Abstract
Crop heterogeneity is a possible solution to the vulnerability of monocultured crops to disease1, 2, 3. Both theory4 and observation2, 3 indicate that genetic heterogeneity provides greater disease suppression when used over large areas, though experimental data are lacking. Here we report a unique cooperation among farmers, researchers and extension personnel in Yunnan Province, China—genetically diversified rice crops were planted in all the rice fields in five townships in 1998 and ten townships in 1999. Control plots of monocultured crops allowed us to calculate the effect of diversity on the severity of rice blast, the major disease of rice5. Disease-susceptible rice varieties planted in mixtures with resistant varieties had 89% greater yield and blast was 94% less severe than when they were grown in monoculture. The experiment was so successful that fungicidal sprays were no longer applied by the end of the two-year programme. Our results support the view that intraspecific crop diversification provides an ecological approach to disease control that can be highly effective over a large area and contribute to the sustainability of crop production.
Many ecological processes are strongly influenced by spatial scale6, 7, 8, 9, causing a major dilemma for experimental biologists, as large-scale field experiments are often prohibitively expensive. For example, there have been increasing calls for ecological approaches to counter the negative environmental impacts of modern agricultural systems10, 11. One such approach, the use of within-field crop genetic diversity, has been shown to reduce disease severity in experimental plots and has been used commercially in some cases1, 2, 3, 4. However, experimental procedure and the nature of pathogen dispersal can cause substantial underestimation of the impact of increased diversity on disease in small-scale experimental plots2, 3, 4. On the other hand, observations at larger spatial scale are few4, and do not allow for unambiguous determination of causal relationships between diversity and disease occurrence.
Our experimental system was blast disease in rice (Oryzae sativa). Rice is the staple crop for about half of the population of the world12. The fungus that causes blast disease, Magnaporthe grisea, spreads through multiple cycles of asexual conidiospore production during the cropping season, causing necrotic spots on leaves and necrosis of panicles. M. grisea interacts on a gene-for-gene basis13, 14 with its host and has a very varied pathogenesis15. It exists as a mixture of pathogenic races, that is, genetic variants that attack host genotypes with different resistance genes. Thus, host resistance genes often remain effective for only a few years in agricultural production before succumbing to new pathogenic races16, 17.
Our experimental site (Yunnan Province, China) favours the development of rice blast epidemics because of its cool, wet climate. Farmers commonly make multiple foliar fungicide applications to control blast. Glutinous or 'sticky' rice varieties are used for confections and other speciality dishes and have higher market value than other rice types, but have lower yields and are highly susceptible to blast. Non-glutinous, hybrid rice varieties are less susceptible to rice blast and are attacked by a different spectrum of M. grisea races. Before 1998, 98% of rice fields in the study area were sown with monocultures of the hybrid rice varieties Shanyuo22 and Shanyuo63. The desirable glutinous varieties were planted in small amounts because of their low yields and vulnerability to blast in this environment. We conducted large-scale tests, made possible through the cooperation of thousands of rice farmers, to determine how the occurrence of rice blast is affected by within-field varietal diversification using mixtures of commonly grown glutinous and hybrid rice varieties. Our approach was based on an observed farmer practice of dispersing single rows of glutinous rice between groups of four rows of hybrid rice at a rate sufficient to meet local demand for glutinous rice ( Fig. 1).
Figure 1: Planting arrangements in rice variety mixture and monoculture survey plots in 1999 and patterned after those used by farmers in Yunnan Province.
Each symbol represents a hill of susceptible (O) or resistant (X) rice. Distances between hills within rows were 15 cm for glutinous monocultures, 30 cm for hybrid monocultures and 30 cm for mixtures. Spacings and arrangements were the same in 1998, except that the distance between rows of glutinous rice in monoculture was 13 cm.
