Introduction

Arthropods were assumed to be resistant to anthropogenic changes and to have large populations worldwide. However, there is now strong documentation of arthropods’ decline around the world. Dirzo et al. (2014) described a decrease in the abundance of certain insect groups. More recently, Hallmann et al. (2017) reported a striking 75% decline in flying insects in natural areas of Germany, and in a global review, van Klink et al. (2020) documented a declining trend for terrestrial arthropods worldwide, highlighting a significant information gap from the tropics. Insect pollinator decline has received significant attention, given the decreased productivity of many agricultural crops as well as the expansion of colony collapse disorder in bees (Potts et al. 2010). Dung beetles, another important insect group related to productive systems, have also been declining (Numa et al. 2020). Insect predator and parasitoid communities’ conservation status have not been evaluated, even though they are crucial for biocontrol in agriculture and forestry (Desneux et al., 2007). On the other hand some herbivorous species (aphids, caterpillars and grasshoppers) with generalist habits that behave as pests in agricultural and forestry systems are increasing in abundance (Lehmann et al. 2000).

Since the green revolution, there has been a dramatic increase in the use of external inputs for agriculture (FAO 2003). Fertilizers, herbicides and pesticides in particular are used without considerations for health or environmental issues or for the particularities of each field (Akanksha et al. 2020). Therefore, the amounts of chemical inputs actually used in agricultural fields are often far above the recommended doses (Bejarano 2017). In recent years in Mexico, approximately 3,000 tons of active ingredients per year have been used to combat the fall armyworm (Spodoptera frugiperda Smith; Blanco et al. 2014). The overuse of pesticides has led to the development of pest resistance and the extermination of arthropod communities, killing not only insect pests but also many beneficial insects such as predators, parasitoids and detritivores (animals that contribute to the decomposition of dead organic material; Attwood et al. 2008; Bengsston et al. 2005; Flores-Gutiérrez et al. 2020; Theiling and Croft 1988). This can result in the loss of the ecosystem services that native predators and parasitoids may provide (Chapin et al. 2000; Desneux et al. 2007; Isaacs et al. 2009; Losey and Vaughan 2006; Zhang et al. 2007).

Mexico has a long tradition of maize agriculture. For centuries, small-scale farmers in Mexico have developed local maize landraces, with 41–65 landraces recognized today (Kato et al. 2009). Until the mid twentieth century, maize was mainly cultivated in the milpa system. This is a highly diversified system that involves high intra- and inter-species diversity and a profound local knowledge of the use of all the species of plants and animals found there to regulate pests and maintain soil health and nutrition, as well as for medical uses and food security (Martínez et al. 2020; Rodríguez-Robayo et al. 2020). Today, maize produced in the context of small farms, for example, covers not only subsistence needs, but also produces a surplus for local and regional animal and human needs (Bellón et al. 2021). However, beginning in the second half of the twentieth century, Mexico’s state policies promoted the green revolution’s technological schemes, including monoculture, mechanization, commercial varieties and synthetic fertilizers and pesticides (Aguilar et al. 2003). Traditional agricultural practices like the milpa are now facing challenges such as a lack of younger generations incorporating into agriculture, low commercial value of milpa products, and state and agroindustry pressure on small farmers to consume synthetic agrochemicals (Ebel et al., 2017; Orozco and Astier 2021).

Most of the beneficial aspects of traditional low-input maize cultivation are not well appreciated, and there are few studies that directly compare traditional versus conventional management. Therefore, the aim of this investigation was to evaluate the effects of conventional versus low-input agriculture on arthropod communities under two irrigation systems in Michoacán, México. Our hypothesis was that arthropod diversity would be higher in low-input rainfed agricultural plots.

Methods

Study sites

This study was performed during the 2011 agricultural cycle in central Michoacán, Mexico. We chose two localities with different maize cultivation practices; one depended on rainfall only (R), and the other used irrigation (I). In each locality we chose plots under low (LI) and high-input (HI) management. The rainfed only locality was in Cherán (19° 41’ N and 101° 57’ W) at 2400 m asl, with temperatures that range between 6–26°C with 1000 mm annual precipitation. The irrigated locality was in Álvaro Obregón (19° 48’ N, 101° 02’ W) at 1800m asl with 918 mm annual precipitation and temperatures between 12 and 27° C (INEGI 2008).

