Do you still do leveling work if you already have GPS?

I have a digital level from Sokkia but sometimes the long rods are difficult to carry into rugged country. When we are doing a topo project our crews are required to close their traverse via loop  and redundunt loops if possible. Then we use our single & dual frequency gps to occupy as much of the control points that are in the open. Usually the elevations from a total station reading and gps are off by 3 cm or less and the largest is less than 10 cm. The entire GPS network is then connected to a nearby MSL benchmark & GPS control point (official) and we ttransfer the elevation & lat/long to our network. We do repeated observations to eleiminate those freak gps signals that throw off some positions.

Question is:

1. Are there other companies that do this method of elevation transfer?
2. Is this a safe or accepted method using dual frequency or RTK gps?

This method is parallel method used in bathymetry survey. In those surveys acceptable variation in RTK elevation readings are 0.10 m.

I'm not sure I understand just how you are performing your control surveys, but let me state that we often are required to establish elevations to a degree of precision that can't quite be done with the total station alone, so we run levels. These occasions are rarely associated with traverses of land for routine subdivision. 

If I was establishing traverse points with GPS alone I would definitely want to run levels to establish elevations unless my accuracy requirement for verticals was very, very crude.

If running a closed loop TS traverse I can do without the levels for most projects.

Also, one of the best things about elevations established by levelling is not the improved precision, but the simplicity and absolute reliability of the method. There is just no arguing with properly done level notes. They are simple enough that a layman (yes, and even an architect) can be made to understand. If I had to explain my elevation derivation in court I'd like my chances with levels much better than with trig or with GPS.  

Most of our data are from total stations. Our network or traverse stations are closed either in loop or from known point to known point. From this setup we already have an error of closure. Most of the time for flat terrain closure in elevation values are very good. 1-2 centimeters most of the time. For topographic mapping an error of 2 cm is negligible for design purposes. 

The purpose of the GPS check was to make sure that indeed our elevations are correct and there was no blunder when the instrument operator encoded the HI or HR. If the difference between GPS and Total Station elevations are more than 10 cm then we knopw that something somewhere is wrong.

But lately we have been using our levels less and less except for work that require 1st/2nd order level loops. The check from the GPS is reliable and sometimes goes into the mm values when we observe for more than 1 hour at 1 sec intervals.

I have read that some companies solely use their gps in rtk mode to do large scale topographic survey. RTK elevation values have an accuracy of 2cm but in reality I see on average a 5-10 cm accuracy from RTK receivers. If data from RTK data are regularly used for topographic mapping using the accuracy of the gps in rtk mode then why can't these same data be used as check on a traverse error of closure?

The big gotcha I see is you are relying on the accuracy of the geoid model in your area.  That accuracy is pretty good in some places, not as good in others, depending on the local geography.

RTK is usually tied to a base station with a known elevation that is near the job site, minimizing impacts of the geoid.

actually we are not relying on the geoid elevation. it is more of  a relative elavation difference between the traverse stations. the final msl elevation is transferred from the msl benchmark to the control network via gps observations made several times for repeatability.

GPS is measuring ellipsoidal height.  Elevations are measured from the Geoid.  In order to get between ellipsoidal height and elevation, you have to apply a geoidal separation, which typically comes from your geoid model.  That's why your accuracy can be impacted using this method.

Over many areas it probably works fine, because the geoid doesn't change much, at least as long as you don't go too far.  But it could have more impact in other areas, where the geoid can change a fair bit over relatively small distances.

Although it also depends to a fair extent on what your task is, as to how much impact this really has.  As with all this GPS stuff, the key is to have a solid understanding of how all these different components fit together, so you can make an educated assessment of the potential impacts for your desired purposes.

My 2-bit opinion on this subject is as follows:

 

IF (big IF) you can express your “elevations” ROUNDED OFF to the nearest tenth-of-a-foot (say 3 centimeters), AND you you fully understand the limitations of GPS technology in general, then you can probably get by with GPS “elevations” for many [practical] applications. Bear in mind however, that the Height Modernization protocols for deriving 2-5 centimeter heights (NGS-58 and NGS 59) are a far cry from what most surveyors seem to think constitutes a “control survey” (lower case).

