Jump to ContentJump to Main Navigation
Darwinian AgricultureHow Understanding Evolution Can Improve Agriculture$

R. Ford Denison

Print publication date: 2012

Print ISBN-13: 9780691139500

Published to Princeton Scholarship Online: October 2017

DOI: 10.23943/princeton/9780691139500.001.0001

Show Summary Details
Page of

PRINTED FROM PRINCETON SCHOLARSHIP ONLINE (www.princeton.universitypressscholarship.com). (c) Copyright Princeton University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in HSO for personal use (for details see http://www.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 15 August 2018

Stop Evolution Now!

Stop Evolution Now!

(p.164) 10 Stop Evolution Now!
Darwinian Agriculture

R. Ford Denison

Princeton University Press

Abstract and Keywords

This chapter considers ongoing evolution, particularly as it relates to control of agricultural pests. It begins with a discussion of how weeds evolved resistance to herbicides, focusing on the case of watergrass. It then examines the high dose/refuge strategy for slowing the evolution of pesticide resistance, along with the experience of Australian cotton farmers with this approach. It shows that cooperation among Australian cotton farmers was key to the relatively successful management of Bt (Bacillus thuringiensis) resistance. The chapter also explores two different ways in which nature can serve as a source of ideas for improving pest control in agriculture: comparing natural ecosystems and studying the pest-defense strategies of individual wild plants.

Keywords:   evolution, pests, weeds, watergrass, refuge strategy, pesticide resistance, cotton farmers, pest control, agriculture

Up to now, I have emphasized agricultural insights from understanding past evolution, but evolution continues today. Weeds and insect pests, in particular, can evolve quickly, with major effects on agriculture. For example, over 180 weed species have evolved resistance to a variety of weed-killing herbicides.316 This ongoing evolution will be the main focus of this chapter.

Wily Watergrass

We have already developed some effective strategies for slowing the evolution of resistance to our pest-control measures. Consistent with our ideas-from-nature theme, we might get additional ideas from studying wild plants and their interactions with the insects that plague them. But let’s start with watergrass, a weed that has evolved resistance to three very different control methods.

Watergrass is a weed that evolved in flooded Asian rice fields, a human-made environment that has been in existence for only a few thousand years. The immediate ancestor of watergrass was barnyardgrass (Echinochloa crus-galli), a weed found in nonflooded fields. Barnyardgrass roots are killed by flooding, but its descendant watergrass has air channels in its stems that supply oxygen to its roots. This is a more impressive innovation than anything biotechnology has done so far, but it’s only the beginning of watergrass’s evolutionary story.

Flooding tolerance was critical to survival in rice fields, but flooding was not the only risk there. One problem for barnyardgrass, even once it evolved flooding tolerance, was that it looks rather different from rice. Both are grasses, so the average city-dweller might have difficulty telling them apart, but to a Chinese rice farmer with a hoe the difference was obvious.

As with most traits, there was some random genetic variation in the appearance of barnyardgrass plants. The plants that escaped the hoe tended to be those that looked more like rice. At first, those survivors probably looked more like barnyardgrass than like rice. However, even a slight (p.165) resemblance to rice was sometimes good enough, if a farmer had poor eyesight or was working in twilight. A few random mutants among the descendants of these first survivors looked even more like rice, fooling a larger fraction of farmers, and so on.

Eventually, some of these weeds hitched a ride to California, maybe on the shoes of travelers. Have you ever been asked, crossing an international border, whether you visited a farm on your travels? That’s why.

In the 1970s, biologist Spencer Barrett and colleagues did a quantitative comparison of visible traits of rice, barnyardgrass, and watergrass in California rice fields. Remarkably, watergrass resembled rice more than it resembled its own recent ancestor, barnyardgrass.317 This was attributed to selection previously imposed, inadvertently, by farmers with hoes in Asian rice fields.

Today, California rice farmers rarely go out and hoe individual plants. They rely mainly on flooding, tractor cultivation, and herbicides. So, how much has the appearance of watergrass plants evolved in the decades since Barrett’s group made their measurements? Has natural selection changed the appearance back toward barnyardgrass? Or has evolution of visible traits wandered off in some new direction? I think it would be interesting to check.

