Therms per Heating Degree Day

Click for interactive graph

In an earlier post I had tried out a few ways to figure out if our energy-saving efforts with respect to natural gas use were paying off; I did a few bar graphs of therms per day, per heating degree day, rolling yearly averages, etc.  I knew that I needed to normalize for the weather using Heating Degree Days, since natural gas is our primary heating fuel, and I probably needed to find a way to separate space heating from water heating, which have different conservation methods, and which may or may not be weather dependent.

I came across this article, Degree Days – Handle with Care! (run by the same folks who provide, a great source of free  HDD data) which listed a lot of the issues involved, some of which I had thought of, and some of which I had not, including:

  • Separating out non-space-heating use (water, cooking, etc)
  • Getting the right base temperature for the building in question
  • Using proper degree day data for the actual monitoring period
  • Differences in occupant behavior over the time period

But the other thing that caught my eye was the idea to do a scatter plot of HDD vs. heating therms used, and doing a linear regression; the slope should give you an idea of your average therms per heating degree day.

The graph above is my attempt at this; I found some old data from 2004-2005, which was before we did any insulation work, and I have more recent data for the past 4 or 5 years as well, so I have a decent data set.  The graph above does not attempt to factor out water heater use, or to perfectly match HDD (which I actually obtained from the Minnesota Climatology Working Group website, as monthly data) to the precise billing period (which can vary from 27 to 35 days).  It simply graphs therms vs. heating degree days, by heating season, for any months which had over 200 HDD.  But even with that basic, raw data, I still think it shows some interesting trends.

(edit: if you click above, the live graph now has all datapoints for all months; this more accurately shows the base gas load at the Y intercept.  It is also more accurately matched to the actual HDD for the billing period.)

If you click the graphic above, you will be taken to an interactive javascript plot of the data (made with my new favorite toy, jqplot).  By clicking on the legend you can turn data sets on or off; most interesting is to compare 2004-2005 (before any work) with 2010-2011 (after all insulation and air sealing, but before the extra heating zone was added in the basement).  All the years in between are pretty clumped together, with roughly the same slope (or, therm per hdd efficiency).  It’s a pretty messy data set so far; I’ll try to come up with a good way to factor out water heating etc, but I still found it to be an interesting way to visualize the data.  What do you think?

Edit: After a brief discussion with the nice folks at /, I have a few more ideas on doing a better analysis.  Hopefully will have some spare time over the holidays to try out some other things, and do another post.

37 thoughts on “Therms per Heating Degree Day

  1. From what I can tell from my summertime natural gas bills, hot water and cooking usage is related to number of persons in the household more than anything else. I’d start by simply assuming that it’s relatively constant throughout the year.

    • Yep, that’s the approach I was going to take, just subtract out the average of the summer months for the year. Not perfect but probably better. I also might play with taking the “shoulder” months out of the scatter plot and just keep the deep-winter plots, where the errors from hot water subtraction would be a smaller percentage of the total.

    • I think perhaps the biggest issues with my data set are that I haven’t used HDD for the exact same days as the billing period (and in some cases I don’t have the exact billing period either), and maybe more significantly, the base temp for the HDD I used might not be right for my home. From looking at the HDD for the shoulder months and how many therms we used, I think our base temp should probably be lower; we don’t turn the heat on when it reaches 65F outside.

  2. I like what you are doing with tracking your energy usage. Keep it up!

    The interactive graph does not work in IE 7 ( but what does?) You might look at for another option for graphing for the less fortunate

  3. Eric,

    You’re headed in the right direction using linear regression. But it will actually give you more information than you think. The intercept when HDD equals zero should be the number of therms used for hot water and cooking. After all, they don’t vary with HDD (actually, hot water does vary a little because ground water is a bit colder in the winter and your basement will be a little colder so conduction losses are greater).

    However, it looks to me like your regression lines are constrained to go through the origin (when x and y are zero). This is an option for certain kinds of regression problems but is not appropriate here.

    Also, it would be helpful to show the regression equations for each line on the graph. Comparing the slopes will show what percentage improvement you got for each change you made.


    • David, thanks for your comments!

      I am a little limited on that interactive graph; the javascript plotting library is neat, but doesn’t have a huge number of options without hacking the library itself. I did confirm, though, that it’s not constraining the line to an origin intercept, by putting in a set of points that shows a regression line which hits well above 0 on the Y axis. Our non-heat gas usage is quite low, though – around 10 therms per month, so I’m not surprised to see intercepts at least near the origin with the raw data.

