Going solar reduces painful fish deaths

By Brian Tomasik

First published: . Last nontrivial update: .

Summary

Thermal power plants require water for cooling. This cooling water often contains lots of small larval fish, which die painfully in the cooling systems. I estimate that a typical American kills perhaps on the order of 1000(?) tiny fish each year by using electricity. Switching to solar electricity helps avert these horrible fish deaths. You can go solar either by putting panels on your property directly or by signing up for "community solar" programs in your area, if you have them.

While I was doing most of this research, I assumed that killing 1000 larval fish every year was a moderately big deal. However, toward the end of this project I tried to figure out exactly how big the larval fish killed by power plants are. I'm unsure of the answer, but plausibly they're quite small, maybe around 2 to 4 mm? A larval fish may have fewer neurons than an ant. If so, then this issue may not be that big of a deal after all relative to other priorities, though it's worth double-checking whether that's correct. The suffering that power-plant cooling systems cause to invertebrates may or may not be more important than the suffering caused to fish.

Work on this topic is probably much less cost-effective than promoting humane slaughter of fish, shrimp, and insects, so I don't want to distract animal advocates from that more important task. Humane slaughter (if it works as intended) replaces a horrible death with a less painful death. In contrast, fish spared from power-plant cooling systems will still endure horrible deaths later on.

Contents

Note

I've been intending to write a page about fish killed by electricity use since 2020, but I never had time, because I wanted to do more thorough research to make sure I understood everything correctly and to read the fine print of various studies on the topic. I eventually concluded I might never get around to doing all of that research and instead decided to write up the outlines of what I already knew, so that perhaps other people could carry the project forward by dotting the i's and crossing the t's. Therefore, I wrote this piece quickly, without citing many sources or doing thorough research.

In addition to what I wrote here, I have some notes and open questions that I could potentially share with someone who planned to do more research on this topic, though I might be slow to assemble those notes.

Killing fish

Fossil-fuel and nuclear power plants that produce electricity typically require water for cooling. They suck up huge numbers of baby fish and other aquatic creatures (Sierra Club 2011), which is called "entrainment". Most of the entrained fish may be killed by rapid changes in temperature, physical damage, and/or chemicals like chlorine (Tetra Tech, Inc. 2008, p. 1-4). (I've seen studies suggesting that like 65-100% of entrained zooplankton may survive passage through cooling systems, but I haven't looked into the details. In this piece I focus just on fish, since they're more sentient than zooplankton.)

Here's one quick attempt to quantify how many larval fish Americans kill per year by electricity use, though I didn't read the fine print about these numbers, so consider this estimate tentative until someone checks whether it's correct:

This number might be too high or too low by an order of magnitude or two. And exactly how many fish you kill depends on where your electricity comes from.

I've also attempted this calculation using estimates of larval fish killed by particular plants and comparing against the number of people served by those power plants. Often a big power plant kills a few billion tiny fish per year and serves a few million people, so this again works out to roughly 103 fish per person per year, though the numbers vary. For example, the former Indian Point nuclear plant in New York state killed up to 1.2 billion fish eggs and larvae per year according to a 2008 article, and according to Wikipedia, it produced enough power for 25% of New York City (whose population in 2008 was about 8 million, so 25% is 2 million people). (I'm not sure what fraction of the 1.2 billion were eggs versus larvae in this case. I've seen two entrainment estimates for other power plants in which the number of larvae killed was about 10 times the number of eggs killed. I also saw an estimate in which more eggs were killed than larvae.)

A bit over 1/3 of US electricity use is residential, while the rest is mostly commercial and industrial. So electricity used at home may only kill like ~1000/3 fish per American per year.

Many effective altruists support (modern) nuclear power because it has many advantages over other non-fossil energy sources. However, nuclear plants actually kill somewhat more fish per megawatt than fossil-fuel power plants.

Electricity conservation

One way to reduce fish killing is to use less electricity. For example, I try to use a fan rather than a window air conditioner during the summer except when it's particularly hot/humid or when I'm exercising. During the winter, to reduce heater use, I wear several layers of clothes when it's particularly cold.

I'm uneasy about the significant amounts of electricity required for new technologies like electric vehicles, cryptocurrencies, large language models, and so on.

