Study site

The study was conducted on the Campus of Biological and Agricultural Sciences of the Universidad Autónoma de Yucatán (21° 51’ N y 89° 41’ W), in Xmatkuil, Yucatán, Mexico (Figure 1), from January to April 2012. The region’s climate is Aw₀ (^{García, 1988}); in the last two years average annual rainfall was 967 mm, most concentrated in the period from June to October. Average high and low temperatures are 37.8 and 24.5 °C (Figure 2). Predominant soils are moderately fertile rendzinas with 1 to 1.5% organic carbon (CO) and pH 7.5 to 7.8 (^{Bautista et al., 2011}).

Fuente: (http://www.seduma.yucatan.gob.mx/ordenamiento-ecologico/index.php).

Figure 1 Location of the study site in Xmatkuil, Yucatán, Mexico. 

Figure 2 High and low temperatures and rainfall in the study site. Arrows indicate harvest dates of forage species biomass. (Data were taken from the mini-weather station of the Campus of Biological and Agricultural Sciences of the Universidad Autónoma de Yucatán, Mexico, from May 2011 to June 2012). 

The area’s vegetation is classified as low deciduous tropical forest, in which bushes reach heights o 4 to 7 m. The most abundant species are Gymnopodium floribundum, Sapranthus campechianus, Bourreria pulchra, Diospyros bumeloides, Neomillspaughia emarginata, Gouania lupuloides, Karwinskia humboldtiana, Exostema caribaeum, Guettarda elliptica, Hamelia patens, Machaonia lindeniana, Randia aculeata, R. longiloba, Thouinia paucidentata, Sideroxylon americanum, Celtis iguanaea, Callicarpa acuminata, Petrea volubilis, and Hybanthus yucatanensis. In addition, there are larger species (8 and 12 m tall) such as Bursera simaruba, Havardia albicans, Lysiloma latisiliquum, Piscidia piscipula and Alvaradoa amorphoides (^{Gutiérrez-Báez and Zamora-Crescencio, 2012}).

Experimental plot

The experimental plot was established in 2004 and experimental units were located; each plot had a woody species: L. leucocephala, G. ulmifolia or M. oleifera. The woody plants were established in rows (three per experimental unit): 2.0 m between rows and 0.5 m between plants. Trees were completely pruned (>90% of the foliage) to 1.0 m above ground level with a use frequency of three months, from the first year of establishment. Auxiliary irrigation was applied during the dry season.

For this study, two crop cycles (CC) were evaluated. In both cycles, maize planting density was 20,000 plants ha^-1, using seeds of the local variety “Nal-Xoy”, a random cross of “Nal-Tel” with the variety PR7822 (^{Graefe, 2003}). This maize cross has been used since establishment of the experimental plot. Seeds were germinated in plastic trays with 70 wells 3 cm in diameter and 5 cm deep. The seedlings were transplanted 15 d after germination. Before transplanting the maize, the woody species were pruned, as was done at establishment of the experimental plot (^{Caamal-Maldonado et al., 2012}).

The experimental design was complete random blocks with three replications. The treatments were the following: L. leucocephala associated with maize (Leucaena + maize), G. ulmifolia associated with maize (Guazuma + maize), M. oleifera associated with maize (Moringa + maize), and a control (maize monocrop).

Weed sampling

To understand weed dynamics, samples were taken at three times: 1) at the beginning of each CC (July 2011 and January 2012) when maize was transplanted; 2) at the middle of each CC after the first pruning of forage tree species (August 2011 and March 2012); and 3) at the end of each CC before pruning forage tree species (October 2011 and April 2012).

Response variables

Total weed cover was estimated in 0.5x0.5 m metal frames, six per experimental unit distributed in zig-zag on both sides of the center row of the respective tree species, using high-resolution digital photographs, 14.0 megapixels, taken with a Canon Rebel EOS XS camera. These images were taken at a perpendicular angle (90°) at a height of 1.20 m. Each image was then analyzed with the software CobCal version 2.1 for Windows^© to estimate cover (^{Ferrari et al., 2006}).

