Published in M. Hackett and S.H. Sohmer (Eds.), 1996. Proceedings of The Ecology of Our Landscape: The Botany of Where We Live symposium, "Exploring the Inerfaces Between Plants, People, and the Environment," Fort Worth, TX. Sponsored by The Botanical Research Institute of Texas and Texas Christian University.

I am not here to debate the overwhelming productivity of modern, industrialized agriculture. If any of you have had an occasion to read Farming In Nature's Image (Soule and Piper 1992), you may recall a statement at the beginning that says something like, "There is a sign along Interstate 70 in Kansas, between Manhattan and Salina, that says one Kansas farmer feeds 96 Americans and you.' Well, here we are a couple of years later and no doubt it is an even higher figure than it was then.

There is no argument that agricultural research has been very successful in increasing the output per unit labor of the typical American farmer. In recent years, however, people concerned about the negative environmental consequences of agriculture, and the sustainability of the natural capital upon which agriculture depends, are recognizing that this high productivity is coming at an ecological cost.

The most critical cost is the loss of our soil. Topsoil is not a resource like air and water, which are renewable over short time frames. Eroding soil to grow food is like spending your savings account to maintain your lifestyle.

Secondly, most modern agriculture depends upon pesticide application. One of the problems associated with large-scale monoculture farming is that if the crop species is a host for a particular pest organism, then the pest organism essentially has a nearly unlimited food supply. So it may require frequent use of toxic chemicals in order to keep those pests at bay, leading to an ironic phenomenon: grain fields unsafe to enter because of the presence of pesticides. We began to recognize the broad-scale effects of pesticide application with the publication of Rachel Carson's Silent Spring (Carson 1962) about a generation ago.

Lastly, much of this agriculture productivity is dependent upon extractable fossil fuel supplies, which we have known for a long time are running out. We now recognize that these fossil fuels also likely pose a large environmental threat in the form of global climate change. Therefore, current levels of agricultural production are sustainable only as long as there is an affordable and environmentally safe source of energy to maintain them.

Definitions of sustainable agriculture abound. Entire conferences could be spent arguing the definition of sustainability. At The Land Institute I use, as a working definition, an agriculture that produces food without a net loss of ecological capital. In other words, there can be no loss of soil faster than the natural rate of formation. Second, a sustainable agriculture must not result in contamination of the environment by chemicals known or suspected of being harmful to people and other non-target organisms. This includes pesticides or fertilizers such as nitrogen that leach into ground water. Third, a sustainable agriculture must not depend upon finite, non-renewables (e.g., fossil fuels or aquifer water that recharges only over geological time).

In stark contrast to our industrialized monoculture systems, natural systems, such as native prairies, display several properties that make them essentially sustainable . They preserve the soil by virtue of a perennial canopy that protects against both wind and water erosion. They provide nitrogen fertility primarily through biological nitrogen fixation in legumes. They do not suffer unduly from weeds, insect pests, and plant diseases �additional benefits of biodiversity. And natural systems run on sunlight and available precipitation. Hence, when we began our quest for some clues as to what sustainability means, our first exploration was to look at natural systems that are inherently sustainable.

This idea, then, truly represents a paradigm shift �using the way natural systems work as an operating model for a permanent agriculture.

What sorts of properties do these systems have that enable them to be sustainable? What are some of the principles �some fundamental properties �of natural systems that we can incorporate into agriculture? Our expectation is that if we mimic the prairie's structure, we will get a large degree of its sustainable function.

Two general aspects of the prairie stand out: perennial cover and biodiversity.

First, the prairie is composed primarily of perennial plants. The perennial canopy protects the soil surface from the ravages of wind and rain. Living roots also protect against soil loss. With time, through the turnover of roots and building up of soil organic matter, soil quality can improve. Some of the benefits of perennial vegetation, then, include protection from soil erosion and improved soil quality with time. In restoration projects, it has been shown that restoring former cropland to perennial vegetation can actually return much of the soil structure and function characteristic of original prairie ecosystems. Perennial systems can promote a diverse community of soil-dwelling organisms, reduce weed growth, and provide habitat for many beneficial insects.

A recent study (Burke et al. 1995) examined a 50-year-old restoration of former cropland that had been planted to perennial grasses. It was found that within a human time frame, about 50 years or so, certain soil properties (e.g., the active organic soil organic matter component) can recover to levels characteristic of virgin prairie. Other properties, total soil organic matter, for example, did not recover to the preplowed condition even after 100 years. The study's results indicate that we can restore several important characteristics of virgin prairie soils with perennial vegetation over a few decades, but other soil properties will take a very long time to restore. All of this probably makes a strong argument for restoring these areas as soon as possible.

