This is coming from a non-surveyor who is using a relatively low-end (L1 only) GPS system to do some surveying. I want to get highly accurate elevations, within 2-5 cm. I found the NGS-58 Technical Memo which describes the guidelines for getting that kind of accuracy. First, I do understand the accuracy limitations of the system we have relative to baseline length, and that I'm actually getting ellipsoid heights which can be converted to orthometric with a geoid model. That said, the guidelines state that first a network of control points must be created for the survey by setting up on HARN/HPGN stations with First Order vertical accuracies for three different 5-hour sessions for each point. I understand this as being to establish a sort of vertical accuracy standard for the survey. Once that is done, then the local points can be surveyed and tied to the control points. Post processing would allow me to basically adjust the local points to the control points, and also bring in CORS for more checks.
However, my superiors wonder why we have to go through all that time (multiple 5-hour sessions) to get accurate results. To them, it makes more sense to set up a base station over a HARN with a published First Order vertical accuracy, set up a rover over the local point, survey them both, and tie them together with CORS. They have come up with that method based on the rationale that the CORS station has a very accurate location and known error at any given time. The HARN point also has a known accurate location and error as well. The the difference between the published values for the HARN and our base station results is then the error within our survey. Therefore, we have known errors for the specific time, duration, and locations of our whole survey (ignoring set-up error/user error and those things for the sake of argument). Those errors can be used to adjust the results from the rover, and voila!
Logically, it seems to make sense because it only looks at the specific errors during the specific time of the survey, but from what I've gathered (again, non-surveyor trying to learn), it's not that simple. I just don't know enough to really explain why. I can see how their way of thinking works, but to me it only gives you relativistic accuracy for that one observation, while multiple long occupations on the control station, followed by base station/rover surveys tied to the control stations afterward takes out the relativism with repeated observations under different conditions (different days, weather conditions, set-ups, ephemerides, etc.). Obviously, the superiors would rather do it their way if it works just as well. Should I keep pushing to follow the NGS guidelines, or are they right and I have been doing a whole lot of work for nothing?
Are you trying to meet a standard or not? Anyone can do anything they would like to generate numbers, but unless the minimum specifications are followed then you will generate a number that is close to the standard only by accident. Unless you follow the specifications you will have no basis to claim the 2-5cm accuracy.The reason to follow the standards and specifications of No.58 is because your job specifications require it and because your job requires compliance with the specifications then you will want to offer the service in a manner that is provable that you followed the standards.
Just because you tie into a high level order control point does not make your new control the same level as the tied control.
Have At It
If you know more than the government agency that has been doing this for hundreds of years, far be it from me to stop you.
Paul in PA
The success you'd have with mixing CORS and L1 will depend entirely one how distant the CORS stations are from the area that you wish to build the network. If you are lucky enough to have a few CORS less than 15km from your site, then I'd say your boss's suggestion might work best. Past 15km, you are pushing the limits of L1 only GPS. Cors was really designed to work with L1/L2 GPS over long distances.
If you have a single CORS within range, you can still have pretty good results mixing the CORS station into your processing/adjustment as a third station. And I highly recommend mixing in observations of HARN or other NGS points as you can, for checks, since you won't get that redundant network check with only one CORS.
Keep in the mind that when you read NGS standards, they are complete overkill. If the network is for private use, I'd take whatever NGS suggests and reduce it by half, as far and occupation times and redundancy is concerned. Lots of redundancy is good, don't get me wrong, but following the NGS specs is not necessary to achieve a high quality network.
Have At It
Ouch Paul, he is asking and wrote the caveat that he is not professional several times.
Paul and Dane are correct, however, the NGS has well-educated scientists that have studied this intensely and for decades. If you want to meet the standards, do it to their specifications.
Otherwise, you can get an elevation that sounds good with lots of numbers behind the decimal place, but you don't want to certify to a certain precision (electronics have given us this false sense of precision). I would be sure and check it from at least one other bench mark; probably two to give you some kind of confidence.
