My county data is drawn from this map on the New York Times. I recommend you have it open as I will be referring to geographical information it presents.
The first thing to note about the map is that Hillary dominated the state. Her areas of weakness were highly concentrated in the northern and panhandle areas, where Florida stops being an extension of New York and starts being part of the old south. There were only three counties where she did not get at least 30% of the vote: Gadsen, Baker and Washington. She was second in all three of those counties. In Gadsen, Obama dominated with 55% of the vote. That was the only county where he received more than 50%. This contrasts with South Carolina, where he had a number of wins with 60+% of the vote. In Baker and Washington counties, HRC was defeated by Edwards. She lost more counties to Edwards than to Obama. The only county where she finished third (and then only by 70 or so votes), Edwards was the winner.
This points to the phenomenon that where Edwards did well, he was highly competitive with HRC. There was only one, very small county where an Edwards win had Obama come in second. Where Edwards did well, so did Hillary, which could indicate that on a political spectrum, more Edwards supporters are going to fall into the HRC camp than into the Obama camp. Edwards did best where Obama polled badly. Four of the six counties where Obama won gave more total votes to HRC and Edwards than to Obama. In three of those counties, had Edwards only polled his usual amount (14%), Hillary would very likley have won.
Where Hillary's margins were thinnest and Obama strongest, it was accompanied by a strong Edwards vote. If Hillary was under 45%, you usually found Edwards above 20%, which then created an opening for Obama if the other two could cancel each other out.
HRC maintained both the highest percentage of votes and the least variation in the percentage she took:
Her highest percentage was 61%, her lowest was 26%, more than double the poorest showing of the other two, the range between high and low was smaller and the median percentage was 42%. Edwards and Obama had the same range, though Edwards' median was 6 points lower.
In the next table, what I have done is look at the counties where the candidate won (horizontal row) and recorded the average % of votes in winning counties. I then looked at the percentages earned by the other two in relation to the winning amounts (vertical columns). Thus in the first row, when HRC won, she averaged 53% of the vote, with BHO pulling 31% in those counties and JRE getting 13%:
What we can see from this is that changes in preference between Obama and Edwards come at HRC's expense. She goes down to 32-33% (still much higher of a base than the other two) and the bulk of that defection goes to one or the other candidate. They do not benefit equally at the same time. Obama gets a little more of the HRC defection in Edwards strongholds than Edwards gets in Obama strongholds, but they don't share the spoils. I estimate that more Edwards voters will go to HRC than to Obama, though it will not be 100%. The crucial measure here is that 10 to 15 times the vote volume was to be found in HRC strongholds vs. Edwards or Obama. They may have some nice percentages, but she overwhelmed them with actual votes.
As an anecdote, at work today I had a colleague come over to my desk and ask me who she should vote for now that Edwards, her first choice, was out. I did not say a single negative thing about Obama except that I did not think he had enough political experience at a national level given the work that lies ahead to clean up the mess left behind by Bush. I focused on the ways in which HRC and Edwards shared many of the same policy perspectives and goals. She admitted afterwards that she really always preferred Hillary, but had not believed HRC was electable - until Florida.
And that is why Obama supporters are so frantic to say Florida is nothing. Democrats who have supported Edwards until now because of the electability factor may make a significantly different calculation next Tuesday.