      I agree that the slope should be shown – again, a limitation of the library, I’ll see if I can get that on there. I think the biggest thing I need to play with is the base temperature used for the HDD, though, and I hope to try that soon.

  4. excellent work! I love this stuff. I would like to do the exact same calculation but I don’t have all my monthly gas readings available, and before that it was oil. long story short, I’m trying to figure out why I seem to be expending about double what I should in gas. How many sf is your house? It looks like you’ve never gone above 200 therms in a month, is that right?

    • About 1400 square feet in the current listings; maybe closer to 2000 if you include the basement. A bit less if you only include the finished part of the basement. Until recently, the basement was not significantly heated.

      Yes, looks like 197 therms was the high point for the years I have data for.

      Your gas company might be able to provide some past data?

    • Natasha,

      Why do you think you should be using half as much gas? The average house in the Northeast uses 50,000 btu per sqft of living space per year. How does your house compare to that?


      • David, is that 50,000 btu/ft^2 statistic for heating?

        FWIW, since 2007 I’ve averaged about 885 therms per year, or 88,500,000 btu, at 1400 square feet that’s 63,214 btu/ft^2 right? Not sure how our heating load compares to your northeastern climate.

        • Eric,

          Yes, that’s for heating.

          I don’t know where you live, but you could adjust the 50,000 number for your climate by comparing the heating degree days for your area with the heating degree days with someplace in the Northeast, like Boston.

          Also, you need to remove the therms for domestic hot water, cooking and other non-heating uses.


          • Ok, I’d conservatively take out about 10 therms/month for non-heat load and get about 54,642 btu/ft^2 on average for the past several years.

            I’m in the twin cities; last year we had 6117 HDD at 60F base temp; Boston had 4071. So we’re about 50% higher on HDD.

      • It is exactly that, 50,000 btu/sf per year, IF i assume I use about a round and a half on the gas meter (the old type, that goes up to 1000 ccfs). There were a bunch of months, especially in the winter, that the meter reader didn’t get to read, either due to weather or me not being home, or the non-working bell… I am not certain but they may have missed a whole round, and estimates were still based on when I only had gas for cooking. ;)

        • Natasha,

          Could you walk us through the numbers? How many therms did you use over what period of time? How many heating degree days per year are there where you live? How many square feet of heated floor space do you have?


          • Hi David,
            ~1500 therms over winter season last year (a guesstimate which I used to calculate my HHI ~10.4btu/sf-dd, since my sf~3000 and hdd=4804). What freaked me out last month was seeing the meter increase 20 ccf’s in one day. After that I replaced a mains vent, added another, and balanced all the radiator vents (more on that on my site’s posts), as well as sealed some drafty areas… Luckily this winter the weather has been mild! After the changes above I’m seeing an average of ~10-15ccfs/day gas consumption. Still monitoring it though, and there’s still a lot more optimizing to do here…

          • Natasha,

            Your HHI of 10.4 sounds pretty high. My leaky old 1835 Greek Revival on an exposed ridge has an average HHI (over the last 4 years) of about 6.
            Eric’s is about 9.

            Your site looks very interesting.


        • (hm I should allow more deeply nested comments… )

          One of the big variables here is the base heating temp used for HDD; base 65F gives me closer to an HHI of 7; 60F gives me closer to 9. That, and the (in)ability to accurately factor out non-heating load… I can torture my numbers to give me almost any HHI I want. ;)

          I should just put all my data on a public google doc and let people draw their own conclusions. :)

          • Eric,

            The base you use for HDD should usually represent your thermostat setting. If you set your thermostat at 60, use a 60 degree base. HDD is not completely arbitrary.

            Having said that, there is no provision in most HDD data sources for thermostat setbacks. I myself set my thermostat to 65 during the day, but to 50 at night. So HDD65 numbers do not reflect that.

            But then it all depends on what you’re using the HHI for. If I want to benchmark one house against another for their energy consumption, you should probably use HDD65 because that doesn’t compensate for differences in thermostat settings. But if you want to compare houses for their heating efficiency, you should probably use an HDD that does. In one method, HHI gives credit for lowering your thermostat; in the other it doesn’t.


