by Brian Tomasik
First written: 2008-2013; last update: 19 Oct. 2017

## Summary

Plant agriculture leaves one of humanity's biggest ecological footprints and hence has major implications for wild-animal suffering. Crop cultivation plausibly reduces populations of large animals, although the sign of impact is less clear for insects, and overall there's extremely high uncertainty in this analysis.

At the end of this piece, I begin an attempt to rank different crops according to how good vs. bad they seem for wild animals. While my method is very crude and imprecise, I come to the general conclusion that beans and nuts are better, while grasses/grains are worse. So try to eat less bread/pasta/rice/cereal and more beans/nuts and (maybe) potatoes.

Translations: Russian

## Introduction

Crop cultivation represents one of humanity's biggest impacts on the planet. Eleven percent of Earth's 13.4 billion hectares of land are used for crop cultivation, according to the Food and Agriculture Organization's article "Crop production and natural resource use." In "Energy Inputs in Food Crop Production in Developing and Developed Nations," David Pimentel's Table 2 shows that in the USA, per-capita cropland consumption is 0.48 hectares.

Determining the net impact of crop cultivation on wild animals is thus an important undertaking, especially since the amount of land we cultivate and in what ways are variables open to intervention by our dietary choices and environmental policies. This investigation has relevance to biofuels, paper farms, etc. as well as food crops.

## Impact on vertebrates

### Reducing population densities

Farm land in general may support less wildlife than wild lands.

In "Human Diets and Animal Welfare: The Illogic of the Larder," Gaverick Matheny and Kai Chan report on findings that bird densities are lower on crop land than in forests. Adding mammals to the equation as well, they estimate that an extra hectare of crop cultivation prevents 30 mammal and bird life-years (p. 585). The impact on other vertebrates is not estimated, but on pp. 587-588, Matheny and Chan explain that the numbers when counting all vertebrates (reptiles, amphibians, etc.) could be orders of magnitude higher. I assume the impact on these animals would have the same sign as the impact on mammals and birds, because crop land tends to be less inhabitable by big organisms. Insofar as these wild animals endure net negative lives, reducing their densities is a benefit to them.

It's also relevant to consider whether crop cultivation increases the degree of r-selection among the remaining animals, with greater r-selection implying more suffering-filled deaths for a given population size. I don't have data on this, but my intuition is that replacing forests with crop land shouldn't cause a significant influence on this variable.

### Killing vertebrates during harvesting

While the reduction in vertebrate populations is net positive, crop farming also entails painfully killing many animals as well. Indeed, there's a whole cottage industry of meat eaters who insist that "vegetarianism is unethical" because growing plants means killing animals in the field. The irony is that a gram of meat protein tends to require many times more farm-grown grains than does a gram of plant protein, but it remains true that even vegans do not totally avoid killing animals for their food.

Even though some of the literature on killing of wild animals via crop cultivation is ideologically motivated to defend omnivory, it can provide useful data when interpreted cautiously. There are many discussions of this topic, which I won't recapitulate here. A few references:

In "The Least Harm Principle May Require that Humans Consume a Diet Containing Large Herbivores, Not a Vegan Diet," Steven Davis details the deaths caused by crop farming and very crudely estimates (pp. 389-90) that around 15 mice are killed by crop cultivation per hectare per year. Given that each American uses half a hectare per year, this amounts to killing ~7.5 mice through crop harvesting per person. (Compare this with ~30 land animals and ~225 fish per year directly from meat eating by the average American omnivore.) Davis cites Barbara Kingsolver's Prodigal Summer (p. 322):

"I've watched enough harvests to know that cutting a wheat field amounts to more decapitated bunnies under the combine than you'd believe."

She stopped speaking when her memory lodged on an old vision from childhood: a raccoon she found just after the hay mower ran it over. She could still see the matted gray fur, the gleaming jaw bone and shock of scattered teeth so much like her own, the dark blood soaking into the ground all on one side, like a shadow of this creature's final, frightened posture. She could never explain to Eddie how it was, the undercurrent of tragedy that went with farming.

These crop-cultivation deaths are indeed tragic, though we have to keep them in perspective. Reducing populations of mammals and birds also prevents animal deaths that would have happened naturally in the wild. Given the high attrition rates of small animals, it's plausible that crop farming prevents net deaths through lowering populations, though this would be good to confirm.

#### Birds kill insects

Also keep in mind that some of the animals killed in harvesting, especially birds, are themselves savage killers:

This piece says regarding moles:

From an examination of the stomach contents of 200 moles taken in all months of the year it was found that earthworms and white grubs constitute the bulk of the food. Beetles and their larvae and other insects that enter the ground, spiders, centipedes, cocoons, and puparia also form a part of the diet. In one stomach were found the remains of 171 small white grubs, in another 250 ant puparia, in another 10 cutworms, and in another 12 earthworms. [...]

Owing to their activity they sometimes consume each day a bulk of food equal to their own weight.

The sheer numbers of insects make them an important, perhaps dominant consideration, so it is to them we now turn.

## Importance of insects

How much might insects matter to the calculations? Consider the following estimates for insect densities on land that could be farmed:

• "There may often be 25 million insects per hectare of soil" on crop land. (Pest Control by Helmut F. van Emden, p. 20)
• "Grape leafhoppers (Erythroneura spp.) reach populations as dense as 30 million per hectare in some vineyards." ("A Class of Distinction")
• "A study of Cupola Basin, in Nelson Lakes National Park, showed that the mass of native grasshoppers above the treeline reached 32.5 kilograms per hectare." ("Insects in New Zealand")
• Assuming an average grasshopper weight on the order of ~0.3 g, this would imply ~105 grasshoppers per hectare above the treeline. This is only one species, so the total of all insects is probably many times greater, especially counting those in the soil too.
• "An expert friend of mine estimates that a hectare of insect-infested crop land would have between 500,000 [...] and 50,000,000 [...] insects, depending on the type of crop and type of insect." (from my piece on humane insecticides)
• Earthworms (not insects, but same general idea) are estimated to number 0.5 to 2.5 million per hectare of forest soil.
• Springtails (formerly classified as insects) may average 10,000 per m2 which is 100 million per hectare.

For simplicity, say the average is ~107 insects per hectare. Then suppose that we can predict that crop cultivation changes this number by an expected 5% either up or down. (In fact, it seems likely that crop cultivation increases or decreases insect numbers by much more than this, but averaged over uncertainty, the expected estimate may be pulled closer to zero.) Consuming 0.48 hectares of crop land per year, the average American would then create or prevent

roughly 107 insect-years/hectare-year * 5% * 0.48 hectare-years = roughly 3 * 105 insect-years.

In fact, the numbers may be even higher than this. In "The Arthropod Population of Pasture Soil," the authors explain (p. 147):

The object of the present work was to estimate the arthropod population of an ordinary pasture in agricultural use. [...W]e collected 42,753 arthropods, representing a density of 263,658 per square metre [...]. Those numbers, large as they are, do not represent a complete, let alone a maximum, population of soil arthropods.

2.6 * 105 per m2 = 2.6 * 109 per hectare, two orders of magnitude higher than what I assumed in the previous calculation, and as the paper's authors explain, even this estimate is too low. Thus, it may be that one American creates or prevents at least tens of millions of arthropod-years per year of food consumption.

As a sanity check of these figures, note that the Earth has 1018 to 1019 insects in total, over a land size of 13.4 billion hectares. This implies a world average of 75 to 750 million insects per hectare. Densities in tropical regions are much higher than this, but those in cold regions are close to zero. Presumably densities in temperate regions suitable for agriculture are relatively close to the world average.

## Does crop cultivation increase or decrease invertebrate populations?

### How to approach the question

In this section I outline some ways we could study the net effects of farming on insects. I haven't done all of this research yet myself and welcome others adding to what I have so far.

For this analysis, we can simplify to the following question: How can we prevent the maximal number of insect deaths per unit time (both natural and human-caused)? This is on the assumption that relative to insects' extremely short lives (maybe just days/weeks on average across all births), the pain of death is probably the dominant factor for considering wild-insect welfare.

If the insect population is P individuals, then we should expect the rate of births to be proportional to P. Since 80+% of insects die soon after birth, this is also basically the death rate. (In a stable population, the birth rate always equals the death rate, but maybe insect populations aren't stable due to periodic insecticide sprayings. But when most babies die shortly after birth, then the death rate approximately equals the birth rate even for an unstable population.) Assuming some nontrivial fraction of those early-dying insect offspring have painful deaths, then this factor dominates anything else (including death rates by the adults due to insecticide exposure, etc., unless insecticides increase the number of offspring that each parent has to compensate for the greater deaths).

