Issues with results
Last Friday was one of many deadlines for which I was set to have a new draft of my manuscript (of my masters research) to my adviser, Dr. Catherine Neish. I was set back by a problem with my newest results that I noticed as I was discussing (i.e. writing) about them. To recap, I have measured crater morphometry (diameters, depths) using SARTopo, a low resolution topography on Titan. Then I compared it to Catherines stereo results (of higher topography) to ensure the low resolution data was reliable. I then replicated her results because the methodology of the measurements were different. Hers being an average of the entire 2D regions. While mine are averages of individual profiles (up to 8). Obviously, these results would differ, especially if there are fluctuations in the topography. However, I noticed that one of my crater measurements was significantly lower than Catherine’s.
At first, I assumed it was differences in the methodology. After all, the point of me doing this was because we feared they would be different, but I studied it closer by looking at the 8 individual measurements I took to make sure the lower result was just a matter of averaging lower areas. This wasn’t the case though. Every profile was significantly lower depths which didn’t add up. It ended up being a simple problem. I was weighting scaled images. As in, I used black and white images where the white region was the max height and black the min height. This will not give precise measurements and will skew the data.
All in all, it wasn’t that off, but for Shikoku it was. That meant I had to go back extracted precise text file matrices of the data. I had another issue in this process where I was using the wrong file for Shikoku to extract the data. Stereo files vs radar images have different resolutions. That’s how I was differentiating them. Except, Shikoku had several 1404m resolution files; this wasn’t the case for the others. I started to use the wrong one. I don’t think this was the problem originally because I used an image already made (by Catherine I assume). Still, the problem is now resolved.
These updates and fixes have significantly improved my results. Before, they all lied below or within error of SARTopo, but some of them disagreed with Catherine’s results. Now they are more in line. That is, Foresti and Hano are unnaturally low compared to SARtopo (~110 and 70m respectively). This has been explained before; there is a problem in the stereo processing where it likely doesn’t have enough representative features in the radar images to identify a difference in the two. Before, Hano was comparable to SARTopo, and I figured Catherine’s results were just because there was significant variance in the topography (some very low and very high) regions. Although, this isn’t what I find. Its around 50-70m in each profile, never reaching 100.
All the others lie slightly above SARTopo results, but they lie well within the margins of error. The highest are observed with Soi and Shikoku (+80 and +90m respectively). All in all, this is the result we were looking for. The results are slightly higher because SARTopo does average the results downward. Dr. Livio Tornabene actually made a point to point this out in my Thesis notes where he pointed out a paper that proved this was happening on Mars with two different topography sources. This is something I’m adding to the manuscript because it is relevant, and they actually take terrain to floor depths too.
Despite being slightly lower, they lie within the error, so these results can be considered reliable. Here is a plot of the values. Neish stereo results with a square, H18 results with triangle, and SARTopo with circle. The 6 stereo craters are color coded along with corresponding SARTopo. If there is no stereo, the SARTopo values are left grey. One last thing I want to point out is that every stereo measurement (save Forseti, which is not reliable) is higher using my (H18) method. This is what I expected. Neish and others used an averaging of a region of floor and rims, but we used peaks. Therefore, you’d expect the highest points in Neish and others to be lower, or at least I did.
Other news and closing thoughts
I hope to have the manuscript sent to Catherine by Friday. Next week I will be gone to go to the meteor crater field camp. As far as future work goes, I’ve emailed McMaster about working with them to do some tests on the production of long chain polymers in a cold wet environment like Titan’s. I may end up emailing Dr. Britney Schmidts student at GA Tech about modeling impurities in a freezing water melt lens, but I have a meeting with a few people in that group tomorrow. I’ll begin my inquiry there.