The weeds inside the metal frame were harvested with roots to determine fresh weight of the total biomass. They were separated by species to determine their diversity (number of species) and counted to quantify their density (individuals m^-2). All samples were dried in a forced air circulation oven at 70 °C until constant weight to determine dry matter content. Weed suppression potential was also determined with the formula of ^{Rippin et al. (1994}):

(1)

where C is weed biomass in the control treatment (g DM m^-2), and X is equal to the weed biomass in the treatment corresponding to the agroforestry systems (g DM m^-2).

Statistical analysis

The data were analyzed with an ANOVA using GLM (^{SAS Institute, 2004}) to examine the effect of agroforestry systems on the cover dynamics, biomass, diversity and density of the weeds in these systems (^{SAS Institute, 2004}). For significant differences, the Tukey test (p≤0.05) was used.

Weed cover

At the beginning of the first CC the Guazuma + maize system had less weed cover (41%), relative to the other treatments. At the middle of this CC (August) the Leucaena + maize system had less weed cover, followed by the plot with Guazuma + maize and the maize monocrop (control). The latter two were similar to the system with Moringa + maíz, which had more weed cover (71%). At the end of this CC, the systems Leucaena + maize, Guazuma + maize, and the monocrop continued to have the lowest weed cover (49%) (Table 1).

Table 1 Weed cover (m²) in agroforestry systems formed by forage trees and maize compared with a maize monocrop in two crop cycles, Xmatkuil, Yucatán, México. 

Means in a row with different letters indicate significant differents for each sampling period (p≤0.05).

At the beginning of the second CC the Guazuma + maize system had the least weed cover (37%) followed by Leucaena + maize and the maize monocrop. The latter two were not different from the system with Moringa + maize. At the middle of this cycle, weed cover was similar in all treatments (51% on average), but at the end of the second cycle, Guazuma + maize and Leucaena + maize had less weed cover (48 and 54%), (Table 1).

In our study, the association of maize with woody species showed positive results for weed control. This agrees with that reported by ^{Rippin et al. (1994}), ^{Caamal-Maldonado et al. (2001}) and ^{Cummings et al. (2012}). This is true particularly in reference to biomass. Even though the trees were pruned at the beginning of each cycle, later branch development provided shade, especially in Guazuma. Moreover, its continuous loss of leaves created permanent cover on the soil, forming a kind of mulch, which limited weed growth. (^{Caamal-Maldonado et al., 2001}). In this way, crop associations with leguminous and non-leguminous species can contribute to reducing weed biomass.

It is important to underline that the methods for measuring cover differ among the studies reviewed; in some cover is estimated using indirect (visual) measurements but in our study digital photographs were used because they offer greater precision in measuring this variable (^{Ferrari et al., 2006}).

Although we used a more precise photographic method to assess cover, the results did not show a direct relationship to weed biomass. In fact, the percentage of cover does not permit discriminating whether there is more or less biomass; it only expresses whether the surface is more or less occupied by weeds. Thus, there may be sampling frames with more or less biomass that have the same cover. Indeed, species with horizontal growth habits (i.e. creepers, procumbent or ascendant) will cover more ground than species with vertical growth (i.e. straight). Likewise, the shape and angle of the leaves are fundamental elements of this variable (^{Rippin et al., 1994}).

Weed biomass

During the first and second CC, the agroforestry systems (i.e. Guazuma + maize, Moringa + maize and Leucaena + maize) had less weed biomass than the maize monocrop (Table 2).

Table 2 Weed biomass (g DM m^-2) in agroforestry systems consisting of forage trees and maize compared with a maize monocrop in two crop cycles, Xmatkuil, Yucatán, México. 

Means in a row with different letters indicate significant differents for each sampling period (p≤0.05).