The question then becomes simply how to reduce the long-term average of insect populations that result from the crop land. Naively one might just compare insect populations on the crop land vs. the counterfactual, but one also has to consider that food grown on crop land is transported elsewhere for consumption, and some of it may be wasted in such a way that it's not eaten by big animals like cows and people but is instead eaten by insects (e.g., cockroaches in dumpsters). In this case, the relevant comparison may be

(fraction of biomass on crop land eaten by insects) * (biomass productivity rate on crop land)
vs.
(fraction of biomass on counterfactual land eaten by insects) * (biomass productivity rate on counterfactual land)

with the goal of choosing the option that minimizes this quantity.

Of course, there are many additional complications, e.g.:

• food may not be the most important growth-limiting factor for insects?
• certain kinds of insects are smaller and so can have much higher numbers with the same amount of food
• seasonality of births (e.g., maybe only the population in the spring/summer matters if it's a yearly insect species?).

There are lots of interesting dimensions to explore, but it's also good to keep it simple, since sometimes considering additional factors introduces more noise than nuance.

The basic questions:

• What kinds of things would grow on the land if it weren't being farmed? Forest? Grassland? Buildings?
• Is there a difference in biomass productivity of crop plants vs. the counterfactual plants? This would give a sense of whether crops produce more or less energy per hectare than if the land weren't farmed.
• What fraction of crop-land food and counterfactual-land food is eaten by insects?

Additional data points that would be interesting:

• What's the average insect population on crop land vs. the counterfactual?
• What's the frequency of insecticide application and what percent of insects are killed each time?
• What additional spillover effects does agriculture have? This analysis gets complicated because it involves water pollution, eutrophication, disruption of land fertility, etc. It would be good to highlight which of these effects are biggest in general without pursuing a full analysis of them.

### Soil-animal density comparison

This book reports:

Compared to virgin areas, cultivated fields are generally lower in numbers and weight of soil organisms, especially the soil fauna. Exceptions may be soils which in the virgin state were originally very acid and which have been well limed and fertilized upon being tilled. Such cultivated soils may contain a higher microflora population than their untilled counterparts.

This page reports: "Agricultural soils generally support less than 100 nematodes in each teaspoon (dry gram) of soil. Grasslands may contain 50 to 500 nematodes, and forest soils generally hold several hundred per teaspoon."

### Reasons crop cultivation might increase suffering

#### Increased primary productivity?

This study of global net primary productivity (NPP) reports that "High-input agriculture in North America and Europe display higher annual NPP than the natural vegetation that would exist in the absence of cropland."

This study reports:

The conversion of forests to agriculture is generally assumed to decrease productivity (Aselmann and Leith 1983, Houghton et al. 1983), yet the NPP estimates we have obtained for the U.S. Midwest are equal to those for temperate forests and approximately double those for natural prairie (Kicklighter et al. 1999).

This study examined crop cultivation in the US Great Plains, "the most heavily cultivated region of the United States" (p. 1863). The study founda (p. 1869)

that cultivation has increased aboveground net primary productivity (ANPP) by 0.066 Pg C/yr, decreased belowground net primary productivity (BNPP) by 0.020 Pg C/yr, and increased net primary productivity (NPP) by 0.046 Pg C/yr. These estimates represent a 26% increase in ANPP, a 10% decrease in BNPP, and a 10% increase in total NPP.

Note that because "Our data indicate that 25.1% of this region is currently cropped" (p. 1866), I think the increase in NPP on cropped land specifically might be more like 10%/0.251 = 40%?? This figure from the paper (p. 1867) shows changes in NPP by region due to crop cultivation:

In reviewing previous studies, the authors also note that "cropping in temperate grasslands typically increases aboveground productivity (Buyanovsky et al. 1987, Kucharik et al. 2001), [and] can decrease belowground productivity (Smith 2003)" (p. 1863).

#### Fertilizer and irrigation

Fertilizer and irrigation might raise yields above what would naturally be the case. More conversion of sunlight to energy means more total food, which may mean more insects can be supported -- if not on the crop lands themselves, then perhaps in other places where the food waste ends up, unless the food is decomposed by less sentient bacteria in landfills.

Fertilization and irrigation seem like some of the strongest arguments for the possibility that crop cultivation increases wild-animal suffering.

The invention of industrial nitrogen fertilizer is credited with enabling a quadrupling of the human population in the 20th century. This page says:

As a result of extensive cultivation of legumes (particularly soy, alfalfa, and clover), growing use of the Haber–Bosch process in the creation of chemical fertilizers, and pollution emitted by vehicles and industrial plants, human beings have more than doubled the annual transfer of nitrogen into biologically available forms.[10] [...]

Ecosystem processes can increase with nitrogen fertilization, but anthropogenic input can also result in nitrogen saturation, which weakens productivity and can damage the health of plants, animals, fish, and humans.[10]

These seem like potentially big impacts, and it would be good to explore further whether they increase plant yields relative to the counterfactual of no farming.b

An intelligent species [...] could apply some of the unused solar energy to redistribute these limiting resources, and so expand life’s footprint beyond what 4 billion years of evolution has managed. Indeed, artificial fertilization and irrigation are exactly this redistribution, and we rightfully identify the development of these methods as a turning point in human history and a hallmark of civilization. Today, humanity has made photosynthetic activity common in many nominally inhospitable parts of the American Southwest and Middle East, for instance.

For more discussion, see "How Irrigation Affects Global Net Primary Productivity".

#### Juicier biomass?

Agricultural crops are often attractive to insects, due to sugar content or otherwise easily digested carbohydrates. I think fibrous plants may be less edible by insects. For example:

• Lignin in trees "makes up 24–35 percent of the oven-dry weight of softwoods and 17–25 percent of hardwoods." However: "Lignin is indigestible by animal enzymes" and instead "Lignin degradation is done by micro-organisms like fungi and bacteria."
• Like trees, grasses may also be harder to eat than food crops: "Grasses normally have higher fibre content than legumes such as clover. The amount of fibre in leaves of grass is twice that of legume leaves, and grass leaves are harder to digest than those of legumes." Reduced digestibility means animals can't extract as much energy from the biomass. Another page agrees: "Legumes—primarily clover and alfalfa (lucerne)—have leaves with less structural material and are generally higher in protein, energy, and calcium than grasses." However, that source adds: "The fiber in grass tends to be more digestible by horses than the fiber in legumes, which have higher lignin content per unit of fiber."

So if we care more about insect suffering than fungi/bacteria suffering per gram of biomass consumed, juicier crop plants seem worse along this dimension.

Of course, certain bacteria/fungi can digest cellulose and lignin, and animals may then eat some of those bacteria/fungi, perhaps several layers up the food chain. But there should be loss of energy in the process of going up the chain.

#### Pest-bug populations

This page reports that "pest outbreaks occur more often on cultivated (disturbed) land." That source doesn't explain why, but I would conjecture this is because pest fields have fewer natural predators and other mortality factors (besides insecticides)? Note that the statement doesn't talk about all bugs, only pest bugs.

Insecticides are a major cause of bug mortality. If high mortality rates due to natural bug predators are net good because they keep bug populations down (is this true?), then presumably pesticides are net good also? And except in rare instances of huge pest invasions that can't be controlled, it would seem that bug populations are pretty low on conventionally farmed crop fields?

#### Fewer big herbivores

Crop fields probably support fewer herbivorous mammals than grasslands or forests do. While this is good news for those mammals who will not be born into lives of suffering, it's plausibly bad news for invertebrates, since without big herbivores, there may be more vegetation left over to be eaten by invertebrates. Of course, it should be explored further whether grazing by larger herbivores actually does reduce invertebrate populations.

#### Would insecticides lead to more r-selection?

In general, environmental instabilities like mass kill-offs encourage animals to have more children rather than trying to invest in a few children. If insecticides cause insects, over the long run, to have more offspring per parent, this could increase suffering even for a given population size. I don't yet have specific data on whether this is true.

#### Eutrophication

Fertilizer from crop fields is a main cause of eutrophication. The sign of net impact of eutrophication on wild-animal suffering is unclear, but a priori, I would expect that eutrophication is net bad because it increases the rate of production of food energy and hence plausibly increases populations of aquatic microorganisms in aggregate.

