The effect of two-component rice cultivar mixtures on the control of rice blast disease was studied in three different experiments under rainfed upland conditions in the Madagascar Highlands. The mixtures involved a susceptible cultivar (either susceptible or very susceptible) and a resistant cultivar in different mixture arrangements (random or row mixtures) and with different proportions of the susceptible cultivar (50, 20 and 16·7%), which were compared to the susceptible cultivar grown in a pure stand. The effect of these mixtures on the incidence and severity of leaf and panicle blast was measured weekly, and on yield and yield components at harvest time. The mixture effect was more efficient in reducing disease with a proportion of 16·7% susceptible component than with a proportion of 50%. Blast epidemic was significantly reduced in all three experiments. However, under high blast pressure, there was no reduction in the disease by the end of the epidemic and yields of the susceptible cultivar were almost zero whatever the mixture. In two other experiments performed under lower blast pressure, disease incidence and severity were significantly lower in mixtures, and yields of the susceptible cultivars grown in mixtures were higher than those of their respective pure stands. Cultivar mixtures are a promising strategy that could contribute to a more sustainable cultivation of rice under upland conditions in the context of subsistence agriculture in Madagascar, where all cropping operations are manual.
The rice–blast pathosystem is of utmost importance. Rice is the staple crop for more than half the world population and blast, caused by Magnaporthe oryzae, occurs in most rice growing areas worldwide and is the most damaging fungal disease in rice (Ou, 1985). In Madagascar, where the present study was carried out, rice is the staple crop. Per capita rice consumption is the highest in Africa and averages 120 kg year−1. As in the main Asian rice growing areas, further expansion in the form of new irrigated lowland rice fields is almost impossible in the densely populated areas of the highlands of Madagascar. One option is to develop new efficient and sustainable rice-based production systems for upland rainfed conditions. Upland rice cultivars adapted for high altitude have been released by a FOFIFA (the national agricultural research institute of Madagascar) and CIRAD (the French agricultural research centre for development) joint breeding programme conducted in Madagascar (Raboin et al., 2011). The first cultivars were released in 1995 and contributed to the rapid development of upland rice in the highlands of Madagascar because of their yield potential at altitude and their good grain characteristics. Fifteen years later, upland rice is part of the landscape of Madagascar’s central highlands. Nevertheless, blast disease has rapidly become a major constraint to rice cultivation in the highlands because of high inoculum pressure and because upland cultivation systems favour rice blast epidemics. In this context, the resistance of the first cultivars released was overcome by new virulent pathogen races within a few years. Although these cultivars have now been replaced by more tolerant ones, this situation is a challenge for the sustainable management of new rice varieties, especially as most Madagascan farmers cannot afford fungicides (Sester et al., 2008). One direction being investigated is the use of cultivar mixtures.
Genetic uniformity in modern agro-ecosystems increases vulnerability to pests and diseases (Stukenbrock & McDonald, 2008) and makes the extensive use of pesticides inevitable. Conversely, in traditional agricultural systems, the cultivation of mixtures of varieties or species helps protect the crops against both biotic and abiotic stresses (Altieri, 1999; Wolfe, 2000). Cultivar mixtures have the potential to limit the development of many airborne diseases (Wolfe, 1985). They have proved to be particularly efficient in controlling specialized polycyclic foliar diseases of cereals such as rusts or mildews (Finckh et al., 2000). Cultivar mixtures operate through three main mechanisms (Wolfe, 1985; Mundt, 2002), each of which has been nicely dissected by Chin & Wolfe (1984). The first mechanism is a dilution effect based on the reduced probability for a spore to produce a new infection due to the reduced density of susceptible plants in the mixture. The importance of dilution effects in heterogeneous environments has been reviewed in a more general context by Keesing et al. (2006). The second mechanism is a barrier effect provided by the resistant plants against spore dispersal. The third effect is induced resistance, triggered by the presence of avirulent spores in the crop that induce the expression of the plant defence mechanisms. This helps prevent infection by pathogenic/virulent spores deposited on the same plants. It is the combined effect of these three mechanisms over several generations of pathogen multiplication that ultimately reduces disease severity (Wolfe, 1985; Mundt, 2002).