Gas heating may cause less animal suffering than electric heating if your electricity comes from thermal power plants, though I haven't done a full analysis of all the different factors that go into such a comparison. If you do use electric heat, a heat pump is more efficient than a normal electrical heater. Good home insulation can significantly reduce heating needs.

Solar electricity

Electricity conservation can only go so far, and electricity needs will increase going forward as the world moves toward electric heating, electric cars, and so on. The other part of the solution is to switch to forms of electricity that aren't generated by fish-killing power plants, such as solar and wind. If you want to produce electricity yourself, the default option is solar.

Solar panels typically receive only 4-5 hours of peak sunlight per day. Production is less when it's cloudy, during the winter, etc. So at some times you'll still be creating demand for electricity from power plants. If you produce more power than you need during daylight hours, this reduces how much power-plant energy other people use. However, once enough people in your area go solar for electricity during the day, there would be less remaining daytime power-plant electricity for you to offset, if I understand this correctly? In the long run, battery storage of electricity produced during the day to use at night can help with this problem somewhat. (I'm not an expert on how all this works. My point is just that the situation isn't as simple as saying that every extra kWh of solar electricity reduces power-plant production by a kWh. The timing of production and what other people around you are doing presumably also matter.)

Residential solar

In the USA, the federal Residential Clean Energy Credit offsets 30% of the cost of installing solar panels through 2032.

For residential solar, most people install panels on their roofs to save space, but I think they should be installed on the ground for several reasons:

If you have solar panels on the ground, you might need to mow the grass around them periodically to prevent trees from growing there. Mowing grass is a violent activity that kills tons of bugs and sometimes even vertebrates like frogs. It seems better to lay gravel underneath your solar panels to obviate the need to mow and to prevent even more photosynthesis (Tomasik "Convert ...").

Community solar

If you can't or don't want to install solar panels on your property, another option is to sign up for a "community solar" provider in your area, if they exist in your state. Community solar allows anyone who pays an electric bill (including renters) to go solar. My impression is that community solar is basically as good as residential solar from the perspective of pushing production away from fish-killing power plants.

Community solar produces electricity at local solar farms. These have some advantages over residential solar. Due to economies of scale, solar farms produce electricity about 2-3 times cheaper per watt than residential solar. Also, solar farms can be disconnected (I'm not sure if that's the right word) if the grid has too much electricity already, whereas my impression is that's less common for residential solar.

Residential solar has some advantages over community solar. In the long run residential solar often saves money if you plan to stay in the same house for a while. Solar panels on your property don't have transmission losses from being sent over the grid. Solar farms often face opposition from NIMBYs in the local community, while residential solar avoids that because you're putting the panels in your own backyard.

Community solar is often guaranteed to reduce your electricity bill by 5-15%. However, community solar sometimes involves more administrative headaches due to communication issues between your utility and a third-party company.

I think community-solar programs generally adjust automatically to your amount of electricity usage, which means you don't have to figure out ahead of time how much electricity you'll use in the future. With grid-tied residential solar, excess electricity production will be sent to the grid for your neighbors to use, which is good, although I've also heard that when too many homes do that in an unregulated way, it can cause problems for the grid.

Alternatives to once-through cooling

Many power plants use "once-through cooling", which means water is sucked in, used for cooling, and put back into the environment. This form of cooling uses the most water and therefore kills the most fish. There are various other ways of cooling that use much less water. Prima facie these alternatives would cause less fish suffering, though I would want to examine the issue in more detail to make sure. My impression is that the water in once-through systems increases by roughly 10-25°F (Tetra Tech, Inc. 2008, p. 1-4) but never gets near the boiling point of water, so maybe the amount of heat isn't extremely painful for fish? Meanwhile, maybe some cooling systems that use less water let the cooling water get a lot hotter(??), which would be much worse for whatever fish might be in the water?

Older power plants could be modified to use alternative cooling systems, but it would be somewhat expensive, and given that renewable electricity is the future, plausibly it makes more sense to focus on that rather than upgrading old power plants? On the other hand, some form of nuclear energy will plausibly also be around for the long run. Even if we switch to nuclear fusion, my impression is that early fusion power plants would still need cooling in some form (Colin 2021).

Even if power-plant cooling water doesn't get that hot, some water in power plants does get extremely hot in order to generate steam that spins turbines. Are there aquatic creatures in that water who get essentially boiled alive? Meanwhile, solar electricity doesn't involve any boiling, except maybe during some manufacturing processes to create the panels and other materials.

