesse BikmanBrian Blanton
Progress since last meeting
• Ran cross validation with different parameters included/excluded (perturbation scenarios)
• Augmented parameter set with DB_latP1/DB_latP2 describing distance along intersection with given latitude parallels
• DB_latP1/DB_latP2 better characterizes the storm population than TH/DB
• Defined some new error metrics to supplement evaluation of cross validation runs
New Distance ParametersLongitude separation between track line intersection point and
reference longitudes on latitudes 34.5 and 36.0 N
DB_latP2
DB_latP1
Parameter perturbation scenarios
Parameters Scenario name CPD RMW H_B TS Vmax DB_latP1 DB_latP21. base 2. base_no_CPD 3. base_no_Vmax 4. base_no_CPD_no_Vmax
We have adopted the 2 latitude distances in lieu of the distance/angle parameters.
Error metrics
Si = original model value at response point iS*i = surrogate model value at response point iEi = error at response point i = S*i – Si
N = total # response pointsM = # response points where 0.5(Si + S*i) > threshold value
Scenario Statistics
RMSE [m] Response Weighted RMSE [m] Threshold RMSE [m]Scenario AR2 Mean Max mean max mean max1. base 0.96 0.12 0.21 0.20 0.43 0.14 0.262. base_no_CPD 0.97 0.10 0.18 0.17 0.34 0.12 0.233. base_no_Vmax 0.97 0.11 0.18 0.17 0.31 0.12 0.224. base_no_CPD_no_Vmax 0.74 0.35 0.65 0.62 1.37 0.41 0.80