When we installed our Geothermal in 2007, there were two 3 ton units that were exactly the same unit. We put one in a 1/2 acre pond with a maximum depth of about 9 feet, and the other we did a ground loop about 6' down in the ground.
Now the pond loop also had a 5 ton unit on it. Our geothermal guy said the pond loop would "work just fine". The systems were in use from August on. In January, I decided to take some measurements to see how the pond vs. ground were doing. Subjectively, it seemed like it was taking longer to heat the place up.
I've attached a .pdf of the measurements. One of the things to notice is that in January the pond water came in at 40 degrees, whereas the ground water came in at 56 degrees. Now I'll be the 1st to say that this is hardly an exhaustive or scientific test. It was hard to get the room temperatures the same, this is just one set of measurements, etc. The pond had the 3 ton and 5 ton units on it, the ground just had the 3 ton. As I did the test, the room temperatures went up which is why the delta is the number to watch. Yeah, it would be better if the room temp was exactly the same but... But, both the space heated with the 5 ton on the pond and the space heated with the 3 ton on the ground were brought up to 68 degrees each morning, so this is close. Also, the 3 ton unit on the pond was the Master Bed Room, and it stayed at 62 degrees all winter long so during the day it basically never ran. So the loads on the pond and the ground were "close".
The heat pumps are two speed compressor units. I took the temperature of the air in and out of the units, and you can see the delta temperature. Of course the temperatures from the pond and the ground didn't change at all during the tests.
What struck me here is that the ground loop did significantly better on just the 1st stage of the compressor, which is where things tend to run once a room is up to temperature. The 2nd stage of the compressor had the temperatures closer, but, the ground loop still won.
I showed this to my HVAC guy, he was surprised. Turns out he's never made these measurements, the "ponds are OK" was a very subjective "no one really complained about the system." As a result of this, he took the temperature of another pond system that had a 13 acre pond. The pond was at 42 degrees (vs. my pond at 40). So the size of the pond is not a big factor in the performance of the system (well, don't use a 1/10 acre pound either).
Based on this, I put the 3 and 5 ton that were on the pond onto a ground loop system. It throws the payback out to something insane like 12 years, but, well, Ok, that's just me

.
The cost for the pond loop was $8K. The cost for the ground loop was $13K. the cost difference is 90% the cost of digging up the ground. It is clear to me that If I had done the group loop up front, the extra cost would have come more than come back to me in the roughly 8 year payback period of the system.
I'm the only customer my HVAC guy has who has a back-up loop. I left the pond pipes and pumps intact, there are valves and switches to kick them in.
Enclosed also are pix of the pond loops before they were sunk, and the ground system before it was covered up. It takes about 18-24 months for those 6' deep trenches to settle out completely, so keep that in mind when doing any final landscaping.
Bottom line: If you can swing the cost, avoid the pond and go with the horizontal ground loops.
Sorry for long post, lots of data...
Pete