It is important to compare different strands of happiness research to figure out if they are measuring the same thing. Kahneman's talk was about how two fairly similar measures - both self-report - yield wildly different answers. Incapacitating depression and especially suicide are another rather different happiness measure. They are very coarse measures, but suicide is fairly objective and thus easy to compare across cultures. It has a clear polar trend, though not within the US. (finer world maps requested! also, cartograms - on the last map is SF deadly?)
We should compare different measures. This failure to match is a big red flag. Kahneman only mentioned weather data incidentally, but it's important. Of course, there are many possibilities, such as SAD being a threshold effect - such a small part of the population should not affect polling. Cross-culturally, what are the self-reports of people with incapacitating depression or suicidal ideation?
People who have had a painful experience remember it as less painful if the pain tapers off, rather than cutting off sharply at the height of intensity, even if they experience more pain overall. I'd heard of this finding before (from Dan Ariely), but Kahneman uses the finding to throw the idea of "experiencing self" vs. "remembering self" into sharp relief. He then discusses the far-reaching implications of this dichotomy and our blindness to it.
The talk is entitled "The riddle of experience vs. memory".