 

The relative (not to mention the absolute) ACCURACY of the geoid model in your area, is (as Richard pointed out above) a serious wildcard in this whole proposition.

 

Below (if I did this right), is a “contoured” version of GEOID09 dropped over the State of Utah. Each “contour line” (isogram) represents 1 decimeter (~0.3 ft.) of geoid height change (the red contours are Meters).

Hmmm...I obviously don't know how to put an image in this post, but I uploaded a jpg (Geoid-09) into the gallery that you "should" be able to look at...

 

So in Utah, the geoidal gradient can range from (pretty much) nothing, to over 2 decimeters (0.66 ft.) in a MILE!

 

Anecdotal BS aside, I don't really expect much better than a about tenth of a foot (or two) from GPS unless the project is VERY (VERY VERY) small, and I have two-three (or more) INDEPENDENT observations on every point.

 

The “LEVEL” is still THE tool for determining Orthometric heights (Helmert Heights in NAVD88)! I would still rather have an old Ni2 and Philly Rod level run than ANY RTK/RTN/OPUS “elevation” on a Point.

 

Just MY 2-bits, your mileage may vary...

Loyal 

benc:
I have a digital level from Sokkia but sometimes the long rods are difficult to carry into rugged country. When we are doing a topo project our crews are required to close their traverse via loop  and redundunt loops if possible. Then we use our single & dual frequency gps to occupy as much of the control points that are in the open. Usually the elevations from a total station reading and gps are off by 3 cm or less and the largest is less than 10 cm. The entire GPS network is then connected to a nearby MSL benchmark & GPS control point (official) and we ttransfer the elevation & lat/long to our network. We do repeated observations to eleiminate those freak gps signals that throw off some positions.

Question is:

1. Are there other companies that do this method of elevation transfer?
2. Is this a safe or accepted method using dual frequency or RTK gps?

This method is parallel method used in bathymetry survey. In those surveys acceptable variation in RTK elevation readings are 0.10 m.




I believe what this gentleman is attempting to convey to us all is that the firm he is with uses static GPS observations, holding the relativity of the elevations only, to confirm manually collected elevation data (total station or level). The GPS data collected, I presume, is adjusted to an occupied MSL benchmark (or, again presumably) a TBM set from a BM for comparison.

I don’t believe the Geoids model comes into the equation using this method. If my understanding is incorrect please correct me.

To answer your question yes other firms use this method for comparing data and implementing quality control, however depending on the type of work you perform this may not be the best way of doing things. I believe someone else already pointed out how difficult to explain GPS observations would be in a courtroom.

I would not use RTK for any type of control work or as a check as the accuracy in RTK is just not there, and god forbid you encounter a freak multipath error, how would you be able to positively discern if the error was in the ground traverse or the GPS network without re-visiting the site? Conversley you mentioned ruggid country and RTK may just provide you with an acceptable tolerance of transfer once adjusted.

Hope this helps
.

Yes, we are only interested in "relative" elevation values between traverse. we do not use the orthometric values from the geoid because we connect to an official benchmark (MSL). the gps is used to transfer elevations from benchmark to control network.
we do it 2x or 3x to get repeatability value checks.

Double_Proportion:

I don’t believe the Geoids model comes into the equation using this method. If my understanding is incorrect please correct me.
 

Yes, the Geoid model comes into play, but it may or may not have any effect, depending on where you are, and the distance covered.  It's a "wild card", and as such, is a bit scary.

The Geoid model ALWAYS comes into play when using GPS.  You can't get an elevation in any other way.  GPS can only calculate the relative heights of points to an ellipsoid.  But elevations follow the Geoid, so you can't get an elevation difference from GPS without a Geoid model.

If the Geoid is pretty much flat over the area you're working in, then no, there would be no real effect.  But if the Geoid is undulating a fair bit, or if you try to go too far, then you'll be introducing an unknown amount of error by ignoring it.  And even when using the Geoid, you would still be introducing slight error anywhere your Geoid model diverged from the real Geoid in your area.  Again, this may be virtually no effect, or it could be a significant amount.  Or it could be introducing some error, but the amount of error is irrelevant for your purposes.