A second feature of many grassland ecosystems that is in stark contrast to monoculture fields is biodiversity. Natural systems are not monocultures. Typically there are many species of warm- and cool-season grasses, legumes important for providing nitrogen fertility to the system, some composites, and other plant groups. Natural systems feature diversity instead of monoculture.

Some of the benefits of plant biodiversity are internal supply of nitrogen, management of exotic and other harmful organisms, soil biodiversity, and overall resilience of the system. Ecologists David Tilman and Sam McNaughton and their colleagues have looked at different grassland systems around the world and found that following a drought or overgrazing period, grasslands that were more diverse in species tended to recover faster than grasslands that were less diverse. Therefore, biodiversity may be an important property that can confer upon ecosystems resiliency to disturbance. (McNaughton 1977 and 1985, Frank and McNaughton 1991; Tilman and Downing 1994. See also Naeem et al. 1994.)

If we were to look at the prairie as a model for a sustainable form of agriculture, the perennial habit and biodiversity are two broad characteristics we ought to mimic with our agriculture system. The prairie consists of herbaceous, perennial plants, instead of annual plants such as virtually all of our grain crops. Such annual plants as wheat, corn, and soybeans die after yielding their seed. After harvest, the ground remains bare for some period of time during which the soil is exposed and vulnerable to loss. The ground has to be tilled again to create a new seed bed that is favorable for these plants, and so the cycle goes on. The first order, then, is to develop perennial grains to replace annuals on erodible soils.

As I have stated, prairies also feature species diversity instead of monoculture. What I have found, in looking at a range of different prairie plant communities, is that four major plant guilds are pretty consistently represented: the perennial warm-season (C4) grasses such as big bluestem (Andropogon gerardii), little bluestem (Schizachyrium scoparius), and Indian grass (Sorghastrum nutans); such cool-season (C3) grasses as Canada wild rye (Elymus canadensis), western wheatgrass (Agropyron smithii), and Junegrass (Koeleria pyramidata); many important nitrogen-fixing legumes; and composites, or members of the sunflower family (Asteraceae) (Piper 1995).

Perennial polycultures would be grain-producing analogs of native prairie systems. If we expect to create an agriculture that mimics the prairie ecosystem, we ought to start with perennial grain candidates that fall within these different plant groups. The research in natural systems agriculture aims to provide the benefits of a diverse prairielike vegetative structure, but with the added feature of a yield of edible grains that could be directly eaten by people or fed to livestock.

We have, through a series of inventories, chosen several species that we are using to explore some of the principles of perennial grain systems. Some of these are probably familiar to you. Eastern gama grass (Tripsacum dactyloides) is a warm-season grass. It is a relative of maize. Mammoth wild rye (Elymus racemosus) is an Old World C3 grass that has been planted extensively in the western United States to stabilize sandy soils. Illinois bundleflower (Desmanthus illinoensis), a nitrogen-fixing legume that is native to the tallgrass prairie, has displayed high seed yield in experimental plots. Wild senna (Senna marilandica) is another legume that has shown very high seed yield. Maximilian sunflower (Helianthus maximilianii) is a native tallgrass prairie composite that, in addition to being a potential perennial grain or oil seed crop, also seems to show allelopathic properties and can be very important in weed control, especially at the early stages of a polyculture system. We are not ruling out also looking at perennial hybrids between some important grains such as milo (Sorghum bicolor) and its wild perennial relative Johnson grass (S. halepense). (See Soule and Piper 1992 for a fuller description).

Four basic questions have guided our research in perennial grain polycultures. First of all, the yield question is very important to us. So the first question is: Can perennial grains yield as well as annual grains? Secondly, can perennial polycultures overyield? Overyielding is simply a way of comparing the yield of a mixture with the yields of the respective monocultures. Because different crop species can occupy different ecological niches, one would expect mixtures to feature more efficient use of the land area than monocultures. Nitrogen fertility is important in a sustainable agriculture. Therefore, the third question is: Can the legume component of these perennial mixtures provide enough nitrogen to support the rest of the system �the non-nitrogen fixing grasses and composites? Finally, because so much energy in agriculture is devoted to managing weeds, insects, and plant pathogens, it is important for us to deal biologically with those organisms. Therefore, we ask: Can a perennial polyculture manage weeds, harmful insects, and plant pathogens?

