Sorry for our delay in replaying.

To calculate the cost of travelling over the ocean surface in relation to wind contidions we used an Anisotropic Cost Analysis to create a friction surface. This surface is a grid where the cost to move over each pixel depends on the angle between the movement heading and the azimuth wind as well as the wind speed.
To calculate the cost to move over a pixel, speed is not considered as an isolate element but in conjunction with the angle between wind and movement vectors: tail winds are considered as a help and head winds as an obstacle to the movement, both proportionally to absolute speed in m/s, see figure at

http://www.esnips.com/doc...

In the absence of wind, the cost of moving through the friction surface is set to 31 (arbitrary units) and it is constant with respect to movement heading (i.e. wind do not help neither obstacle the movement). In the case of movement following the exact wind vector azimuth (tail wind, corresponding to 0º between vector flight and vector wind) and maximum wind speed (30 m/s as defined by the sensitivity of the QuikSCAT), we assigned the minimum resistance value of 1 (arbitrary units). Deviations from this minimum cost angle and/or decrease in wind speed are increasingly costly until reching a maximum of 61 (arbitrary units) at an angle of 180º (head wind, opposite to bird flight) and a wind speed of 30 m/s (see above figure). This is the interpretation of the function on PLoS ONE paper; we avoided changes of sign as the ArcInfo “pathdistance” module only works with positive values.