What does a basic DGPS set cost for improving accuracy, especially next to tall buildings?

DGPS works by taking a reference station of known position that constantly recieves data from satalites, and uses it to work out the details of any erros over a wide area, these erros are usually caused by slight mis positioning of a satalite, or slight mis timings in it signals. This data is then sent to your GPS set so it can compensate.

I have found from personal experiance that GPS can be quite inacurate when next to large metal structures, this seems to be due to signals reflecting off them, and them blocking parts of the signal. DGPS is unlikelly to do very much for this as it's a local problem not a wide area problem.

It's possible that a better GPS set with a better antenna could help as lower signal intensities would be picked up better (from shielding) and dirty signals would be sorted out better (from reflections). I have however only played with 2 GPS sets, and I can say that a SIRF3 powered BGT31 is reasonably resiliant, but still has issues at times. I've never really tried it around proper highrise jungles though.

The only other thought that I have is that giving the set a few minutes to get it's bearings could help, it's hardly practical in many instances but it could improve things in some situations.

DGPS was developed when the GPS network had Selective availability witch made the accuracy less then 100m. DGPS could reduce this to 10-5m by using gps observatories to correct the error in the signal. After 2000 SA is turned of which makes accuracy 20-10m without DGPS.

DGPS are quite cheap. You probably have DGPS in your standard GPS unit, but might not be included in cheap chips often found in cell phones.

But you will probably not be able to use DGPS much in Ethiopia because DGPS depends on ground stations transmitting the GPS errors. Most of the transmitters are placed at the shores as DGPS was most used by ships that needed better accuracy then GPS with SA to safely navigate.

You could try to make your own GPS with DGPS with sources from the internet or a control GPS at a stationary position, but the gain would probably not be worth it.

Would it make sense to purchase a DGPS system?

It helps to reduce the inaccuracy that is caused by the "weather" in the ionosphere. It does not help at all to reduce the inaccuracy caused by high buildings nearby. I agree that a better antenna might help.

How does that work?

You find a good article about DGPS on wikipedia.

What would it cost?

DGPS devices are from 0 € (because included in mobile phone already) up to thousands of €. The price depends on the accuracy you want to reach and the system by which you want to receive the DGPS correction data for your area.

How to import data to JOSM?

Just the same like without DGPS.

If a building as good straight sides you could produce some points well in the open but on the same sight line do this for all sides and you could draw the box. cheap but time consuming.

If these points are averaged (garmin vistas can take a number of readings and average a mark/waypoint) this would improve accuracy a bit more. If you don't have this facility on your GPS you could walk towards the building while staying on this sight line. Then if you draw a way through the average of the trace to and past the building this way can be edited two form the wall of the building. hope you get the idea