Maize intercropping in the milpa system. Diversity, extent and importance for nutritional security in the Western Highlands of Guatemala

Data

Data were collected in 2015 through a survey deployed as part of the Feed the Future Buena Milpa project (http://www.cimmyt.org/project-profile/buena-milpa/). The project’s goal was to reduce food insecurity and malnutrition by fostering sustainable, resilient, and innovative maize-based farming systems in the WHG.

We conducted the survey, with support from local researchers and agronomy students of the University of San Carlos (USAC), in summer of 2015 in 989 maize-growing farm households in 64 communities of 16 municipalities of the WHG. The number of surveys at the departmental level was: Totonicapán (226), Quiche (350), Quetzaltenango (187) and Huehuetenango (226) (See Supplementary Table 1 and Supplementary Fig. 1 for details). Criteria to select communities were: location within the targeted municipalities in the Buena Milpa project, the active work of project partners within the communities, and their distribution within four selected watersheds that included farming systems at different altitudes, ranging from 1400 to 3200 masl. For household selection, enumerators walked radial transects to survey household members, choosing only households that explicitly agreed to participate and had agricultural land on at least part of which maize was grown. The survey was a closed questionnaire with 139 questions in 18 sections including 5 project themes: milpa-maize germplasm improvement, natural resource conservation in farming system, farming system diversification, agricultural innovation systems and social inclusion.

Milpa diversity and extent

We used survey findings regarding the previous year’s crop production (2014) for all 989 surveyed households to understand the importance and diversity of the milpa system. The 989 households reported crop production on a total of 1,541 plots, 1,324 of which included maize. The other 217 plots grew potato (85 plots), coffee (50), vegetables (31), bean (19), fruit trees (12), faba bean (7), forestry trees (6), pea (2), oats, wheat, etc. Given the study’s focus on smallholder farmers, we discarded other 119 plots that had unrealistically large values for plot size or maize production levels for the region (i.e. more than ten times the average plot size or calculated maize yield, > 2 ha and > 11 ton ha⁻¹ year⁻¹).

For the resulting 1205 plots we constructed a tree depicting maize-system diversity in the WHG, with each node indicating the main cropping associations. The tree was nested, so the first node displays all maize plots, with successive splits for monocrop vs intercropped maize and with crop names presented according to how often they are grown (overall: maize > common beans > potatoes > squash > faba beans > fruit trees > vegetables). So, in plots where maize is grown with potatoes and common beans, the association is termed maize-bean-potatoes.

Milpa and food security

We assessed maize yield differences for monocropping and intercropping to detect a yield penalty or advantage for maize (see below), including survey information only for households from which we had available yield data for all crops. In the survey, total production for each crop was recorded regardless of the number of plots on which it was grown and therefore, for 398 households (40.2% of the sample) that had more than one plot under maize or any of the milpa crops, it was impossible to calculate the yield and had to be discarded from the analysis due to a lack of available plot-level information. The results were further screened for complete crop information and unrealistic values on crop production levels (i.e. ten times higher than the average yield for the different crops—See Supplemental Table 2), resulting in a usable sample of 368 plots.

For statistical analysis, only those crop combinations with a sample size equal to or greater than 9 plots were selected, resulting in 357 plots with, in descending order, the following numbers of 300 plots sown to each cropping combination: maize monocrop (163), maize-bean (109), maize-bean-squash (30), maize-bean/faba (12), maize-potato (13), maize-squash (11), maize-bean-potato (10), maize-faba (9). Other crop combinations with very low sample sizes were maize-squash-faba (4), maize-potato-faba (3), maize-bean-squash-faba (2), maize-bean-potato-faba (1) and maize-bean-potato-squash (1), making it difficult to include them in further statistical analyses.

Maize yields for plots under monocropped maize (163) were first compared to maize yields from intercropped plots (194). To choose the most appropriate statistical test, we checked if the outcome variable, maize yield, met the assumptions required for a parametric test. Although we had independent samples, large sample sizes, and homogeneous variances (F = 1.1809, p-value = 0.267), maize yield proved to be non-normally distributed after Shapiro-Wilks normality test was significant (W = 0.911, p-value < 0.000 (1.398e−13)). Thus, we choose a non-parametric test, the Wilcoxon rank-sum (also known as the independent 2-group Mann–Whitney U Test), to compare maize yields in monocrop and intercrop. Results are shown in boxplots for yield of both groups in Fig. 3-A. Maize yield was also compared between all eight crop associations, as maize yield presented a non-normal distribution and we have several groups to compare, a Kruskal–Wallis test, was performed. While in general, omnibus test like Kruskal–Wallis are used to detect the existence of at least one significant difference across groups, sometimes they fail in detecting significant differences between pairs of groups, and hence we decided to perform a post hoc test. The Dunn’s test of multiple comparisons was selected as the most appropriate one as it is suited for groups with unequal number of observations²⁷ and because it retains the rank sums from the omnibus test, this case Kruskal–Wallis²⁸. Results of maize yields for the different cropping systems and Dunn’s test significant groups are presented with boxplots in Fig. 4.

Potential adequacy of milpas for human nutrition

We applied a Functional Diversity approach used in ecology research^29,30,31 to assess the nutritional performance of milpa systems. We calculated the number of people who can obtain recommended daily allowances (RDA) of 14 different nutrients from different milpa systems. We then calculated the Potential Nutrient Adequacy (PNA)¹⁸ to determine the number of persons (male adult equivalents) who could obtain RDAs of a full set of nutrients from 1 ha of a milpa system, including monocropped maize and associations, using the following equation:

where (N) is the number of nutrients considered (14), (s_{i}) is the fraction of potentially nourished male adults by a given nutrient, and (overline{s}) is the average of potentially nourished persons for all nutrients. To calculate s we multiplied the yield of each crop reported in the survey (kg ha⁻¹) by the nutrient composition of each crop, using the content per 100 g of the 14 different nutrients for each crop, as per the INCAP Food Composition Table for Central America³² and using the INCAP RDA for an adult male³³ (See Supplementary Tables 3 and 4). All values were calculated per hectare and year. PNA levels across cropping associations were also compared. First, we checked if PNA data met the assumptions required for a parametric test. PNA variances proved to be non-homogeneous (F = 14.5, p-value =  < 2e-16) and also have a non-normal distribution (Shapiro–Wilk , W = 0.93742, p-value = 4.107e-11). Thus, a Kruskall-Wallis and Dunn’s tests were performed to assess PNA differences between cropping combinations. Results are presented with boxplots in Fig. 6.

Source: Ecology - nature.com