          • David I realize it’s not abitrary, but isn’t equating base temp to thermostat setpoint going to result in something a bit too high? The rule of thumb I’d heard was “the outside temp at which you start heating your building.” I have my thermostat at 67F-68F during the day, but I certainly don’t fire up the furnace when it’s 68F outside.

            It seems like using your thermostat setpoint ignores other heat sources like solar heat gain, appliances, and warm bodies in the house.

            I had settled on 60F base because that seems like about when we fire up the heater, and it also seemed to match my HDD regression graphs the best (highest R^2 for the dataset).

          • Eric,

            Right, if your thermostat setpoint is 68 and it is 68 outside, your furnace will not fire. But if the outside temperature drops to 67.99, eventually your furnace will fire. Only at exactly 68 is your house in equilibrium with the outside. That’s why 68 would be the most sensible HDD base temperature.

            You’re also right that the HHI doesn’t take into account solar gain and internal heating sources. To that extent, HHI is inaccurate. More detailed modeling of a home takes these heat sources into account. It may not be obvious but the HHI is actually a simple model of the heat loss from a house based on Fourier’s Law.


          • David, but that’s not really quite how it works, right…? at 67.99F outside I still have a refrigerator, an espresso machine, a dryer, a TV, a dog, etc all putting heat into the house, never mind the solar heat gain. But I may be over thinking this; for common comparisons, 65F base for HDD is probably the way to go.

          • “David I realize it’s not abitrary, but isn’t equating base temp to thermostat setpoint going to result in something a bit too high? The rule of thumb I’d heard was “the outside temp at which you start heating your building.” I have my thermostat at 67F-68F during the day, but I certainly don’t fire up the furnace when it’s 68F outside.

            It seems like using your thermostat setpoint ignores other heat sources like solar heat gain, appliances, and warm bodies in the house.”

            Eric, you’re right; I’m wrong. Internal gains affect the balance point. See the discussion at:


  5. Hi Eric, I just came across your site, awesome stuff and the discussion is great. I just recently tried the same HDD/heat energy regression analysis after reading the same degreedays article a while back. I also built an interactive tool to visualize the results. I record outdoor temps so I can calculate my own HDD at any base temperature I want, the same way degreedays does. I have 11 months of temperature data and 9.5 months of circuit level data (via eMonitor, we use a air-source heat pump for all our heat) but I used the HDD data to estimate what our heating energy might have been for the months we don’t have circuit level data. I just wrote about it on our blog. The visualization tool is on my hobby site,
    (No guarantees it works in all browsers, sorry.) Wish I had found this post a year ago. Nice work!

    • Thanks Larry, I had seen your site before, and am envious of your house. :) I’d love to be able to build from scratch and try to do it all “right” – meanwhile I do the best with our embodied-energy 1931 home.

      I like your graphs! I kind of struggled with what our base temp “should” be, too. I guess the regression should tell me, but when I see it change a bit from year to year, I’m not sure what to make of that.

      As I sit here at 10am working from home on a freezing cold Minnesota day, I realize that the other thing I don’t account for at all in my data is the simple matter of occupancy. :)

  6. Eric,
    After adding several years of data the monthly chart gets too busy to read. It would improve the chart to ignore the ‘months’ and aggregate per year, perhaps after subtracting the water heating baseload. That should minimize the noise and error from DD data vs heating data boundaries. If you divide by square footage then your newer data should fit right in.

    The ‘live’ chart shows a 0 (Y) point on one 2014 data point.

    Ted Kidd’s website changed, his energy calculator is now at:

    Get a thermal imager and look for opportunities.

    You might also want to look at DC.js ( which allows you to ‘drill’ into your data when you want to, but otherwise makes it easier to see aggregated data.


    • Hey, thanks for the feedback.
      I agree that it is pretty busy; you can click on a data series in the legend to make it disappear, though, FWIW. The “year” value would be interesting, that’s true. I found the slope to be interesting, and the Y intercept also should show the water heating baseload, but yeah, you have a point. Both might be good.
      And I did divide by square footage, and found that while our absolute use has gone up, therm/hdd/ft^2 has gone down. Win some, lose some…
      I have done thermal imaging on the house twice now; there is one bypass to the attic that the renovators missed. I’m honestly just about out of things I can do short of replacing the boiler – which actually needs to be done for safety reasons.
      Thanks for the suggestions! (and I’ll look into the 0 datapoint & fix it)

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.