We have a set of 36 research plots at The Land Institute in which I have been looking at most of these basic questions. To summarize, I will briefly present some results that support the perennial polyculture model.

To answer the question of whether a perennial can yield as much seed as an annual grain crop, we can use the benchmark yield for Kansas winter wheat of 1,800 pounds per acre. A yield of 30 bushels per acre at 60 pounds per bushel multiplies to 1,800 pounds. Converted to international units that works out to be 1,960 kilograms per hectare, or 196 grams per meter. Already wild senna, some accessions of Illinois bundleflower, as well as perennial sorghum hybrids without any selection and any inputs other than some hand weeding, have produced yields within the ballpark of this benchmark yield for Kansas winter wheat (Piper 1992, 1993a, Piper and Kulakow 1994).

We have evidence, then, that perennial species can yield as much seed as an annual crop. Part of the yield question concerns these absolute yields.

There is another aspect that has to do with possible resource trade-offs in plants as we select for higher seed yield. An argument sometimes raised against the feasibility of perennial grain development is this question of resource reallocation. A perennial may never yield as well as an annual plant, because an annual plant is able to devote all of its available resources to seed, whereas a perennial has to allocate resources to its perennating organs: roots and rhizomes. Some very nice work done by Laura Jackson and Chet Dewald, however, found no detectable trade-off between increased seed production and plant vigor between high-yielding (T. dactyloides forma prolificum) and normal, low-yielding forms of eastern gama grass (Jackson and Dewald 1994). Similarly, in some of the hybrid crosses we made between milo and Johnson grass in which we saw variability in rhizome expression, there was not an apparent strict trade-off. In other words, plants that produced more rhizome biomass did not necessarily produce less seed. So the energy for increased rhizome production was not coming at a direct cost of lowered seed yield within that generation (Piper and Kulakow 1994).

Can a perennial polyculture overyield? The use of the Relative Yield Total is a simple way of measuring the yield of a mixture relative to the yields of the respective monocultures. If the RYT is equal to 1.0, then there is no difference in yield between the mixture and the monocultures. In eastern gama grass-Illinois bundleflower mixtures, we obtained an RYT of 1.19, or 19 percent overyielding. And in a three-species mixture of eastern gama grass, Illinois bundleflower, and the C3 mammoth wild rye, we observed 26 percent overyielding. In addition to these representative high RYTs, we have observed, in some cases, overyielding maintained for several years.

Can a perennial polyculture provide its own nitrogen fertility? This is a question for which I have mainly indirect evidence. Illinois bundleflower is an important nitrogen-fixing species in the prairie. We have planted it on what was originally a high-nitrogen, good agriculture soil as well as a low nitrogen, eroded soil that had been planted to row crops for several decades prior to being abandoned. In years of adequate rainfall, Illinois bundleflower yielded similar amounts of seed between the two soil types. In other words, Illinois bundleflower was able to compensate for the lower soil nitrogen without reducing its seed yield (Barker and Piper 1995). Another piece of indirect evidence is that gama grass yielded and grew better, and the yield did not decline over successive growing seasons, when it was grown with Illinois bundleflower compared with gama grass in monoculture.

Moreover, we have been monitoring soil nitrate levels in plots containing Illinois bundleflower. Available soil nitrate has actually increased in initially low-nitrate soil three to five years after bundleflower establishment. Here then is some evidence that Illinois bundleflower can improve available soil nitrogen status with time, despite our removing seed from the plots every year.

The fourth and final question has to do with management of weeds, insect pests, and plant disease in perennial polycultures. Effective weed control has occurred in two separate experiments at The Land Institute. In one study, a plot containing rows planted with various densities of Maximilian sunflower, a probable allelopathic species, and a control (no sunflowers), weed biomass was significantly reduced in the sunflower plots relative to the control. In the second year, sunflower plots reduced weed biomass by 50 to 75 percent. In the third year, weed biomass in sunflower rows was 44 percent of weed biomass in the control during May. Here, effective weed control was maintained across years despite changes in the weed community from predominantly annuals in the first year to perennials by the third year. Eastern gama grass, by virtue of its thick canopy, also appears to reduce weed biomass significantly (Piper 1993b).

In general, densities of host-specific plant-feeding insects are reduced in annual polycultures relative to monocultures (Risch et al. 1983). Insects in diversified perennial systems have been for the most part unexamined, however.

In our system, Illinois bundleflower is fed upon by the specialist chrysomelid beetle, Anomoea flavokansiensis. The Chrysomelidae are a family of beetles that contain our familiar friends the striped cucumber beetle and the Colorado potato beetle. We have observed that the density of this specialist beetle is lower, and the beetle appears later, in polycultures relative to monocultures of Illinois bundleflower (Piper 1996). This pattern occurred even several years after establishment, which is an additional favorable result.