In both CC the association of forage tree species with maize decreases weed biomass, relative to the monocrop. The leaves of these trees intercepted sunlight, so that weeds did not germinate or did not develop to their reproductive stage. This agrees with ^{Jama et al. (1991}), who showed that the association of L. leucocephala with maize can reduce weed biomass up to 90%. The broader leaves of G. ulmifolia may also have contributed to reducing the biomass of these species. It is important to consider that pruning the trees at the beginning of each CC, to obtain forage in a cut and carry scheme, permitted the maize seedlings to receive the light necessary to begin their development. Regrowth of tree leaves gradually shaded the soil surface, affecting weed germination and establishment. We can add the fact of the litter of the trees, among which G. ulmifolia is especially relevant, covered the space like a true carpet with greater effect on weed development.

At the beginning of the experiment, we proposed assessing forage production and maize yield in agroforestry arrays. That is, the trees were not established as a method of weed control. Nevertheless, the effect of woody species on herbaceous plants is important and can contribute to success of the system and reduce the use of external inputs such as herbicides.

Weed diversity

Weed diversity at the beginning and middle of the first and second CC was similar in the evaluated treatments. However, at the end of both cycles, the agroforestry system Guazuma + maize was outstanding because it had fewer weed species than the maize monocrop (Table 3). This agrees with the dynamics of the biomass of these species in the respective treatments: lower values of this variable coincide with fewer species in the treatment.

Table 3 Number of species (in 0.5 m^-2) in agroforestry systems consisting of forage trees and maize compared with a maize monocrop in two crop cycles, Xmatkuil, Yucatán, México. 

Means in a row with different letters indicate significant differents for each sampling period (p≤0.05).

Forty-seven species belonging to 18 families were identified. Of these, more than 21% were leguminous (Fabaceae), followed by the families Poaceae (12.8%), Asteraceae, (10.6%), Malvaceae, (8.5%), y Euphorbiaceae (8.5%), which together account for more than 61% of all the species found (Figure 3).

Figure 3 Botanical families of the weeds associated with agroforestry systems consisting of forage trees and maize in Xmatkuil, Yucatán, México. 

Diversity of the weed species at the end of the two CC was lower where G. ulmifolia was present than with the presence of other tree species and the maize monocrop. This can be explained by G. ulmifolia’s architecture: its leaves form a dense crown that limits entry of sunlight. Although cover of forage plants was not measured, we could observe that, unlike G. ulmifolia, L. leucocephala, even when branches sprout from the trunk base, permits entry of light, which is necessary for heliophilous plants such as weeds. Moreover, M. oleifera has peculiar growth after pruning: the trunk lengthens and the branches sprout in the upper part, similar to the shape of a feather duster. This, together with the fact that its branches do not form a dense crown, permits the emergence of weed species. In this regard, ^{Segura-Rosel et al. (2012}) indicate that G. ulmifolia has large simple leaves that, after falling, persist on the ground as litter for more than four weeks. Besides, ^{Casanova-Lugo et al. (2014}) report that the biomass composition of G. ulmifolia has an influence in the number and seasonality of the species under its canopy. Thus, it can create a moist shaded microclimate that would limit the presence of weeds.

Fabaceae, Poaceae and Asteraceae were the predominating botanical weed families in the agroforestry systems formed by forage species and maize. This is in agreement with ^{Caamal-Maldonado et al. (2001}), who found that the most important weeds were Asteraceae, Poaceae, Fabaceae and Acanthaceae in their study of the allelopathic effects of four leguminous species, Mucuna deeringiana, Canavalia ensiformis, L. leucocephala and Lysiloma latisiliquum, on weed control in maize crops in Yucatán. But in Havana, Cuba, in a maize crop preceded by a transitory fallow after growing potatoes (Solanum tuberosum), the dominant weed families were Poaceaes and Euphorbiaceaes (^{Blanco and Leyva, 2010}). This suggests that the diversity of weeds is related to edaphic-climatic conditions of the region, as well as to preceding cultural practices related mainly to residual damage by selective herbicides.

Weed density

At the beginning of the first CC, the treatments were similar in terms of weed density. In contrast, at the middle of this cycle, the highest weed density occurred in the maize monocrop. In the same way, at the end of this first cycle, the maize monocrop accumulated a higher weed density; the lowest density corresponded to the agroforestry systems Guazuma + maize and Moringa + maize, with statistically similar values (Table 4).

