From: Juan Horrillo Date: Mon, 07 Mar 2011 15:19:00 -0600 Please refer to: http://nctr.pmel.noaa.gov/benchmark/index.html B) Laboratory benchmarking 1-Solitary wave on a simple beach Matlab Script : LabBM_SW_01 Data Files : SW_01_00185.mat SW_01_03.mat Example(Model result file): NEOWAVE_SW_01_00185.dat NEOWAVE_SW_01_03.dat 4-Tsunami runup onto a complex three-dimensional beach; Monai Valley Matlab Script : LabBM_IW_04 Data Files : IW_04_gages.mat Example(Model result file): TSUNAMI_IW_04.dat 5-Tsunami generation and runup due to three-dimensional landslide Matlab Script : LabBM_SL_05 Data Files : SL_05_025.mat SL_05_100.mat Example(Model result file): FUNWAVE_SL_05_025.dat FUNWAVE_SL_05_100.dat C) Field benchmarking 1- Rat Islands tsunami Matlab Script : FieldBM_CS_01 Data Files : CS_01_gage.mat Example(Model result file) : ADCIRC_CS_01.dat Matlab script-functions are self-containing and easy to use. They load lab/field data automatically to facilitate user comparison, peer review and determine model performance. For instance, by just running the script eg. > LabBM_SW_01 directions will pop-up with preloaded fake files pretending model results. To see that try this >FieldBM_CS_01('ADCIRC_CS_01.dat') . Users just have to change the model result file name or names for theirs in the argument of the script function according to a pre-established format indicated in the direction, then they can obtain their model comparison and performance. I adopted the normalized root mean square deviation (NRMSD) or normalized root mean square error to measure the numerical model precision. However, other statistical quantities like the correlation coef., index of agreement, scatter Index or absolute error are as well valid and can be implemented with little effort (more discussion is needed here and I open to suggestions) . The NRMSD measures the differences/desviations between values predicted by the numerical model and the values actually observed in the experiment or in the field. In some benchmark problem the NRMSD is plotted in time to visualize model performance in a particular moment. As it is known tsunami models usually predict fairly well the leading waves but perform poorly in predicting subsequent waves. This feature allows users quantify model performance for the first, first two or first three waves and so forth. Notice there are some benchmarks still in need of a Matlab Script-function. 1- Single wave on a simple beach 2- Solitary wave on a composite beach (no available on the webpage) 3- Solitary wave on a conical island 4-Okushiri Tsunami Sincerely, JUAN