I know mattdm's answer has already been accepted and that focal plane shutter effects are real, but that is not what is happening here. As mattdm said, a focal plane shutter causes distortion because different parts of the frame are exposed at different times. For a spinning blade that moves a fraction of a rotation during the entire exposure, the blade would appear curved and its thickness wrong. However, that is not what we are seeing. In no case will this effect cause two distinct images of a smoothly moving object like that blade.
I see two other artifacts in this picture:
- Horizon diffraction. The subjects are 35 miles away very low to the horizon. The air near the ground (or water in this case) has various temperature gradients and hot and cold pockets. Hot and cold air has a slight difference in its index of refraction. The hot and cold air layers and pockets then bend light as a result. This is a slight effect, but over 35 miles can certainly cause these kinds of artifacts and much more. This is the same effect that causes mirages over much shorter distances in deserts where the ground is warmer than the air. It would be very suprrising if the see temperature and air temperature were well matched that day, especially over such a long distance.
Since there are pockets of hot and cold (relatively) air, distant object can apear multiple times or not at all at any one instance. It is no stretch at all to believe that the double image of the blade could be caused by this.
- Compression artifacts. There is a lot of stuff around the windmills that look like JPEG compression artifacts. This image is at full resolution, so it's no surprise these artifacts are visible even at a "hiqh quality" setting in the camera. I don't think these artifacts are so strong as to cause the double blade image, but they certainly don't help. I really don't understand why people shoot in anything but raw if they are going to pixel peep afterwards. It makes no sense. Also, you want as much information as possible going into post-processing, which certainly does NOT mean adding compression artifacts and squasing the image to only 8 bits per color per pixel. However, that is another discussion.
For those who are still thinking this might be a focal plane shutter artifact, here is some math. The picture was taken at 1/640 second shutter speed, which means each pixel was exposed for about 1.6 ms.
I don't know what the X-sync speed is of that camera (the maximum shutter speed at which the whole shutter is open at once), but let's be generous and say it is 1/100 second. That's pretty slow by today's standards. Even if so, it means the shutter traversal time accross the whole image is no more than 10 ms. In other words, the center of exposure time varies by 10 ms accross the picture. This will be true regardless of shutter speed.
However, this 10 ms time lag is spread accross the image. The center is only 5 ms off from either edge, for example. I looked at the picture above, and the windmill blade in question is only 12 pixels wide at most. A full resolution picture on that camera is about 3900 pixels accross, so the time skew across the width of the artifact is 10ms(12/3900) = 215 ns. That is tiny compared to the exposure time, but adding the two still comes out to less than 1.6 ms.
Now let's say the windmill was rotating at 1 Hz. That would be pretty darn fast for such a large machine. 1.6 ms would be only .6 degrees of rotation. Considering a radius of 12 pixels, that would mean a motion of .12 pixels at the tip of the blade, which represents the maximum size that any focal plane shutter artifact can be in this case. That's simply not what's going on.