Table 4 Weed density (individuals m^-2) in agroforestry systems consisting of forage trees and maize compared with a maize monocrop in two cropping cycles, Xmatkuil, Yucatán, México. 

Means in a row with different letters indicate significant differents for each sampling period (p≤0.05).

At the beginning and end of the second CC, the maize monocrop had the highest density of weeds. Nevertheless, at the middle of this second period, there were no significant differences among the assessed systems (Table 4).

Regarding weed density, at the middle and end of the first cycle, the maize monocrop had the highest value, as was expected, given that the presence of the trees in the other systems can reduce shade-intolerant weeds. In the treatments with trees, pruning increases the presence of weed species, but when the trees recover, they again limit the conditions for growth of herbaceous plants. ^{Barbier et al. (2008}) indicate that weed density decreases because of the variation in availability of resources such as light and other effects caused by the characteristics of tree species.

In the second CC, the maize monocrop had a larger quantity of weeds, and since there were no tree species in this treatment, weeds exerted more pressure. In this sense, ^{Ayala-Sánchez et al. (2007}) mention that in the traditional systems, there is greater presence of weeds because, when there is activity in the soil, the latent seeds germinate and their competitive potential is greater. This was more evident in our study since the site had been cultivated for more than eight years (^{Caamal-Maldonado et al., 2012}). In addition, the trees associated with maize limited weed development. It is also important to mention that in another experiment in the same plot, yield of maize monocrop increased when L. leucocephala and G. ulmifolia foliage was applied as cover (^{Gallegos-Pérez et al., 2013}). This coincides with ^{Jama et al. (1991}), who sustain that the use of leguminous trees (i.e. L. leucocephala) as green organic fertilizer in maize cultivation increases yield by 24 to 76%. This was also corroborated by ^{Rippin et al. (1994}), who used other leguminous trees, such as Erythrina poeppigiana and Gliricidia sepium. In this context, it is possible to point out that agroforestry systems have the potential to generate sustainable production systems by promoting better weed control.

Potential weed suppression

Potential weed suppression varied from 23% to 36% in the agroforestry systems evaluated, relative to the maize monocrop. The Guazuma + maize agroforestry system had the highest potential, followed by the Moringa + maize and Leucaena + maize systems. Consequently, the maize monocrop had the highest relative proportion of weeds (Figure 4) of the four treatments.

Figure 4 Potential suppression and relative proportion of weeds in agroforestry systems consisting of forage trees and maize compared with a maize monocrop, in two crop cycles in Xmatkuil, Yucatán, México. 

The weed-suppressing effect of trees reported by ^{Jama et al. (1991}) was also observed in our study, although to a lesser degree; the agroforestry systems we evaluated produced 23% to 36% less weed biomass than the monocrop. This suppressing effect was more evident with maize associated with G. ulmifolia for the reasons given above: leaf size limiting the entry of light, permanence of leaf litter on the soil, and modification of the microclimate. It is also feasible that the nature and content of chemical compounds, especially of secondary metabolites in the plant leaves, have significant effects on weed germination and growth. For example, L. leucocephala leaf litter contains compounds that limit weed development (^{Caamal-Maldonado et al., 2001}). The amino acid mimosine is possibly the main compound responsible for this effect. However, few studies focus on determining whether said compounds act individually or in conjunction, in order to promote them for weed control, particularly in agroforestry systems.

The search for sustainable alternatives for weed management involves maximizing beneficial ecological processes in agroecosystems to maintain weed populations at low thresholds so they do not negatively affect crop growth and development (^{Caamal-Maldonado et al., 2001}; ^{Cummings et al., 2012}). To this end, it is necessary to identify the type of biodiversity that is desirable to maintain or broaden in order to propose better practices (^{Lin et al., 2004}; ^{Barbier et al., 2008}) and reduce the use of herbicides to minimal levels (^{Cummings et al., 2012}). Although agricultural systems are simpler than natural ecosystems, there are options for redesigning and managing agroecological systems and reduce weed populations (^{Charudattan, 2001}). Among these alternatives are included agroforestry systems, which offer benefits for both the farmer and the environment.