### Reasons crop cultivation might decrease suffering

#### Disruption of evolved ecosystems

Nature tends to fill itself with as many organisms as can fit into it. When an ecosystem is disrupted, there may temporarily be a decrease in how many organisms can be supported.

As a visual analogy, imagine a jar of pebbles and sand as being an ecosystem. If you stir the water in the jar, the pebbles and sand fly around, and the density of rocks per unit volume decreases. Over time, the big pebbles settle down, and later the smaller ones fill the cracks (niches) left by the bigger ones. That is, as stability returns, the ecosystem is able to support a higher density of life.

The downside to this, as noted above, is that ecological disruption favors the small organisms that have lots of offspring, which may actually mean more total suffering before the larger, longer-lived animals take their place.

Vogel (2017) quotes ecologist Dave Goulson as saying that "If we turn all the seminatural habitats to wheat and cornfields, then there will be virtually no life in those fields." And Carrington (2017) quotes Goulson as saying: "Farmland has very little to offer for any wild creature". These quotes strike me as hyperbole, but maybe the trend for crop fields to have fewer insects is accurate?

#### Reduced primary productivity?

This page says "In most cases, land use reduces NPP, at least over larger areas, as few croplands achieve the NPP of the native forests or grasslands they replace [...]. In regions where precipitation or nutrient availability limit NPP, irrigation and fertilization can also increase NPP (e.g. industrial agro-ecosystems in The Netherlands or irrigated drylands in Egypt or Pakistan)."

This paper adds: "most croplands have lower NPP due to the shorter growing period of crops and to the inability of one or a few crops to use the total solar radiation and other productive resources as fully as a mix of native species (17). However, by increasing yields over the last 50 y, farmers brought cropland closer to replicating the productivity of native vegetation".

This study estimates (Table 2) that in the year 2000, the potential NPP of global cropland averaged 611 g C per m2 per year, but the actual NPP was only 397 g C per m2 per year, a reduction of 35%. However, many Western nations have crop NPP levels not much lower than the NPP of native vegetation. In contrast, regions like "Eastern and Southeastern Europe", "Central Asia and the Russian Federation", and "sub-Saharan Africa" have large losses of native NPP due to farming.c

This piece gives NPP values for different land types. In the Jackson and Jackson (2000) table of that piece, we can see that cultivated land has roughly the same NPP (650 tons per km2 per year) as temperate grassland and considerably lower NPP than forest. So on average, farming may have either no impact or some impact to reduce productivity per hectare. Of course, a tiny amount of farmed land would have gone toward roads, buildings, etc. instead, in which case the crop land has higher productivity than the counterfactual. A lot of farm land in the US is converted from grassland.

Some studies report cropland NPP values higher than 650 metric tons per km2 per year = 6.5 Mg per hectare per year. Prince et al. (2001) report (p. 1200):

The total range of net primary production (NPP) estimated in this study, 4-20 Mg * ha-1 * yr-1, is consistent with the estimates of cropland NPP reported by Whittaker and Likens (1975). It includes values considerably greater than cropland NPP estimated by Sharp et al. (1975) for counties in North Carolina, even taking into account of the fact that their study only considered aboveground biomass production.

However, I think the NPP estimates by Prince et al. (2001) were expressed in terms of total dry mass, not just carbon content. The authors write (p. 1194) that one should multiply their numbers by 0.5 to get carbon mass. Are the Jackson and Jackson (2000) numbers expressed in terms of total dry mass or just carbon mass? I don't have the source at the moment and don't know the answer....

##### Cropland NPP: correlation vs. causation

Another consideration is that arable land may be more fertile on average than non-arable grassland? So even if farmed land has higher productivity per hectare than non-arable grassland, maybe this is partly because that land is particularly fertile and would have had higher productivity anyway.

For example, the tallgrass prairie of central North America had been very productive, with

the deepest level of topsoil recorded anywhere. Animals such as buffalo, elk, deer, and rabbits added nitrogen to the soil through urine and feces. Prairie dogs, a ground squirrel-like rodent considered to be a keystone species, dug tunnels that "aerated the soil and channeled water several feet below the surface."[1]

Now 96-99% of that prairie has been lost, much of it converted to farmland. This loss of native habitat has resulted in declines of mammals and birds. But that farmland is "one of the richest agricultural lands in the world", consistent with the idea that productive crop lands would have been productive grasslands.

This paper notes:

Annual NPP in North Carolina, USA, may exceed 16 Mg * ha-1 * yr-1 (Sharp et al. 1975), a value that is within the range of widely accepted estimates for temperate forests and is only exceeded by humid tropical forests (Kicklighter et al. 1999). These high NPP values are, no doubt, partly a consequence of the fact that throughout the world the most favorable soils and climates are the first to be exploited for crop production.

This study says:

In the global average, areas currently under forestry are most productive, followed by areas used today as cropland and infrastructure. The potential productivity of grazing land is lower than that of cropland, reflecting the fact that fertile areas are used for cropping rather than for grazing, but its current productivity is slightly higher. This stems from a substantial reduction of productivity [...] on croplands that can be explained on the one hand by the prevalence of low-yield agriculture in developing countries and on the other hand by the low belowground productivity of crops (25).

#### Wetland loss

This section has moved to an upcoming piece.

Most farming practices severely erode topsoil and quickly remove its nutrients. Prof. John Crawford estimates that "soil is being lost at between 10 and 40 times the rate at which it can be naturally replenished" and "A rough calculation of current rates of soil degradation suggests we have about 60 years of topsoil left." This should lead to very long-term declines in land productivity. On the other hand, how many of these nutrients find their way to other places and increase productivity there?

#### Desertification

WWF claims that "unsustainable agricultural practices are seeing 12 million hectares of land lost each year to desertification."

#### Monoculture

This article suggests monoculture crops as one cause for insect declines: "In large parts of Europe, the U.S., and South America, monocultures cover vast areas of the landscape, creating 'biological deserts' devoid of hedges or ponds where insects could reproduce." However, a reduction in aboveground invertebrates might mean an increase in belowground invertebrates, since aboveground invertebrates would be eating less plant vegetation, leaving more for the belowground invertebrates??

#### Herbicides

These help reduce non-crop plant biomass, meaning that insects on crop fields mostly only have the crop plants to eat.

Herbicides can also inhibit aquatic autotrophs. Mother Earth News (1984): "The herbicides paraquat, atrazine, and MSMA have been found to inhibit the growth and productivity of algae in streams, which can affect the overall bioproductivity of the water."

Mother Earth News (1984) also says: "There is widespread concern that herbicides may kill soil microorganisms, those bacteria and fungi that decompose organic matter and make the earth fertile. In fact, many herbicides may inhibit microorganism growth, and a few are very destructive. Others, such as 2,4-D, seem to have no effect at all." Death is plausibly painful for those microorganisms killed, but on the positive side, slower nutrient cycling may reduce plant growth.

#### Insecticides

Insecticides seem to decimate insects for short periods of time, possibly long enough that the prevented future births and deaths outweigh the acute pain caused by the insecticides themselves. Crop land is mostly barren of insects, both on the food and in the soil. The parts of the plants that aren't harvested decompose by bacteria because there aren't many bigger insects. (Source: A friend who works in agriculture.)

### How much primary productivity is eaten by insects?

Assuming crop fields and counterfactual land (e.g., grassland) have roughly comparable levels of primary production, an important further question is whether bugs eat a bigger fraction of the primary production on one land type vs. the other. For instance, if bugs eat, say, X% of primary production on crop fields but 2X% on grassland, then crop fields would support fewer bugs.

A priori it's not obvious whether the fraction of vegetation eaten by bugs is lower on crop fields or in grasslands. One reason to think the fraction is lower in crop fields is that the juiciest parts of food plants (corn ears, tomatoes, apples, peanuts, etc.) are eaten by humans or livestock (or by bacteria in landfills, or are incinerated, for the 30-40% of food that's wasted). One reason to think the fraction is lower on native grasslands is that big herbivores (bison, elk, deer, etc.) graze on native lands but are kept off of crop fields. Big herbivores may eat vegetation that would otherwise feed bugs.

What happens to crop-plant material that humans don't eat, like stalks and leaves? Crop residues end up in different places depending on the crop and farm. Possibilities include

• tilling back into the soil or used as mulch (bad because it restores nutrients and feeds soil organisms)
• bedding for livestock (somewhat bad because it will eventually feed small organisms, though maybe mostly bacteria?)
• fed to livestock (good because it's not fed to smaller animals)
• burned (probably good because it's not fed to animals, but burning is probably awful for whatever bugs are trapped inside the material)
• used for biofuel (good because it doesn't feed small animals and because it reduces soil fertility).