In each locality, we located six 1 ha. maize fields—three managed with conventional high-input management (HI; chemical fertilization, herbicide, and insecticide applications) and three with low-input management (LI; green manure and composted animal manure for fertilization and manual weed removal, Table 1 ). Thus, in total we had twelve plots (six per locality). Plots were separated by at least 500 m in each locality, and the farmers had followed the same management strategy for at least three consecutive years in the selected plots. After maize harvest, five composite soil samples were collected (25 cm depth) from all plots. Soil samples were sieved at 2 mm and air-dried until constant air-dried weight was achieved. We characterized soil texture and determined phosphorous content determined following the Olsen and Dean (1965) method, total Nitrogen content by the Kjeldahl method, and soil organic matter using the Walkley and Black method (1934) and the Cation Exchange Capacity following SEMARNAT (2002).

Table 1 Characterization of plot management including external inputs and soil characteristics
Full size table

Arthropod and maize sampling

In each maize field, we sampled arthropod diversity three times during the agricultural cycle in July, August and September 2011. We used nine pitfall traps located in the center of each plot, arranged in three lines separated by 3 m, as recommended by Duelli et al. (1999). Pitfall traps consisted of buried 250 mL plastic cups half filled with soapy water and 10 ml of ethanol. To prevent the water from evaporating and rainfall from accumulating, the cups were covered with plastic plates supported on metal legs 3 cm above the rim of the cup. Pitfall traps were left open for 96 h in each sampling period. The arthropods collected were sorted and identified in the laboratory using taxonomic keys and regional arthropod guides (Borror et al. 1989; White and Peterson 1998; Eaton and Kaufman 2007). We identified all taxa as morphospecies to the highest degree of detail possible, a technique known as “taxonomic sufficiency” (Ellis 1985) or “lowest practical taxonomic level” (LPT) (e.g., Hanula et al. 2009). A morphospecies can be defined as a group of biological organisms whose members differ from all other groups in some aspect of their form and structure, or species that can be distinguished from other species by their external morphology (Hale et al. 2005). We also assigned a trophic guild for each morphospecies, considering the Order or Family to which they belonged using Insect identification guides (Borror et al. 1989; White and Peterson 1998; Eaton and Kaufman 2007) and the Naturalista (https://www.naturalista.mx) and Enciclovida (https://enciclovida.mx) web pages. When a morphospecies belonged to a Family where several trophic guilds have been reported, the guild was defined as “Other” and was not considered for the statistical analysis.

Since we were interested in linking agricultural practices with arthropod diversity and maize production, we measured cumulative herbivore damage to maize in October 2011. We quantified herbivore damage on the fourth developed leaf from the top of 20 plants per maize field using a 10 × 10 cm transparent acetate with a 1 × 1 cm grid, assessing the percent leaf area damaged as the number of grid squares presenting some damage. We also estimated maize production by collecting 20 ears of corn per maize field and measuring the dry weight of 100 grains following Pérez- de-la-Cerda et al. (2007).

Statistical analyses

We pooled all of the arthropod data from different sampling periods per plot. Total arthropod diversity per maize field was calculated via the effective number of species using the coverage-based integrations of rarefaction and extrapolation of Hill numbers. This method has been recommended as the diversity measure of choice to compare species diversity across multiple assemblages that differ in sample size (Ellison 2010; Chao et al. 2014; Hsieh et al. 2016). We assessed the more widely used Hill numbers, species richness (which does not consider species abundance), Shannon diversity (which counts species in proportion to their abundances, thus assessing the effective number of common species) and Simpson diversity (which discounts all but the dominant species; Chao et al. 2014; Hsieh et al. 2016). To calculate these indices for each maize plot, species abundances were pooled for the three sampled periods per plot. We used the R package iNEXT (R Development Core Team 2008) to compute rarefaction and extrapolation sampling curves (Hsieh et al. 2016).

