Introduction

The global cacao bean production between 2017 and 2022 averaged approximately 5.4 million tons by year, and exhibits a consistent upward trend (https://www.fao.org/faostat/en). South American nations contribute roughly 16% of this output, with Brazil, Ecuador, and Peru being the key producers in the region. Moreover, cacao is one of the main non-woody products cultivated in the tropical regions of South America (Imbach et al. 2015). Although Bolivia accounts for only a 6% share of South America´s cacao production, this crop is the chief source of income for smallholder farmers in the “Alto Beni” region (Ortiz and Somarriba 2005, te Velde et al. 2006). These small farmers are highly vulnerable to climate change, and the use of agroforestry for cultivation of cacao could enhance their resilience, allowing better conditions for the sustainable development of the region (Jacobi et al. 2015).

In the wild, cacao trees (Theobroma cacao) develop as understory plants in tropical rainforests, being adapted to growth with low levels of light. Therefore, cacao is an ideal crop for cultivation in agroforestry systems. On the other hand, the phenotypic plasticity of this species allows its cultivation under full-sun conditions (Jaimes-Suárez et al. 2022). Cultivation under full-sun conditions can give higher yields in the first years and can simplify the use of agrochemicals (Andres et al. 2016). However, the lifespan of full-sun monocultures tends to be shorter compared to shaded environments, probably due to photooxidative stress (Jaimes-Suárez et al. 2022). Moreover, working conditions for farmers in full-sun monocultures are very harsh due to high environmental temperatures (Andres et al. 2016). In agroforestry systems, either cacao is sown beneath existing forest canopy, or specific shade trees are deliberately placed between cacao trees. The cultivation of cacao in agroforestry systems can compensate lower yields by allowing a longer life span of the trees, higher carbon sequestration in the biomass of shade trees, a decrease in temperatures and an increase of humidity, and maintenance of higher biodiversity (Niether et al. 2020). Work conditions for farmers are better and additional income from products such as timber and fruit can be obtained (Andres et al. 2016). A special form of agroforestry is called dynamic or successional agroforestry. In this cultivation system, diverse crops spanning various strata are densely planted, simulating natural regeneration. The natural regrowth of forest plants is also included. In addition to the benefits offered by traditional agroforestry, successional agroforestry enables farmers to harvest various products starting from the first year (Andres et al. 2016).

Further investigation is required to fully understand the impact of shade trees and accompanying plants on soil characteristics within agroforestry systems. Some authors report that agroforestry systems generally have higher soil organic matter contents, better nutritional status, more dynamic soil biota and improved ecosystem services in comparison to other cultivation systems (Dollinger and Jose 2018, Rodríguez-Suárez et al. 2021). On the other hand, Blaser et al. (2017) found that the effect of shade trees on soil fertility was not significant at the plot scale. Moreover, since there are many variations in the design and management of cacao-agroforestry systems (i.e., shade tree species selection, diversity and density, pruning intensity, conventional or organic farming) there is a need to generate knowledge about the effect of these different managements on soil quality, which is one of the main drivers of the health of the whole ecosystem (Dollinger and Jose 2018).

The Research Institute of Organic Agriculture (FiBL) and local partners established a long-term field trial in the Alto Beni region (http://www.systems-comparison.fibl.org) with the aim of investigating the environmental and socioeconomic aspects of cacao-based agroecosystems. In this trial, which started in 2008, five different cacao production systems differing in plant diversity and management (conventional monoculture, organic monoculture, conventional agroforestry, organic agroforestry and successional agroforestry) are compared (Schneider et al. 2017; Marconi and Armengot 2020). Figure 1 summarizes the characteristics of the different systems and of fallow plots.

