I’ve been collecting various sets of data with my Arduino-based RIMU.1 environmental data logger. In particular, I have seven nights worth of overnight records showing both indoor and outdoor temperatures. The data were all taken with the recorder in the same locations. Some nights are contiguous, but the 7-night set spans about 10 days.
The indoor temperature, during the “night” setting on the thermostat tends to drop quite linearly until the furnace runs again around 6 am. The following graph shows the data, which looks like a black staircase, along with a linear fit of the data. The fit is quite good.
I capture only one important datum from this overnight trend—the slope. Later I’ll show how I use the rate of temperature decline, measured in degrees Fahrenheit per hour of decline.
The outdoor temperature sensor is a thermistor, and the measurement is markedly noisier. The temperature varies more too, presumably due to wind turbulence near the house. The linear trend is much less clear, though it is clearly cooling some during the night. I do measure the average difference between the indoor and outdoor temperature which I call the “mean temperature contrast”. This is done in degrees Fahrenheit, though I probably should have done it in Kelvin. The contrast is measured as a difference, not a ratio.
If I take the seven days of data and plot the temperature decline rate versus the temperature contrast I see the picture you might expect, below. Remember that bigger numbers mean faster falling indoor temperature. You can see the outlier caused by the windy night. There is another outlier (~28, ~0.60) which, if I recall correctly, is the one night I closed the blinds. A hint of the follow-on experiment.
The next question is, does closing the blinds cause this rate to be materially different?