Choosing 'best' coordinates of base station based on multiple days of surveying
I'm new to surveying and have just been doing some GPS work as part of a larger project. So I have very basic questions about finding the 'known' location for a base station. I've gone out and surveyed (RTK) at this one location three different days in the last two weeks. I'm walking with a Trimble backpack to survey a large area of beach, with my base station set up on a fixed-height tripod in the same location each day. The occupation times at my base station were just over 2 hours for day 1, just over 5 hours for day 2, and just under 4 hours for day 3. I've submitted the base station data for each day to OPUS (my OPUS reports were generated, respectively, 9 days, 2 days, and 1 day after each survey).
I've now got three slightly different coordinates for the base station, and I'm wondering which one to 'choose' as the 'real' base location to adjust all my data (so the three days worth of data line up) and use for future surveying at this site. According to the OPUS reports, horizontal coordinates are all within 22mm of each other. Errors are all less than 0.05m. Ambiguities are all above 85% and RMS is around 0.012. For the two longer occupations, the percent observations used is around 88% (for the shorter occupation, it's about 96%).
Do I just choose the longest occupation? Is there a way to combine the three different results to find the best solution? Should I resubmit the long occupation in a few days to OPUS to get a better solution (maybe get percent observations used above 90%) and then use that or will it not make much of a difference after a day?
Any guidance is appreciated!
You could use any of the 3. They are all "correct" - whatever that means. The results are given as a position with a +/- error value. Those plus or minuses will overlap.
If you want to get rigorous use an average of the three. Or even more rigourous, a weighted average. You could give your 5 hour observation a weighting of 5, the 4 hr a weighting of 4, and the 2, a 2.
The extended OPUS reports include covariance data on vector quality. That data can be used to weight the results based on data quality rather than occupation time. I would not attempt that with some software made to deal with it, like StarNet.
It is always very desirable to wait until OPUS will use "rapid" orbit data rather than "ultra-rapid" as there is often significant improvement. That means waiting until UTC midnight after the end of the data in the file (USA evening), and then waiting until afternoon of the following day to submit.
You may get a small further improvement by waiting 2 or 3 weeks to get "Precise" orbit data, which is usually available by Monday 15 days after the Saturday/Sunday UTC midnight after your data ends. Taking data on Saturday evening will stretch the wait time by a week compared to Saturday early afternoon data.
If all of your results are using "rapid" data or better, some sort of weighted average will usually improve the accuracy. Choice of weights can get deep in math and philosophy, but a simple start would be to use weights that are 1/pk-pk from the reports. You can also consider the rms reported. Experts or software like Star*Net may use the covariance matrices given in the OPUS extended reports to better apply the relationships in the errors.
It is tempting to average all of the vectors from each CORS to your point with weightings, but note that each OPUS report gives vectors to the averaged point coordinates and not the vectors to each of its three solutions so that just gives you back the same average.
Edit: I see that I mostly agreed with Mark. I would, however, suggest that if you use the lengths for weights, you use square root of hours, as that is the typical curve of accuracy improvement.
Did all three OPUS solutions use the SAME three CORS?
If not, resubmit and make sure that they do!
Are all three CORS "good" ones?
Some of the CORS coordinate estimates STINK, and if one of those CORS found its way into one or more of your solutions, then that/those positions will also stink.
Even some of the "reasonably good" CORS coordinate estimates are on CORS that are NOT particularly well behaved (either in the past, or say this week).
there is more to using OPUS than just mashing the send button.
Once you get GOOD coordinate estimates on your remote station (and the extended output OPUS report), you can use any good LSA program to really zero in.
Just my 2-bits
Thank you all for the valuable answers! I'll try some of these out and see what I get!