Fig. 1
figure 1

Scheme of the five cacao production systems and the fallow plot (secondary forest) in the long term trial in Sara Ana (Alto Beni, Bolivia). In all cultivation systems, cacao plants were placed in a 4 × 4 m grid

Full size image

In order to determine the effect of different production systems on soil quality, indicators that respond rapidly to management are the most appropriate. Parameters such as soil microbial biomass (Joergensen and Emmerling 2006) and soil enzymatic activities (Nannipieri et al. 2018) are often used with this aim, since they show faster response compared to more classical physico-chemical analysis. The effect of different land uses and management on these indicators has been reported by many authors, for example Acosta-Martínez et al. (2007), Araújo et al. (2013), Moghimian et al. (2017). Generally, soils under intensive management, monocultures or degraded soils showed lower values of microbial biomass and enzymatic activities than soils under forests or under more complex agroecosystems.

In a previous study in the same trial, we reported data on carbon and nitrogen microbial biomass and cellulase activity from soils sampled in 2011 and 2012, i.e. about three years after the establishment of the trial (Alfaro-Flores et al. 2015). The results showed high spatial and seasonal variability, but no clear effect of the production system on the assessed variables was found. For example, cellulase activity was not affected by the production system and, only in the dry season, soils under monocultures showed lower microbial biomass than non-cultivated soils, i.e., fallow plots with the same age than the cultivated systems. To overcome the limitations of the short time elapsed between the start of the trial and the sampling time, and the limited number of measured variables, a new sampling was performed five years later, in 2016. For these new samples, we analysed more microbiological and biochemical variables: carbon and phosphorus microbial biomass, microbial activity by evolution of CO2, acid phosphomonosterase (phosphatase), urease, protease and β-glucosidase. Chemical variables pH, organic carbon and available phosphorous were also determined, as these parameters could contribute to explain the values obtained for microbiological and enzymatic parameters. Finally, plant inputs from pruning, herbaceous strata and natural litterfall were quantified, since they could also influence the assessed variables. We worked with the following hypothesis: (1) after eight years of establishing different production systems, if higher plant biodiversity positively affects soil quality, then the soil under agroforestry systems would have higher microbiological and enzymatic parameter values compared to soil under cacao monocultures (2) if organic management has a positive effect on soil quality, then systems with organic management would have higher microbiological and enzymatic parameter values compared to their conventionally managed counterparts (3) if soil quality parameters are affected by the quantity of plant inputs to the soil, there would be a relationship between the quantity of plant inputs and the soil's microbiological and enzymatic parameters.

Methods

Study area and experimental design

The long-term trial is located at “Sara Ana’’, in the Alto Beni region of the department of La Paz, Bolivia. It is an alluvial terrace at the north-eastern side of the Andes (15° 27′ 36.60″ S, 67° 28′ 20.65″ W). The soils are Lixisols and Luvisols, according to FAO classification (Schneider et al. 2017). The altitude is between 380 and 400 m. a.s.l. The mean annual temperature of the area is 26 °C, and the mean annual rainfall is 1519 ± 220 mm, mostly accumulated between October and March.

The trial started in 2008 by clearing a twenty years’ secondary forest characterized by Guadua spp., among other species. Six systems are compared: (1) conventional monoculture (CM), (2) organic monoculture (OM), (3) conventional agroforestry (CA), (4) organic agroforestry (OA), (5) successional agroforestry (SA) and (6) fallow (F) in a randomized block design with 4 repetitions. For the definition of blocks, soil physico-chemical analyses were conducted in 2008, including the assessment of texture, pH and organic matter contents. Subsequently cluster analysis was used to assign five plots to four different blocks (Schneider et al. 2017). The plot size is 48 m × 48 m and in all systems, cacao trees were planted in a 4 m × 4 m grid (625 trees ha−1, see supplementary material S1). In the agroforestry systems, shade trees, mainly the leguminous Erythrina spp. and Inga spp., but also some additional fruit and timber species, were planted between the cacao trees in a 8 m × 8 m grid (total density of about 304 trees ha−1, see Figure S1 in Supplementary Information). Successional agroforestry (SA) includes the same species arrangements as the two other agroforestry systems with additional cultivated crops (e.g., ginger, curcuma), trees (e.g., peach palm) and spontaneous species from the natural regeneration. The total tree density in SA is about 1180 trees ha−1. The fallow plots (F) were left unmanaged after the same land preparation as the other treatments. Eight years after the land preparation, the fallow plots have developed into a young secondary forest, characterized by species such as Cecropia sp. (Urticaceae), Senegalia loretensis, Triplaris americana and Guadua weberbauri and by a very dense understory.