Crop residues can also be composted. If composting is done in sufficiently big and active piles, the number of invertebrates involved in the process may be limited during the thermophilic phase of decomposition, though I presume that invertebrates may be present during later decomposition phases. For further discussion, see "Invertebrates Created by Composting".

Manitoba Agriculture (n.d.): "The majority of farmers do not burn. They prefer to handle straw in other ways, for example, by tilling it into the soil, and by chopping and spreading the straw so it does not plug up seeding equipment. It has been estimated that in Manitoba, province-wide, only about five percent of producers burn unwanted straw."

This page says "Stover can be grazed as forage or collected for use as fodder but is commonly not utilized. It can also be used as a fuel for bioenergy or as feedstock for bioproducts."

This page says "Wheat straw is the leftover canes after the wheat grains are harvested and is treated mostly as waste. As such, in some countries (like China) farmers burn it". Burning organic matter releases stored energy without creating sentient creatures.

Crop cultivation explicitly for biofuels rather than food seems like it might have even less crop-residue waste. Most of the biomass should be burned in a way that doesn't provide energy for suffering animals.

According to Wikipedia's article on "Primary production":

there is a NPP reduction due to land changes (ΔNPPLC) of 9.6% across global land-mass.[20] In addition to this, end consumption by people raises the total human appropriation of net primary production (HANPP)[21] to 23.8% of potential vegetation (NPP0).[20] It is estimated that, in 2000, 34% of the Earth's ice-free land area (12% cropland; 22% pasture) was devoted to human agriculture.[22] This disproportionate amount reduces the energy available to other species [...].

### How does this compare with grassland, rainforest, and organic crops?

Based on the above, it seems plausible that conventional crop farming creates the lowest insect populations per hectare. How does this compare with other land types?

#### Grassland may have some more insects

• Grassland is a typical counterfactual land type relative to farming in the USA. Such land has some insects directly (e.g., locusts). One page says "insects, seeds, or rodents [...] are plentiful in a healthy prairie." A lot of the fiber may be decomposed by bacteria, and some insects eat these bacteria.
• That said, the food may not be extremely juicy for insects in general.

#### Organic farms may have higher insect populations

• The soil can support more insects than on conventional farms, and the food may be more desirable than grassland plants.
• Still, given that organic food doesn't usually have insects in it, the densities on the core food itself must be pretty low.

See here for further discussion of the pros and cons of organic farming relative to conventional with respect to insect suffering.

#### Tropical forests have very high insect densities

If the counterfactual land use is tropical forest, then it seems plausible that farming prevents a lot of insect deaths. Soy from South America might be an example of a crop that displaces rainforest.

Moreover, because markets for commodity crops are somewhat global, an increase in food prices in the US can slightly increase food prices around the world, creating slightly more pressure for deforestation in tropical regions.

### Considerations regarding bacteria

Because insecticides leave conventional-crop fields fairly barren, conventional farmed crops may have lower insect populations than counterfactual land even if they don't have lower bacteria populations. Indeed, it's possible (if not necessarily likely?) there would even be more bacteria to decompose the biomass on crop land than on counterfactual land. Bacteria matter extremely little in my moral views, but they might matter to a nonzero degree, and the astronomical numbers of them involved in decomposition might begin to arouse some moral concern. My guess is this consideration wouldn't compete with the reduced suffering by insects, but I remain partially agnostic on this question. In general, preventing plant biomass in the first place is morally preferable to creating it and then decomposing it with bacteria (for the sakes of both the plants and the bacteria).

### Climate change

More food cultivation generally means more climate change. Producing the food itself requires energy. "Cultivation also oxidizes 25-30% of the organic matter in the upper meter of soil and releases that to the atmosphere" (p. 13). And clearing forests to create room for crop land releases carbon too (p. 13):

Tropical deforestation, including both the permanent conversion of forests to croplands and pastures and the temporary or partial removal of forests for shifting cultivation and selective logging, is estimated to have released on the order of 1-2 PgC/yr (15-35% of annual fossil fuel emissions) during the 1990s.

The sign of climate change for wild animals is unclear but might be negative.

### Net impact

Overall, it seems plausible that crop cultivation prevents more insects than it causes, relative to most types of counterfactual land. (Obviously this wouldn't be true if the counterfactual land use was buildings or parking lots. Greater demand for farmed crops does imply some very slight increase in land used for farming relative to buildings, by making farming marginally more profitable than it would have been.) On the other hand, irrigation and fertilization are worrisome. I continue to have high uncertainty on the net impact of crop cultivation on short-term wild-animal suffering.

According to one author: "There’s been a definite decline in both variety and quantity of many species associated with farming."

See here for discussion of the net impact of human activity overall on wild-animal populations, although the sign of humanity's overall impact needn't be the same as the sign of human impact from crop cultivation specifically.

## Implications for future stability

The impact of crop farming on wild animals is extremely significant, but it's not the only important input to this analysis. Growing food also has implications for the stability of human societies in the coming decades.

### Food security

As a general rule, less crop cultivation now probably implies more food stability in the future.

One clear example is in the area of topsoil loss as discussed above. John Crawford explains:

water will reach a crisis point. This issue is already causing conflicts in India, China, Pakistan and the Middle East and before climate change and food security really hit, the next wars are likely to be fought over unsustainable irrigation. Even moderately degraded soil will hold less than half of the water than healthy soil in the same location. If you're irrigating a crop, you need water to stay in the soil close to the plant roots. [...]

Soil erosion is most serious in China, Africa, India and parts of South America. If the food supply goes down, then obviously, the price goes up. The crisis points will hit the poorest countries hardest, in particular those which rely on imports: Egypt, for example, is almost entirely dependent on imports of wheat. The capacity of the planet to produce food is already causing conflict. A lot of people argue that food price hikes caused the Arab spring, and may even have contributed to the recent violence following the release of an anti-Islam film.

In general, consumption of more food crops implies higher prices on the world market. From "Food Insecurity and Violent Conflict: Causes, Consequences, and Addressing the Challenges" by Henk-Jan Brinkman and Cullen S. Hendrix (p. 4):

is food insecurity itself a cause of conflict? Based on a review of recent research, the answer is a highly qualified yes. Food insecurity, especially when caused by higher food prices, heightens the risk of democratic breakdown, civil conflict, protest, rioting, and communal conflict. The evidence linking food insecurity to interstate conflict is less strong, though there is some historical evidence linking declining agricultural yields to periods of regional conflict in Europe and Asia.

That said, the effects of these rebellions on democracy can be both negative and positive (p. 7):

Food insecurity, proxied by low availability of calories for consumption per capita, makes democratic breakdown more likely, especially in higher-income countries, where people expect there to be larger social surpluses that could be invested to reduce food insecurity (Reenock, Bernhard and Sobek, 2007).

Though statistical evidence is lacking, rising food prices have been implicated in the wave of demonstrations and transitions from authoritarian rule to fledgling democracy in some countries across North Africa and the Middle East in 2011. There are some historical precedents for this: a bad harvest in 1788 led to high food prices in France, which caused rioting and contributed to the French revolution in 1789; and the wave of political upheaval that swept Europe in 1848 was at least in part a response to food scarcity, coming after three below-average harvests across the continent (Berger and Spoerer 2001).

Most of these conflicts occur in poor countries and so are less likely to influence AGI arms races among major world powers. Still, it seems plausible that the destabilizing consequences of environmental degradation are net harmful for compromise prospects among the big players in AGI development in the long term.

### Climate change

Crop cultivation contributes to climate change, which may be net bad on balance for global stability.

### How much could increased instability matter?

The contributions of increased agriculture to global instability are rather small. Could they really matter compared with the immediate suffering that might be prevented? In this section I suggest an illustrative back-of-the-envelope calculation to show how the adverse consequences on future stability might compete with the immediate potential benefit to insects.

