Estimating homelessness via too-clever thinking and long train rides

Grar. Too much to do and think about and too much too much too much. So here's an idea I had on the train this weekend...

While riding the train, sometimes I play the game 'where would I sleep if I were homeless?' I've recently started playing a variant called 'How many of those spots are already taken?'

The resulting number is far too high. I remember when I took the train into the Bay Area, starting about two hours north of the Oakland stop I didn't see a SINGLE PLACE that looked like a good sheltering area that did not have signs of being currently or formerly inhabited. Riding the train into Portland, the line at which the density of good shelters vs. occupied shelters ratio gets bad seems to be about 30 minutes away from the station.

Since this is "distance outside of town", and area covered goes up with the square of distance, then since the bay area became bad 4x farther out of town it follows that they have 16x the number of homeless people (all assuming a uniform distribution of homeless people of course - maybe not a good assumption).

The greater Portland area has 1.5 million (-ish) people. Does the bay area have 24 million people? Answering this question should help determine whether homelessness per capita is higher around SF bay or PDX. Both of them seem like pretty good places to be homeless, so I would expect them both to be higher than, say, Fairbanks.