Here in the UK we're lucky - a geoid model at 1km resolution over most of the country. Recently I had a couple of opportunites to comfirm it's relative accuracy, the first was a GPS survey  around a reservoir picking up the waterline and the second was a survey of a road route along a lake shore, again picking up the waterline.
We knew the water level was constant during the surveys, variation in the levels was a max of 4cm and some of that was due to estimating a still water level for the rod point. 
In the UK your never too far from a legacy bench mark and away from mining areas they are a useful comfort check even though they haven't been maintained since the 1970's.

Yes, the Geoid model comes into play, but it may or may not have any effect, depending on where you are, and the distance covered.  It's a "wild card", and as such, is a bit scary.


I am not sure if we are on the same page of understanding here. The ellipsoid height from a gps reading is just another way of saying that the level reading is taken from MSL datum. The 2 have different values for the same point but if your base is occupying a point with known MSL value then you know the common difference between the wgs84 ellipsoidal value and a msl value. the ellipsoidal height reading that one gets for a rover is accurate when measured from the base. if the base line gets longer then just like level operation you begin to apply for curvature/refraction corrections. in the gps case you try to use more base points for each rover point or try to move your base so that your base line distances do not get too long or out of range of the radio.

from the answers that i am getting, it would seem that you are saying that the elevation values that i get from the gps is dependent on the geoid value? remember that i am using a base with known msl value. based on your responses it would seem that the entire bathymetry algorithm in most softwares is wrong! in bathymetry softwares the basic assumption is that the values from rtk/dual gps are accurate into the mm or cm level. this is the reason why the rise/fall of the boat (x,y,z) is captured from the rover position. this z value is then added/subtracted from other constants including depth readings to get the bottom seabed elevation. prior to rtk/dual gps sea water level observations are taken at a nearby port at regular intervals as the aasumption is that this is also the level of water at the boat. during this time there is always an "unknown" measurement in the rise/fall of the boat.

i am making the bathymetry analogy because this is my background. i only do topographic or boundary survey once in a blue moon. i am just transferring the technology from bathymetry survey to topography survey. in bathymetry survey the aaceptable accuracy is 0.10m because some older echo sounders can only measure up to 1/10th of a meter. but in topography instruments you do not have this limitation.

 

Ben,

 

In a nut shell...the 'N' value (difference between the ellipsoidal surface and the geoidal surface) VARIES from one point to another (except in those RARE cases when it doesn't).

 

How MUCH it varies...well...VARIES (depending on where you are).

 

The premise that the delta-height (ellipsoid) is the same “thing” as delta-height (orthometric/Elevation), is simply a false premise from the get go. Granted...in some areas, and within limited distances, this DIFFERENCE may be trivial for SOME applications, BUT it is DIFFERENT.

 

The surface of the Geoid is an undulating surface that can vary (relative to the ellipsoidal surface) quite a bit (like .2 meters in a mere mile down in Utah). The are many places on Earth where this variation is significantly LARGER (steeper geoidal gradient). The ONLY way to KNOW what the variations are in your area, is to run the calculations and see what you see. You might be QUITE surprised!

 

Loyal 

Here's a 3D image of the Geoid:

http://www.esa.int/esaCP/SEM1AK6UPLG_index_0.html

This is approximately the surface that Mean Sea Level follows.  As you can see, it looks very different from the nice, even WGS84 Ellipsoid, and diverges by quite a bit (by as much as 100 meters or so in either direction).  The bulk of Indonesia is in a relatively low-gravity area, so MSL bulges from the Ellipsoid by quite a bit through most of that area.  You're kind of in a "mountain" in the Geoid, and it looks like there's some pretty large gradiants in the Geoid in that area.

Again, though, you may or may not be introducing much error by ignoring the Geoid, depending on precisely where you are, and on how far you're going between your port base station and your rover.  The problem is that, unless you look into it, you have no idea how much error you might be introducing.  And that's the part that would give me a bit of concern.