Evolution continues. Although evolution has had millions of years to improve photosynthesis and water-use efficiency, as discussed in earlier chapters, evolution need not take millions of years. For example, weeds can evolve herbicide resistance in only a few years, as shown in a photo I took at a UC Davis “Weed Day” event.

The photo shows watergrass from various California rice farms. Each vertical column of pots in figure 10.1 was treated with a different herbicide, except for the middle column, which was not sprayed. Notice that the watergrass plants grew poorly or died in the two front rows, where sprayed with the herbicides Whip or Abolish, whereas plants in the back two rows of those same columns survived those same herbicides.

The difference is that soil in the two front rows came from two rice fields being farmed organically. No herbicides are used on those farms, so the watergrass seeds in those soils came from plants with no recent evolutionary history of herbicide exposure. Pots in the back rows have soil from farms where herbicides are used. In those fields, only the few mutants that were resistant to those herbicides survived and reproduced. So weed populations evolved resistance on those farms, sometimes in only a few years.

The evolution of herbicide resistance is a widely recognized problem. But it’s important to remember that weeds can also evolve resistance to other, nonchemical methods of weed control. For example, watergrass also evolved resistance to flooding and to detection by farmers with (p.166)

Stop Evolution Now!

Figure 10.1. Evolution of herbicide resistance in weeds. Research by Albert Fischer and James Hill, of UC Davis, found that watergrass from populations not previously exposed to an herbicide were more susceptible to it.

Photo by R. F. Denison.

hoes. Similarly, yellow foxtail growing in alfalfa fields, which are mowed several times a year, evolved to escape mowing damage. It did this through genetic changes that made it grow along the ground, below the cutting height.318

Slowing Evolution

We probably can’t stop evolution of resistance to our pest-control measures, but slowing it down can give us time to develop new control methods. To do this, we need to understand the factors that control the rate of evolution.

The rate of evolution depends on how much genetic variation there is within a population. It also depends on the intensity of selection. The intensity of selection for a given allele is the extent to which individuals with that allele have greater survival and reproduction than those with other alleles. More intense selection leads to faster evolution.

On the organic farms that were the source of the herbicide-susceptible watergrass in figure 10.1, the intensity of selection for herbicide resistance was zero. It’s not that weeds had 100 percent survival there, but (p.167) that, without herbicides, resistant weeds were no more likely to survive than nonresistant ones. (An herbicide-resistance allele might even impose some fitness cost, in the absence of the herbicide.) The intensity of selection for herbicide resistance was higher on the conventional farms, where herbicides were used. There, herbicide-resistant weeds were much more likely to survive than susceptible ones were. This concept is widely understood.

What is less-well understood is that increasing herbicide use may sometimes reduce the intensity of selection. This can be true, for example, for resistance genes that protect against low but not high doses of herbicide. With a high enough dose, such genes confer no increase in survival, so they don’t become more common over generations. Higher herbicide doses may endanger humans or wildlife, so it’s not necessarily a good option, even if it would slow the evolution of herbicide resistance. But remember this example, when I discuss evolution of insect resistance to the chemical defenses of transgenic crops.

Another reason that increased herbicide use could sometimes slow the evolution of herbicide resistance is that the rate of evolution depends on the amount of inherited variation within a population. (One plant may be more resistant than another because it’s in a wetter or drier spot, but it’s hereditary differences that matter to evolution.) Larger populations usually include more inherited variation in every trait than small ones do. A one-in-a-million mutation is much more likely to occur in a population of a billion plants than in a population of a hundred plants. So, if a farmer can keep weed populations in a field low enough, perhaps using a combination of herbicides and mechanical cultivation, herbicide resistance is less likely to evolve.

The evolutionary importance of population size also has implications for testing new pesticides or pest-resistant crops. Small-scale tests will have lower total pest populations, so fewer potentially resistant mutants, than large-scale commercial production will. So even after a pesticide or transgenic pest-resistant crop has passed initial tests and been approved for sale, ongoing monitoring to detect resistant pests is needed.