The characteristics of the rice/blast host–pathogen system are optimal for an efficient host-diversity effect on blast epidemics (Garrett & Mundt, 1999), with a small genotype unit area (GUA) for the host, a shallow dispersal gradient, a short generation time and a strong host specialization for the pathogen. Multiline cultivars of rice have been developed and used in Japan to control blast epidemics (Koizumi et al., 2004; Kojima et al., 2004; Ashizawa et al., 2007). Rice cultivar mixtures have been successfully implemented on a large scale to control blast epidemics in irrigated rice fields in Yunnan Province of China. In this study, cultivar mixtures were composed of one row of a susceptible traditional glutinous cultivar planted within every four or six rows of a resistant non-glutinous hybrid cultivar. Susceptible cultivars yielded 89% more and their panicles were 94% less severely attacked when cultivated in a mixture than when cultivated in monoculture (Zhu et al., 2000; Leung et al., 2003). To the authors’ knowledge, only two studies have addressed the question of the efficacy of mixing cultivars in reducing blast under upland conditions (Bonman et al., 1986; Prabhu, 1990), where rice blast is more severe than in irrigated or lowland conditions (Moormann et al., 1977). These studies used complex mixtures (of three and five cultivars, respectively) and obtained a rather modest control of panicle blast. Both concluded that further investigations were needed to optimize the approach for upland rice.
The objective of the present study was to contribute to the construction of a sound genetic diversification strategy through the use of cultivar mixtures that could help reduce rice blast epidemics in the stringent epidemiological conditions of upland rice in the highlands of Madagascar. Two-component mixtures, involving a susceptible and a resistant cultivar, in different proportions and different spatial arrangements, were tested and compared to the susceptible cultivar in a pure stand over three cropping seasons. The effects of these mixtures on leaf and panicle blast severity, as well as yield and yield components, were analysed.
Materials and methods
Three cultivars expressing a range of resistance levels were used in this study. Fofifa 152 (F-152) and Fofifa 154 (F-154) are highly valued upland cultivars but are susceptible (S) and very susceptible (VS) to blast, respectively. Inoculation with a collection of 100 isolates collected in different places and different years in Madagascar suggest that F-152 and F-154 have very similar resistance spectra (both are compatible with 58 of these isolates) but also that F-152 bears partial resistance factors that F-154 does not have. Fofifa 172 (F-172) is resistant (R) under field conditions and was shown to be resistant to nine differential blast strains isolated from upland rice in Madagascar, after inoculation in the greenhouse (data not shown). All are upland rice cultivars, adapted for high altitudes, released by the FOFIFA and CIRAD joint breeding programme (Raboin et al., 2011; Table 1). In the following, F-154, F-152 and F-172 will be referred to as VSC (very susceptible cultivar), SC (susceptible cultivar) and RC (resistant cultivar), respectively.
Table 1. Upland rice cultivars, released by FOFIFA for the High Plateau region of Madagascar, used in this study
Grain shape and colour
aStill cultivated, especially in blast free areas. Good cultivars with long grains.
Slender white pericarp
Slender white pericarp
Medium red pericarp
Experimental plots were all located close to the village of Andranomanelatra, 130 km south of the town of Antananarivo in Madagascar (19°47′ S, 47°06′ E). Altitude is 1635 m a.s.l. and average annual rainfall 1460 mm. The wettest months are December and January. Mean temperatures range from 18°C in October, the beginning of the rice sowing period, to 20°C during the reproductive stage. Minimum temperatures can fall below 10°C during the early vegetative stage and are below 15°C during the reproductive and grain filling stages. The night/day thermal amplitude is high (10–12°C) throughout the rice-growing season. The soil is classified as a ferralsol according to the FAO classification (FAO, 2006).
Planting method and crop management
Fields were ploughed manually and secondary tillage ensured a fine seedbed. Before sowing, seed holes were dug every 20 cm in both directions, in a regular pattern. A mixture of Zebu (humped ox) manure (5 t ha−1), dolomite (500 kg ha−1) and NPK fertilizer (11-22-16, 300 kg ha−1) was mixed with soil in each seed hole, with even distribution at the plot scale. Six to 10 seeds were sown per hole (referred to as a ‘hill’). Before sowing, seeds were treated with an insecticide (Gaucho®, 35% imidacloprid +10% thirame, at 2·5 g kg−1 of seeds) to protect the plants from white grubs (larval stage of scarab beetles) and adult black beetles that feed from rice roots. Two or 3 days after sowing, a pre-emergence herbicide was applied (Pendimethalin at 1·2 kg ha−1 a.i.). Further weeding was manual. Urea was applied at a rate of 80 kg ha−1 during the crop cycle: half at the beginning of the tillering stage, and the other half at the end of the tillering stage or at the beginning of the booting stage.