Using the calculated Hill numbers and different abundances, we analyzed the effect of management type and site on arthropod community attributes using nested ANOVAs. The response variables were species richness, Shannon diversity, Simpson diversity, total abundance, morphospecies abundance, Order abundance and guild abundance. The explanatory variables were management (low-input or high-input) nested within site (irrigated or rainfed). Herbivory and maize production per plot were also analyzed using nested ANOVAs. Abundances and maize production were log-transformed to comply with ANOVA assumptions.

Similarities in arthropod composition per maize field were analyzed using non-metric dimensional scale analyses (NMDS), with morphospecies abundance per plot. We calculated a Bray–Curtis dissimilarity matrix between plots. This ordination method is recommended since it can detect gradients without assuming linear relationships between variables (Quinn and Keough 2002) and produces an ordination based on a distance or dissimilarity matrix. We used the metaMDS and adonis functions from the vegan package for R (R Development Core Team 2008). To obtain a probabilistic statement of statistical differences in the community composition across the sampling sites, we used a permutational non-parametric multivariate analysis of variance (PERMANOVA, Anderson 2001; McArdle and Anderson 2004) using the Bray–Curtis distance metric. This test allows the evaluation of the null hypothesis that groups (in this case, management types) do not differ in their species compositions. First, F statistics are recalculated after a random shuffling of the labels on the rows that identify them as belonging to a particular group. This is repeated for all possible re-orderings of the rows relative to the labels, to create a distribution of pseudo F values, which is then used to compare the F value calculated with the original ordering of the data, yielding a P value to test the null hypothesis (Anderson 2001).

All statistical analyses were performed in the R environment (R Development Core Team 2008).

Results

Arthropod abundance and diversity

We collected a total of 14,315 individual arthropods belonging to 5 Classes (Arachnida, Malacostraca, Insecta, Diplopoda and Myriapoda), 12 Orders, 42 Families and 204 morphospecies (Additional file 1: Table S1). The Orders with the most morphospecies were Coleoptera (77), Diptera (31), Araneae (28), Hemiptera (36) and Hymenoptera (16). The arthropod abundance was highest for Coleoptera (3870 individuals), Diptera (3775 individuals), Hymenoptera (1841), and Collembola (1780 individuals) which together accounted for 79% of all of the individuals collected during the study. Four very abundant species—one Collembola, one Coleopteran, one Dipteran and one Hymenopteran—accounted for 39% of all individuals (1780, 1463, 1424 and 904 individuals respectively). We also trapped one vertebrate in one of the sampling periods, which was identified as the rat Oryzomys couesi (Alston, 1877) (Fig. 1).

The most abundant insects considered to be pests were Nicentrus testaceipes (Coleoptera), Macrodactylus sp. (Coleoptera), Diabrotica sp. (Coleoptera), Dalbulus maidis (Hemiptera), Rhopalosiphum maidis (Hemiptera), Gryllus rubens (Orthoptera) and Spodoptera frugiperda (Lepidoptera). The most abundant predators were Forficula sp. (Dermaptera), Orius sp. (Hemiptera), Calosoma sp. (Coleoptera) and four species of Formicidae.

Total abundance was highly variable among plots. On average there were 999.5 ± 373.3 arthropods/plot in high-input plots and 1386.3 ± 373 arthropods in low-input plots; there was no significant difference in abundance between management types (Low-input vs. High- input: F(1,8) = 2.76, p = 0.13) or sites within each management type (Rainfed vs. Irrigated: F(1,2) = 9.67, p = 0.09, Fig. 2), although low-input irrigated plots tended to have higher abundances. Similarly, the diversity estimators did not show differences between management types or sites either (Richness: management: F(1,8) = 2.25, p = 0.17, site F(1,2) = 0.2, p = 0.7, Shannon est: management F(1,8) = 0.3, p = 0.6 and site: F(1,2) = 0.09, p = 0.78, Simpson est: management F(1,8) = 0.001, p = 0.99 and management/site: F(1,2) = 0.0001, p = 0.99, respectively). The abundance of different arthropod Orders per plot was similar between sites and management types, except for Hymenoptera, which was more abundant in the low input plots (Table 2).