With this background, how can we explain the herbicide resistance shown in figure 10.1? At some point, an herbicide-resistant mutant arose (probably, but not certainly, in a field with a relatively high weed population) and survived long enough to make seed. Would a higher herbicide dose have killed this mutant? Possibly, although adding a different herbicide would be even more likely to work. (Similarly, a life-threatening bacterial infection in humans can be brought under control by a combination of antibiotics that work in different ways.)

A nonchemical method, such as flooding (or cultivation, if feasible) would be even better. One reason is that a single mutation sometimes (p.168) gives resistance to two or more different herbicides, a phenomenon known as cross-resistance. Sometimes we can figure out why, after the fact, but it isn’t always possible to predict cross-resistance in advance.

Decisions about herbicide use depend on economic and ecological considerations as well as evolutionary ones. A higher herbicide dose may sometimes slow the evolution of resistance, but it will cost more, and it will often have more negative effects on the environment. Using less herbicide isn’t the only way to reduce those environmental effects, however. For example, rice farmers in California decreased herbicide pollution of rivers by 98 percent, not by reducing use, but by slowing the movement of water from their fields into rivers, giving herbicides more time to break down.319

Slowing Resistance with the High Dose/Refuge Strategy

To summarize, pests tend to evolve resistance to all of our control measures, not just to chemical pesticides, and using less pesticide is only one of the possible ways to slow evolution of pesticide resistance. Here’s a detailed example of a more-sophisticated approach.

The transgenic pest-resistant crops that have been most widely used and studied up to now are those containing protein toxins from the bacterium Bacillus thuringiensis, or Bt. This protein kills insects that eat crops containing it. Because it mainly affects caterpillars and has to be eaten to have much effect, it is mostly harmless to other species.

Mostly, but maybe not entirely. One possible ecological risk from Bt crops comes when the pollen they produce lands on other plants, such as the milkweeds that are eaten by Monarch butterfly larvae. Negative effects of pollen from Bt plants on Monarchs and swallowtail butterflies have been reported,320,321 although these results have been disputed,322 and any negative effects may be less serious than the alternative of sprayed insecticides.

To the extent that Bt crops have economic and possibly ecological benefits, those benefits would be undermined by the evolution of Bt-resistant pests. Evolution of Bt resistance could also affect those organic farmers who haven’t grown Bt crops, but who may sometimes use low-toxicity Bt sprays.

So how can we slow the evolution of Bt resistance? Evolutionary biologists have worked on this problem. They have developed a resistance-management strategy that has been widely adopted and appears to be working reasonably well: the high dose/refuge strategy.323

The high dose/refuge strategy is based on certain assumptions. The most important of these assumptions is that an insect pest needs two copies (p.169) of a resistance allele to be resistant to Bt. Like humans, insects have two versions of each gene, one from each parent. Insects with two copies of a resistance allele are designated RR and are assumed to be resistant. Most insects whose ancestors have not been exposed to Bt will probably have two copies of the susceptible version of that gene, making them SS. These are readily killed if they eat a Bt crop. Newly arisen mutants are assumed to be RS, because it’s so unlikely that both copies would mutate at the same time. RS insects are assumed to be resistant to low doses of Bt, but killed by high doses.

So the key assumption is that RS insects can be killed by Bt, just like SS insects. So far, this appears to be true, but only if Bt concentrations in the transgenic crop are high enough. Hence the high dose part of the strategy. As discussed earlier for herbicides, a higher dose can actually result in lower-intensity selection for resistance, if there is no inherited variation for resistance to high doses.

But even if that assumption is true—that is, no RS mutant has genetic resistance to a high dose of Bt—relying on high doses alone would be risky. What if Bt kills only 99 percent of the RS mutants? Maybe 1 percent of them survived by eating weeds, or something. These RS survivors can’t eat the Bt crop without dying, so they aren’t really a problem … until two of them mate. Then they produce a mixture of 50 percent RS, 25 percent SS, and 25 percent RR offspring. Each RR offspring and its many RR descendants would be resistant to even high doses of Bt, making the Bt crop useless.