Experimental design in 2007
This trial was sown on 19 November, 2007 at the FOFIFA breeding station. There were four treatments. Three arrangements of two-component mixtures of VSC and RC were compared to the VSC monoculture: alternate rows of VSC and RC (1:1 row mixture), one row of VSC to every five rows of RC (1:5 row mixture) and a 1:5 VSC:RC random mixture. All seeds in one hole were of the same genotype. In row mixtures, rows were oriented in an east–west direction, perpendicular to the dominant wind. The trial was arranged according to a randomized block design with four replications. Individual 6 × 6 m plots (30 × 30 rows) were separated from each other by a 4 m buffer band of RC to limit interplot interactions.
Experimental design in 2008 and 2009
As no difference between row and random 1:5 mixtures was observed in 2007 (see below), and as sowing and harvesting are technically difficult in random mixtures, only row mixtures were used in 2008 and 2009. Moreover, the inoculum pressure was extremely high at the breeding station in 2007, which probably masked the mixture effect. To be in more realistic conditions, the 2008 and 2009 experiments were carried out in farmers’ fields located outside the breeding station and the less susceptible cultivar (SC) was used along with the VSC. These trials were sown on 30 October 2008 and 27 October 2009, at a distance of 4 km (2008) or 200 m (2009) from the breeding station. The experimental design was the same in both experiments. The two susceptible cultivars VSC and SC were used in pure stands or in 1(susceptible):4(resistant) row mixtures with RC in a randomized block design with four replications. In row mixtures, the rows were oriented east–west as in 2007. The size of individual plots was 5 × 5 m (25 × 25 rows) and the plots were separated from each other by a 5 m buffer band planted with RC. To reinforce this barrier, a row of maize was planted in the middle of the buffer band (but did not grow well in 2008). Compared to 2007, the size of the buffer band was increased and, because of space constraints, the plot size was slightly reduced. For this reason, a 1:4 row mixture (20% of susceptible cultivar) was used rather than 1:5 (16·7% of susceptible cultivar) in order to keep five rows of the susceptible cultivar in each plot. Mixture plots included 5 m2 of the susceptible cultivar and 20 m2 of the resistant cultivar.
Blast infection occurred naturally in all experiments. Disease incidence and severity were estimated for both leaf and panicle blast on 10 hills of the susceptible cultivar in each plot. In mixtures, disease was assessed on the susceptible plants only. The 10 hills were chosen on the diagonal of each plot. Over time, the assessment changed from one diagonal to the other between consecutive assessments of a given plot. Leaf blast was evaluated up to the flowering stage. Afterwards, only panicle blast was evaluated. In the 2007 and 2008 experiments, blast was scored weekly after observation of the first symptoms. In the 2009 experiment, leaf blast was scored at 98 and 113 days after sowing and neck blast was scored at 127 and 141 days after sowing.
At each observation date, the total number of tillers and the number of diseased tillers (presenting at least one leaf with one blast lesion) were counted for each of the 10 assessed hills per plot, which gave an estimation of disease incidence. In addition, the percentage of leaf area covered by blast lesions was estimated for each leaf of three tillers randomly chosen among the diseased tillers, for each hill assessed. In the following, only the disease severity value of the third leaf from the top of each tiller (F-2 leaf) was considered as an estimation of plant disease severity. Older leaves were often necrotic and younger leaves presented too few symptoms. Considering all three experiments, disease severity estimated based only on the F-2 leaf was highly correlated with severity estimated from all rated leaves (Pearson correlation coefficient R = 0·873 and P <0·001).
To measure panicle blast after flowering, the total number of panicles and the number of diseased panicles were counted for each of the 10 hills assessed per plot, giving an estimation of disease incidence in the plot. For five panicles per hill randomly chosen from the diseased panicles, the percentage of spikelets affected by the disease was assessed, giving an estimation of disease severity at the panicle level. The product of disease incidence and disease severity was an estimation of the global disease severity (percentage of diseased spikelets at the plot scale). When the neck nodes are infected by the fungus, the entire panicle dries and becomes white. When the infection only occurs on panicle branches or spikelets, only a part of the panicle is affected.