Fig. 1
figure 1

Map showing study sites in Cherán and Alvaro Obregón, Michoacán, Mexico

Full size image
Fig. 2
figure 2

Arthropod diversity indices in different management regime and site combinations, IHI (Irrigated High Input), ILI(Irrigated Low Input), RHI (Rainfed High Input), RLI (Rainfed Low Input). Shown in red: Species Richness, green: Shannon diversity and blue: Simpson diversity

Full size image
Table 2 Nested ANOVA of the effect of site and management on different arthropod orders, showing the degrees of freedom (d.f.), F and P values
Full size table

Arthropod community composition and trophic guilds

Arthropod community composition was very similar among plots in one locality, the rainfed site (R locality, Cherán, Fig. 3). At the site with irrigation (I) (Alvaro Obregón), however, low-input plots were strongly significantly different from high-input plots (PERMANOVA: r2 = 0.9532 p = 0.001). This analysis shows that there are differences between the centroids.

Fig. 3
figure 3

Non-multidimensional scaling analysis showing different plots from different sites and management practices. A clear cluster was formed with the plots from the rainfed-Cherán site (Low and High input, RHI and RLI) while irrigated-Alvaro Obregón showed two different communities depending on management regime (Low and High Input, LHI and ILI)

Full size image

The analysis of arthropod guilds showed that morphospecies that could be categorized as predators or herbivores from the literature were more abundant in low-input plots (F(1,8) = 7.74, p = 0.02 and F(1,8) = 6.39, p = 0.03, respectively, Fig. 4). For herbivores, this difference was more pronounced in irrigation plots, while detritivores were particularly variable between plots, with no apparent differences between management types (F(1,8) = 4.33, p = 0.07). There were no differences in guild abundances between sites (p > 0.05).

Fig. 4
figure 4

Total arthropod abundance per functional guild (mean per plot) in four different management treatments: IHI (Irrigated High Input), ILO (Irrigated Low Input), RHI (Rainfed High Input), RLO (Rainfed Low Input)

Full size image

Herbivory and maize production

Maize leaf damage was 10.24 ± 6.4% on average in all management types; there were no significant differences between management types or sites (management: HI = 9.85 ± 4.6% damage and LI = 8.27 ± 3.06% damage, F(1,8) = 2.67, p = 0.11; and sites: R = 9.8 ± 1.95% damage and I = 10.24 ± 6.43% damage, F(1,2) = 1.13, p = 0.29). Maize production estimated as the dry weight of 100 grains per plot did not differ between management types or sites, with an overall average of 26 ± 8.6 g /100grains (management: HI = 29.2 ± 5.5 g /100grains and LI = 22.7 ± 10.7 g /100grains, F(1,8) = 2.42, p = 0.15; and sites: R = 29.3 ± 7.2 g /100grains and I = 24.7 ± 8.8 g /100grains, F(1,2) = 1.13, p = 0.29).

Discussion

This investigation found that agricultural maize management practices have some significant effects on arthropod communities. In particular, it was evident that high-input management involving the frequent use of chemical inputs (insecticides, herbicides and fertilizer) had a negative effect upon Hymenoptera (mainly ants) and therefore a negative effect on predators.

