localization residuals

> What is a localization residual?

I expect it's the difference between the calculated value and the record value for each of the localization points. If so, then yes, smaller is better.

> Silly question.
>
> What is a localization residual? I'm working on setting up a localization right now and am somewhat clueless as to what that means.

From the Carlson manual:

When localizing, it is advisable to use at least three points for
horizontal control and four points for vertical control, in order to get a measure of
residuals (or accuracy). The program will “best fit” a plane through all activated
(H=On and/or V=On) control points. The residuals are how much each activated
point is off of the plane surface. Because multiple elevation points may create a
slightly tilted plane, some surveyors will verify that the vertical control has low
residuals and is accurate using multiple vertical control points, then turn off all
but one (V=Off) and use only the nearest vertical (elevation) as they progress
through the job.

(Page 171 of V3 manual)

This is essentially my best guess as well. Glad someone might be thinking the same thing.

Tomorrow I'm going to run back out there and store a point on the monument and inverse between the recorded value and what I'm staking and seeing if the difference is the same number as the "residual"

[sarcasm]..Manual.. We don't need no stinkin manual, WE'RE MEN![/sarcasm]

Localization residuals are the differences between the record value you provide and the computed values using the localization parameters. They are intended to quantify distortion, not a relationship with reality.
I would urge you to read up on how your software 'treats' data in localizations. Of particular note would be scale and plane 'tilting'. I have seen software assign a slope north of 1 foot in 4 without raising any alarms...

The program will “best fit” a plane through all activated
(H=On and/or V=On) control points. The residuals are how much each activated
point is off of the plane surface. Because multiple elevation points may create a
slightly tilted plane, some surveyors will verify that the vertical control has low
residuals and is accurate using multiple vertical control points, then turn off all
but one (V=Off) and use only the nearest vertical (elevation) as they progress
through the job.

YIKES, too bad.

Trimble software has the option to NEVER TILT THE VERTICAL PLANE, it will best fit it.

"YIKES, too bad."

Please elaborate...

I'm not sure what he means, but I will say that tilting the plane has consequences. Of course not tilting it doesn't work at all.

I don't recommend calibrating, or localizing, but that's just me.

And if for some reason I'm forced into it I will not use control points to tilt the plane, only a geoid model for me.

It's much better to project your coordinates and apply the model to an elevation point;-)

> I'm not sure what he means, but I will say that tilting the plane has consequences. Of course not tilting it doesn't work at all.
>
> I don't recommend calibrating, or localizing, but that's just me.
>
> And if for some reason I'm forced into it I will not use control points to tilt the plane, only a geoid model for me.
>
> It's much better to project your coordinates and apply the model to an elevation point;-)

In Carlson, you can also use a GEOID model, and assign a single point of elevation as the calibration.

That is better, best is not to calibrate either the horizontal or the vertical, however I know it's done all the time, some have told me they calibrate each day on the same job:-(

I looked into how to get a geoid file into carlson. looks pretty easy.

I kind of don't understand the full implications of what a geoid file will do.
I guess its essentially a ground correction right?

so if I go get a GSF file loaded in there I would only need a single vertical point to be localized?

I since opening this thread have made a accurate working localization for the job.
(my adverage residual is about.18 ft but everything is checking in great)

What will happen if I continue using this localization with a geoid file?

Thanks for your help.

Maybe it would help if you knew just what kinda coordinates you are dealing with a number like 18 doesn't sound good.

Are you tying into state plane coordinates, some type of ground coordinates, some combo, do you have any metadata at all.

ALL localization software should sound an alarm, when certain perimeters are exceeded, especially inclined planes.
N

0.18 ft not 18 ft. haha, thats a big difference. sorry

NAD27 on Nevada state plane West.

.18 would be too much for me also, I would project it to state plane, hold a point then check the results that way first.

If that doesn't work, then I would consider calibration, but if there are points that far out, I wouldn't use them, that is too far........

Could very well be that the control is kinda junk, putting a couple of tenths into a calibration would be iffy........

Imagine what's it's doing: it's stretching and fitting one set of numbers to another, usually you would like to see under .05' when doing this.