Using a Geoid model keeps you closer, but only within the accuracy of the Geoid model.  So if you're using the latest Geoid model in your observations, you'll be more-likely to stay within your error limits.  Although to tell you the truth, I'd actually be surprised if you were ever really mapping any submarine surface to within  .1m accuracy, unless your underwater geography happens to be very flat and barren...

yes and no

My following comments are based on the knowledge of how "good" Geoid 09 is in the state I survey in. I am comfortable in believing the Geoid model is relatively accurate at the 2 cm level over 10 to 15 miles due to numerous statewide observations. I would not blindly accept the process we use in other areas of the country without testing (or hiring Loyal). We do transfer Orthometric heights to a project using GPS. These are projects where third order leveling accuracy is acceptable. We publish the project datum the same way NGS does in the OPUS solution ...NAVD88(Computed using Geoid 09). We observe published benchmarks in and around the projects site and compute the orthometric height using Geoid 09. We the difference the observed computed height of the published mark to the project marks. We apply this difference to the published height to calculate a project height. So if you do this on four published marks the first thing you can do is check the observed differences with the published differences. This will give you a sense of how "good" the Geoid is in and around the project area at the published marks. If you are lucky these results will comfort you. If you are unlucky, the results will alarm you and may force you into more observations on other published marks to prove or disprove the pulsed heights or re observations. Once you have identified the published marks you are comfortable with you transfer the observed difference from the published height to the project. So if you are "holding 4 published heights you will get 4 answers at the project. If these 4 answers have a peek to peek result height that is in your error comfort range you can get cute with a  weighted least square adjustment based on distance from the project to the published mark. The closer the published mark is to the project, the more weight it gets.

Of course each manufacturer has software that will post process for you.

And one more thing. We will only transfer heights using this method to a project every 5 miles or so. All control at smaller intervals are surveying in conventionally traversing between the primary control.

The real question isn't either/or but when to and when not to. It really means we must understand the client's project needs.

Mike Falk:
The real question isn't either/or but when to and when not to. It really means we must understand the client's project needs.

I agree.

Not only that, but it appears that the answer also changes geographically. Here it isn't too bad for reasonable distances (only a few miles to the nearest control point)

Around here for a general topo, I can get by with GPS, but for any kind of critical operation (like staking curb, sewer, or storm), I'm using the level. No to the TS also, except for checked-in side shots.

I used to do a lot of monitoring wells, and I'd run a mile or so with my NA-28 and close right on.

Comes in handy at 3 AM when you don't wake up in a sweat. You just dead know it's good.

The easy answer is "Yes", I still use my level on occasion.   They are different tools for different purposes, I still use my ballpeen hammer even though I have a claw hammer. 

I am lucky being close to sea level and many benchmarks nearby.    On occasion I will transfer a bench to a traverse point that is "GPS'able"  and then RTK to my site.  I use a base/rover unit and do not use a network, yet.  Networks are a little sparse out here, and pricey.  Of course the elevation transfer is within a tenth which is fine fro most of my work, but I'll break the level out when necessary...

William Wenzel:
Around here for a general topo, I can get by with GPS, but for any kind of critical operation (like staking curb, sewer, or storm), I'm using the level. No to the TS also, except for checked-in side shots.
 

I suppose that depends on what you have for a TS...

If you're using a decent robot, and you keep your shots short (for many instruments, within about 400 feet or so), then there's no problem staking curb or sewer with the TS.  We've even managed to prove to the FAA that we can use this method for airport runways, and the FAA has some insane tolerance limits.  Occasionally they make us shoot stuff with the digital level, too, but it's almost always within 0.01 or 0.02.  And a lot of times, you can get that much difference simply by moving the rod slightly.  Wherever four concrete panels meet, there's often 0.02' floating between the corners.  Pick a different corner, get a different elevation.

Don Poole:
I am lucky being close to sea level and many benchmarks nearby.
 

Careful... Being "close to sea level" has no bearing on the primary topic we've been discussing in this thread.  You can have large geoid gradients at any elevation, including sea level.  But having many benchmarks helps.

I thought the subject was whether we still use a level even if we have GPS....  My relation to sea level wsa in contrat to some one like Loyal who might see thousans of vertical feet difference in a day.  I rarely see 100' of vertical change anywhere unless i happen to be driving down cape and go over shoot flying hill.

Having many benchmarks and NGS stations fairly close (within a mile at least), and working in an area that is fairly low in elevation and without very much deviation leads me to think that the geoid issue is NOT that SEVERE as other places.  

my original question was something in the line of : can i use my gps to  check on my traverse elevations instead of having to run a level network.

the scenario is that the gps checks on the values from the total stations.  from the answers that i am getting some are suggesting that elevations from gps readings vary from place to place from time to time etc. so now the question is :

how do you check on your gps results if you are using rtk for a topographic survey? i was reading several months or years ago and i saw a post wherein a surveyor tied (literally) a rover rtk gps onto the back of his all terrain vehicle and roamed the area to be surveyed. are his gps data (z value) off? how will he know which data are off?

