Radial Curves

You can create tangent curves for the whole length. Get your plumb bob and a 100 foot rag tape then go out with your plumb bob and locate the PI's for each curve then measure the apparent Tangent length to the PC and PT and record the average for each curve, then tie the PI's and as a check, measure from the PI to C/L, "external" and record it. You can locate and define compound curves, reverse curves and curves with a very short tangent between them this way. Once located and tied, compute the curves using the TL, PI and the Curve angle which you have from the PI ties. You can get very close this way, it is quick and you have checks. Took me years to figure out the best way to get this job done using a method that no one will find problems with the results of and provides a good C/L description of the as built C/L. You can do it the hard way or give this a try, you can get a very close match to existing curve data if that info is available, try one as a check on what I have recommended.
jud

I try and locate the PC's and PT's in the field (these locations will "float" some from the final solution) and also collect enough info to "create" the curves (externals, etc.). I then (in CAD) create lines and tangent curves from that info that fall close to the physical centerline. Try not to make non-tangent curves if at all possible, you may have to make some relatively short lines between the curves.
My pet peave is bearings to the second and distance to the hundredth for field located centerlines. Round off the bearings and try to make the distances a nice round number, just remember, in most cases you aren't building a piano.

I hold the PI's as I tied them, they were located by eyeball sighting both ways along a plumb bob string and the intersection then marked on the ground to tie later. Tieing in the apparent PC's and PT's, then letting them float to locate the PI was how I used to do it, I get better results holding the PI and letting the PC's and PT's be adjusted if needed. For about 90 percent, the data collected as noted is used as is. The remainder sometimes need tweaking a bit, usually overlapping tangent lengths, sometimes I will need to add another PI to the mix to keep C/L close to the computed data, usually a compound curve where the radius needs adjusting as part of the mix, that is where a workable plot helps.
jud

I like to locate the roadway in the field, and in my cad intersect the straight stretches to make a PI. I come back the tangent lengths on both directions and experiment around averaging the two tangents. If that makes a "close enough" curve, that contains the roadway I'm pretty happy. (say I get a 50' from pi to where the road curves at the back-tangent, and 73' to the forward tangent I just average them for about a 61.5' tangent. delta is the change in the tangents, so I calculate the radius based on using 61.5 and the delta. If it's a 20º delta, I would calculate 348.8 Radius. I would then use 350'radius, hold the delta, and calculate a new tangent length). I don't think it's necessary to do all that and get an even radius, but I like to do that. 61.5' tangent might not be close enough in a 350' radius curve, and I might have to do something else.

I like to round the radius to 10' or even 50' intervals if I can, and hold tangency if at all possible. Sometimes none of the above works. I just try to keep it relative simple. A few feet in a prescriptive right is not that wide of an error. The road where some cars or goats wandered, changed daily. Paving on the road location changed every time they overlaid. exactly where the top layer of asphalt ends, and where the underlying asphalt ends varies. A close approximation in this kind of application is adequate to me.

Edit
Oops, I reread your question. When I can't get anything to work I will do different things for different scenarios depending on how bad I fit. I would rather have nontangent curves or more than one curve rather than short tangents. But if I saw the road and thought the short straight tangents were the best representation on what it looked like, I'll do that.

We use a similar approach to Jud to make all our curves tangent.

I do a lot of these.

It's suprising to find that almost everytime there are curves that "fit" well to a driven roadway. I drive the road and locate it as often as the tightest curve needs. Then import the points into autocad with a circle node that is twice the size of what I accept for a maxium distance from centerline. Then design tangents and tangent curves to create a centerline that will pass through the points or touch the nodes.
Sometimes you will end up with a boatload of courses; but, if that's what the road is doing you're just stuck with it.

I always use the degree of curve definition, I just can't help myself.

I just finished one today of about 3/4 of a mile. 28 tangent lines and 7 curves. It was just for a pasture access and my road wouldn't hold up as a subidivision road which is what the grantor and grantee wanted.

When working with a 12ft roadway that is to fit into a 25ft wide R/W, I just come up with a situation that the road will fit inside of my plan. It may meander back and forth from the R/W, I am happy to keep it inside.

I will tie the actual road in and get every bend and do a best fit curve and adjust the PC and PT to have tangent curves.

If it is not a crucial thing, I don't have any curves at all. Simply all PIs for the entire easement.

Another field expedient method to check curvature and radu=ius for your design is to measure a 62 foot chord along centerline in the curve. Then measure perpendicular from the midpoint of the chord (at 31 feet) to centerline. This distance in INCHES is the degree of curve.

I like the rodman to locate their estimates of centerline PC's POCs, PTs , POT's and also PRC's and PCC's if necessary. Makes calcing curve data a lot easier in the office.

Plus Or Minus 0.5' Is Close Enough

It can easily move that and more on each repaving.

Paul in PA

I have always field located PC's, PT's, POC's and tangents. Sometimes for non-crucial roadways (drives, gravel roads, etc), I will field shoot a PI and measure an external. I then connect the tangents to create a PI. I usually have 2 or 3 POC's for each curve. I then hold the delta and vary the radius to do a best fit between the POC's. I then usually round the radius and go with it. Tangent bearings are almost always rounded to the nearest minute. I hate seeing seconds on a road or R/W tangent.

Perry,
we have done similar in field using total station as to locating the PC's, PT's and enough POL's and POC's to define the tangents and curves. Have not tried to get off roadway to locate PI's - seemes to take too much time and often there are too many obstructions (buildings, trees, hills). fit it all together in office and I agree with the post about rounding off stations and deltas as much as possible to nearest whole number.
With RTK we can drive the roadway and locate the approx. centerline of the road at 5' intervals and bring that info into the office and best-fit a centerline from that.