The wells – which had helped fuel the growth of South Florida by eliminating the need for expensive water treatment plants -- had passed rigorous EPA and state evaluation throughout the 1980s and 1990s. Inspections showed they were structurally sound. As Class 1 wells, they were subject to some of the most frequent tests and closest scrutiny.
Yet they failed.
The wells' designers would have calculated what is typically called the "zone of influence" — the space that waste injected into the wells was expected to fill. This was based on estimates of how much fluid would be injected and under what pressure.
In drawings, the zone of influence typically looks like a Hershey's kiss, an evenly dispersed plume spreading in a predictable circular fashion away from the bottom of the well. Above the zone, most drawings depict uniform formations of rock not unlike a layer cake.
Based on modeling and analysis by some of the most sophisticated engineering consultants in the country, Florida officials, with the EPA's assent, concluded that waste injected into the Miami-area wells would be forever trapped far below the South Florida peninsula.
"All of the modeling indicated that the injectate would be confined in the injection zone," an EPA spokesperson wrote to ProPublica in a statement.
But as Miami poured nearly half a billion gallons of partly treated sewage into the ground each day from the late 1980s through the mid 1990s, hydrogeologists learned that the earth – and the flow of fluids through it – wasn't as uniform as the models depicted. Florida's injection wells, for example, had been drilled into rock that was far more porous and fractured than scientists previously understood.
"Geology is never what you think it is," said Ronald Reese, a geologist with the United States Geological Survey in Florida who has studied the well failures there. "There are always surprises."
Other gaps have emerged between theories of how underground injection should work and how it actually does. Rock layers aren't always neatly stacked as they appear in engineers' sketches. They often fold and twist over on themselves. Waste injected into such formations is more likely to spread in lopsided, unpredictable ways than in a uniform cone. It is also likely to channel through spaces in the rock as pressure forces it along the weakest lines.
Petroleum engineers in Texas have found that when they pump fluid into one end of an oil reservoir to push oil out the other, the injected fluid sometimes flows around the reservoir, completely missing the targeted zone.
"People are still surprised at the route that the injectate is taking or the bypassing that can happen," said Jean-Philippe Nicot, a research scientist at the University of Texas' Bureau of Economic Geology.
Conventional wisdom says fluids injected underground should spread at a rate of several inches or less each year, and go only as far as they are pushed by the pressure inside the well. In some instances, however, fluids have travelled faster and farther than researchers thought possible.
In a 2000 case that wasn't caused by injection but brought important lessons about how fluids could move underground, hydrogeologists concluded that bacteria-polluted water migrated horizontally underground for several thousand feet in just 26 hours, contaminating a drinking water well in Walkerton, Ontario, and sickening thousands of residents. The fluids travelled 80 times as fast as the standard software model predicted was possible.
According to the model, vertical movement of underground fluids shouldn't be possible at all, or should happen over what scientists call "geologic time": thousands of years or longer. Yet a 2011 study in Wisconsin found that human viruses had managed to infiltrate deep aquifers, probably moving downward through layers believed to be a permanent seal.