High resolution image and legend (67K)In the first year of the experiment, four different mixtures of varieties (Fig. 2) were planted in a 812-ha area consisting of all rice fields in five townships of Shiping County, Yunnan Province. Because of the excellent blast control provided by the variety mixtures, only one foliar fungicide spray was applied. Mixtures were compared to monoculture control plots at 15 survey sites. Unlike standard experiment station fields, control plots of monocultures were small relative to the total area of mixtures planted by farmers in the surrounding area, reducing the potential impact of spore dispersal from the more heavily infected monocultures to the mixture plots2, 3, 4. The study was expanded to 3,342 ha of rice fields in 1999. This area consisted of all rice fields in 10 townships that spanned Jianshui and Shiping Counties, with five participating townships and 15 survey sites per county. Procedures were the same as in 1998, except that no foliar fungicide applications were made. In addition, some farmers chose to plant mixtures in a ratio of 1 glutinous: 6 hybrid rows, rather than 1:4.
Figure 2: Panicle blast severity (mean percentage of panicle branches that were necrotic due to infection by Magnaporthe grisea) of rice varieties planted in monocultures and mixtures.
a, The susceptible, glutinous varieties Huangkenuo and Zinuo. b, The resistant, hybrid varieties Shanyuo22 and Shanyuo63. S98, Shiping County, 1998; S99, Shiping County, 1999; J99, Jianshui County, 1999; open bar, blast severity for a variety grown in monoculture control plots; black bar, blast severity of the same variety when grown in mixed culture plots in the same fields. Error bars are one s.e.m.; n, number of plot means that contribute to individual bars for each of the four combinations of susceptible and resistant variety. All differences between pairs of monoculture and mixture bars are significant at P < 0.01 based on a one-tailed t-test, unless indicated by 0.05 (significant at P < 0.05), 0.10 (significant at P < 0.10) or n.s. (not significant at P = 0.10).
High resolution image and legend (55K)Diversification had a substantial impact on rice blast severity (Fig. 2). In 1998, panicle blast severity on the glutinous varieties averaged 20% in monocultures, but was reduced to 1% when dispersed within the mixed populations (Fig. 2a). Panicle blast severity on the hybrid varieties averaged 1.2% in monoculture and was reduced to varying degrees in mixed plots, though only the larger differences were statistically significant (Fig. 2b). Results from 1999 were very similar to the 1998 season for panicle blast severity on the susceptible varieties (Fig. 2a), showing that the effect of diversification was very robust among mixtures and between seasons and counties. In contrast, effects of crop diversification on blast severity of the hybrid varieties were larger in 1999 than in 1998. Panicle blast severity on these varieties averaged 2.3% in monoculture and was reduced to 1.0% in mixed populations (Fig. 2b), despite the fact that hybrids were planted at the same density in both mixture and monoculture survey plots.
Several mechanisms may reduce disease severity in genetically diverse plant populations2, 4, 18. Increased distance between plant genotypes, which dilutes inoculum of a given pathogenic race as it is dispersed between compatible host varieties, has been considered the most important mechanism contributing to disease reduction in variety mixtures2. Such dilution effects almost certainly had a role in reducing blast disease on the susceptible, glutinous varieties in this study. In addition, canopy microclimate data collected at one survey site in 1999 indicate that height differences between the taller glutinous and shorter hybrid varieties resulted in temperature, humidity and light conditions that were less conducive for blast on glutinous varieties in the mixtures than in the monocultures. Disease reductions on hybrid varieties in the mixtures are more difficult to explain. Dilution and microenvironmental modifications are unlikely mechanisms, as the hybrids were planted at the same density in mixtures and monocultures ( Fig. 1). The taller glutinous varieties may physically have blocked spore dispersal and/or altered wind patterns compared with the hybrid monocultures. In addition, induced resistance may have some contribution to disease suppression in the hybrids. Induced resistance occurs when inoculation with avirulent pathogen race(s) induces a plant defence response that is effective against pathogen races that would normally be virulent on that host genotype. This has made significant contributions to disease reductions in variety mixtures of other small grain crops19, 20.