Determination of yield and yield components (2008 and 2009 experiments)
At maturity, 10 hills of the susceptible cultivar were randomly selected in each plot and uprooted to estimate yield components. This sample represented an area of 0·4 m2. The mean density of plants, the mean density of tillers and the mean density of panicles per plot were evaluated from these samples. Panicles were then clipped from their stems, sun dried for 3 days and threshed manually. Filled and unfilled grains were separated and weighed for each sample from the 10 hills. Subsamples of 200 filled grains and 200 unfilled grains were weighed to estimate the number of filled grains (NF = 200× weight of filled grains of total sample/weight of 200 filled grains) and unfilled grains (NUF = 200× weight of unfilled grains of total sample/weight of 200 unfilled grains). These values were used to estimate the fertility ratio (FR = 100 × NF/(NF+NUF)). The following parameters were also calculated: number of plants per m2, number of panicles per plant, number of panicles per m2, number of grains per panicle, number of grains per m2, number of filled grains per m2 and 1000-grain weight.
After the sampling was completed, each plot was hand harvested with sickles. In mixtures, the rows of each cultivar were harvested separately. Threshing was performed in the field with a small manual thresher. After sun drying for 3 days, the harvested grains were winnowed with an electric fan and weighed to estimate plot yield.
All statistical analyses were performed with Statistical Analysis System (sas) for Windows v. 9.1 (SAS Institute Inc.). Uncorrected means and standard errors were generated with proc means. Disease incidence was analysed using proc genmod with a binomial distribution and a logit link function. Disease severity ratings were transformed using the arcsin of the square root transformation to better meet the homogeneity of variance assumptions, and analysed with proc glm. However, the severity data in the figures are untransformed means with their respective standard error. The effect of the mixture arrangement (2007) and of the mixture, the cultivar and the interaction between mixture and cultivar (2008 & 2009) on disease incidence and severity were tested. The effect of mixture on yield and yield components was analysed for each cultivar by analysis of variance (proc glm).
Effect of mixture arrangements on the progress of rice blast (2007 experiment)
Blast epidemic was slowed down in the mixtures although disease severity reached its maximum before harvest time in all treatments (Fig. 1). At the end, nearly all panicles of VSC were attacked whatever the dilution. As a consequence, the final yield of this cultivar was almost zero and there was no statistical difference between treatments (113, 185, 158 and 150 kg ha−1 for VSC grown in pure stand, 1:1 row mixture, 1:5 random mixture and 1:5 row mixture, respectively). By comparison, RC yielded an average of 4660 kg ha−1. In this experiment, blast pressure was very high because the plot was located in the middle of the breeding station, where spreader rows of susceptible cultivars are routinely used for the purpose of selection. Nevertheless, the mixtures significantly reduced leaf blast incidence at 66, 79 and 86 days after sowing (DAS), leaf blast severity at 79, 86 and 93 DAS, panicle blast incidence at 108, 114 and 129 DAS, and panicle blast severity at 114 and 129 DAS (Table S1). The 1:5 mixtures, either sown in rows or randomly, appeared to be more efficient for disease reduction than the 1:1 row mixture, but no difference was observed between row and random 1:5 mixtures.
Effect of 1:4 row mixtures on the progress of rice blast (2008 and 2009 experiments)
Disease incidence and disease severity on leaves as well as on panicles were always lower in 1:4 mixtures than in the susceptible pure stands of both VSC and SC (Figs 2 & 3). At all recording times, a significant statistical difference was detected either in incidence, severity or both (Table S1). No significant interaction between the main effects, cultivar and mixture, was detected (except at 131 DAS in 2008, but the main effects were much stronger), which means that the mixture had roughly the same blast reduction effect for both cultivars.
As expected, the disease level was much lower in the 2008 and 2009 experiments than in 2007. The disease level was also lower in 2008 than in 2009. For SC, at the last recording time, the disease severity at the plot scale was 7·7% of blasted spikelets in pure stands and 2·8% in the 1:4 row mixture in 2008, and 45·1% in pure stands and 7·2% in the 1:4 row mixture in 2009 (Figs 2 & 3). The final severity on VSC was 40·7% and 4·7% in pure stands and in the mixture, respectively, in 2008 and 68·9% and 13·8% in pure stands and in the mixture, respectively, in 2009 (Figs 2 & 3).
