I came to the conclusion recently that I needed to start understanding and being concerned about imagery offset. This started when Bing imagery changed and I began to notice that buildings had shifted. I also began to use other imagery such as USGS Large Scale Imagery, which also did not necessarily align with Bing imagery.

Having decided to learn about imagery offset and use it I came to discover there had been no offsets created in the areas I typically mapped. Now I needed to understand how to create offsets.

I had read discussions of using GPS traces to establish an offset based on roads. In looking at the traces that were being downloaded within JOSM those traces seemed less than ideal. This was due to randomness of lanes driven in, possibly poor sampling resolution, as well as possibly poor GPS accuracy or bounce. It was going to take a larger sample to reduce that. So I started recording and uploading traces myself.

The slowness of creating a decent sample caused me to think about how to do this quicker. I considered finding a fixed and visible feature and doing some longer GPS averages on those sites over a number of days. It then occurred to me I should just find some benchmarks or survey marks as the National Geodetic Survey (NGS) already provides the coordinates for those.

I was familiar with benchmarks from the geocaching I do. It is a subtype of geocaching, at least in the U.S. In doing a little more searching on offsets I found this was a suggested approach so I proceeded on with it. I thought I would document and share my techniques to the extent they may help someone else or could be improved by someone else’s suggestions.

The benchmark datasheets that are posted on Geocaching.com are older. The coordinates are not the full resolution that is available. Not all benchmarks are on Geocaching.com. It was just what had been recorded at the time of the one and only import. So I start my search with the National Geodetic Survey Data Explorer. I can search an area of interest and look at the datasheets for benchmarks of interest.

The first step is selecting suitable benchmarks. Some benchmarks are large structures, like water towers or broadcast towers. Although there may have been a precisely measured point on the structure, it is generally not clear where that is and it would likely be obscured by the vertical structure.

There are benchmarks that are in grass and usually easy to find in person, particularly when accompanied by a witness post. However, they are not easily seen from satellite unless the concrete or dirt around them prevents grass from overgrowing the benchmark.

I have had the most success with benchmarks that are in sidewalks. These are often at a road intersections and in an intersection of sidewalks as well. The contrast of a dark and roughly eight inch benchmark cover against a lighter sidewalk makes for relatively reliable spotting. Even so, sometimes shadows or other things reduce the contrast such that the cover is not seen.

One thing that has helped me in improving the reliability of the benchmark placement is reviewing photos of the benchmark area. I have visited a number personally and have photos that I took. Others I have been able to verify position for by reviewing photos of the benchmark posted on Geocaching.com. In one case, the spot on the satellite imagery I thought was the benchmark turned out not to be.

In then looking at the datasheets it is important to look at the type of coordinates that are available. Typically, I have found the coordinates to be described as one of those below.

• “The horizontal coordinates were established by GPS observations and adjusted by the National Geodetic Survey.” These are what I have chosen to limit my efforts to. I use the full coordinates listed to create a survey_mark in OSM to then use to create an imagery offset.
• “The horizontal coordinates were established by classical geodetic methods and adjusted by the National Geodetic Survey.” Classical geodetic methods are generally speaking triangulation measurements. I do not know what increased level accuracy came with any adjustments made by NGS. As such, I have chosen not to use these.
• “The horizontal coordinates were scaled from a topographic map and have an estimated accuracy of +/- 6 seconds.” These are not suitable for offset purposes. That 6 seconds can be over 600 feet or 180 meters off. In practice, I have found the measurements to often be 30-60 feet off.

The NGS datasheets describe the physical characteristics and location of the benchmark which may aid in finding it on satellite imagery. The PID listed in the datasheet can be used to look up the benchmark on Geocaching.com. For any survey_marks I have created in OSM I have included the PID and Designation names. I have also included the GPS coordinates in a note for future reference. Coordinates do get updated by the NGS.

I have found it a good idea to read the entire description. The benchmark may be gone. This would be noted in Station Recovery information at the bottom as can be seen in this datasheet where I reported one missing.