In the systems with conventional management, mineral fertilizers (Ntotal-P2O5-K2O-MgO) were applied twice (March and December) at a dose of 18–12–24–4 kg ha−1 in the case of the CM, and half of the dose in the CA. Herbicides were applied 4–5 times per year, mainly glyphosate. For systems with organic management, compost was applied around the cacao trees in December, 8 Tn ha−1 for monoculture and 4 Tn ha−1 for agroforestry. The composition of compost is detailed in Alfaro-Flores et al. (2015). The SA system did not receive any external input. Other details of the study area and of the characteristics of the trial are described in Schneider et al. (2017).

Soil sampling

In each plot, five cacao trees (cultivar ICS1) were randomly selected. Soil was sampled at 0–20 cm depth with a steel borer at distances of 50 cm, 1 m and 1.5 m from the trunk of each chosen tree. We chose a 0–20 cm depth because we considered that this depth influences cacao nutrition since the highest cacao fine root densities are by far in the first 20 cm (Rajab et al. 2018). In the fallow plots, 15 soil subsamples were taken from random points. The 15 subsamples from each experimental plot were pooled in one composite sample and sieved (2 mm mesh). The composite samples were kept at 4 °C until analysis. Sampling was carried out in March 2016 (rainy season) and August 2016 (dry season).

Laboratory analysis

Soil pH was determined in aqueous solution with a relationship dry soil/water 1:2.5. Microbial biomass carbon (MBC) and microbial biomass phosphorous (MBP) were determined in extracts obtained by the fumigation-extraction method proposed by Joergensen (1995a, b). MBC in extracts was quantified by a colorimetric adaptation of the Walkey Black Method for organic carbon and a calibration curve with glucose solutions of known concentration (Alfaro-Flores et al. 2015). Microbial biomass phosphorus (MBP) was quantified following the method described by Joergensen (1995a, b). The microbiological activity of the soil samples was determined by quantification of the evolution of CO2 from soil samples (Alef and Nannipieri 1995a). The enzymatic activities acid phosphomonoestherase, β-glucosidase, protease and urease were evaluated by following the methods described in Alef and Nannipieri (1995b). Organic carbon in soil samples was quantified as described in Allison (1965), available phosphorus was determined following Cochrane and Barber (1993). Samples were tested in triplicate for microbiological and enzymatic properties. For the case of chemical analyses, since these are well stablished routine analysis, occasional replicates were performed for control purposes.

Estimation of plant matter inputs to soil

We considered that plant inputs to soil near the sampling points came from three sources: tree pruning, natural litterfall and inputs from herbaceous strata. The quantification of the plant material inputs to the soil due to the pruning of trees was carried out from June to September 2016, since the major pruning activities were carried out in these months. As part of the maintenance activities at the field trial, the residues produced by pruning were placed around the trunk of the pruned tree. Since soil sampling was carried out near cacao tree trunks, we assumed that the sampled soil in all systems was more affected by cacao pruning residues than by residues from other species. Therefore, we quantified cacao pruning residues with detail. For each plot, two cacao trees (cultivar ICS 1) were selected. During the pruning activities, all the material that fell down was collected in bags placed at the base of the trees. The total fresh weight of the material was registered in the field. A sample of 500 g of pruning residues was placed in paper bags and taken to laboratory for the determination of the dry-to-fresh weight ratio. With this ratio, the fresh weights of material were converted to dry weight. For each plot, the complete input of pruning material was estimated by multiplying the average of dry weights that were registered in the plot by the total number of cacao trees in the plot and then converted to Kg of dry weight by hectare. In the case of agroforestry systems, we also estimated the inputs due to the pruning of Inga sp. and of Erythrina spp, since these are the more numerous accompanying trees (pruning residues of eight Inga sp trees and six Erythrina spp trees from different plots were quantified by using the same method as for cacao trees).

