As dusk turned to night, I parked in the Tennesee Beach trailhead lot. The last stragglers made their way back to the car as I checked my pack and laced my hiking boots. I walked the couple of miles to the beach in the cool darkness, spotting an occasional owl and listening to the sounds of other nocturnal creatures.
By the time I got to the beach, it was a dark moonless night. I climbed the cliff to the platform on the southern end of the beach using my headlamp. Here’s the view from my destination with some daylight, and a fisheye daylight view from the spot.
This climb isn’t that long, but it is not for the faint of heart even during the day. A narrow path tends steeply up along the verge of erroding cliffs with a straight drop down to the pounding surf. At night, I felt it was good that at least I knew where I was going.
At the top, I unpacked my camera and tripod, put on the fisheye lens, and pointed the camera north. The pole star is in the center of the star trails in this image. I programmed my setup so that I automatically took 10 four minute exposures, and lay back to watch the sky and listen to the surf and the monotonic drone of warning buoys guarding the approaches to the Golden Gate. The forty minutes of exposure time was enough so that I was glad that I had brought warm clothes and a sandwich.
My plan was to “stack” the ten exposures to create a single image. Before packing up and carefully heading down the cliff in the dark I took a longer exposure for the foreground, and some “conventional” night photos.
Back at home, I used the Statistics script (found on the Actions > Script menu in the Extended version of Photoshop CS3) to stack the ten exposures. This means placing one on top of the other as a layer, and combining them using a statistical method specified.
I’ll write more about this later, but the Photoshop documentation doesn’t explain the results of the various options, and I’ve forgotten most of what I once may have known of formal statistics, so the vocabulary describing the various options (for example, “skewness”, “kurtosis”, “entropy”) didn’t mean much to me. This left my usual approach, empirical trial and error.
The default setting, Mean, produced some kind of average sampling, not a very striking result. Standard deviation was interesting, but not ultimately satisfying. Range was good, but Maximum was best. My assumption is that this blended in the maximum value for every sampled point, so it makes sense that it produced the brightest star trails.
With the stacked images of the sky in place, I then used Fluid Mask to quickly cut out the foreground from my brighter exposure, and blend it with the vortex in the night sky.
[Nikon D300, 10.5mm digital fisheye, eleven exposures, all ISO 100, tripod mounted. Foreground exposure: 6 minutes at f/4. Ten stacked sky exposures, 4 minutes each at f/5.6, combined using Photoshop Statistics set to Maximum.]