My company came up with this data for the Wall Street Journal, so I should explain that the title here is a bit of an over-simplification.
The main point is that if you sort in ascending order on e.g. the "Average number of degrees temperature deviates from 60-80 range in last five years" column you should have a good indication of how the cities compare in terms of the time and extent to which their air temperature deviates from a comfortable range.
But if you're interested in more detail, the figures in that column are actually derived from:
- The heating degree days (HDD) with a base temperature of 60 F summed for each day over the last 5 years (measuring the time and extent to which the temperature dropped below 60 F in that 5-year period).
- The cooling degree days (CDD) with a base temperature of 80 F summed for each day over the last 5 years (measuring the time and extent to which the temperature rose above 80 F in that 5-year period).
We summed the HDD and CDD together to get a relative measure of how much and for how long the temperature was outside the optimal range over the 5-year period. Then we divided by the number of days in that 5-year period to get the total degree days per day. The day bits cancelled out, effectively giving the average number of degrees that the temperature deviated from the optimal range. (I think the WSJ sensibly thought that expressing the data in terms of average temperature deviation would make it more accessible than if it were expressed in terms of degree days, which most people probably aren't familiar with.)
Anyway, irrespective of whether the meaning of the figures themselves is intuitive, I think they're ideal for ranking the cities in terms of how favourable their climates are in terms of temperature.
The main body article also has some interesting discussion of the ins and outs of using temperature as a measure of a given location's comfort levels.
Probably of more interest to me than most as I'm quite chuffed that my company's data was used in the WSJ :) But putting my shallow ego aside I do think the numbers are cool as I've long been intrigued by the idea of finding the locations on earth with the perfect climates.
I see you just launched an API. Looks like you're working off a dump of WU's data? Nice dataset there.
For the back-of-the-envelope calculations I always do, I would probably just write a script to scrape WU's data myself and do the HDD/CDD calculations manually. It's a shame that there's no free trial of your data (~50 request/day or similar). It would make me far more likely to fall in love with your API. Then, on the not-inconceivable chance that I write energy software in the near-future, I would be far more likely to query your API instead of rolling my own solution.
I definitely agree with you that we need to bring more computation to energy. I'm gifting someone a Nest thermostat later today, which embodies the kind of innovation we need. The fact that the Nest is considered as innovative as it is shows just how lacking we've been in this area. It's just an Android on the wall, after all!
Thanks for posting. I take it info@energyauditsoftware.com reaches you?
Thanks for your thoughts there! Yes we are working off of WU's data. They have an awesome database of weather data and they've been very kind in allowing us to use the temperature part of it for our degree days system.
We do have a free API trial actually, though it's limited in the sense that you can only get data for the Cape Cod area: http://www.degreedays.net/api/test
We'll happily give most people a full free trial if they ask. We don't offer that publicly, partly because a lot of people are just looking to assemble a database of 5-year-average data and then quit, and partly because the number crunching to calculate degree days properly is pretty expensive so we have to be careful about watching our costs. A proper free tier could kill us, particularly if a manufacturer of popular installable energy-related hardware or software decided to wire things up so all their customers fetched degree days from our system whether they wanted them or not.
But perhaps we should re-consider the current setup for trials... We only launched the API recently so at the moment we're still playing it by ear somewhat.
If you mention "bromley" in an email to that address or just email info at degree days dot net it'll get through to me - and I'd love to hear from you or anyone else developing software for energy-related applications.
>particularly if a manufacturer of popular installable energy-related hardware or software decided to wire things up so all their customers fetched degree days from our system whether they wanted them or not.
I always check the TOS of any "public" API I use, and I suspect most developers are the same. Perhaps this could be mitigated by some carefully-worded clauses? If someone does it anyway, you could protect yourself by blacklisting their API key for TOS violation.
It's possible that could work. Our current testing API is rate limited only on IP address but if we required a sign-up we'd have more options. A free tier couldn't go too far out of control if we could stop taking new sign-ups at any time.
But at the same time we already have a full free tier in the sense that anyone can download free data from the website. At the moment we're seeing the API purely as something for those that want more data than the website practically allows. My hunch is that it's probably best to keep the API paid only... Free accounts with really low rate limits would be frustrating for the people using them (as they'd hit their rate limit all the time) and would just increase our support costs. In the end we'd have to increase the costs of new paid accounts to support them, which is not something I'd want to do.
Never say never though... My thoughts on that could change over time as we get more customers and feedback.
I got your email, thanks, will get back to you shortly :)
My problem with this is that I think degrees over are much worse, and humidity is also such a major factor in temperatures over 80. 85 and dry is perfect; 85 and humid is miserable. Also, with typical NYC humidity, 90 is a lot worse than 50 (assuming you have access to appropriate clothing). You don't want to do anything outdoors in NYC 90 unless you know what you're doing, whereas biking or running in 50 is pretty easy. NYC 90 is equivalent to 35, and 100 is equivalent to 10.