In 1999, we determined the genetic composition of the pathogen populations derived from inter-planting and monoculture fields using polymerase chain reaction (PCR) fingerprinting21 of pathogen isolates. Preliminary results indicate that fields with mixtures supported diverse pathogen populations with no single dominant strain. In contrast, pathogen populations from monoculture fields were dominated by one or a few strains. The more diverse pathogen population from the mixed stands may have contributed to greater induced resistance from incompatible interactions. In the longer term, this increased pathogen diversity may also slow adaptation of the pathogen to resistance genes functioning within a given mixture. Clarifying the mechanisms by which host diversity influenced disease in our study will be helpful in extending these results to other agro-ecosystems. These mechanistic studies are currently underway.
Grain production per hill of glutinous varieties in mixtures averaged 89% greater than that in monoculture (Table 1). As a result, glutinous rice in mixtures produced 18.2% of monoculture yield, on average, though it was planted at rates of only 9.2 and 9.7% that of monoculture in 1998 and 1999, respectively (see also Fig. 1). Reduced disease severity certainly had a role in this yield response, though other factors (for example, improved light interception) may also have had an influence. Despite the increased overall plant density in mixtures (see Fig. 1, bottom), grain yields per hectare of the hybrids in mixture were nearly equal to the corresponding monocultures. Thus, mixed populations produced more total grain per hectare than their corresponding monocultures in all cases (Table 1). Land equivalent ratios22, which estimate the ecological efficiency of mixed populations, indicate that an average of 1.18 ha of monoculture crop land would need to be planted to provide the same amounts of hybrid and glutinous rice as were produced in 1 ha of a mixture (Table 2). After accounting for the differing market values of the two rice types, the gross value per hectare of the mixtures was 14% greater than hybrid monocultures and 40% greater than glutinous monocultures (Table 1).
Though disease reductions are theoretically maximized in random mixtures of plants23, row mixtures provided the most practical approach in our specific application. As rice is hand-harvested in Yunnan Province, farmers can easily separate the hybrid and glutinous grains, which are used for different purposes. However, many other approaches can be used to attain within-field genetic diversity of crops3, 24. For example, wheat (Triticum aestivum) mixtures are grown in the Pacific Northwest of the USA under highly mechanized conditions4. In this case, varieties are chosen to be similar in height, maturity and market quality, planted as random mixtures, and harvested and marketed as bulk populations3.
Commercial-scale use of crop diversity has provided observational support for the disease-suppressive effects of crop diversity in a limited number of cases4, 25, 26, most notably the control of barley (Hordeum vulgare) powdery mildew (caused by Erysiphe graminis) in the former East Germany26. However, the varietal diversification program in Yunnan Province provided an unusual opportunity to determine causal relationships between crop diversity and disease, as replicated monoculture controls were available for comparison within a substantial expanse of mixed culture. The impact of crop diversification on blast severity in this study was greater than that reported from small-scale experimental plots with this disease3, although we do not have proof that this difference is due only to the spatial scale. By the second year of the project, no foliar fungicides were needed for blast control in the diversified area, though this may not be possible in all seasons. The Yunnan diversification program has resulted in great interest by farmers, and the practice has expanded to more than 40,000 ha in 2000.
The 'Green Revolution' has provided remarkable increases in crop productivity over the past four decades27. However, this agricultural transformation has also resulted in problems, including loss of crop genetic diversity11. The current world population of over six billion does not allow us to return to agricultural production practices of the past. Rather, we need to maintain the benefits of modern agriculture while addressing its drawbacks. In this regard, it is significant that the diversification program described here is being conducted in a cropping system with grain yields approaching 10 Mg ha-1, among the highest in the world. The value of diversity for disease control is well established experimentally and diversity is increasingly being used against wind-dispersed pathogens of small grain cereals4. Recent experimental results indicate other applications of diversity, for example, against soil-borne pathogens and for tree crops4. The effect of varietal diversification will vary among diseases and agro-ecosystems4. Further, one can not expect all variety mixtures to provide functional diversity to a given plant pathogen population24, 28, nor can one predict the time for which they may remain effective. Indeed, we have identified variety combinations that provide little or no blast control in Yunnan Province. Nonetheless, our results demonstrate that a simple, ecological approach to disease control can be used effectively at large spatial scale to attain environmentally sound disease control.