What is the slope of the geoid for the point you are interested in determining the elevation? Only way to get a reasonable estimate is to surround that point by observing the ellipsoid/geoid/elevation differences in that geographic area. That's why NGS came up with their procedure.
The "geoid" of choice is theoretical. I don't know many people willing to "stamp" someone else's theory without proving it. Using L1 only is a precarious way to do it unless you have extraordinary experience with the geoid in your local area of practice.
Have At It
That's basically my thought- I could get some numbers that may or may not be close, but I couldn't really certify that I actually had achieved that level of accuracy. To clarify, we don't have any regulatory reason why we need 2 cm vertical accuracy. We want it, and we want to be able to get it and use it. Since I didn't know how to go about ensuring I got that accuracy, I found the NGS memo and used it as the standard because I figured it is the NGS, and I had nothing else to really go on. The other hurdle for me is my lack of experience at post-processing. I don't know enough to understand if what I'm doing is actually making my data more robust, or just arbitrarily making it look more accurate. One thing I'm thinking, and correct me if I'm wrong, but it has to do with the fact that the errors are mostly independent at each site. I can do the survey one day with a rover/base/CORS set up and get one measurement, then do it again the next day and get something completely different because the errors at each site might be completely different. I guess I look at it as the redundant observations are meant to average out those errors to try to get somewhere near the actual value.
> The success you'd have with mixing CORS and L1 will depend entirely one how distant the CORS stations are from the area that you wish to build the network. If you are lucky enough to have a few CORS less than 15km from your site, then I'd say your boss's suggestion might work best. Past 15km, you are pushing the limits of L1 only GPS. Cors was really designed to work with L1/L2 GPS over long distances.
>
> If you have a single CORS within range, you can still have pretty good results mixing the CORS station into your processing/adjustment as a third station. And I highly recommend mixing in observations of HARN or other NGS points as you can, for checks, since you won't get that redundant network check with only one CORS.
>
> Keep in the mind that when you read NGS standards, they are complete overkill. If the network is for private use, I'd take whatever NGS suggests and reduce it by half, as far and occupation times and redundancy is concerned. Lots of redundancy is good, don't get me wrong, but following the NGS specs is not necessary to achieve a high quality network.
Unfortunately, the nearest CORS is about 75 km away.
Where could I find data on the slope of the geoid? I suppose that question tells you how much experience I have with the geoid. I understand the concept of it and what it does, I just lack skills in knowing how to really apply it.
In that case, forget CORS, it will never reliably work with L1 GPS over 75km. You really need to build out the network from HARN and local NGS benchmarks (or get an L1/L2 GPS).
Simple Does Not Mean It Is Easy Or Quick
Simply use 3 L1 receivers. (Means out errors.)
Simply record nightime observations of 12 hours each on three different dates. (L1 cannot account for atmospheric correction like L1/L2, but atmospheric corrections are generally less at night.)
Simply process from at least 3 CORS such that your 3 L1s are within the triangle. (Being inside the triangle means out atmospheric corrections. It is key to the OPUS-RS methodology.)
Simply rotate each receiver/antenna position. (Means out receiver errors.)
Simply traverse through all 3 observed points. (Verifies post processing precision between field points.)
Paul in PA
Simple Does Not Mean It Is Easy Or Quick
Simply stated, can you take your current equipment, local control and your level of expertise and derive 2-3cm accuracy with confirmed or guaranteed results? Simple answer: No.
Simple Does Not Mean It Is Easy Or Quick
Thanks for the suggestion, Paul. We actually have 3 receivers, so that is feasible. I know about the ionospheric delay problem with L1-only vs dual frequency and have suggested night observations, but that is probably not going to happen just from a safety standpoint. Unfortunately with it being so close to the solstice, we don't even have 12 hours of night right now. Would be do-able in the fall or early winter, though. A quick check shows me the nearest CORS sites are 88, 92, and 93 km away. Would it still be worthwhile to attempt to use them at those distances?
