Just a quick update on investigating alternatives to Calsky (also thanks for the comment above!). For those just joining this conversation, in a strange twist of fate, the website Calsky.com (which was very useful for predicting planetary transits of the ISS, such as my Mars image) shut down just weeks after my image was published. There are other resources available for predicting solar and lunar transits, but these don't include planetary predictions. I found a program written by Ed Morana that does include planets. This is available as a Java application below (also as an Android app, although I haven't used that):
The source code is also published. This program has been around for a number of years, but never achieved the level of recognition as Calsky and others. Getting it to work in Java Runtime Environment requires bypassing some security features, and so there are some annoyances there. I contacted Ed, and he told me he hasn't updated the Java program in a while. I've conducted some preliminary investigations about the accuracy of the program, and it works great, although the details involved here are beyond the scope of this post. The short version is that you have to decide which TLE data you want to use, the two main sources being NASA and NORAD. The program is based on the SGP4 model of simple perturbations, that have been in use for monitoring satellite orbits for many decades. The catch here is that TLE data, which are used to define orbital state vectors, are associated with up to 1km error (in absolute xyz coordinates), which will always be reflected by up to several hundred meters of error in any predicted transit ground path. Ed told me that he has no indication of which TLE data would be more accurate (NASA or NORAD) and that he frequently makes maps of each, and then splits the difference. I did some retroactive studies on previous transits I have imaged, for which I have the "true" position of the ground path centerline (based upon the image). It appears that splitting the difference between the two predictions is not a bad idea, although in some cases, one set of data is almost nearly perfect, and the other is ~500m in error. So, in the end, each imaging attempt is a gamble. Sometimes you win, and sometimes you lose.
Ed's program gives you .kml files that you can drag and drop into Google Maps or Earth. Shown below is a map drawn in Google Earth, depicting my Mars transit ground path from September 14, 2020. Note that Ed's program doesn't let you set the date (for testing past transits), but I found a way to bypass this by changing the time and date on my computer. Also, I had to download the historical TLE data corresponding to the epoch immediately prior to the transit on September 14, 2020. In conclusion, planning for planetary transits of the ISS is still possible even without Calsky, but it now requires more work on your part. The silver lining is that you have the potential to become much more aquatinted with the models used, and therefore can better understand sources of error. For example, you can conduct numerous experiments using different TLE data and creating maps from each, which is especially interesting for comparison to an actual transit event that you have imaged.