So we need to keep any RS survivors (presumably rare) from mating with each other. This is where the refuge part of the strategy becomes important. The purpose of the refuge is to supply enough SS insects that the few RS survivors each end up mating with an SS rather than another RS. This is an example of gene flow via flying insects.

For refuges to succeed, they must meet three requirements. First, there must be no selection for resistance there, so that almost all insects in the refuge will be SS. Therefore, the refuge can’t have Bt plants or be exposed to Bt spray. Second, survival and reproduction of SS insects in the refuge must be high enough that they outnumber other potential mates in the Bt area. Therefore, the refuge must have plants that can support reproduction of the insect pest, and it can’t be sprayed with insecticides that would drastically reduce their numbers. Third, the refuge must be close enough to the Bt areas that SS insects from the refuge will swamp the mating market in the Bt area.

Ideally, refuges will still produce some economic or ecological benefit, despite supporting reproduction of Bt-susceptible pest insects. For example, the refuge could consist of wild plants that the insects can eat, or a crop they can eat without affecting yield too much. Differences between (p.170) refuge and nonrefuge crops in the timing of planting may sometimes reduce yield losses in the refuge.324 Still, refuges will often represent a short-term economic loss to farmers.

In recent years, Bt crops have been planted on millions of acres worldwide, with substantial regional differences in land allocation to Bt-free refuges. Not all refuge areas have been deliberately planned. An organic or a conventional farm that chooses not to grow the Bt version of a crop, for whatever reason, may act as a refuge for Bt crops on other farms nearby. Or a different crop can act as a refuge, if it also supports reproduction of the particular pest species.

When these factors are taken into account, the rate of resistance evolution in the field appears to be consistent with mathematical models developed by evolutionary biologists. For example, Helicoverpa zea evolved significant resistance to Bt in Arkansas and Mississippi, where refuge area was about 39 percent of Bt-crop area, but not in North Carolina, where the effective area of refuges was about 82 percent.325

Even a 39 percent refuge area could represent a significant economic cost to farmers, however, unless the refuge can be planted to a crop that is profitable, while still producing significant numbers of Bt-susceptible insects. So one might expect farmer opposition to refuge requirements. This isn’t always true, however, as shown by the history of resistance management in Australia.

Australian Experience with the High Dose/Refuge Strategy

Previous experience with pests evolving resistance to sprayed insecticides apparently helped convince Australian cotton farmers that they needed to slow the evolution of Bt resistance in cotton pests when they started growing Bt cotton. A committee of farmers, researchers, and government-agency and industry representatives developed a resistance management plan. Initially, farmers themselves called for a 70 percent refuge requirement, although this percentage was later reduced.

Impressively, this reduction was based on science, not shortsighted political pressure from farmers.326 The scientific justification for reducing refuge requirements came with the arrival of a new variety of cotton that made two different versions of the Bt toxin, rather than just one. Assuming no cross resistance—a big assumption!—the chances of simultaneous mutations giving resistance to both toxins in the same individual insect are very low. A supply of susceptible insects from refuges is still needed, to make sure that individuals with different resistance mutations mate mainly with susceptibles rather than with each other. Otherwise, their (p.171) offspring could have some resistance to both Bt versions. But, with two resistance genes, the refuge size could be reduced.

The overall economics of Bt cotton in Australia have varied among farms and regions, but there has been an interesting trend over time. The first year that Bt cotton was used, the higher price of seed for Bt cotton outweighed the benefits of reduced pest damage, for a net economic loss. That was followed by three years of break-even net benefits and then by two years with a strong positive economic return.