Assumptions:

• In the Introduction, we saw that an American consumes 0.48 hectares of crop land, against a world total of (11%) * (13.4 billion) = 1.47 billion hectares. Suppose the marginal effect of an extra hectare of crop cultivation on resource conflicts of the future is constant. (Indeed, it might even be increasing if conflict rises sharply as resources become more and more scarce.)
• Maybe market or political forces help to offset the resource shortages that would otherwise result from an extra hectare of crop cultivation, so multiply the effects by a dampening factor, conservatively set at 0.1.
• The destabilizing effects of crop cultivation are cumulative over humanity's next few decades. Say that at most the next 300 years of crop cultivation matter to prospects for compromise on AGI. I'll assume the effect of a single year of crop cultivation is equal over all years, though in practice, the next few decades probably matter the most.
• Say the probability is only 1% that resource conflicts in general would lead to a nuclear-level war that wouldn't otherwise have happened. (This seems too low in view of conflicts over water between India and Pakistan, for example.)
• Furthermore, say the contribution of all the world's crop cultivation to those resource conflicts is only 5%. (Again, this is probably too low given that water scarcity is mainly an issue for agriculture.)
• Suppose that an additional nuclear-level war would increase chances for a future AGI arms race rather than global cooperation by an absolute 0.0001 probability.
• Say that a future with AGI arms race contains 2% more expected suffering than one with calm compromise. For example, in the absence of compromise, maybe the AGI builders would not take measures to avert suffering subroutines used for computational purposes.
• Let there be only a 1% chance that the compromise arrangements reached in the near future matter to the very far future, i.e., that goal preservation is successfully implemented.
• Suppose that only 0.01% of experiences in the future entail suffering, such as by suffering subroutines, sentient simulations of wild animals, etc.
• Nick Bostrom estimates potential future hedonic experience as 1038 humans surviving for ~1010 years in the Virgo Supercluster, or 1048 experience-years. (Given that insects have ~105 fewer neurons than humans, this could be ~1053 insect-level experience-years, but the brains of the future might not be insect-sized, and there's some possibility we don't want to count the bigger brains as more morally valuable than insect brains. So sticking with 1048 keeps the calculation conservative.)
• Say the probability that a colonization scenario with this many minds actually happens is 10‑8. Alternatively, we could say that the probability is 10-6 that a colonization future with this magnitude of computational power has 1% hedonically relevant computations. Any combination of possibilities that leads to an expected 10‑8 fractional multiplier is equivalent.
• Apply a significant discount factor to account for the anthropic consideration that there may very well not be a far future for us to influence. Given model uncertainty, probably this value shouldn't be too small. 10-7 seems conservatively low.

The result is that one year of an American's crop-cultivation impact causes

[0.48/(1.47 billion)] * 0.1 * (1/300) * 1% * 5% * 0.0001 * 2% * 1% * 0.01% * 1048 * 10‑8 * 10‑7 = ~105 life-years of suffering,

which is basically on par with the magnitude of insects affected in expectation by one person's crop-cultivation effects. In both calculations I was aiming to be conservative, but this latter one has many more parameters and so is arguably more conservative overall.

Needless to say, there are major problems with extreme Fermi calculations like these. I've only considered one extremely narrow possible pathway, while in fact there are many going in both directions. This calculation is not proof that the far-future effects outweigh the near-term ones, or even that the far-future effects are necessarily net bad (though it seems somewhat more likely they are than not); I'm merely aiming to show that far-future considerations could compete with near-term ones even just through the single causal chain suggested here.

In practice, our assessments should be more robust to a broad range of possible scenarios, but it seems intuitively that greater environmental instability does lead to more conflict and possibly worse futures across a broad range of scenarios. Moreover, many people are upset by the environmental and economic side effects of excess farming, so supporting more plant agriculture too aggressively could increase disdain for the effort to reduce wild-animal suffering, or even the effort to reduce suffering in general.

In view of these considerations, it may be best not to support greater crop cultivation even if we thought it had net short-term benefits to wild animals.

## All crops are not created equal

In the discussion so far, I've mostly been considering crop cultivation as an aggregate activity. Sometimes this alone is actionable, such as in our appraisal of growing more grain to feed to livestock, of recycling paper vs. harvesting new fiber from tree farms, and of growing new plants for biofuels. But in other cases, the differences among crop types are most important, especially in the choice of which foods we consume.

### Different animal impacts

From "The Collateral Damage of Vegan Foods" by Erik Marcus:

I strongly suspect there are a handful of vegan foods -- this list is probably led by rice and sugar -- that entail vastly more vertebrate killing than other foods. I'd love to see this information thoroughly researched and widely disseminated, along with a list of comparable foods that entail far less killing. For instance, I wouldn't be at all surprised if it emerged that a given quantity of millet could be produced with one-tenth the killing of the same quantity of rice. Nor would it surprise me if it turned out that many of today's vertebrate deaths could be prevented given the introduction of improved planting and harvesting methods. In both cases, people concerned with compassionate eating -- vegans and omnivores alike -- can be counted on to vote with their food dollar, once presented with reliable information.

I find it unlikely that different cereal grains would differ by an order of magnitude in impact when all factors are considered together (rather than just when looking at a single dimension), but I agree with the general sentiment: that those concerned with animal welfare should put more effort into exploring the impact of big-picture issues like crop cultivation relative to many other more minor tragedies that occupy them.

### Different sustainability impacts

Foods differ in their implications for global stability as well as their impact to wild animals. As just one small example, rice has nontrivial implications for water and climate change:

Rice requires slightly more water to produce than other grains.[86]

Long-term flooding of rice fields cuts the soil off from atmospheric oxygen and causes anaerobic fermentation of organic matter in the soil.[87] Methane production from rice cultivation contributes ~1.5% of anthropogenic greenhouse gases.[88]

The environmental community has already looked at differential impacts of foods from a sustainability perspective. For example, "Fruit and Vegetables & UK Greenhouse Gas Emissions: Exploring the Relationship" has as its third section: "Assessing Greenhouse Gas Emissions By Product Type." Similarly, Table 4.1 of A.Y. Hoekstra's "Virtual water: An introduction" compares water use by crop type. (In general, the environmental community is more quantitatively minded in its analyses than the animal-welfare community has been so far. Partly this reflects the greater numerosity of environmental advocates than animal advocates, but it may also reflect a fundamental difference in outlooks between the two fields. The animal-welfare community would benefit by adopting more of the methodology of the environmental community, without also adopting its ethics.)

### Sample research questions

In assessing the animal impacts of foods, there are many dimensions to explore further. For example:

• Conventional vs. organic? (conventional probably results in less insect suffering?)
• Beans vs. rice? (I personally buy only beans and avoid rice because of the factors discussed above)
• Which fruits/vegetables are best?
• How does country of origin matter? (What kinds of ecosystems are being replaced? How much does farming affect long-term food security in that country?)

An ideal scenario would be to find particular foods that have low sustainability impact and also high impact on reducing wild-animal suffering.

### Policy and technology changes

That said, it's important not to become obsessed with individual consumer choices. Probably most of the gains in the end lie in shaping government policy or developing better agricultural technologies. There can be a tendency to hyper-focus on the thought that "this is something I can control, so I have to choose ethically" without feeling similar ethical obligation to contribute toward big-picture changes that ultimately have more expected impact. Of course, it can be good to think about both personal decisions and broader contribution to policy at the same time. For some people, the personal ethical choices help keep their minds on the bigger questions. As an example, vegetarians are often reminded of animal suffering because their everyday purchasing decisions are shaped by it.

## No-till farming

Mother Earth News (1984) explains:

In conventional tillage, the earth is turned to a depth of 8 to 12 inches with a plow, most commonly one of the moldboard variety. Subsequently, the plot is disked at least twice more to prepare the seedbed before planting takes place. In no-till, however, the first three steps in conventional cultivation are dispensed with. Planting is done right through the residues of previous plantings and weeds with a device (usually a coulter) that cuts a slot a few inches wide, followed by equipment that places the seeds and closes the trench.

Following are some pros and cons of no-till farming from the perspective of wild-animal suffering.

Pros:

1. This page says: "no-till management results in fewer passes with equipment". Fewer passes through the field with mechanical equipment mean fewer vertebrates and invertebrates killed. This page says no-till farming has a "reduced chance of destroying ground nesting birds and animals (plowing destroys all of them)."
2. Less need for irrigation means fewer zooplankton are killed by removing irrigation water from rivers and reservoirs.
3. Herbicide use to kill weeds reduces weed net primary productivity (although tilling weeds into the soil would also reduce weed growth).
4. Mother Earth News (1984): "One of the main problems with conventional agriculture's heavy use of nitrogen fertilizers is the leaching of these compounds into surface water during runoff. By retaining rainfall, the untilled field also better holds the chemicals that have been applied to it, thereby decreasing their pollution potential." Reduced eutrophication may mean lower populations of aquatic organisms.
5. This page suggests that no-till fields might store more soil organic carbon, although it's not clear how reliable this finding is, because "a growing body of research is showing that no-till systems lose carbon stocks over time." Insofar as no-till fields do store more soil organic carbon, they delay decomposition of organic matter and thereby delay feeding populations of soil heterotrophs with that organic matter. That said, assuming the organic matters gets eaten eventually, it's not clear whether this consideration is actually relevant to total long-run suffering.

