4 ms·

> increased variance is very much in line with systemic climate change Not really, it's also in line with the limited amount of high-quality data we have. E.g

by tomp 6y ago

> increased variance is very much in line with systemic climate change Not really, it's also in line with the limited amount of high-quality data we have. E.g. viewed in isolation, recent Gulf of Mexico hurricanes were pretty bad (2019 in particular, IIRC). But if you look into history, 1932,1933 was also pretty bad. Arguably, if we had even more recorded history, such outliers wouldn't seem that exceptional - there still is a possiblity that climate change is affecting hurricanes (making them worse) but you can't just look at the variance (outliers) - trend is also (more?) important. https://en.wikipedia.org/wiki/Atlantic_hurricane_season https://en.wikipedia.org/wiki/Atlantic_hurricane_season

titzer 6y ago
Let's use math and statistics, not anecdotes. If an increase in variance means that you get a +5 outlier 10x more often, then looking at a 10x longer timescale will mean that you see the +5 outlier roughly as often, making it seem "less exceptional". The problem is that the exceptional thing is now happening 10x more often. For example, Texas getting three "500 year" floods in 3 years. These were floods that were estimated to be a 0.2% probability in any given year. The fact that they are happening at much higher frequency is because the entire probability distribution has shifted (and expanded). Now if you improved your ability to peer into the past and found another "500 year flood" a few hundred years back, you'd be tempted to think everything is fine, when everything is decidedly not fine.
- koheripbal 6y ago
  You cannot take prove a global phenomenon with local data, unless you compare global occurrences of local events. Taking only Texas as an "example" means you're cherry picking data. I'm sure we could find other parts of the world that had multiple rare events in every decade of the last century as well.
  - titzer 6y ago
    It was just an example. If you look at all the data together, the trend is clear. Both median temperatures and variance are increasing. Same for other extreme weather events like hurricanes, wildfires, and droughts.
    - koheripbal 6y ago
      Is there a data source that says that temperature variances have increased? I have never seen meaningful evidence of this. > Same for other extreme weather events like hurricanes, wildfires, and droughts. Extreme weather events are extremely hard to define, and have such a low occurrence rate that you cannot extrapolate an increase in the timeframe we're discussing on a global level.
craftinator 6y ago
> Arguably, if we had even more recorded history, such outliers wouldn't seem that exceptional This holds true if you assume a system is static, which is one of the hardest communication points the scientific community has to convey to the general public. A good part of climate science is not just mapping the statistics of a system, but analyzing those in terms of physics and chemistry to figure out what causes dynamism in the climate system. The driving mechanic behind climate change theories is that we are changing the chemistry of the atmosphere, which is making the climate unstable. If you just look at the statistics through the lens of data science, it's easy to write off effects as "not enough data", but this is a simplistic view, stemming from not spending huge amounts of time learning and studying climate science.
- koheripbal 6y ago
  > The driving mechanic behind climate change theories is that we are changing the chemistry of the atmosphere, which is making the climate unstable. This is a strange characterization of the increase in CO2 and resultant heat retention. It's not itself "unstable". > If you just look at the statistics through the lens of data science, it's easy to write off effects as "not enough data", but this is a simplistic view, Selecting only a specific set of events in specific areas is, by definition, cherry picking data.
  - titzer 6y ago
    > Selecting only a specific set of events in specific areas is, by definition, cherry picking data. You used the same phrase elsewhere in the thread. I am not sure if we agree on what "cherry picking" really means. There is a difference between isolating relevant variables and cherry picking data. For example, if you wanted to track the long term trend of lion weight in Africa, you wouldn't also include in your dataset the weights of monkeys in South America. That'd just add irrelevant noise because, even though they are measurements of animal weight, they are not what we are studying. It absolutely would not be "cherry picking" to exclude monkeys from your sample. Real cherry picking would be things like excluding particular years of measurements or finding other reasons to exclude data that goes against a trend you are looking for.
  - 6y ago

by tomp 6y ago

titzer 6y ago

Let's use math and statistics, not anecdotes. If an increase in variance means that you get a +5 outlier 10x more often, then looking at a 10x longer timescale will mean that you see the +5 outlier roughly as often, making it seem "less exceptional". The problem is that the exceptional thing is now happening 10x more often. For example, Texas getting three "500 year" floods in 3 years. These were floods that were estimated to be a 0.2% probability in any given year. The fact that they are happening at much higher frequency is because the entire probability distribution has shifted (and expanded). Now if you improved your ability to peer into the past and found another "500 year flood" a few hundred years back, you'd be tempted to think everything is fine, when everything is decidedly not fine.

koheripbal 6y ago
You cannot take prove a global phenomenon with local data, unless you compare global occurrences of local events. Taking only Texas as an "example" means you're cherry picking data. I'm sure we could find other parts of the world that had multiple rare events in every decade of the last century as well.
- titzer 6y ago
  It was just an example. If you look at all the data together, the trend is clear. Both median temperatures and variance are increasing. Same for other extreme weather events like hurricanes, wildfires, and droughts.
  - koheripbal 6y ago
    Is there a data source that says that temperature variances have increased? I have never seen meaningful evidence of this. > Same for other extreme weather events like hurricanes, wildfires, and droughts. Extreme weather events are extremely hard to define, and have such a low occurrence rate that you cannot extrapolate an increase in the timeframe we're discussing on a global level.

craftinator 6y ago

> Arguably, if we had even more recorded history, such outliers wouldn't seem that exceptional This holds true if you assume a system is static, which is one of the hardest communication points the scientific community has to convey to the general public. A good part of climate science is not just mapping the statistics of a system, but analyzing those in terms of physics and chemistry to figure out what causes dynamism in the climate system. The driving mechanic behind climate change theories is that we are changing the chemistry of the atmosphere, which is making the climate unstable. If you just look at the statistics through the lens of data science, it's easy to write off effects as "not enough data", but this is a simplistic view, stemming from not spending huge amounts of time learning and studying climate science.

koheripbal 6y ago
> The driving mechanic behind climate change theories is that we are changing the chemistry of the atmosphere, which is making the climate unstable. This is a strange characterization of the increase in CO2 and resultant heat retention. It's not itself "unstable". > If you just look at the statistics through the lens of data science, it's easy to write off effects as "not enough data", but this is a simplistic view, Selecting only a specific set of events in specific areas is, by definition, cherry picking data.
- titzer 6y ago
  > Selecting only a specific set of events in specific areas is, by definition, cherry picking data. You used the same phrase elsewhere in the thread. I am not sure if we agree on what "cherry picking" really means. There is a difference between isolating relevant variables and cherry picking data. For example, if you wanted to track the long term trend of lion weight in Africa, you wouldn't also include in your dataset the weights of monkeys in South America. That'd just add irrelevant noise because, even though they are measurements of animal weight, they are not what we are studying. It absolutely would not be "cherry picking" to exclude monkeys from your sample. Real cherry picking would be things like excluding particular years of measurements or finding other reasons to exclude data that goes against a trend you are looking for.
- 6y ago