This is reminiscent of the yield trends some farmers report after switching to organic methods. Lower yields at first are sometimes followed by improvements, which have been attributed to improvements in soil microbial communities.327 Similarly, trends of decreasing pest populations may have been a factor for Bt profitability in Australia. In both cases, however, increasing farmer experience may also be important. At LTRAS, we showed that first-year organic yields can be as high as those in established organic fields, if both are managed by the same people.328

More than 80 percent of Australian farmers listed “protection of the environment” as their main reason to use Bt cotton. Use of sprayed insecticides decreased by 20 to 80 percent. From a short-term ecological perspective, any reduction in sprayed insecticides is probably good. From a longer-term evolutionary perspective, however, spraying any Bt fields that have unusually high pest populations (with a non-Bt pesticide) may be a good idea, as those large populations may contain resistant mutants. Spraying refuges, on the other hand, would reduce the pool of Bt-susceptible insects, making the refuge less effective.

Current refuge options in Australia are flexible, reflecting scientific understanding of differences between refuge and Bt fields. For each 100 hectares of Bt cotton, a farmer can plant either 10 hectares of unsprayed non-Bt cotton, or 100 hectares of sprayed non-Bt cotton (with fewer surviving insects per hectare, more hectares are needed), or as few as 5 hectares of unsprayed pigeon pea. Pigeon pea produces a large number of Bt-susceptible insects per hectare. As long as farmers have to set some land aside as refuges anyway, they often position refuges to avoid spraying pesticides near houses or streams.326

The involvement of farmers in development of this resistance-management strategy was key to its success. Still, farmers don’t always like being told what to do, even by other farmers. So there are some individual financial incentives, provided by the company that sells the Bt-cotton seed, to encourage everyone to follow the rules. As in the case of the legume sanctions and symbiotic rhizobia,298 individual rewards or punishments can help to maintain cooperation, when collective benefits fail.

(p.172) Area-wide Management Requires Cooperation among Farmers

Cooperation among Australian cotton farmers was key to the relatively successful management of Bt resistance. Cottony cushion scale, an insect pest of California citrus crops, shows how a more individualistic approach can fail.329

Beneficial vedalia beetles, predators that feed mainly on the cottony cushion scale insect, were introduced to California from Australia and were very successful at controlling the pest, for a while. Unfortunately, these beneficial beetles are killed by insecticides used against other pests. Farmers who weren’t using insecticides often lost their beneficial beetles anyway, when the beetles visited a neighbor’s farm and were killed there.

Those insecticide-using neighbors didn’t see an immediate increase in their own cottony cushion scale damage, because scale insects were controlled fairly well by the insecticides they were using for other insect pests. But many farmers who had been relying on biological control by the vedalia beetles found that they weren’t working anymore, so they started spraying insecticides themselves. Those insecticides then killed still other beneficial insects, which had been controlling other pests, leading to even more pesticide use.329 This downward spiral illustrates the point that insect pests are often an area-wide problem. Pest-management activities by individual farmers can have big effects, positive or negative, on their neighbors.

In Australia, biological control of citrus scale insects has been more successful. First, one major citrus grower started using biological control. Some Australian farms are huge, and maybe this one farm was big enough that pesticide use by neighbors didn’t kill too many of its beneficial insects. Neighbors copied this successful practice, until 75 percent of citrus growers were using biological control, saving money and reducing overall pesticide use by 75 percent.330 Biological control of scale insects in Australia apparently involved little formal coordination, in contrast to the Bt-cotton example.

Area-wide management of cabbage-family pests, also in Australia, is something of an intermediate case. Diamondback moths were evolving resistance to available insecticides, and public concern about pesticide use was increasing, both of which contributed to widespread farmer interest in better approaches. As in the Bt-cotton case, cooperation among farmers, government agencies, and industry was important.

Farmers switched to Bt sprays, which kill specific pests of cabbage without killing predatory insects that control other pests. But the overall pest-management strategy also included a 3-month “break,” when no crops that the diamondback moth can eat were to be grown. Our 2003 (p.173) paper on Darwinian agriculture proposed a more extreme version of this approach, where no farmer in a region would grow a crop that a particular pest needs to survive, perhaps for a year or more, driving that pest to local extinction.28 We used this as an example of how humans could design an ecosystem-level pattern not found in nature but potentially beneficial for our purposes.