Yield components (2008 and 2009 experiments)
For the susceptible cultivars and in both experiments, there was no significant difference between mixtures and pure stands in the number of plants per m2, the number of panicles per plant, the number of panicles per m2, the number of grains per panicle and the number of grains per m2 (Table 2). This indicates the absence of competition between the susceptible and the resistant component in the mixtures as well as the absence of an effect of the disease on plant development before heading. Moreover, the yield of RC was not different when mixed with VSC or SC, which also indicates the absence of competition between RC and the susceptible cultivars (Table 2).
Table 2. Comparison of yield and yield components in two upland rice cultivars grown in 1:4 row mixtures or in pure stands
1:4 row mixture
P > F
1:4 row mixture
P > F
aSusceptible cultivar used in the mixtures.
bSignificant effect of mixture on the yield component at P <0·05 threshold for a given cultivar (figures in bold).
Plants per m2
Panicles per plant
Panicles per m2
Grains per panicle
Grains per m2
Filled grains per m2
Susceptible component yield (kg ha−1)
Resistant component yield (kg ha−1)
The yield of SC was almost the same in the pure stand and in the 1:4 mixture in 2008. It increased by 56% in the mixture in 2009, but this difference was not statistically significant. Nevertheless in 2009, the number of filled grains per m2 and the spikelet fertility of SC were significantly higher in the mixture than in the pure stand (Table 2). The yield of VSC was significantly higher in the 1:4 mixture than in the pure stand both in 2008 and 2009 (Table 2). Accordingly, spikelet fertility was significantly higher in the mixture in both 2008 and 2009, and the 1000-grain weight was significantly higher in the mixture in 2008.
The yield and the disease severity estimated at the plot level were negatively correlated (Pearson correlation coefficient R = −0·91 and P <0·0001 in the 2008 experiment; R = −0·85 and P <0·0001 in the 2009 experiment; Fig. 4). Spikelet fertility and disease severity were also negatively correlated (R = −0·94 and P <0·0001 in the 2008 experiment; R = −0·88 and P <0·0001 in the 2009 experiment). Yield and spikelet fertility were positively correlated (R =0·81 and P =0·0001 in the 2008 experiment; R =0·92 and P <0·0001 in the 2009 experiment). Thus, the yield increase observed in the 1:4 mixtures can be explained by an increase in spikelet fertility that resulted from a reduction in the disease.
In these experiments, blast disease epidemics in upland rice were significantly reduced when a susceptible cultivar was grown with a resistant cultivar in a two-component mixture, compared to the same susceptible cultivar grown in a pure stand. This extends the results obtained on cultivar mixtures of irrigated rice (Zhu et al., 2000) to upland rice systems. In conditions that were close to farmers’ fields, the disease reduction provided by the mixture was very effective: in 2008, the yield of VSC was 46% higher in the 1:4 row mixture than in the pure stand. For SC, the mixture did not provide a significant advantage over the pure stand but the impact of blast was lower. In 2009, the yields of VSC and SC were 247 and 56% higher in the mixture than in the pure stand, respectively. Mixtures may thus have a good potential for limiting yield losses due to blast epidemics in upland rice.
The mixture was ineffective in 2007, but that year, the experiment was set up in the FOFIFA breeding station where a dense network of spreader rows of susceptible cultivars provided a continuous source of inoculum near the trial. The dependence of the mixture efficacy on inoculum pressure has been observed in other situations (Mundt, 2002); because the mixture efficacy mainly results from a reduction in spore exchanges between plants it can be impaired by a strong input of exogenous spores that directly infect most of the plants. However, when the mixtures were tested outside the experimental station, their performance appeared relatively stable in different situations; they significantly reduced the disease on VSC as well as SC and they were effective both in 2008 and 2009, even though inoculum pressure was higher in 2009. The comparison of the 2007 experiment with that of 2008 and 2009, performed in farmers’ field conditions, suggests that growing cultivar mixtures is a robust strategy with regards to varying inoculum pressure and host susceptibility.
