Should be "E.coli at worst-ever levels in Toronto Harbour and Lakeshore". It's an issue, but the entire lake is not teeming with unsafe amounts of E.Coli. In the summer, there are updating signs about water quality on every beach.
Just the parts where the most people live and want to swim. Sad state of affairs when the land we value most [both
fundamentally (water is life) and culturally (waterfront land most expensive)] is toxic to humans.
Toronto beaches are way too cold for most people to swim at. They're mostly for sun tanning.
I occasionally practice snorkeling at them over the summer. The beaches are usually full, and there are about 3 other people who actually submerge above their waists.
I attended a few meetings in which the city of chicago is attempting to predict the E.coli levels in the water for the next day using data science techniques.
It was a frustrating experience since the current day's E.coli level measurement can vary wildly depending on various factors such as time of day, water depth at site of measurement, month in year, weather, etc...
or maybe sometimes it's just a bunch of birds that all pooped in the same area on a given day.
Plus, each beach has it's own pattern and correlations; some folks started adding rules
the USGS has some method of averaging readings, but I was very tempted to try to find the ideal model that could handle the non-iid beach data. someone on the net suggested multilevel hierarchical modelling, which i don'f fully understand.
i think part of the issue is that data is already semi-structured, as well as time based. one approach is to flatten the structure (assume iid) and make the data stationary (remove the time component).
but i wonder if something like an RNN would handle the multiple time steps better.
I swim in Lake Ontario pretty regularly. Think I went until late November this year. They'll generally put up advisories at the beaches when there is an uptick, so if you're in the GTA / Niagara region don't keep this from letting you get out there.
Yeah I generally swim at Hanlan's Point. The beach is pretty nice and if you're ok with some butts the lack of children or large crowds is kinda nice. Though during pride its pretty packed.
13 comments
[ 0.26 ms ] story [ 42.6 ms ] threadJust the parts where the most people live and want to swim. Sad state of affairs when the land we value most [both fundamentally (water is life) and culturally (waterfront land most expensive)] is toxic to humans.
I occasionally practice snorkeling at them over the summer. The beaches are usually full, and there are about 3 other people who actually submerge above their waists.
Lake Ontario Waterkeeper was only founded in 2001.
I wouldn't worry until there's official confirmation and we're close to swim season.
Yes, this is the problem. Anyone in the GTA can tell you about the wonders of our Combined Sewer systems:
https://en.wikipedia.org/wiki/Combined_sewer
The storm drain should have the overflow-to-river mechanism before it's combined with sewage?
It was a frustrating experience since the current day's E.coli level measurement can vary wildly depending on various factors such as time of day, water depth at site of measurement, month in year, weather, etc...
or maybe sometimes it's just a bunch of birds that all pooped in the same area on a given day.
Plus, each beach has it's own pattern and correlations; some folks started adding rules
the USGS has some method of averaging readings, but I was very tempted to try to find the ideal model that could handle the non-iid beach data. someone on the net suggested multilevel hierarchical modelling, which i don'f fully understand.
i think part of the issue is that data is already semi-structured, as well as time based. one approach is to flatten the structure (assume iid) and make the data stationary (remove the time component).
but i wonder if something like an RNN would handle the multiple time steps better.
It's like a few hundred feet in some cases and a world of difference.