I blogged about this book a few weeks ago, when Hurricane Ike was headed toward Galveston, and speculated it would be a good read. It was.
But one thing bothered me about it (and I've come across it at least once in another Erik Larson book, "Devil in the White City."). He is meticulous about his research; but he also will write as fact something that he has deduced from evidence uncovered during his research.
For example, a guy walking through Galveston smells manure and sawdust. There's no record that he did, no letter to a relative talking about it, or journal entry, or whatever. But Larson knows there was a block-long stable, and construction workers sawing wood to build homes nearby, thus ... the man smelled manure and sawdust.
In another case, he has his main character reading a particular story in the local paper. Again, no direct evidence that he read it. But: the story was the most prominent in the paper that day, and it was documented that this guy read the paper every day, so ... he read it.
What do you think -- is that OK in a historical non-fiction narrative? I will (possibly) influence any discussion by saying: I'd rather he establish the facts (stable existed, construction work existed) and say something like, "He must have smelled manure and sawdust." To me, you are being honest and saying, I can't possibly know this for sure, but it had to be the case.
Steven Johnson, in The Ghost Map, takes that idea a step further in a couple places in that book by saying something like, 'There's no way to know exactly what he saw walking down that street. But a block away there would have been a building ...' He puts you in the scene and suggests what may have happened while admitting he can't know for sure.