Estimation of Sampling Errors

VOS - VOS-like WAM

	Wind Sea	Swell	Significant Wave Height
January
July

WAM - VOS-like WAM

	Wind Sea	Swell	Significant Wave Height
January
July

Random Sampling Errors

	Wind Sea	Swell	Significant Wave Height
January
July

    It is very difficult to analyse the reasons of the "VOS minus WAM" differences from a direct comparison of the climatologies. On one hand, these differences result from the performance of WAM, from the uncertainties in visual wave observations. On the other hand, these differences are strongly influenced by an inadequate sampling of VOS reports, especially in poorly sampled regions. COADS provides between zero and several thousand of samples per month in a 2x2 degree box, while the WAM model always gives 120 samples for a 30-day month. In order to quantify sampling biases, we simulated a VOS-like sampling of the WAM data. The WAM individual data were interpolated in space and time onto the VOS reports. If several VOS reports were available for the same time moment, the corresponding WAM wave parameters were repeated to simulate the oversampling of the VOS data in comparison to the WAM sampling density.

    We also simulated VOS-like sampling density in the WAM, using random generator. For each month and 2°*2° box 20 simulations provided estimates of differences between monthly means of wave parameters, taken from the original WAM and VOS-like randomly sampled WAM    S_i, where i=1,…20. The value <S_i²>^1/2, where <...> is the averaging operator, gives the estimate of monthly random sampling error.

Return to "Atlas Maps" page