Cons:

1. Mother Earth News (1984): "There is no question that wildlife is damaged or killed by certain herbicides. Birds can be injured if they get in the way of spraying (small amounts of paraquat can kill eggs)".
2. Mother Earth News (1984): "Yields are generally at least as good with no-till agriculture as they are with plow techniques." Sadly, higher crop yields probably imply more wild-animal suffering.
3. Reducing soil erosion and otherwise improving soil health mean higher yields in the long run.
4. Like herbicides, tilling also kills weeds and reduces their net primary productivity.
5. This page says regarding no-till: "because of the higher water content, instead of leaving a field fallow it can make economic sense to plant another crop instead." This would presumably imply higher net primary productivity per year.
6. This page reports: "No-tilled fields often have more beneficial insects and annelids," and "No-till farming increases the amount and variety of wildlife."
7. Mother Earth News (1984): "in some cases, where crop residues harbor insect pests, the use of insecticides may be greater, as well" in no-till farming. If no-till crop fields allow more insects to survive the winter and/or hatch in spring, this seems bad because it increases total insect populations.
8. Mother Earth News (1984): "no-till farming goes a long way toward eliminating water runoff, so" there may be less runoff of algae-stunting herbicides into water bodies.

Unclear:

1. Reduced greenhouse-gas emissions. Plumer (2013): "The UNEP estimates that no-tillage operations in the United States have helped avoid 241 million metric tons of carbon-dioxide since the 1970s. That's equivalent to the annual emissions of about 50 million cars."

On balance, I would guess that no-till farming sadly causes an overall increase in wild-animal suffering assuming it does increase crop yields and soil invertebrate populations, but I haven't explored this topic thoroughly.

Logsdon (2010) claims that some ostensibly no-till farming isn't really no-till:

most of the time, “no till” is a big fat fib. What it really means is “no moldboard plow” and because of that we are supposed to believe that farmers are controlling soil erosion. Instead of the plow, farmers work up the soil with a variety of disks, chisel plows, field cultivators and turbo tillage tools. When I point out that they are tilling the soil as much as they ever did, in fact more in some cases, with just about as much subsequent soil erosion, they look at me blandly, like I am speaking a foreign language. They don’t want to hear that. They are determined to believe, along with their university and USDA partners, that they are controlling erosion simply because they quit using moldboard plows and use no-till planters.

The pretension reaches hilariously ludicrous proportions. For instance, in “Farm and Dairy” magazine in the latest issue, there is an article titled: “No-till All the Way.” Immediately above it is a photo of the farm where no-till is being practiced “all the way.” In the photo, behind the farmstead buildings, stretch acres and acres of soil as tilled and bare as a desert.

Some comments on the Logsdon (2010) blog post dispute this picture.

## Cover crops

Cover crops seem likely to increase animal suffering for two reasons:

1. Cover crops improve soil fertility and help the main food crops grow better. Higher yields for food crops imply more total plant matter and thus larger animal populations.
2. When grown over winter, cover crops increase the total amount of plant production that occurs during the year, compared against covering farm fields with non-living material or leaving them uncovered.

In one study, researchers compared arthropod and songbird species composition and field use between conventionally and cover cropped cotton fields in the Southern United States. [...] During the migration and breeding season, they found that songbird densities were 7–20 times higher in the cotton fields with integrated clover cover crop than in the conventional cotton fields. Arthropod abundance and biomass was also higher in the clover cover cropped fields throughout much of the songbird breeding season, which was attributed to an increased supply of flower nectar from the clover.

## Indoor farming

Growing crops indoors would seem to reduce collateral damage of crop cultivation and harvesting on small vertebrates and invertebrates. Because the crop-growing buildings still occupy space, indoor crop cultivation may still reduce wild-animal suffering via habitat appropriation and diminished primary productivity per hectare of land (unless the farms are underground).

However, insofar as indoor farms can be more space- and resource-efficient than outdoor farms, they may not reduce as much wild-animal suffering per unit of food produced as conventional farms do. This is particularly plausible for vertical farms, which dramatically reduce land use per unit of output food and might even increase primary productivity per hectare of land. More primary production per hectare means less primary production needs to be appropriated from other land, which means more production is available to wildlife, which is generally bad. This piece argues that relative to a(n overly) simplistic framework, lower crop yields always reduce insect populations (and hence insect suffering).

Of course, some land freed from use by farmers could instead be used for building houses and parking lots rather than being restored as wilderness. But the fraction of land that would be covered over in this way seems small to me because

1. Relatively little of Earth's land area is covered by buildings and roads at present. This discussion estimates that human structures cover 1-3% of the Earth's land area, while crop cultivation covers 11%. If the global land area devoted to crop production decreased from 11% to 10% of Earth's land area, the amount of land freed up might almost equal the total amount of land already covered by buildings and roads.
2. In rural areas, land is a small fraction of the cost of constructing new buildings. In big cities, land might be expensive because of "location location location". But in the countryside, which is where no-longer-needed farmland would be located, land is cheap. For instance, in 2016, "The Iowa Realtors Land Institute farmland survey through March shows the average value of cropland fell to $6,732 an acre". I browsed on Zillow for homes in Iowa to look at their lot sizes and list prices. Lot sizes varied from ~0.2 acres to ~10 acres, and house prices from ~$100K to ~$500K (with little obvious correlation between lot size and house price). Suppose you wanted to build a$340K house on 5 acres of farmland. It would cost only $6,732 * 5 =$34K to buy the land, just 10% of the selling price. The other 90% of the cost comes from actually building the house. Now, to take an extreme illustration, suppose we had infinitely efficient indoor farms that yielded all the food the world needed on a trivial amount of land area. Suppose farmland prices then drop to \$0 because that land no longer needs to be farmed. Even then, house prices would only decline ~10% (ignoring effects on other component costs, like wages for house builders), and assuming the price elasticity of demand for houses is not vastly far away from 1, the quantity demanded of houses would only increase ~10%.d Suppose that buildings currently cover 1% of Earth's land area. A 10% increase in buildings would mean only an extra 0.1% of Earth's land covered over, which is very small compared with the 11% of Earth's land that would be freed up from crop cultivation if we had hypothetical crops with infinitely high yields.

Most land can't be used profitably in ways that cover it over (buildings, parking lots, etc.), because if it could be, wilderness land wouldn't be as cheap as it is.

## Ranking foods

### General criteria for ranking foods

Following are some criteria that affect how good/bad different food types are per calorie:

• Country of origin. Crops grown in locations that would have otherwise been lush habitat prevent more suffering than those grown in less fecund areas. So, for example, rainforest-grown soybeans are better than locally grown corn.
• Quantity. If crop cultivation is net good (which is unclear but maybe more likely than not), then crops that require more land per calorie are preferred (e.g., cucumbers are better than bread?).
• Biomass density in the fields. Crops where the fields contain less biomass in total (from both the crops and also weeds) imply less food for insects to eat. Crops where herbicides are used may be good to prevent weed biomass.
• Insecticides? The sign of these is less clear. Insecticides reduce insect populations in the short run, but in so doing, they leave around biomass that may later be eaten by other insects. Of course, much of this biomass will instead be eaten by humans, cows, or bacterial decomposers, so it may be best to kill an insect eating a crop and hope the uneaten biomass is removed in another way than by some other insect eating it. This suggests that insecticides could be net good, especially if they're broad-spectrum and also inhibit soil insects. Still, they may cause higher turnover of insects and so may cause slightly higher death rates in the short term, though probably not a comparable amount to what happens naturally due to r-selection. It's also important to remember that a plant type where higher insecticides are used may not be better because this may suggest that the crop requires more insecticides because it's so juicy and attracts so many bugs. So don't pick a crop just because it requires more insecticides. However, for a given plant type, higher insecticide use seems plausibly better (e.g., conventional better than organic).
• Collateral damage to mice/insects/etc. Maybe shorter, less dense crops are less likely to be inhabited by small mammals, leading to reduced mortality at harvest time. I would conjecture that wheat and other grains kill lots of mammals because these crops are easy to nest in? Note that given a crop type, it's unclear whether more collateral damage would be net good (by reducing future populations) or net bad (by increasing the number of painful deaths per unit time), but it seems better to have crops that support fewer total mammals/birds in the first place, and the extent of collateral damage is one reasonable proxy for total animal populations.
• Above or below the ground? Anecdotally, I've noticed with crops grown in my housemate's garden that those above ground can end up with lots of bugs on the leaves (especially broccoli, kale, collards, brussel sprouts, etc.), while those below ground may be bug-free (e.g., beets). This weakly suggests that below-ground plants might support fewer insects in general, which would be good. However, this trend isn't universal: my housemate's below-ground carrots also had bugs in them, while above-ground tomatoes and squash were pretty bug-free.e Of course, which particular pests are problematic for which crops will depend on many details of the farm, but some crops are probably in general more tasty for bugs than others.

