Evaluation of GEV model for frequency analysis of annual maximum water levels in the coast of United States

被引:38
作者
Huang, Wenrui [1 ]
Xu, Sudong [1 ]
Nnaji, Soro [1 ]
机构
[1] Florida State Univ, Florida A&M Univ, Civil & Environm Engn Dept, Tallahassee, FL 32310 USA
关键词
coastal; annual maximum water level; frequency analysis; GEV; 100-year flood; return period;
D O I
10.1016/j.oceaneng.2008.04.010
中图分类号
U6 [水路运输]; P75 [海洋工程];
学科分类号
0814 ; 081505 ; 0824 ; 082401 ;
摘要
The three-parameter general ized-extreme-value (GEV) model f, as been recommended by FEMA [FEMA (Federal Emergency Management Agency of the United States), Z004. Final Draft Guidelines for Coastal Flood Hazard Analysis and Mapping for the Pacific Coast of tl e United States. http://www.fema.gov/ library/viewRecord.do?id=21881 for frequency analysis of annu il maximum water levels in the Pacific coast of the United States. Yet, the GEV model's performance n other coastal areas still needs to be evaluated. The GEV model combines three types of probabilit3 distributions into one expression. The probability distributions can be defined by one of the three par, meters of the GEV model. In this study, annual maximum water levels at nine water-level stations witl long history data (more than 70 years) were chosen for analysis in five coastal areas: Pacific, Nord east Atlantic, East Atlantic, Southeast Atlantic, and Gulf of Mexico coasts. Parameters of the GEV i iodel are estimated by the maximum likelihood estimation (MLE) method. Results indicate that probi bility distributions are characterized by the GEV Type III model at stations in the Pacific, Northeast, ani 1 East Atlantic coastal areas, while they are described by GEV Type 11 in stations of the Southeast Atlant c and Gulf of Mexico coastal areas. GEV model predictions of extreme water levels show good correl ition to observations with correlation coefficients of 0.89 to 0.99. For predictions of 10% annual n aximum water levels, the GEV model predictions are very good with errors equal to or less than ! % for all nine stations. Comparison of observations and GEV model estimations of annual maximun water levels for the longest recorded return periods, close to 100 years, revealed errors equal to or le s than 5% for stations in the Pacific and Northeast Atlantic coastal areas. However, the errors range fron 10% to 28% for other stations located in the East and Southeast Atlantic coasts as well as Gulf of Mexic( coastal areas. Findings from this study suggest caution regarding the magnitudes oferrors in applying i e GEV model to the East and Southeast Atlantic coasts and Gulf of Mexico coast for estimating 100-1 ear annual maximum water levels for coastal flood analysis. (c) 2008 Elsevier Ltd. All rights reserved.
引用
收藏
页码:1132 / 1147
页数:16
相关论文
共 25 条
[1]  
[Anonymous], REGIONAL FREQUENCY A
[2]  
[Anonymous], 1997, REGIONAL FREQUENCY A, DOI DOI 10.1017/CBO9780511529443
[3]  
Boon J.D., 1978, HARMONIC ANAL TIDAL, DOI [10.21220/V5PF2T, DOI 10.21220/V5PF2T]
[4]  
Chow T.V., 1988, APPL HYDROLOGY
[5]   Return periods of extreme water levels estimated for some vulnerable areas of Buenos Aires [J].
D'Onofrio, EE ;
Fiore, MME ;
Romero, SI .
CONTINENTAL SHELF RESEARCH, 1999, 19 (13) :1681-1693
[6]   Some hydrological applications of small sample estimators of Generalized Pareto and Extreme Value distributions [J].
De Michele, C ;
Salvadori, G .
JOURNAL OF HYDROLOGY, 2005, 301 (1-4) :37-53
[7]  
DEAN RG, 1986, COMBINED TOTAL STORM
[8]  
FEMA, 2004, FIN DRAFT GUID COAST
[10]   EVALUATION OF VARIOUS DISTRIBUTIONS FOR FLOOD FREQUENCY-ANALYSIS [J].
HAKTANIR, T ;
HORLACHER, HB .
HYDROLOGICAL SCIENCES JOURNAL-JOURNAL DES SCIENCES HYDROLOGIQUES, 1993, 38 (01) :15-32