Edit: The study seems to have a major flaw in that it ignores the effect of the block subsidy in its calculations: Though it mentions the block reward halvenings, it ignores them in the CO2e calculations. They basically calculate the CO2e emissions per transaction in 2017, then multiply that by a predicted number of transactions that will be made in future years [1]. Currently, transaction fees only account for 1-2% of the miner reward - (in 2017 on average 17% of profit was the fees, in 2018 6% and in the past 6 months only 1.5% [2] )
The study makes predictions in a very long term (15+ years) - but they ignore that what is currently 95+% of the reward (and thus 95% of the energy used) automatically halves every four years. This ratio will forcibly change over time - in 20 years bitcoin is either dead or fees will account for almost all of the reward - which means you have to divide the predicted emissions in the paper by at least 5.
In the long term the amount of energy one BTC transaction burns is pretty much the same as the fee paid for that transaction times the cheapest electricity in the world.
[1] "The CO2e emissions of projected Bitcoin usage were estimated using the CO2e emissions for Bitcoin transactions in 2017 as a reference. We randomly sampled blocks mined in 2017 until their total number of transactions were equal to the projected number of transactions, then we added the CO2e emissions from computing such randomly selected blocks. The approach was repeated 1,000 times."
Assuming we don't find better ways to generate electricity. This ignores that a lot of mining power is concentrated in China hooked up to cheap renewable sources like hydro-electric power. If we can generate electricity without emissions and run miners off that (wind, nuclear, solar, hydro-electric...) then there is no need for Bitcoin to warm the planet at all.
https://www.nature.com/articles/s41558-018-0321-8
https://github.com/moracamilo/Bitcoin/
Edit: The study seems to have a major flaw in that it ignores the effect of the block subsidy in its calculations: Though it mentions the block reward halvenings, it ignores them in the CO2e calculations. They basically calculate the CO2e emissions per transaction in 2017, then multiply that by a predicted number of transactions that will be made in future years [1]. Currently, transaction fees only account for 1-2% of the miner reward - (in 2017 on average 17% of profit was the fees, in 2018 6% and in the past 6 months only 1.5% [2] )
The study makes predictions in a very long term (15+ years) - but they ignore that what is currently 95+% of the reward (and thus 95% of the energy used) automatically halves every four years. This ratio will forcibly change over time - in 20 years bitcoin is either dead or fees will account for almost all of the reward - which means you have to divide the predicted emissions in the paper by at least 5.
In the long term the amount of energy one BTC transaction burns is pretty much the same as the fee paid for that transaction times the cheapest electricity in the world.
[1] "The CO2e emissions of projected Bitcoin usage were estimated using the CO2e emissions for Bitcoin transactions in 2017 as a reference. We randomly sampled blocks mined in 2017 until their total number of transactions were equal to the projected number of transactions, then we added the CO2e emissions from computing such randomly selected blocks. The approach was repeated 1,000 times."
[2]: https://gist.github.com/phiresky/a6755c650b3bbb340c4d641f89f...