I believe soil shrink/swell issues are probably bigger issues on large topo surveys for quantities (e.g. earthwork balancing on developments) than GPS elevations. We use and trust our RTK values much more when we use a base or network with an OPUS solution.

One of the interesting things we have learned since the advent of GPS is how much things really do move. The same can be said for using Laser Trackers in industrial metrology. We are showing our clients how much their foundations/structures move and what the effects of this movement are on the operation of their production lines.

I would suggest to anyone that longer OPUS solutions on base stations are much more important than leveling or traversing to an NGS monument that has been there for 60 years.

Richard Sincovec:

William Wenzel:
Around here for a general topo, I can get by with GPS, but for any kind of critical operation (like staking curb, sewer, or storm), I'm using the level. No to the TS also, except for checked-in side shots.
 

I suppose that depends on what you have for a TS...

If you're using a decent robot, and you keep your shots short (for many instruments, within about 400 feet or so), then there's no problem staking curb or sewer with the TS.  We've even managed to prove to the FAA that we can use this method for airport runways, and the FAA has some insane tolerance limits.  Occasionally they make us shoot stuff with the digital level, too, but it's almost always within 0.01 or 0.02.  And a lot of times, you can get that much difference simply by moving the rod slightly.  Wherever four concrete panels meet, there's often 0.02' floating between the corners.  Pick a different corner, get a different elevation.

"No to the TS also, except for checked-in side shots."

I do have a decent robot, and after I check in my TS to a leveled station, I do stake curb and sewer with the TS.

I've done that for literally miles of the stuff, and I like it better than side shot level shots (which you can blow with a math mistake) and it's faster.

I actually did a side-by-side comparison with all three methods on a string of curb hubs. The results are what you'd expect: GPS with good geometry (the time I did this) averaged 0.05, TS averaged 0.01, and level was dead on and used for control.

So, I wanted my curb (and sewer) tighter than 0.05 with an occasional 0.08 or so. Therefore I use my Leica robot for speed and acceptable accuracy.

Again, I level through control to keep my net tight, and then use a TS to stake.

Don Poole:
I thought the subject was whether we still use a level even if we have GPS....  My relation to sea level wsa in contrat to some one like Loyal who might see thousans of vertical feet difference in a day.  I rarely see 100' of vertical change anywhere unless i happen to be driving down cape and go over shoot flying hill.

Having many benchmarks and NGS stations fairly close (within a mile at least), and working in an area that is fairly low in elevation and without very much deviation leads me to think that the geoid issue is NOT that SEVERE as other places.  

It sounds like you are confusing the issue we've been discussing in this thread with the "Grid to Ground Problem".  They are two unrelated issues.

This thread started when benc said he was using the difference in ellipsoid heights to transfer elevations.  This is an error in procedure.  Elevations are measured to the Geoid, not to the Ellipsoid.  So you can't transfer elevations between points if all you do is use your GPS to measure ellipsoidal height.

The distance between the Geoid and the Ellipsoid changes constantly.  In some areas, this difference changes very slightly over fairly large horizontal distances.  In other places, it changes a lot over relatively short horizontal distances.  It depends on how steep the geoid gradient is in your area.  And the steepnes of this gradient is not related to your elevation.  You can have steep geoid gradients at sea level.

Using a Geoid model in your GPS reductions helps with this problem.  And the latest Geoid09 is pretty accurate in most places, which means you can get pretty accurate elevations using your GPS, in most places.  However, your accuracy is limited by the accuracy of the Geoid model.  It might be right on.  It might be out .1 feet.  Or you might be in a "bad area", and it might be out by a fair bit more than that.  For a lot of purposes, you can use GPS and the Geoid model to transfer elevations between points, and you're perfectly fine.  But for some purposes and in some areas, you may not be.

Really, the only way to know is to use a level.  Do you NEED to use a level?  Maybe, maybe not.  Once again, it depends on your purposes.  As usual, this stuff comes down to understanding what you're doing, and using common sense.