Multistatistics metric evaluation of ocean general circulation model sea surface temperature: Application to 0.08° Pacific Hybrid Coordinate Ocean Model simulations

Abstract

This study is a multimetric statistical evaluation of interannual and climatological mean sea surface temperature (SST) over the Pacific Ocean (north of 20°S) simulated by an ocean model. The evaluation procedure is outlined using daily and monthly SSTs from eddy‐resolving (0.08°) Hybrid Coordinate Ocean Model (HYCOM). Satellite‐based products and buoy measurements are used for model‐data comparisons. Three are three principal findings. (1) Using monthly mean climatological atmospheric forcing with the addition of a 6‐hourly wind component can yield realistic simulations of monthly mean climatological SST in comparison with observations and interannually forced simulations. (2) Nondimensional skill score can be a very useful metric for validating SST from an ocean model in a large region, such as the Pacific Ocean, where the amplitude of the SST seasonal cycle has large spatial variations. The use of skill score is extensively discussed along with its advantages over other traditional metrics. Interannual model‐data comparisons (1993–2003) using satellite‐based SST give basin‐averaged yearly mean skill score values ranging from 0.35 to 0.58 for HYCOM. (3) A comparison of HYCOM to 804 yearlong daily buoy SST time series spanning 1990–2003 gives a median root mean square value of 0.83°C. Relatively small SST biases and high skill values are essential prerequisites for SST assimilation using an ocean model as a first guess and for SST forecasting. The validation procedures presented in this paper include a variety of statistical metrics and use a comprehensive observational buoy data set. Such procedures can be applied to any global‐ or basin‐scale ocean general circulation model that predicts SST.