### A specific attempt at ranking

Following is just an illustration of how an analysis of these factors might begin. It's not rigorous and merely reflects some extremely rough guesses of mine based on watching videos of how crops are harvested. I've evaluated only two factors to start with:

1. biomass density on the crop fields (lower productivity on crop fields is plausibly always better)
2. likelihood of killing rodents, worms, etc. during harvesting (I assume this is bad and is more likely for dense, grassy crops).

I use green to mean "good" and red to mean "bad". The following table is ordered based on a very rough guess of best foods at the top and worst at the bottom.

 Food Crop density per hectare Collateral damage Videos Comments Peanuts very low low 1, 2 Almonds [might be bad due to irrigation] very low very low 1, 2 The nuts are harvested in a forest with no underbrush and hence probably few insects. It also looks very easy for any mammals present to escape. However, almonds require lots of water to grow, and insofar as irrigation plausibly increases net plant growth, this could make almonds a bad food choice all things considered. Kidney beans very low low 1, 2 Lentils very low low 1, 2 Carrots low low 1, 2 Digging in the soil hurts plenty of critters there, but probably this cost isn't huge compared with the benefits of the low biomass density of carrot fields. Also: anecdotally, carrots in my housemate's garden sometimes had lots of larvae in them, suggesting that bugs like eating carrots if they can, which is bad. Cucumbers low low harvest: machine, by hand Apples, oranges, bananas, etc. low/medium low apples, oranges, bananas These fruits seem to be harvested either by hand or by a machine that touches the tree branches. A few apple-harvesting videos showed brushes scraping the ground, which might harm more bugs. I'm slightly more worried about fruit cultivation than vegetable cultivation because fruits are very juicy and so seem perhaps more attractive to bugs? Lettuce low/medium low? 1, 2 Most videos show lettuce being harvested by hand, which may imply low collateral damage for mammals and moderate damage for bugs that get stepped on. Some harvesting is being done by machines. The mechanization trend might reduce net insect crushing because fewer human feet are stomping on the field, although a machine harvester may crush more surface area than the machine that's used to collect lettuce harvested by hand. Crushing by machine tires is probably less painful than crushing by human feet because the machine is heavier. Machine harvesting presumably hurts mammals in the fields, but there probably aren't large numbers of them. Tomatoes medium medium 1, 2 Note: Tomatoes may cause harm to bugs by killing them for nutrients. This might increase our assessment of tomato harm a bit, though (1) I'm not sure how big the number of insects killed this way is; it may be small on insecticide-sprayed tomatoes and (2) it's not clear if tomato killings are more painful than insecticide killings, and insecticide killings may be net good anyway. Tomatoes would presumably be worse if they attracted more bugs to their fields than other crops do, since this would imply more insect killings without necessarily producing long-term reductions in insect populations away from the tomato plants assuming the regions neighboring the tomato crops still reach insect carrying capacities. Chick peas medium medium 1, 2, 3 Soybeans medium medium? fields: 1; harvest: 1 Broccoli medium medium/high? harvest: by machine, by hand Anecdotally, in my housemate's garden, broccoli ends up having lots of caterpillars on its leaves, and I find that Brassica oleracea plants in general (broccoli, kale, Brussels sprouts, etc.) host tons of aphids. Of course, on commercial farms, such bugs will be killed, but it's still unfortunate that so many bugs show up at all and therefore will get killed.f Sunflower seeds low high 1 This article explains how tons of seed-eating birds may be poisoned in the cultivation of sunflower seeds. Oats medium/high medium? 1, 2 At this point in one video, you can see a rabbit run away for safety :( Wheat high high 1, 2 Rice very high high 1, 2, 3 Corn very high high?? growing: 1, 2; harvest: 1, 2 Corn has one of the highest numbers of calories of edible food per acre: ~15 million. This suggests it probably also has high total biomass per acre, although the ratio of total biomass to human-edible biomass is not constant among food crops.

My assessment for the collateral damage of crops roughly correlates with my assessment of crop densities because (a) taller plants seem more likely to be home to bugs and bunnies and (b) it's harder for animals to escape tall, grassy plants. I saw no mammals in the fields of any of the sparse, low-density plants in any of the videos.

Note that I haven't considered how much land is required to produce a calorie of the different foods. I've been basically assuming it's constant for all crop types. This assumption is obviously inaccurate, but it may not be wildly wrong given that, e.g., vegetables have many fewer calories per gram but vegetable plants probably produce many more grams of food per hectare than is the case for nuts and beans. If crop cultivation is net good, we'd prefer foods that produce fewer calories per hectare so as to increase the number of hectares farmed; if crop cultivation is net bad, we'd prefer foods that produce more calories per hectare.

### General trends

The most striking pattern in the above table is that nuts and legumes (peanuts, kidney beans, chick peas) have much lower densities than cereal grasses (oats, wheat, rice, corn). So one practical recommendation that seems to emerge from this analysis is to eat more beans and nuts, rather than bread/cereal/pasta/rice. Of course, if you have time, try looking up videos for the specific foods you're considering.

One caveat is that pure biomass on the fields may not be a fully adequate metric. For example, peanuts and beans are more dense with utilizable energy than grass is. Assuming bugs can't digest fiber very well, a kilogram of peanuts would feed many more bugs than a kilogram of corn stalks. That said, most peanuts and other energy-dense parts of plants should be harvested and fed to mammals rather than insects. Another reason above-ground biomass is not a perfect measure is that some plants may have more substantial below-ground biomass than others. This is obvious in cases like carrots and potatoesg, but perhaps it applies to other plants as well.

A caveat about legumes is that "legumes will often increase the nitrogen content of nitrogen-poor soils." In fact: "As a result of extensive cultivation of legumes (particularly soy, alfalfa, and clover), growing use of the Haber–Bosch process in the creation of chemical fertilizers, and pollution emitted by vehicles and industrial plants, human beings have more than doubled the annual transfer of nitrogen into biologically available forms.[10]" This productivity-boosting effect of legumes may be especially relevant in poorer regions without as much access to chemical fertilizers. Increasing plant productivity seems to be one downside of legumes; I don't know if, or how much, it should detract from the benefit that legumes seem to have lower productivity per hectare. Plus, it's unclear if nitrogen fixing matters that much for legumes grown in richer countries that can just use artificial fertilizers? (That said, nitrogen-fixing crops could reduce use of artificial fertilizers and might thereby reduce energy use.)

Another caveat about legumes is that, as noted earlier, they tend to be more digestible than grasses, so even if legume biomass is lower, maybe edible biomass isn't?

One more worry is that, as a rough generalization, many fruits and vegetables are grown in California using irrigation, while many grains are grown in the Midwest with less(?) irrigation. So it may be that, for example, wheat and corn grown in the Midwest don't really increase NPP relative to the NPP of native grassland, while fruits and vegetables grown in California do increase NPP relative to the minimal vegetation that would exist in the absence of irrigation? If so, then the generalization that fruits/vegetables are less bad than grains would be reversed!

These questions highlight the importance of further investigation. It would also help to find studies that actually measure crop and native-vegetation NPP by crop type and region.

### NPP of different crops

This book reports (without citation) the following rough figures for a good yield of the following crops, which I converted from pounds/acre to Mg/ha:

 Commodity Aboveground dry NPP (Mg/ha) Oats 6 Corn 9 Sugar cane 17

#### Prince et al., 2001

This study reports the following NPP values by crop type for Richland, North Dakota in 1992:

 Commodity NPP (Mg/ha) Soybean 4.639 Corn silage 5.158 Sunflower 6.958 Hay 7.114 Oats 7.448 Corn grain 9.514 Wheat 9.914 Barley 10.940 Woodland 12

This study of the US Great Plains did a linear regression of a region's change in NPP due to crop cultivation (dependent variable) against the proportion of the county devoted to each of several crop types (independent variables). As the study explains:

The crop coefficients produced by this process allowed us to identify both the direction of the influence that each crop has on productivity (positive sign indicates that the crop increases productivity and negative sign suggests that the crop decreases productivity), and the magnitude of the influence (since the crop variables are all proportions of the county, the size of the coefficients indicate the magnitude of the effect).