Is killing fish bad or good?

Fish killed in power-plant cooling systems endure horrible deaths. Of course, those fish would have later died in other horrible ways. "natural mortality rates [are] assumed to be 99% for [...] larval stages of most marine fish species" (Tenera Environmental 2008, p. 3-5).

Cooling systems also suck up and destroy numerous unhatched fish eggs, thereby preventing those eggs from becoming more sentient after they hatch. Killing fish larvae and eggs has complicated effects on population sizes of various aquatic organisms. The thermal pollution from heated cooling water returning to water bodies also has ecological effects.

Thus, it's not obvious whether power-plant cooling systems actually increase total suffering in the world on balance. This could be a reason to prioritize other interventions where the sign of impact is clearer, such as humane slaughter of farmed and wild fish. Still, I err in the direction of assuming it's net bad to kill fish in cooling systems because the ways of dying in cooling systems seem very painful.

How sentient are larval fish?

Tetra Tech, Inc. (2008) (p. 1-4) says entrained organisms are less than 3/8 of an inch (i.e., ~10 mm). The screens on water intakes have slots 3/8 to 1 inch in size (Tetra Tech, Inc. 2008, p. 1-3). Does this mean that almost no fish longer than those sizes can fit through? What if the fish approached the hole headfirst rather than sideways? A fish who is 3/8 of an inch in length is less wide and less tall than that.

Tenera Environmental (2008) studied entrainment of fish larvae and other plankton at the Encina Power Station in California. As is common, that power plant used screens with a 3/8-inch mesh size (p. S-4). The researchers sampled fish larvae in front of the cooling-water intake, assuming that they all would have been entrained into it (p. 3-6). The study measured the distribution of lengths for samples of the most noteworthy types of entrained fish larvae. For three goby species, most of the larvae were between 2 and 4 mm, though a tiny fraction of them were up to 6.5 mm (Figure 3-9). For combtooth blennies, most larvae were between 2 and 3 mm (Figure 3-13). For anchovies, most larvae were between 1 and 4 mm, with a tiny fraction between 8 and 18 mm (Figure 3-17). Garibaldi larvae were typically between 2 and 3 mm (Figure 3-21). You can check out the study to see size distributions for a few more fish types, but the numbers are pretty similar as what I just listed.

Bates (2021) reports that a 3-mm larval zebrafish has 100,000 neurons. 3 mm is roughly the size of many of the entrained larvae mentioned in the previous paragraph, so presumably this is a reasonable estimate of how many neurons those larvae had as well? A larva that's 6 mm long would be twice as long as the zebrafish larva and might be twice as wide and twice as tall too, meaning it could be 8 times bigger, and maybe it could have a correspondingly bigger brain? But the fraction of larvae 6 mm or larger in the Tenera Environmental (2008) figures was probably less than 10%, so this wouldn't make a huge difference to a calculation of the average number of neurons across larvae of different sizes.

Fruit flies have 150,000 neurons, while ants have 250,000. So are larval fish less sentient than fruit flies and ants? But we tend to think that adult fish are a lot more sentient than insects, so maybe we should assume that larval fish are more sentient than their raw number of neurons would suggest?

If a larval fish only matters as much as a fruit fly or ant, then the issue of power plants killing fish seems less significant. Someone who doesn't seal his food scraps and has fruit flies breeding in his trash can may kill a dozen or more flies every time the trash is picked up. It's easy to drive over several ants on a single outing with a car. Mowing the grass underneath your solar panels might kill quite a few bugs comparably large as a fruit fly.

Adult zebrafish have roughly ~10 million neurons, which is ~100 times more than zebrafish larvae do. For someone who weighs sentience linearly in brain size, killing 1000 larval fish would then be like killing 10 small adult fish. If we give small brains more weight than their proportional number of neurons, the badness would be worse than killing 10 small adult fish.