At first, 70 percent of Australian farmers were using the recommended 3-month break. But then some of them found that a 1-month break reduced pest levels enough on their individual farms, yet let them produce a crop at times when other farms were on break, leaving less competition for markets. Even 1-month breaks might have worked fairly well, if everyone had their break at the same time, but they didn’t. So diamondback moths could usually fly from one farm to another and find food.

It has been argued that one key to the success of area-wide pest management is “how to keep the majority of people acting toward the public (that is, their own) long-term good.”330 If there is clear evidence that each individual farmer really increases his own long-term welfare when he acts in ways that promote the long-term public good, then presenting that evidence to farmers may be enough.

But “if everyone did X, we would all be better off” is not always the same as “if I do X, I will be better off.” This was one of the key insights in Hardin’s paper on the tragedy of the commons.287 Individual incentives, like those provided to encourage compliance with the Bt-resistance-management program, may be needed. These incentives could include peer pressure.

Although farmer compliance with currently recommended practices is an important challenge, developing new and better methods of sustainable pest management may be just as critical. Once again, nature may be a useful source of ideas.

Slowing Pest Evolution—Tricks from Nature

There are two different ways in which nature can serve as a source of ideas for improving pest control in agriculture. First, we may get ideas from comparing natural ecosystems, while recognizing that pest control in any particular natural ecosystem has not been improved by competition among ecosystems. Second, we may want to copy some of the pest-defense strategies of individual wild plants, which have been improved, by competitive natural selection.

We know that insects that attack wild plants are often eaten themselves by other insects, by birds, or by bats. Similar food webs occur on farms,331 although we don’t always recognize their existence until an (p.174) insecticide used to kill one pest kills the natural enemies of another pest, as in the preceding citrus scale example.

Because no natural process has optimized food-web structure to maximize the overall productivity or stability of natural ecosystems (chapter 6), uncritical mimicry of natural food webs on our farms would be foolish. However, comparing food webs and their effects on pest damage in a variety of natural ecosystems might help us identify particular food-web features that minimize losses to pests.

The pest-defense strategies and tactics of individual wild plants, on the other hand, have been repeatedly tested competitively and improved by natural selection. Individuals using various genetically programmed defenses competed, directly or indirectly, against other members of their species. Today’s wild plants are descended from the winners of many such past competitions, and they inherited the winning strategies and tactics. (Our crops are also descended from wild plants with winning strategies, but they may have lost some of those strategies after humans took more control over their evolution.) So if we copy the pest-defense strategies of individual wild plants, we know we are copying something that worked better, at least in past environments, than a wide variety of alternatives.

In particular, strategies that were quickly defeated by pest evolution would not have persisted, so the strategies that did persist in wild plants must have been at least somewhat evolution-proof. Here are some examples.

Wild potatoes have two types of glandular hairs, or trichomes, on their leaves. One type releases a sticky substance when disturbed by insect pests, gluing them in place and preventing them from doing much damage. This type is sometimes found on cultivated potatoes as well. I remember a seminar at Cornell, years ago, reporting that some sprays used to control fungal pathogens inactivate this glue, making potatoes more susceptible to insect damage.

The second kind of trichome is even more interesting. These trichomes release gases similar to the gaseous alarm signals of aphids, which aphids use to warn relatives nearby that they are under attack by predators. Aphids smelling plants’ false alarm signals suddenly decide they have urgent business elsewhere, and leave the potato in peace.332

Wild-potato genes for making these alarm signals have been identified. This information might be useful to traditional plant breeders trying to transfer this trait from wild to cultivated potatoes. A breeder might want to keep most of a cultivated potato’s traits, but incorporate a few genes from a wild potato. One way to do this is to cross wild and cultivated potatoes, then repeatedly back-cross to the cultivated parent, keeping only the progeny that have the desired gene(s) from the wild plant. Knowing what those genes are may make it easier to identify which progeny to keep.

