One of the questions raised by several of the commenters was that while one can estimate the calories burned by bicycling, it is hard to estimate what change in calories the cyclist requires (over what would have burned by driving to work).
I designed an experiment using a Heart Rate Monitor to help me estimate just that.
Notes and assumptions
Calories were determined by a Heart Rate Monitor, which infers calories burned based on heartbeat count. Assumption: number of heartbeats are directly proportional to actual calories burned.
Calories were burned either 1) While driving to a bus stop in a car and then taking the bus to or from work or 2) while bicycling to or from work. Assumption: the cyclist will eat more on the day he cycles, corresponding to the larger quantities of calories being burned.
The car travels mostly at highway speeds to the bus park-and-ride. Distance 5.7 miles. Assumption: Nominal 30 mpg and 1 passenger were used to estimate 0.16 gallons of unleaded fuel.
Using http://www.eia.doe.gov/oiaf/1605/coefficients.html; a car fueled with unleaded gasoline produces 8.87 kg of CO2 per gallon.
The bus takes a (mostly) express route over highway. Distance 10.8 miles.
Using http://docs.wri.org/wri_co2comm_2002_commuting_protected.xls and medium haul of .20 kg CO2 per passenger mile.
CO2 is a proxy for environmental impact.
The commuter is a vegan male, weight 68 kg (150 pounds).
Design of statistical test
A difference between means (of total CO2 production) was the chosen hypothesis test.
$\mu1 represents the mean CO2 emissions required to fuel the bus, car, and the subject's body while commuting.
$\mu2 represents the mean CO2 emissions required by the foods needed to sustain the subject's body while bicycling.
Null hypothesis: $\mu1 - $\mu2 <= 0
This corresponds to the OP's friend’s assertion that bicycling produces as much or less CO2 as driving.
Alternative hypothesis: $\mu1 – $\mu2 > 0
Will choose a significance level of 0.01
Will use a one-tailed test.
Measurements were taken over a period of about one month. There were 18 measurements taken: 9 by bus and car, 9 by bicycle. Here is the data:
Date Mode Cal Food Bus Car Total
CO2 CO2 CO2 CO2
(kg) (kg) (kg) (kg)
24-Aug Bike 781 0.109 0.00 0.00 0.11
24-Aug Bike 830 0.116 0.00 0.00 0.12
28-Aug Bus/Car 177 0.025 2.16 1.42 3.60
30-Aug Bus/Car 326 0.046 2.16 1.42 3.62
31-Aug Bike 1148 0.161 0.00 0.00 0.16
31-Aug Bike 770 0.108 0.00 0.00 0.11
11-Sep Bus/Car 117 0.016 2.16 1.42 3.60
11-Sep Bus/Car 77 0.011 2.16 1.42 3.59
12-Sep Bike 842 0.118 0.00 0.00 0.12
12-Sep Bike 802 0.112 0.00 0.00 0.11
13-Sep Bus/Car 117 0.016 2.16 1.42 3.60
13-Sep Bus/Car 102 0.014 2.16 1.42 3.59
17-Sep Bus/Car 274 0.038 2.16 1.42 3.62
17-Sep Bus/Car 97 0.014 2.16 1.42 3.59
20-Sep Bike 850 0.119 0.00 0.00 0.12
21-Sep Bike 871 0.122 0.00 0.00 0.12
21-Sep Bike 712 0.100 0.00 0.00 0.10
24-Sep Bus/Car 381 0.053 2.16 1.42 3.63
Row Labels Average of Total CO2 (kg) StdDev of Total CO2 (kg) Count
Bike 0.118315556 0.017288929 9
Bus/Car 3.605146667 0.015794736 9
The data indicates that the null hypothesis is rejected at p < 0.01.
This weakens the OP’s friend’s argument that bicycling causes as much environmental damage as commuting.
The environmental cost of heating the water for a shower was not included, as it seems to have minimal impact (see dancek's fine answer on this point).
Even with a substitution of a more typical American diet with more meats and fewer vegetables, the “bicycling is as bad as a car” hypothesis would be rejected.