Copepod neuron totals

In terms of total neuron count of organisms affected by entrainment into power-plant cooling water, invertebrates might be more significant than vertebrate fish larvae. Above I mentioned an estimate of ~1 larval fish per cubic meter of water, which is 0.001 larval fish per liter. Meanwhile, crustacean zooplankton (such as copepods and cladocerans) may number roughly ~10 per liter in lakes and rivers, with lots of variance (Tomasik "Water ..."). I did some cursory Googling of copepod abundance in ocean water and found similar numbers, typically between 1 and 100 per liter, though I didn't read the relevant studies carefully. If 10,000 times as many crustacean zooplankton as fish larvae are entrained in power-plant cooling water, then even if only like 10% of the zooplankton are killed, that's still 1000 times as many deaths. (Plus, maybe some zooplankton are injured without being fully killed, which would be an additional source of suffering.) A crustacean zooplankter might have ~103(?) neurons (Tomasik "Water ..."), which is two orders of magnitude less than a larval zebrafish. 1000 times as many deaths of creatures with 0.01 times the neurons would still imply 10 times more total neurons for the killed creatures. This doesn't even count invertebrates besides crustacean zooplankton.

Tenera Environmental (2008) reports on a 1979 study at the Encina Power Station (EPS). That study found the following:

I'll focus on the smaller net size because that captures more organisms. The total number of copepods entrained per day, adding up the two copepod rows, was 56 million. Fish larvae were 2.5 million. The ratio of copepods to fish larvae here was only 56/2.5 = 22, not 10,000 as I assumed above. So using these numbers, total neurons of the killed animals would presumably be higher for larval fish than for copepods. That said, maybe the copepod counts in the study were too low because presumably individuals smaller than 335 μm = 0.335 mm wouldn't have been caught? But some copepods can be as small as 0.2 mm, and copepod nauplii may be 0.05 to 0.1 mm.

The balance of eggs versus larvae may be important

Unfertilized fish eggs presumably have very low levels of sentience. Fertilized eggs become increasingly sentient as the fish develop inside the eggs. If you search for pictures of fertilized fish eggs, you can see that some of them look like amorphous balls, while others contain larvae that are nearly fully formed. So how sentient an egg is depends on its developmental stage. What fractions of entrained fish eggs are at which stages?

It's plausible that it's better to kill a fish egg than to let it develop further because the egg is probably less sentient than what it will soon develop into. For simplicity, suppose a typical entrained fish egg has 1/3 as much sentience as a typical entrained larval fish. Let's also assume that dying by entrainment is 1.5 times as painful as dying naturally. Suppose that the main effects of entrainment on suffering are to kill the entrained creatures in more painful ways than they would have died otherwise and to prevent eggs from hatching into larvae. In that case, we can make the following calculation:

Given these numbers, whether entrainment of fish eggs and larvae increases or decreases suffering on balance depends on whether more larvae are entrained than eggs (in which case suffering increases) or more eggs are entrained than larvae (in which case suffering decreases). In Table 3-3 from Tenera Environmental (2008) as shown above, ~6 times more fish eggs were entrained than larvae (looking at the column for the smaller net size), but I've also seen studies in which ~10 times more larvae were entrained than eggs (such as the Bay Shore Power Plant numbers here), so I don't know which case is more typical.

Of course, the above analysis (using made-up numbers) is overly simplistic. Following are some complications:

As is often the case, the analysis descends into cluelessness when too many factors are considered. It could be interesting to develop more sophisticated models of all these different effects, though I also suspect those models would depend a lot on input assumptions, so that you could prove basically any conclusion depending on what parameters and model structure you chose. Cluelessness about the net effect on suffering of changing ecosystems at levels above primary production is a reason to prefer interventions like humane slaughter (where population sizes don't change) or reducing net primary productivity (where animal population sizes should mostly go down rather than up).

Other environmental impacts of solar

The main reason most environmentalists push solar electricity is not because of fish killing but to address climate change. Reducing the burning of fossil fuels also decreases air pollution, acid rain, and so on. All of these other environmental considerations should be included in an analysis of the pros and cons of switching to solar from the perspective of wild-animal suffering.

The effects of climate change on total animal suffering are extremely complicated—some effects seem net good and some net bad—but at the moment I slightly err in the direction of opposing CO2 emissions if your goal is reducing wild-animal suffering, in which case this is a further consideration in favor of solar electricity.

When I began this research, I assumed the larval fish killed by power plants were larger than it appears they in fact are. Now that I realize how small fish larvae tend to be, it seems likely that killing of larval fish is one of the smaller considerations in an overall analysis of the net impact of going solar. Fortunately, it seems plausible that going solar is also net good in order to reduce atmospheric CO2 concentrations, though the error bars on that belief of mine are huge.