I think running the numbers raw, then comparing it all would be the first step.

Now the question is why NAD27? Is the control really old stuff from instrument surveys?

To anyone who cares, I got this expanded explanation from tech support at carlson.
essentially I think what I'm learning here is that the residual is the amount its correcting to the ground, so if you know the monument that your localizing to is good then you should use it if the residual is big or small.

at some point I would think though that if the residual is a pretty substantual measurement then you would want to do some double checking to make sure everything is correct.

Thanks to everyone for your input.

"In Carlson SurvCE or SurvPC under Equip / Localization in the Points tab the Hres and Vres values are shown to represent the difference in the Calculated Solution (Horizontal scale factor and Vertical tilted plane) compared to your Control Point coordinates.

It takes the raw GPS latitude and longitude measurements typically taken from 4 to 6 points around the outside edge of the jobsite and converts them into a Nad83 Grid Coordinate.

Then it uses a rubber sheet or least squares formula to create a horizontal scale factor and tilted Vertical plane to fit your highly accurate Control (0.01 feet or tighter) to your RTK GPS Nad83 grid coordinates.

Essentially if you use all the points the Hres and Vres for each point is the difference you will see in Survey / Stake Points with the "Calculated Solution" compared to the physical control point coordinates.

We always recommend staking out your 4 to 6 control points after you complete the Localization routine in the Points tab to see the results from it.

Here is a more detailed description from our Development Team on how our Carlson SurvCE localization routine works:

First the GPS latitude, longitude and ellipse are projected to cartesian coordinates of x,y,z such as a Nad83 or UTM state plane projection.

Then the localization is done in 2 parts: x/y horizontal and z elevation.

For the x/y horizontal solution, the plane similarity applies a transformation of translate, rotate and scale. The best-fit transformation parameters are calculated by least-squares. No single control point is held. They are all weighted equally. Two control points are enough to define the transformation. If you use two, then the residuals are zero. If you use more than two, then all control points will have a bit of residual.

For the z, the elevation difference is modeled at each of the control points and a best-fit elevation difference plane is calculated. If only 1 control point is used, then this plane will be level (ie. delta z of 5.0 feet everywhere). If 2 or more control points are used, then the plane will be tilted. Three points define a plane. So the program can calculate a plane that exactly matches the control points when 3 or fewer are used and the residuals will be zero. When more than 3 points are used, it makes a best-fit plane and each control point will have some residual. No single control point is held in this case."

Its an old freeway project. NAD27 is what they were using when the first phase of the project was started and its what all the control is for the job.

Every other job I do is in NAD83 normally with way newer control.

The localization I have now ties in with exsisting construction from the last job great (less than .1 in most cases), so even if I'm not doing it right, I guess I'm doing it simular to the way the last dummy did 🙂

The control is only going to be so good on a job like this. Old stuff.

so... what should I do with this geoid? should I utilize it?

"I think running the numbers raw, then comparing it all would be the first step."

you mean taking a shot with no localization on the control and comparing it to what its supposed to be?

thats essentially what the residual means, I tried doing that and got the same number.

since it's NAD27, I would set up a state plane projection in NAD27, then go out and locate all the control.

I would do that static, using as many units as I have access to, then I would compare the located state plane numbers with the given ones. I wouldn't expect the raw numbers to be very close but I would look at the inverse distances and interior angles formed by the control. I would figure out what was good and what was not so good in the control. Then I would do a calibration after figuring out what to hold.

The raw state plane distances and interior angles should match pretty good, not the coordinates or azimuths.

There is probably a rotation but the calibration should take care of that. For verticals I would locate everything with a GEOID applied and compare the differences in the elevations. With a GEOID they should be off close to the same amount at each point, if you are .25' low, .22' low, .21' low, .24' low it's telling you that the GEOID model is adjusting the slope very well.

Then I would apply the GEOID to one of the points that match vertically (holding the elevation, not the ellipsoid height) in the middle and horizontally well.

It's probably NAD27 control measured with instruments, and adjusted with a compass rule adjustment. You really are stuck doing a localization, but do it with as much raw info as you can measure.