The study's results table (p. 1870) for total (aboveground + belowground) NPP was as follows:

The authors explain:

Total NPP change shows a very large positive correlation with corn, and smaller positive effects of sorghum and wheat [...]. On the other hand, NPP is negatively related to soybean and hay, with soybean indicating a large negative effect and hay a much smaller effect.

These trends roughly match what I guessed by eye based on videos of crop harvesting as described above.

### Yield and harvest index

Suppose you buy 1 kilogram of a given food product per year. Different crops have different yields (kg of food produced per hectare per year). Suppose a given crop has economic yield of EY kg of food per hectare-year. Then the food you eat of this crop in a year requires 1/EY hectare-years.

The harvest index (HI) of a crop is (mass of food output) / (mass of total aboveground biomass). So an annual food crop with yield of EY kg/(hectare-year) has aboveground net primary productivity of EY/HI kg/(hectare-year).h

Let RS be the root-to-shoot ratio for the crop. Then belowground net primary productivity is RS * EY/HI.

Total net primary productivity for the crop (belowground + aboveground) is then (RS + 1) * EY/HI.

HI is the fraction of aboveground production eaten by humans or livestock (ignoring food waste). So the amount of productivity eaten by big animals is (aboveground net primary productivity) * HI = (EY/HI) * HI = EY kg/(hectare-year). Thus, the amount of net primary productivity not eaten by big animals is (RS + 1) * EY/HI - EY.

Now let's switch to considering native vegetation. Suppose the total net primary productivity of native vegetation that would grow if the crop weren't planted is N kg/(hectare-year). Suppose that a fraction f of this vegetation is eaten by small organisms (while 1-f is eaten by big grazing herbivores).

Assume that insect populations are roughly proportional to net primary productivity not eaten by big animals. Then the change in insect populations that results from growing a crop instead of leaving the vegetation native is proportional to [(RS + 1) * EY/HI - EY] - N * f insect-years/(hectare-year). And since your eating of 1 kg of the food per year causes 1/EY hectare-years per year of crop cultivation, your total impact is proportional to this many insect-years per year:

([(RS + 1) * EY/HI - EY] - N * f) * 1/EY = [(RS + 1)/HI - 1] - N * f/EY.

Let C be the total net primary productivity of the crop. Then the aboveground net primary productivity is C/(RS + 1), and EY = HI * C/(RS + 1). Then the above expression becomes

(RS + 1)/HI - 1 - N * f * (RS + 1)/(HI * C).

Defining P := (RS + 1)/HI, that becomes

P - 1 - P * N * f / C.

If our goal is to reduce insect suffering, we want this expression to be as negative as possible, since that means the reduction in insect-years of suffering is as big as possible.

The expression is negative (i.e., insect suffering is reduced by crop cultivation) iff

P - 1 - P * N * f / C < 0
P - 1 < P * N * f / C
N * f / C > 1 - 1/P

This page gives values of RS and HI for various crops (where "Allocation" is RS and "Harvest Index" is HI). For example:

 Crop RS HI 1 - 1/P Barley 0.50 0.49 0.67 Beans 0.26 0.55 0.56 Corn 0.18 0.52 0.56 Soybeans 0.15 0.42 0.63

This means that, e.g., in order for growing barley to reduce insect suffering, it must be that N * f / C > 2/3. If f is, say, 2/3, then crop cultivation is good iff N / C > 1, i.e., if native vegetation would be at least as productive in total (aboveground + belowground) as the crop is. This may or may not be the case, as is discussed previously in this piece.

A further refinement of these calculations would consider whether different fractions of biomass not eaten by large herbivores end up feeding insects between native vs. crop vegetation. For example, if some crop residue is burned, that fraction doesn't feed insects. On the other hand, native grassland may also burn naturally.

## Acknowledgements

Thanks to a few friends for pointing out to me some of the sources cited in this piece.

## Footnotes

1. While I'm not an expert, I have a few comments on the study methodology.

The authors apparently got native-vegetation NPP estimates from a USDA STATSGO database (p. 1864), while crop NPP estimates were based on economic-yield numbers from the USDA National Agricultural Statistics Service, combined with several parameters taken from the literature (p. 1865). Since the two input data sources are so different, and since calculating NPP from economic yield probably has reasonable error, it's not at all obvious that the results of this study are correct rather than just resulting from different data sources? But maybe the authors and reviewers of the paper have reason to think that the numbers are actually quite accurate? And in any case, noisy estimates are better than no estimates.

Secondly, I don't think the study controls for correlation vs. causation? Maybe the land where crops are planted is naturally more productive than land where crops aren't planted, and this partly explains the higher NPP of some crops than the NPP of native vegetation. I would guess that this effect wouldn't be huge, though, since I assume that soil productivities don't vary dramatically within a single county.  (back)

Too much of a nutrient can also have a limiting impact on a community. Recent studies have shown that excess nitrogen from human activities such as agriculture, energy production, and transport have begun to overwhelm the natural nitrogen cycle. [...] In terrestrial ecosystems, nitrogen saturation can disrupt soil chemistry, leading to loss of other soil nutrients such as calcium, magnesium, and potassium. This means that while the nitrogen is not a limiting factor, it causes other nutrients in the soil to become limiting factors.

3. The study makes this point using its own terminology (which you can read about in the study itself):

in Western Europe, the high total HANPP of 40% coincides with only a small ΔNPPLC because of its high-yielding, intensive agricultural systems. By contrast, in Eastern and Southeastern Europe, with similar ecological conditions, land use has caused a large ΔNPPLC and harvests are low. In Central Asia and the Russian Federation, most HANPP is actually due to a reduction in productivity; the situation is similar in sub-Saharan Africa. The situation in Eastern Asia (including China, Japan, and Korea), in contrast, is characterized by negligible ΔNPPLC but large total HANPP.

4. Actually, assuming the supply curve of houses is somewhat inelastic, a decline in house-construction costs of 10% would translate to a decline in equilibrium house prices of less than 10% because increased quantity demanded for houses would push prices back up somewhat. Hence, the equilibrium increase in quantity demanded of houses would probably be less than 10%.  (back)
5. Another problem with bug-ridden plants is that it can be hard to get all the caterpillars, aphids, etc. off, and if you don't, you might boil them -- which seems like one of the worst ways to die. Or even if you don't boil the bugs, you might eat them alive.  (back)
6. This killing isn't collateral damage during harvesting specifically, but I didn't have anywhere else to include this speculation about insect-pest deaths, so I slightly increased the rating in the "Collateral damage" column relative to what one would conclude from the harvesting videos alone.  (back)
7. What are my thoughts on potatoes compared with other crops? I haven't studied the matter in detail, but my first-pass guess is that they're at least better than corn, wheat, oats, etc. Aboveground biomass for potatoes is quite small, but that's because most of the biomass is belowground, so I don't know how total (aboveground + belowground) NPP compares with NPP of other crops. Potatoes have a high harvest index, which is good because it means a lot of the primary productivity of potatoes is eaten by humans or livestock rather than bugs:

Among cultivated plants, potato is characterised by the highest values of harvest index. Contemporary potato cultivars, grown in temperate climate zones, in favourable agrometeorological conditions reach the most often HI values, for the final crop, within the range of 0.70 to 0.85 [Jefferies and MacKerron 1989, Beukema and Zaag van der 1990, Vos 1997, Belanger et al. 2001]. In literature, it is possible to find reports even about 90% (HI = 0.90) biomass distribution to tubers [Victorio et al. 1986, Beukema and Zaag 1990]. For comparison, in cereals HI values between 0.3 and 0.6 dominate [Hay 1995]. Results presented in this work, contained within the range of 0.7-0.8 (Table 4-5), are thus consistent with the ones presented in the above quoted literature.

My housemate grows potatoes in her small garden, and when they're harvested, they don't seem to have huge numbers of bugs on them, which is another good sign.  (back)

8. This equation and that in the next paragraph is based on p. 1865 of this study.  (back)