I would investigate other possible sources that are contributing to the resulting u and v currents. Probably the wind alone is not the only forcing producing currents, although it is sufficient to produce a noticeable upwelling from Ekman transport.

Hi Ricardo,

I agree that there could be other sources of  u and v variability. There is a buoyancy current in the inner shelf but I am not looking at current data from that region. I am also looking at timescales longer than one day, so the major tidal effects are not there.  Also, it's not just u and v that are not  correlated with the local wind. Temperature is not correlated with the local wind either.  Current and T are coherent with remote wind in a way that suggest coastal trapped wave influence. The remote wind accounts for about half the variances in the current and T.  So, one would think the remaining half of the variance should be accounted for by local forcing, one of which is wind forcing. 

But my question remains what  physically based reasoning  should be used to include the contribution from local winds. In short, I want to determine "factor" in the equation below based on physical arguments. From pure statistical analysis, "factor" is about 0.5. Are there possibilities that factor =0, which means Ekman theory does not apply ?

w_local=factor*(wind stress/density*Coriolis parameter*Length Scale).

Hmm,

If I understand correctly, you want to know if the Ekman transport caused by local winds could be neglected in your region of study.

On physical arguments, you can neglect one or more forcings of a system by scale analysis.

What probably happened is that during the currentmeter data sampling, conditions were such that the contribution of local wind forcing was very small in comparison to the other combined forcings - for instance, caused by pressure or density gradients - to the resulting u and v currents. Hence, the low correlation found.

Hi Ricardo,

Very good point on the scale analysis. I was thinking about something like that. Thanks for the stimulating discussion.

I should have mentioned a bit about the record length of the current. It is  over two decades.  So, the poor correlations suggest local wind forcing was persistently  very small, but probably not zero. I hope that the non zero contribution will be represented by the number "factor", after a scale analysis.

Thanks Guillermo. Yes, I do have wind data and estimated the Ekman upwelling strength, which I think is acceptable (5-10 m/day).  The problem is that I don't have the evidence for the local wind effect on upwelling to accept the theoretically calculated Ekman upwelling strength. Hence, I am searching for a sound reason to include only a fraction of that upwelling in a model.  The current meter data comes in hourly time resolution but I low pass filter them to get rid of tidal components.  I am in a coastal region (southern Vancouver Island) and yes there are wind speed and direction gradients  along the west coast of North America with less upwelling favorable winds near my region. It appears now the poor correlation between current (from depths shallower than mine) and local wind in my study region has been documented in another work . I know some upwelling regions where observed water properties were satisfactorily explained without Ekman upwelling (Best example is a paper by Allen J Clarke, http://onlinelibrary.wiley.com/doi/10.1029/JC084iC07p03743/abstract).  When it comes to my region, the issue of whether Ekman upwelling is negligible is not settled. Therefore, including at least a fraction of it in a model is what I am convinced to do.

ps. I use reanalysis wind with 32 km grid spacing and wind location is not a problem. Some grid points are right next to the mooring.

Hi Guillermo,

Thanks so much for the papers. I have read two of them and they were very useful to me, fun to read as well. But they left me with more questions:

Let's say, for the sake of argument, the coherence between the current and local wind has a peak in the period band 2-4 days. Does that mean the wind stress should be 2-4 day band pass filtered  before it can be used in calculating the Ekman upwelling? Because, in the end that is what I want to calculate. If the answer is yes, wouldn't that be assuming the forcing and its response  have the same frequency range? Is this always true? What if my system have a low frequency response to a high frequency forcing? I suspect an upwelling system has this kind of response due to features like stratification.