The final part of the question concerns levels of plant disease in perennial mixtures versus monocultures. Eastern gama grass is subject to two different strains of Maize dwarf mosaic virus, an important aphid-vectored pathogen of corn and Johnson grass. When we observe viral symptoms on gama grass, we are actually seeing the consequences of insect colonizing and movement within plots. We have found that disease can be delayed two years on gama grass plants when grown in association with Illinois bundleflower relative to gama grass plants grown in monoculture (Piper et al. 1996).

So, in summary, we have evidence of the benefits of perennial polycultures in reducing weeds, a specialist insect, and viral plant disease.

In introducing the notion of natural grassland ecosystems as models for a sustainable grain agriculture, I have discussed two general properties. One characteristic is that prairie plants are nearly all herbaceous perennials. The second characteristic is that natural grasslands are not monocultures but consist of diverse assemblages containing many species.

A third characteristic of all natural plant communities is that these complex systems each have a history. If you inventory any natural system, whether it be a forest, grassland, or desert, identifying the species present and their relative abundances, you are really only looking at the final stage of a long assembly process. Such accurate descriptions represent only half of the story, a "snapshot" of an ecosystem if you will. The history of a system's development can play a crucial role in determining the final community. Moreover, it may not even be possible to identify the processes that led to the community we see today (Post and Pimm 1983, Robinson and Dickerson 1987, Drake et al. 1993). Aspects of this history include the size and composition of the initial species pool at the beginning and the sequence in which species enter the community during its assembly.

Stuart Pimm and Jim Drake, University of Tennessee ecologists who study complex systems, have suggested to us that we would have a very difficult time establishing a diverse, persistent group of perennial grain species without constant intervention. Theoretical work, studies using laboratory microcosms, and literature reviews of ecological restoration have shown that the likelihood of achieving a diverse, persistent community increases when communities are allowed to assemble over time. In other words, when they are allowed a history. In addition, there may be so-called "Humpty Dumpty" effects (Drake 1990, Luh and Pimm 1993). Once the system is broken, you can't simply put it back together again using only the species present at the final endpoint.

So the third aspect of our natural systems agriculture research concerns discovering assembly rules that will allow us to construct diverse, highly productive systems. This work represents an alternative approach to the construction of persistent perennial grain polycultures.

Earlier, I introduced several species we are examining as potential perennial grains. In 1994, we undertook a study looking at incorporating some other species as part of an initial species pool that will be assembled into a perennial polyculture containing the species that we want. We are incorporating, in addition to mammoth wild rye, such cool-season grasses as western wheatgrass; the legumes �purple prairie clover (Dalea purpurea), bird's-foot trefoil (Lotus corniculatus), and lead plant (Amorpha canescens) �and composites in addition to Maximilian sunflower, such as grayhead prairie cone flower (Ratibida pinnata) and Kansas gayfeather (Liatris pycnostachya). We call this our community assembly experiment. It represents a possible alternative approach to a stable perennial polyculture.

Treatments consist of four incrementally diverse mixtures of herbaceous perennial species that represent four guilds that predominate on North American tallgrass prairie: perennial C4 grasses, C3 grasses, nitrogen-fixing plants, and composites (Piper 1995). The initial seed mixtures comprise four, eight, 12, and 16 species (Table 2) chosen for their adaptation to grassland environments. Each lower diversity treatment is nested within its higher diversity counterpart. The treatments vary the size of the species pool, while keeping guild representation constant.

We are starting with one treatment that consists of the four species we want to end up with: eastern gama grass, mammoth wild rye, Illinois bundleflower, and Maximilian sunflower. This is our "control." Treatment II consists of these four species plus four more representatives of the perennial warm-season grasses, the cool-season grasses, the nitrogen-fixing legumes, and the composites. The next diversity treatment (III) contains those eight plus four more within each of the plant guilds. Finally, the most diverse treatment (IV) contains the 12 species of the previous incrementally diverse treatments, plus four more. The question is: If we start with a larger species pool, say an excess abundance of species initially, are we more likely to end up with a stable end point than if we simply start out with the four species that we want?