An improvement on the analysis might be to adjust all the temperatures for humidity and wind speed etc. before calculating the deviations below 60 F and above 80 F. My company's system doesn't do that because those other factors are much less important for energy consumption calculations (which is what degree days are traditionally used for). But I think they probably would be good factors to include for a more thorough analysis of human comfort. (I say "probably" because the complexity would increase considerably and also the figures would become less clearly defined in a sense as well. So it wouldn't be 100% positive.)
7 comments
[ 4.4 ms ] story [ 24.7 ms ] threadThe main point is that if you sort in ascending order on e.g. the "Average number of degrees temperature deviates from 60-80 range in last five years" column you should have a good indication of how the cities compare in terms of the time and extent to which their air temperature deviates from a comfortable range.
But if you're interested in more detail, the figures in that column are actually derived from:
- The heating degree days (HDD) with a base temperature of 60 F summed for each day over the last 5 years (measuring the time and extent to which the temperature dropped below 60 F in that 5-year period).
- The cooling degree days (CDD) with a base temperature of 80 F summed for each day over the last 5 years (measuring the time and extent to which the temperature rose above 80 F in that 5-year period).
We summed the HDD and CDD together to get a relative measure of how much and for how long the temperature was outside the optimal range over the 5-year period. Then we divided by the number of days in that 5-year period to get the total degree days per day. The day bits cancelled out, effectively giving the average number of degrees that the temperature deviated from the optimal range. (I think the WSJ sensibly thought that expressing the data in terms of average temperature deviation would make it more accessible than if it were expressed in terms of degree days, which most people probably aren't familiar with.)
We did that calculation for each of the 66 "global cities" listed in the 2012 list that you can find on Wikipedia at http://en.wikipedia.org/wiki/Global_city
Anyway, irrespective of whether the meaning of the figures themselves is intuitive, I think they're ideal for ranking the cities in terms of how favourable their climates are in terms of temperature.
The main body article also has some interesting discussion of the ins and outs of using temperature as a measure of a given location's comfort levels.
Probably of more interest to me than most as I'm quite chuffed that my company's data was used in the WSJ :) But putting my shallow ego aside I do think the numbers are cool as I've long been intrigued by the idea of finding the locations on earth with the perfect climates.
For the back-of-the-envelope calculations I always do, I would probably just write a script to scrape WU's data myself and do the HDD/CDD calculations manually. It's a shame that there's no free trial of your data (~50 request/day or similar). It would make me far more likely to fall in love with your API. Then, on the not-inconceivable chance that I write energy software in the near-future, I would be far more likely to query your API instead of rolling my own solution.
I definitely agree with you that we need to bring more computation to energy. I'm gifting someone a Nest thermostat later today, which embodies the kind of innovation we need. The fact that the Nest is considered as innovative as it is shows just how lacking we've been in this area. It's just an Android on the wall, after all!
Thanks for posting. I take it info@energyauditsoftware.com reaches you?
We do have a free API trial actually, though it's limited in the sense that you can only get data for the Cape Cod area: http://www.degreedays.net/api/test
We'll happily give most people a full free trial if they ask. We don't offer that publicly, partly because a lot of people are just looking to assemble a database of 5-year-average data and then quit, and partly because the number crunching to calculate degree days properly is pretty expensive so we have to be careful about watching our costs. A proper free tier could kill us, particularly if a manufacturer of popular installable energy-related hardware or software decided to wire things up so all their customers fetched degree days from our system whether they wanted them or not.
But perhaps we should re-consider the current setup for trials... We only launched the API recently so at the moment we're still playing it by ear somewhat.
If you mention "bromley" in an email to that address or just email info at degree days dot net it'll get through to me - and I'd love to hear from you or anyone else developing software for energy-related applications.
I always check the TOS of any "public" API I use, and I suspect most developers are the same. Perhaps this could be mitigated by some carefully-worded clauses? If someone does it anyway, you could protect yourself by blacklisting their API key for TOS violation.
We should continue this over email.
But at the same time we already have a full free tier in the sense that anyone can download free data from the website. At the moment we're seeing the API purely as something for those that want more data than the website practically allows. My hunch is that it's probably best to keep the API paid only... Free accounts with really low rate limits would be frustrating for the people using them (as they'd hit their rate limit all the time) and would just increase our support costs. In the end we'd have to increase the costs of new paid accounts to support them, which is not something I'd want to do.
Never say never though... My thoughts on that could change over time as we get more customers and feedback.
I got your email, thanks, will get back to you shortly :)
An improvement on the analysis might be to adjust all the temperatures for humidity and wind speed etc. before calculating the deviations below 60 F and above 80 F. My company's system doesn't do that because those other factors are much less important for energy consumption calculations (which is what degree days are traditionally used for). But I think they probably would be good factors to include for a more thorough analysis of human comfort. (I say "probably" because the complexity would increase considerably and also the figures would become less clearly defined in a sense as well. So it wouldn't be 100% positive.)