The estimation of plant inputs due to natural litterfall was carried out from October to December 2016, once the pruning activities had finished. Two collecting traps of 1 m diameter were placed under the two selected cacao trees. One of the traps was placed from the base of the tree to 1 m distance. The second trap was placed at a distance between 1 and 2 m from the base of the tree. In the fallow plots, the traps were randomly located. In the agroforestry systems, the traps collected a mixture of leaves from cacao trees and from other species but no distinction was made between them. The fresh weight of the leaves collected in the traps was registered in the field and the dry weigh was determined in laboratory. We added up the weight of the litterfall obtained in each of the four traps of each plot, divided the sum by the surface of the four traps and expressed the result in Kg dry weight per hectare.

The herbaceous strata was sampled by placing two quadrants of 50 × 50 cm around 1 m distance to the two chosen cacao trees in each plot (4 quadrants in total, 1 m2 area). In the fallow plots, the four quadrants were randomly placed. All the herbs of each quadrant were cut and the dry weight was determined. The results were expressed in Kg dry weight of biomass per hectare.

Statistical analysis

The soil chemical, microbiological, and biochemical analysis results were analysed using generalized linear models. The initial complete model incorporated block, season, biodiversity level, management (organic vs. conventional), and pairwise interactions between season and biodiversity level, season and management, and biodiversity level and management as predictor variables. The biodiversity level was assigned as follows: level 1: monoculture (for CM and OM), level 2 simple agroforestry (CA and OA), level 3 successional agroforestry (SA), level 4 fallow (F). The best model was determined by using the dredge() function of the MuMin version 1.47.5 package and the effect of the chosen predictor variables was considered significant at the 0.05 level. In case of significant differences between biodiversity levels, a post hoc analysis (Tukey´s HSD) was conducted on marginal means. Additionally, within each season, Pearson correlation analysis (using the Hmisc package) was conducted for all variables, and principal component analysis was performed. For principal component analysis, we used all results from chemical and microbiological analysis but only phosphatase for enzymatic analysis due to high correlation found between values obtained for enzymatic activities. We also evaluated the effect of the block, biodiversity level and management on the plant biomass inputs to soil through generalized linear models. All analyses were performed with the program Rstudio version 4.3.1. Graphics were generated with packages ggplot2 version 3.4.3. and ggbiplot2 version 0. 55.

Results

Chemical parameters

Soil pH of the different plots varied between 5.7 and 7.4 in the dry season and between 6.1 and 8.3 in the rainy season (Fig. 2A, B). The results of generalized linear models showed that the only significant predictor variable was season, with higher pH values measured in rainy season (p < 7.38e−06, Table S2 in Supplementary Information).

Fig. 2
figure 2

Soil pH in dry (A) and rainy (B) season, soil organic carbon in dry (C) and rainy (D) season, and available phosphorus in dry (E) and rainy (F)) season for each system. In the case of organic carbon, significant differences between systems are indicated by small letters. The system abbreviations follow as: CM conventional monoculture; OM organic monoculture; CA conventional agroforestry; OA organic agroforestry; SA successional agroforestry; F fallow. Circles represent mean values for each system, with lines indicating standard deviation. Diamonds denote values obtained for different plots within each system

Full size image

Soil organic carbon values registered in the different plots varied from 1 to 2.8% (Fig. 2C, D), with no significant effect of the sampling season. The best model for organic carbon comprised only the biodiversity level as a predictor variable. Organic carbon values were significantly higher in fallow plots compared to monocultures (p = 0.0012) and to simple agroforestry (p = 0.0010, Table S2).

Available phosphorous contents varied from 2 to 6 mg/Kg in dry season and from 1 to 11 mg/Kg in rainy season (Fig. 2E, F). Very large differences between replicate plots were observed for this nutrient in the rainy season, as can be observed on Fig. 2F. None of the tested predictor variables had any effect on phosphorous contents (Table S2).