One component of this experiment then is looking at these incremental diversity treatments to determine if starting with a more diverse species pool reduces the amount of time it takes to reach a persistent, perennial polyculture end-point. The other aspect of this experiment has to do with the sequence in which species enter the community. It is possible that some species may not establish at the beginning, when competition with annuals is fierce, but may be able to enter the community at some later point during the assembly process. To help us address both of these aspects, half the plots sown with the different species mixtures will be left alone to assemble without further intervention. In the other half of the plots, any species that fails to establish or that disappears from the plots will be reseeded. This reseeding will provide additional opportunities for species to enter the community.

This experiment has been running for two years so I have only a few data to show you, but we are beginning to see some pretty interesting results. Here, diversity is simply species richness or the number of species sampled per plot. In both years, 1994 and 1995, species diversity increased with diversity treatment . Evenness, a measure of equability among species, ranges from 0 to 1. A community where all the species are equally represented will have an evenness value of 1.0. A community that has very high dominance by one species and several rare species will have a very low evenness value. We saw a tendency for evenness to increase with diversity treatment, but the differences were not significant. There were no treatment effects on total percentage cover in these early years.

presents the results for five prominent plant functional groups or guilds. Annuals, including many weedy species, are essentially species that we do not want in the final community. The other categories represent the four guilds that predominate in natural prairie ecosystems. I left the woody plant guild off the table because in all cases its percentage cover was much less than 1. By 1995, percentage cover by annuals differed between the least and most diverse treatments. In treatment IV, annuals were in the minority (61 percent versus 107 percent for all perennials). This may be an important benefit relevant to prairie restoration or establishing conservation reserve ground. By starting out with more species in the seeding mix, one may be able to overcome the weed community faster. Cover by perennial C3 grasses and composites increased with diversity. It already appears that there are some benefits of starting with a higher species diversity in terms of getting some of the properties that we want.

An important question is: Where do we go from here? In September 1995, we hosted several representatives from the United States Department of Agriculture's Agricultural Research Service. An organization like The Land Institute, which is attempting to bring about a fundamental shift in the prevailing agricultural paradigm, becomes effective when others begin to adopt its research agenda. Several of the other authors in these symposium proceedings have mentioned the need to form partnerships. We think that a significant partnership would be a link among The Land Institute, one or more land grant universities such as Kansas State University, and the agriculture research wing of the federal government, the ARS. I would like to call your attention to a quote from the Senate Agriculture Committee working on the Research Title of the 1995 Farm Bill:

"The committee recognizes that there have been exciting and promising advances made in Natural Systems Agriculture. This includes perennial grain polyculture ecosystems and:

1. high seed yield;
2. management of insects injurious to plants, plant pathogens, and weeds;
3. nitrogen fertility managed by legumes;
4. minimizing of soil erosion, use of fossil fuels and synthetic chemicals;
5. enhancement of soil quality."

"The Committee is aware of breakthroughs in long-term natural systems agriculture research and wishes to have these research breakthroughs further examined. Therefore, the Committee expects the Secretary to make an analysis of the feasibility, productive potential, and environmental benefits of long-term natural systems agriculture and to identify associated near-term research needs."

We consider this language to be a very important early step in the direction of a paradigm shift for agricultural research. What is needed is a transdisciplinary team to focus on all of the various aspects of perennial agriculture systems . Such a team should include ecologists, geographers, landscape ecologists, geologists and soil scientists, entomologists, plant pathologists, environmental historians to present us with an historical picture of the landscape, and several plant breeders, and developmental biologists to study the biological mechanisms underlying increased flowering and seed production in perennial species. Biotechnology and computer modeling should be included within an overall ecological framework. Computer modeling has become an extremely important tool in ecological research.

We have proposed central Kansas as the logical "alpha site" for this type of research. The Land Institute is centrally located in the Great Plains and has nearly a 20-year history of research in the "nature-as-measure" paradigm. Seventy miles away is Kansas State University, in Manhattan, which oversees the Konza Prairie, an 8,500-acre tallgrass prairie natural laboratory, and operates a USDA Plant Materials Center.

So as we explore this paradigm shift from annual grain monoculture to something that more resembles a natural grassland ecosystem, we acknowledge that the work is inherently long-term. It is also necessarily interdisciplinary, not only within the plant sciences but between the plant sciences and other sciences, as well as between the "hard" sciences and the social sciences. It is, therefore, going to require a great deal of cooperation among people who may not have interacted outside their specialties traditionally. A major difficulty in presenting this type of research to funding agencies is that it is still somewhat risky, and short-term payoffs may be few. As we continue to deplete our soils and contaminate the environment, however, the long-term riskiest path for agriculture may well be the one we are on currently. It is time to explore some innovative, creative, and, yes, "wacky" alternatives.