Simple Does Not Mean It Is Easy Or Quick
Here's a way to test it. Find two CORS sites 90 miles apart, get 12 hours of night data, process it in an L1 only processor holding one fixed, and see if the elevation of the second point comes up within 5cm of what's published. I have my doubts ... I don't think that adding redundant "failed" vectors to an adjustment is going to mean out the error and give you results that exceed 5cm vertically, and I certainly wouldn't certify that.
Here's the crux of what is being put to me by my superiors:
If we have HARN points with published elevations of First Order vertical accuracy and known errors (I'm thinking of the network and local accuracies on the NGS datasheets), why do we need to survey the HARNs in advance and establish a control network when we already know what the elevations and accuracies are? Can't we just set up the base station over those points and use them as control for the rover, with maybe two or three occupations for redundancy? Then in post-processing, the software adjusts the measurements collected by the rover in accordance with the amount of error between the base station measurements and the actual published values for the HARNs. I think they are thinking of it like an equation where the only unknown variable is the measurements of the point being surveyed. Why go through the trouble of solving for the known values if they're already known? This is a highly simplified way of stating it, but when it comes down to it, that is the question.
Again, my understanding of the answer to that question is that the errors in the base and rover measurements are independent. They may be observing the exact same satellites on the exact same day at the exact same time, but that doesn't mean the error in position measurement is exactly the same for both points. It's the relative difference between the two that helps resolve the errors. But if that is the case, then why once the errors are resolved (more or less) in post-processing can't the results just be adjusted by the amount of error between the known HARN and the base station? That is what I'm struggling to explain. Sorry if I just repeated the question. I'm looking for validation as to whether my understanding is correct, and if it's not, what is the correct way to explain it?
Night Is Relative
Go for 8 hours then.
If those three CORS provide a triangle around the site, give it a daytime try. It may be necessary to use a CORS farther away to get such a triangle, or use 4.
Hold one CORS fixed and solve your 3 points. Hold the second CORS fixed and solve all 3 again. Hold the third CORS fixed and solve the 3 again. Mean each of your L1 site positions. Pick one L1 site as your control, preferably the one with the smallest variance. Hold your control point fixed and solve for the other 2 L1s and the 3 CORS. Since you most likely do not have an L1/L2 post processor always exclude CORS to CORS vectors.
Processed to published CORS values gives you a number for your project site accuracy.
The L1 to L1 to traverse comparisons give you a number for your site precision.
BTW, I started out and used 3 L1 units for years, sometimes up your CORS ranges.
Paul in PA
> However, my superiors wonder why we have to go through all that time (multiple 5-hour sessions) to get accurate results. To them, it makes more sense to set up a base station over a HARN with a published First Order vertical accuracy, set up a rover over the local point, survey them both, and tie them together with CORS. They have come up with that method based on the rationale that the CORS station has a very accurate location and known error at any given time. The HARN point also has a known accurate location and error as well. The the difference between the published values for the HARN and our base station results is then the error within our survey.
This is an example of why you follow guidelines. Errors in this case will not be consistent. GNSS produces random errors. This means it will be up 0.2' on one, perfect on another and down 0.15' on another.
The only reason your manager's idea seems logical is due to their ignorance.
This won't be the only thing like this you run into. If this is critical, hire an expert, if not, who cares, hack away.
I don't mean a non - surveyor can't measure. I mean an expert that can train you and be on call for your questions.
Kind of off the topic, but can I ask the question...If you are non-surveyor as you state, are your superiors licensed surveyors? And you are expected to certify to the accuracy of your measurements to who and for what purpose?
What is your financial and/or legal exposure for not meeting your stated accuracy goals?
Not really off topic. I've gotten that same question before when I've posted other times on here. We're not a surveying agency. We just need location data on the projects we do, but we're not certifying the accuracy of our measurements to any regulatory or legal standard, although being able to at least qualify or explain the quality of our data is what we're going for here.