Microbial parameters

Values observed for MBC in all plots were between 85 and 396 mgC/Kg soil in dry season (Fig. 3A) and between 57 and 762 mgC/Kg soil in rainy season (Fig. 3B). For MBP, the range of values went from 10 to 37 mgP/Kg soil in dry season (Fig. 3C) and from 18 to 51 mgP/Kgsoil in rainy season (Fig. 3D). The values for biological activities were situated between 0.6 and 2.6 mg CO2/Kgsoil in dry season (Fig. 3E) and between 1.3 and 8.1 mg CO2/Kgsoil in rainy season (Fig. 3F). Season had a strong and significant effect on MBC (p = 0.0187), MBP (p = 5.84 × 10–5) and microbial activity (p = 0.0167), with higher values in the rainy season. No significant effect of the block, the biodiversity level or the management was found for any of the microbial parameters (refer to Table S2 in supplementary material).

Fig. 3
figure 3

MBC in dry (A) and rainy (B) season, MBP in dry (C) and rainy (D) season, and biological activity in dry (E) and rainy (F) season for each system. The system abbreviations follow as: CM conventional monoculture; OM organic monoculture; CA conventional agroforestry; OA organic agroforestry; SA successional agroforestry; F fallow. Circles represent mean values for each system, with lines indicating standard deviation. Diamonds denote values obtained for different plots within each system

Full size image

The percentage of soil carbon contained in microbial biomass and the values of the respiratory quotient are presented on Fig. 4. In dry season, the values obtained for the percentage of carbon in soil microbial biomass across the different plots was between 0.5 and 3%, while in the rainy season, observed percentages were between 0.6 to 5%. The respiratory quotient ranged from 0.04 to 0.19mgCO2/gC.h in dry season, while reported values for rainy season were between 0.09 to 0.54 mgCO2/gC.h. The values were significantly higher in rainy season than in dry season for the respiratory quotient (p = 2.6 × 10–6) but season had no significant effect on percentage on soil carbon contained in microbial biomass. No significant effects attributable to block, to biodiversity level or to management were observed for either parameter.

Fig. 4
figure 4

Percentage of microbial carbon for dry (A) and rainy (B) season, and respiratory quotient in dry (C) and rainy (D) season. The system abbreviations follow as: CM, conventional monoculture; OM, organic monoculture; CA conventional agroforestry; OA organic agroforestry; SA successional agroforestry; F fallow. Circles represent mean values for each system, with lines indicating standard deviation. Diamonds denote values obtained for different plots within each system

Full size image

Enzymatic activities

For soil phosphatase activities, the measured values were comprised between 134 and 495 µgPNP/gsoil.h in dry season (Fig. 5A) and 134 and 906 µgPNP/gsoil.h in rainy season (Fig. 5B). Glycosidase activity values were between 5 and 28 µgPNP/gsoil.h (Fig. 5C) in dry season and 35 to 250 µgPNP/gsoil.h in rainy season (Fig. 5D). For urease activities, values went from 6 to 22 µg N-NH4 /g soilh in dry season (Fig. 5E) and from 35 to 250 µg N-NH4 /g soil h in rainy season (Fig. 5F). The range for protease activities went from 0.1 to 2.0 µg tyrosine /g soilh in dry season (Fig. 5H) and from 0.2 to 5.0 µg tyrosine/g soilh in rainy season (Fig. 5I). The season had a significant effect on phosphatase (p = 0.0187), β-glucosidase (p = 1.52 × 10–7), urease (p = 0.0141) and protease activity (p = 0.0055), with higher values in the rainy season (Table S2, Supplementary Information). No effect of the block or the management (conventional vs organic) was observed on any enzymatic activity. Fallow plots showed significantly higher phosphatase activity than monocultures (p = 0.0001) and simple agroforestry systems (p = 0.011). In the case of β-glucosidase and urease activities, the interaction between season and biodiversity level was found to be significant. During the dry season, biodiversity level did not affect these enzymatic activities. However, in the rainy season, fallow plots exhibited significantly higher β-glucosidase activity compared to monocultures (p = 0.0143) and simple agroforestry (p = 0.0059), along with significantly higher urease activity compared to monocultures (p = 0.0209). In the case of protease activity, no significant effect of diversity could be observed.