Most maize agriculture in Mexico is now strongly dependent on external inputs (Hernández-Antonio and Hansen 2011). However, these changes in management practices do not necessarily translate into higher yields and economic profits. The application of external inputs without technical guidance can cause pest resistance, soil depletion, increase of herbivore populations due to lack of land rest, and other detrimental effects (Arnés et al. 2013; León-García et al. 2012). In our study case, herbivores and predators showed higher abundances in low-input plots. On average, predators were twice and three times as abundant in low-input plots compared with high-input ones in the rainfed and irrigated plots, respectively. This pattern suggests that chemical insecticides are not only affecting pest species, but also their predators. Other investigations have also found this pattern; Letourneau and Goldstein (2001) found greater predator abundances on organic farms compared to farms under conventional management, and in a recent global synthesis, Lichtenberg et al. (2017) concluded that organic farming provides better conditions for predators. In another recent study, Rosas-Ramos et al. (2020) found that organic management of cherry orchards benefited parasitoids and pollinators, though not predators. When comparing organic versus high-input farming, it has been difficult to determine the effect of stopping insecticide and/or herbicide use for arthropod communities. Most studies have found that increasing plant diversity within agricultural plots by intercropping or using cover crops is beneficial for predators (Philpott et al. 2006; Geldenhuys et al. 2021; Mhlanga et al. 2020; Del Pedro et al. 2020; Saenz Romo et al. 2019; Rivers et al. 2016; Otieno et al. 2019). Also, increasing plant diversity along field margins has proven to benefit predator abundances in some sites (Mkenda et al. 2019, Rusch et al. 2016). One aspect that warrants further investigation that is highlighted in the study by Tschumi et al. (2018) and Flores-Gutiérrez et al. (2020) and warrants further investigation is that the effect of seminatural habitats surrounding crops on the services or disservices of arthropods depends on the ecosystem type and region (e.g. dry versus wet forests).

In our study, ants were the predators that benefited the most from low-input agriculture; at the irrigated site they increased 100-fold, while in the rain-fed they increased twofold. This result is very significant, since ants have been shown to be important pest controllers for several (Thurman et al 2019; Philpott and Armbrecht 2006) including maize agroecosystems (Perfecto 1991; Perfecto and Castiñeiras 1998). Ants are known to be important egg-predators (Wills et al. 2019) and they can also prey on larvae, pupae and adult insects (Perfecto 1990, 1991). Other studies have also found that management affect ant community composition and predation effects, in particular maize fields sown surrounded by forested areas have shown to have higher ant predation rates than clear maize fields (Risch and Carroll 1982). In our study, in addition to avoiding the use of insecticides, low-input fields also have more diverse vegetation, which may have provided more prey for ants.

Insecticides were developed during the green revolution to control pest damage to crops (FAO 2003). Since then, insecticides are used heavily throughout the world with benefits to production but without consideration for insect diversity or other services provided by insects (Akanksha et al. 2020, Dirzo et al. 2014, Hallman et al. 2017; Klink et al. 2020). Insecticides currently have highly variable effects on crop production and pest reduction (Rosenheim 2021; Emery et al. 2021). In our study sites, high-input farmers cultivating in irrigated or rainfed regimes spend a considerable amount of money buying insecticides, which apparently does not translate into a considerable reduction of herbivores or herbivory levels on plants (Arnés et al. 2013), since herbivore damage to plants was similar between management regimes (high-input vs. low-input), so the expense of insecticide did not translate into protection of plants from herbivores. Furthermore, although we do not have an estimate of total maize production per plot, we found that maize grain weight was similar between treatments, suggesting that the differences on external inputs did not result in a strong increase in maize production. Similar results were found in papaya cultivation in western Mexico (Flores-Gutiérrez et al. 2020) and in corn cultivated in northern California, USA (Clark et al. 1998). Despite evidence to the contrary, conventional farmers often feel that it is too risky to stop using insecticide, so the transition towards more sustainable agriculture has to be gradual, and results from this type of investigation should be shared with farmers.

Another aspect to take into consideration is that herbivore impacts are normally considered to be directly and linearly related with plant productivity, but this not always the case (Perez-Alvarez et al. 2018; Poveda et al. 2003, 2010). We need more studies looking into real herbivore population thresholds that affect crop production to allow sustainable pest management based on local data. In our study system, maize in low-input management plots appears to cope with the levels of herbivory without a decrease in productivitt, suggesting some kind of compensation.

Conclusions

Agricultural management regime had a significant effect upon arthropod communities in both rainfed and irrigated maize farms. Given that arthropod species are experiencing significant declines worldwide, low-input management could contribute to conservation. The plots under conventional maize management at both irrigated and rain-fed sites used several costly external inputs, but did not have significantly reduced herbivores or maize damage, suggesting that alternative solutions would be able to control pest damage.