Even though the mixtures were globally effective in 2008 and 2009, their effectiveness level varied, probably because of differences in environmental conditions between both years. To evaluate whether or not cultivar mixtures are appropriate and efficient for rice cropping in farmer’s field conditions in the Madagascan Highlands, this approach will have to be further validated through multi-location and multi-year tests in local production conditions. It is nevertheless anticipated that the efficacy of mixtures for reducing disease may be higher in farmers’ fields than in the present experiments for three reasons. First, the high blast pressure observed in the experiment run in 2007 was not representative of conditions found in farmers’ fields. Secondly, nitrogen-based fertilizers (NPK and urea) were used that most Madagascar farmers cannot afford. It is well known that the severity of blast increases with the rate of nitrogen application (Kurschner et al., 1992; Long et al., 2000). Thirdly, the small size of the experimental plots may have led to underestimation of the efficiency of the mixture (Wolfe, 1985; Zhu et al., 2000).
The design of cultivar mixtures must take both practical and efficacy criteria into account. In particular, the number and proportion of the components and their spatial distribution have to be optimized. In the 2007 experiment, the 1:5 mixture was more efficient than the 1:1 mixture and, in 2008 and 2009, the 1:4 mixture provided a good control of rice blast. Interestingly, the randomized arrangement of susceptible and resistant plants was not more efficient than the row arrangement in controlling the disease. This is surprising because the effectiveness of cultivar mixtures is expected to decrease with increasing Genotypic Unit Area (area occupied by a same susceptible genotype; Mundt & Leonard, 1986). An explanation of this result may be the row orientation. The east–west orientation of the susceptible cultivar, i.e. perpendicular to the dominant wind, may have prevented the disease from spreading along the rows. However, such a dispersal effect would need to be better quantified. In two-component mixtures in which one component is resistant to all pathogen strains, only the barrier and dilution effects are exploited. Differences of architecture and height between cultivars in the mixture may also create micro-environmental conditions that are less favourable to blast infection (Zhu et al., 2005), although in these experiments all cultivars were very similar in height. An effect of induced resistance, which has been demonstrated for rice resistance to blast (Manandhar et al., 1998), could be obtained by mixing more cultivars with complementary resistance factors to rice blast. In addition to providing an induced resistance effect, increasing the number of cultivars may also increase the dilution effect (Newton et al., 1997). However, designing more complex mixtures would require the characterization of the interaction between M. oryzae populations and new upland rice cultivars, which has not yet been done. It would also require the analysis of the mixing ability of the different cultivars to avoid any unwanted effects of competition between the components of the mixture that, in some instances, may counter the advantage obtained from disease reduction (Alexander et al., 1986; Finckh & Mundt, 1992). The efficiency of mixtures using a cultivar with a high level of partial resistance instead of a completely resistant cultivar could also be investigated to increase the options.
In Madagascar, breeding for resistant cultivars is currently the main strategy for controlling blast disease. Upland rice development is recent (<20 years) and few varieties are available for rice grown at high altitude. The most recent cultivars brought great progress in terms of blast resistance as well as yield potential and were readily adopted by farmers because they cannot afford chemical control. Nevertheless, to ensure their durable efficacy, resistant cultivars should be used as part of an integrated strategy. For instance, cultivar mixtures could be used in combination with cropping systems designed to reduce diseases (Sumner et al., 1981; Sester et al., 2010). The effect of these integrated strategies would be amplified if they were deployed at the regional scale (Zhu et al., 2000; Mundt et al., 2002). In the short term, simple two-component mixtures, such as those tested in this study, could be used to protect highly valued upland cultivars such as F-152 (SC) and F-154 (VSC) that have become susceptible to blast. The latest resistant upland cultivars released (including RC F-172) have medium grains with red pericarp that contrast with the highly appreciated longer grains and white pericarp of VSC and SC. In the future, when more upland cultivars with a well-characterized resistance spectrum are available, more complex cultivar mixtures could be promoted.
The use of varietal mixtures is a promising strategy for the sustainable cultivation of rice under upland conditions in Madagascar. In the context of subsistence agriculture, it may enable more stable production, reduce the damage caused by the disease, and allow intensification of production despite farmers’ limited resources (Smithson & Lenné, 1996). In Madagascar, as elsewhere (Zhu et al., 2000), no practical problems are posed by the use of row mixtures because most of the cropping operations are manual. The separate harvesting of cultivars with distinct qualities is consequently not a serious problem for farmers. In Madagascar, farmers already have long experience in the use of mixed cropping and landrace diversity and thus could rapidly adopt this strategy.
This research was conducted in the FOFIFA Research Station of Antsirabe, Madagascar. The authors thank all those who contributed.