Fig. 5
figure 5

Phosphatase activity in dry (A) and rainy (B) season, β-glucosidase activity in dry (C) and rainy (D) season, urease activity in dry (E) and rainy (F) season and protease activity in dry (G) and rainy (H) season for each system. Significant differences between systems are indicated by small letters. The system abbreviations follow as: CM conventional monoculture; OM organic monoculture; CA conventional agroforestry; OA organic agroforestry; SA succeddional agroforestry; F fallow. Circles represent mean values for each system, with lines indicating standard deviation. Diamonds denote values obtained for different plots within each system

Full size image

All enzymatic activities showed significant correlation between them except phosphatase and protease activity in the dry season (see Figure S3A in Supplementary material). In the rainy season, all enzymatic activities correlated between them (Figure S3B). In the dry season, phosphatase, β-glucosidase and urease correlated with microbial biomass carbon. Significant correlation of microbial biomass phosphorus with protease activity and with pH in dry season was also observed. In the case of the rainy season, phosphatase, β-glucosidase and urease were correlated with total organic carbon but no significant correlation was observed with microbial biomass or microbial activity.

In the dry season, principal component analysis revealed no clear distinctions between samples belonging to cultivation systems with different biodiversity levels (Figure S4A, Supplementary Information). During the rainy season, while some points were distinctly separated between fallow and monoculture, the majority of points corresponding to simple agroforestry and successional agroforestry were intermixed with points from monoculture (Figure S4 B, Supplementary Information).

Plant biomass inputs

Input of plant materials to soil due to cacao pruning varied from 1039 to 8000 kg dry weight/ ha (Fig. 6A). High variation was observed between plots and even between trees from the same plot (data not shown). Since no pruning activities were carried out in the fallow plots, we compared the mass of cacao plant pruning residues between monocultures, agroforestry systems and successional agroforestry. Significantly lower inputs from cacao pruning were observed for simple agroforestry (p = 0.024) and successional agroforestry (p = 0.0475) in comparison to monocultures. No significant effect of block or management was observed. In agroforestry systems, however, an additional input of around 1839 kg dry weight /ha due to pruning of Inga and 504 kg dry weight /ha due to Erithrina sp was estimated.

Fig. 6
figure 6

Plant biomass inputs to soil (Kg dry weight/ha) due to cacao tree pruning (A), natural litterfall (B) and herbs (C). The system abbreviations follow as: CM conventional monoculture; OM organic monoculture; CA conventional agroforestry; OA organic agroforestry; SA successional agroforestry; F fallow. Significant differences between systems are indicated by small letters. Circles represent mean values for each system, with lines indicating standard deviation. Diamonds denote values obtained for different plots within each system

Full size image

Dry weight of biomass due to herbs varied from 222 to 2279 kg dry weight/ha. Both the biodiversity level and the management had significant effect on biomass due to herbaceous strata. Agroforestry systems (p = 0.003651), successional agroforestry (p = 0.001084) and fallow plots (p = 0.000267) had significantly less plant biomass in herbaceous strata than monocultures. Organic management also gave rise higher plant biomass from herbs than conventional management (p = 0.013).

Natural litterfall, measured during three months after the pruning activities had finished, accounted for 65 to 298 kg dry weight/ha (Fig. 4B). No significant effect of block, diversity or management was observed on the litterfall dry weight, although in the monocultures it only contained cacao leaves while in the agroforestry it contained leaves from cacao and some from the shade trees, and in fallow plots, leaves from different species.