Methods
Study sites
In 1998, townships participating in the diversification experiment were Baxing, Baoxiu, Songchun, Maohe and Xincheng of Shiping County. These townships are contiguous, and all rice fields in the five townships were involved in the diversification program. In 1999, the study area consisted of all rice fields in ten contiguous townships: Chenguang, Dongba, Mianding, Nanzhuang and Xizhuang in Jianshui County; and Baxing, Baoxiu, Songchun, Maohe and Yafanzi in Shiping County.
Disease assessments
To monitor disease, survey plots were established at 15 sites per county, three in each of the five townships participating in the diversification program (15 sites in 1998, 30 sites in 1999). Seedlings were transplanted into the field in April or May in hills of 4–5 plants for glutinous varieties and 1 plant per hill for the hybrid varieties, which produce a greater number of tillers per plant. All plots were managed by farmers and treated in the same manner as the surrounding mixed variety plantings, including fungicide application. In each of the survey sites, a field was divided into three plots. One plot was planted with the mixture grown most commonly by local farmers, and the remaining two plots were monocultures of the glutinous and hybrid variety included in that mixture. For mixtures, the same row spacing of hybrid rice was used as in monoculture, but one row of glutinous rice was added between each group of four rows of hybrid rice, in an 'addition' approach (Fig. 1 ). Each of the four mixtures was evaluated for disease severity in 3–5 of the 15 survey sites in each county, depending on the popularity of a given mixture with farmers. Plots ranged from 100 to 450 m 2 each, depending on field size.
Survey plots were assessed in late August for the severity of blast symptoms,
expressed as the percentage of panicle branches that were necrotic due to
the effects of M. grisea. Disease was assessed at five sampling points
in each plot, distributed in a uniform pattern. Twenty hills resulting from
the transplanting process were evaluated at each sampling point, with each
hill containing about 10 panicles per hill, to give a total of approximately
1,000 panicles evaluated per plot. Each sampled panicle was visually examined
by experienced personnel to estimate the percentage of branches that were
necrotic due to infection by M. grisea. Each panicle was given a rating29 from 0 to 5, where 0 is no disease; 1 is less than 5% of panicle
branches necrotic; 2 is 5–30% necrotic; 3 is 30–50% necrotic;
4 is greater than 50% necrotic; and 5 is 100% necrotic. Disease severity was
summarized within each plot as {[(n1 
1) + (n2 2) + (n
3 3) + (n4 4) + (
n5 5)]/
n0...
n5} 100, where n0...
n5 is the number of culms in each of the respective disease
categories. Thus, a disease severity of 0% would indicate no disease and 100%
would indicate that 100% of panicle branches were necrotic.
Yield evaluation
Plots were hand-harvested, threshed and weighed to determine grain yield. Individual varieties were evaluated separately in mixtures. Land equivalent ratios22 were calculated as (yield ha-1 of variety A in mixture/yield ha-1of variety A in monoculture) + (yield ha-1 of variety B in mixture/yield ha-1 of variety B in monoculture).
Statistical analyses
Each survey plot was considered to be an experimental unit, and analyses were based on mean disease severities and grain yield for each plot. Statistical analyses were conducted separately by year and county owing to differences in disease level. One-tailed t -tests were used to determine if blast severity for each of the two varieties in each of the four mixtures differed significantly from its corresponding monoculture control.