Nathan Myhrvold defends the solar section of Superfreakonomics:
The point I was making to Dubner and Levitt is the following: when you build a solar plant it costs you energy. Lots of energy. Pacca and Horvath, in a 2002 study, found that the greenhouse gas emissions necessary to build a solar plant are about 2.75 times larger than the emissions from a coal plant of the same net power output (1.1 * 1010 kg of CO2 to build the solar plant versus 4 * 109 kg of CO2 per year for coal). The numbers vary depending on the specific technology, but there are dozens of “Life Cycle Assessment” papers on solar photovoltaic cells that document a similar effect.
So building the solar plant hurts global warming, at least during the construction period. Once you turn it on and are able to throttle back a coal plant because you get electricity from the solar cells, you gradually earn back the deficit through CO2 emissions that are saved. You need to operate the solar plant for at least 2.75 years before you break even versus the coal plant at least versus CO2 emissions. This is very much like the old adage “you need to spend money to make money.” You need to “spend” some carbon emissions in order to create a carbon-free infrastructure which will ultimately yield a carbon emission “profit.”
Solar cells pretty much have to be “black” in the energetic side of the solar spectrum because they absorb sunlight! Of course no material is a perfect absorber, so when I say “black,” what I mean is very high absorption of light 90 percemt [sic] or more. Solar cells often have a bluish tint to them because they reflect a tiny bit more blue light than other colors, but that is small enough that it doesn’t matter for our purposes here.
Unfortunately, solar cells are not very efficient. Efficiencies of 9 percent to 13 percent are typical for current widely deployed technology. In the future that will change, and some laboratory examples are better, but this is what people deploy now. So for every watt of electricity they generate, current solar cells throw about 10 watts into the climate as heat. Some of this heat would have occurred anyway when the light was absorbed by the ground, but the most effective solar cell installations are in deserts where the albedo is pretty high (.4 to .5) and there is little cloud cover, so the solar cells cause a bunch of heating that would not have otherwise occurred. A typical coal power plant gives off about 2 watts of thermal heat for each watt generated, so the direct thermal heating from solar plants is likely to be as large or larger than that from coal plants.
[W]e need to build out lots of renewable energy if it is going to make a difference. If we finish one plant today, it takes it three years to break even. Three years may not be the exact number, but let’s use it for simplicity...the three-year break-even times start to overlap and pile up as we build more and more plants.
The net result is that we may not get much CO2 benefit from the solar plants until we are past the rapid-growth phase of building out new plants. If we go hell-bent for leather in building solar plants for the next 50 years or so, it is entirely possible that we won’t see much small benefit for 30 to 50 years. In the long run, we still get benefit from the solar plants lots of benefit (hence the “great carbon-free infrastructure”) but in the near term, we may get little or no benefit. I say “may” because the details matter, and it is beyond the scope of what I can do here to calculate and explain them all; but the basic effect is that the time to get real benefit is delayed. A large part of this is due to the energy it takes to make them, and some is due to their blackness.