(p.175) But would cultivated potatoes that make this alarm signal really suffer less damage from aphids? If so, how quickly would aphid populations evolve to ignore the false signal? The fact that natural selection has maintained this trait in wild potatoes suggests a benefit to individual plants in the wild, but things might be different on a farm covered with cultivated potatoes.

If all the plants on which the aphids could feed make the alarm signal, that might impose strong selection on aphids to ignore it and to keep feeding on alarm-producing potatoes. This seems similar to plants making natural insecticides (or Bt), thereby imposing selection on pests for resistance to those insecticides.

But maybe not. Aphid responses to their own genuine alarm signals have been maintained by strong natural selection. Aphids that ignored the warning got eaten by predators. So if we could maintain fairly high populations of aphid-eating predators in fields of potatoes that make false alarm signals, the aphids would be caught between two bad options: ignore the signal and get eaten, or continue to respond to the signal and stay away from our potatoes.

Identifying an alarm-signal gene also raises the possibility of genetic engineering to transfer that gene to other crops, perhaps protecting them from aphids. A group at Rothamsted (the agricultural research station mentioned in chapter 2, for pioneering research on long-term sustainability, also known for earlier research on wild potato trichomes332) put the wild-potato alarm gene into Arabidopsis.333 As expected, aphids exposed to air around transgenic alarm-producing plants seemed agitated, moving around much more than those exposed to air around control plants. Moving around isn’t quite the same as fleeing in terror, however, so additional research is needed to determine whether transgenic plants that make the alarm signal actually repel aphids in the field.

Biotechnology is not the only way to repel pests from crops they would normally attack. In East Africa, farmers have mixed corn or sorghum, which normally attract stem borers, with molasses grass, which repels borers, apparently by producing gaseous chemicals. Borers might evolve to ignore those chemicals, except that farmers also plant Napier grass nearby, which strongly attracts stem borers. Borers therefore lay most of their eggs on Napier grass, but few survive on that host. (How long will it take borer evolution to eliminate this preference for an unsuitable host?) Using such combinations of repellent crops and attractive but lethal “sinks” is known as the push-pull strategy.334

A somewhat similar approach has been used, also in Africa, to control the parasitic plant, Striga, or “witch weed.” When Striga seeds in the soil detect a host plant nearby, they germinate, grow toward it, attach, and start drawing nutrients from the hapless host, sometimes killing it. Cleverly, (p.176) someone discovered a crop that released similar chemicals as Striga’s usual host, triggering germination and attack, but without allowing actual attachment. Without connecting to a compatible host, Striga seedlings soon die. So growing this “trap crop” tends to deplete the supply of Striga seeds in the soil, decreasing Striga problems over years.

Returning to our alarm-signaling potatoes, there are other ways in which crops might benefit from making aphid alarm signals, aside from just scaring the aphids away. Some wasps that parasitize aphids “eavesdrop” on alarm signals, and use them to track down aphids. Rothamsted researchers found that one parasitic wasp species spent much more time on the alarm-producing transgenic Arabidopsis plants than on control plants, looking for the aphids the wasps smelled, but couldn’t find.


Mention evolution and agriculture together, and many people think of the rapid evolution of agricultural pests and the development of resistance-management strategies that slow this process. At least, that has been my experience. But ongoing evolution of pests has been only a minor theme of this book, for several reasons.

First, pest evolution has received considerable attention elsewhere.316,325,326 Second, even perfect pest control would not be enough to meet increasing demand for farm products, driven by population growth and greater use of farm products as chemical feedstocks. We need to increase yield potential for actual yields to keep up with demand. Last, understanding the contrast between long-term, competitive improvement of individual adaptations and the lack of such long-term improvement in overall ecosystem structure is as important to evolution-resistant pest control as it is to improving water-use efficiency.

Recognizing that natural ecosystems have not been perfected by any natural process, however, does not diminish their value as a source of ideas for improving agriculture, and pest-control in particular. This is particularly true when three species (potato, aphid, and wasp, for example) interact. So the next chapter considers some more-complex interactions and what we can learn from natural ecosystems.