Discusion

Chemical parameters

The pH values obtained in the dry season (September 2016) were similar to those reported by Alfaro-Flores et al. (2015) for the samples of 2011–12 and Lori et al (2022), for samples taken in 2019 in the same trial, but measured at a depth of 0–10 cm. However, in the rainy season, our average value is slightly higher than the ones reported by the mentioned authors. Increases in pH in rainy season, and mostly in late rainy season, have been previously observed (Yamashita et al. 2011), and have been attributed to rapid microbial decomposition of litter, which frees Na+, K+ and Mg2+ that adsorb on the soil exchange complex. Values reported in Alfaro-Flores et al. (2015), correspond to samples taken in February, while, in the present work we report values from samples collected in March (later in the rainy season), which can explain higher values.

Mean organic carbon values in our study were consistent with those reported by Lori et al. (2022) for the same field trial. However, while Lori et al. (2022) found significantly higher organic carbon levels in plots under organic management, compared to conventional management, our results did not indicate management as a significant predictor variable. This discrepancy could be explained by (a) the different sampling depth used in the two studies, and (b) the sampling year. Lori et al. (2022) sampled soil at a depth of 0–10 cm, since topsoil layers are often more responsive to management interventions (McCallister and Chien 2000). In addition, samples analysed by Lori et al. (2022) were collected three years later than our samples. In these additional years, the management practices could have led to more pronounced differences, due for example to increased accumulation of compost.

The strong variations observed for pH, organic carbon and, even more, available phosphorous between replicate plots (see Fig. 2) are an indication of high spatial variability. From the outset of the long-term trial in 2008, the field exhibited inherent variations in soil parameters, such as soil texture. The experimental block delimitation was designed to eliminate, at least partiality, the variability coming from the spatial heterogeneity. However, the trial is situated is an alluvial plain with two different soil types identified and occupies a large area (around 7 ha) where each plot has a surface of almost half a hectare. This land extension complicated the minimisation of the soil heterogeneity. Niether et al. (2018) documented significant differences in soil texture and water availability even among different sampling points within individual sampling plots in the same trial. Although the variable “block” showed no significant effect on any of the measured variables in the generalized linear models, the presence of variability within plots implies that future studies should consider higher sampling density and possibly the use of geostatistical methods.

Spatial heterogeneity can intensify over time. Although mean values for soil organic carbon tend to respond slowly to changes in land use, variations in quantity and quality of plant inputs, light, temperature, soil characteristics and soil moisture regimes can give rise to different rates in carbon sequestration (Lorenz and Lal 2014) and this could cause spatial variation over short ranges (less than 50 m). Rates of organic matter decomposition can also affect pH and available phosphorous. In the case of available phosphorus, there are additional factors that can cause spatial variation such as the rate of phosphorus immobilization in the soil and the rate of plant uptake. This accumulation of factors that influence phosphorous availability could explain the fact that no significant effect of management was found for this parameter, although phosphorus fertilization was markedly different in each production system. Spatial variability was generally higher in rainy season, i.e., when spatial variation in soil moisture can be more pronounced. In addition, the rainy season also corresponds to faster growing season of plants, which have different rates of nutrient uptake. The spatial variation of pH, organic carbon and available phosphorous at short distances has previously been observed in forest soils, agroforestry systems or even in tree plantations (Gonzalez and Zak 1994; Yavitt et al. 2009; Dos Santos et al. 2017, 2023).

Microbial and enzymatic parameters

The mean values obtained for MBC for dry season (June 2016) in the present work were around 85% lower than for samples taken in September 2011 (Alfaro-Flores et al. 2015). However, values obtained in the rainy season were similar in both studies. Besides year to year variability, the difference between both studies in the dry season can be explained by the different specific time of sampling, June and September, which have different climatic conditions. For both data sets, values were significantly higher in rainy season than in dry season. Seasonal variation for MBC has been observed by several authors (see for example Lungmuana et al. (2017); de Oliveira Silva et al. (2019)), but the reasons for this variation are not totally elucidated and probably should be analysed in a case to case basis. Microbial biomass depends on soil pH, humidity and temperature, as well as substrate availability (Zhao et al. 2022). The higher MBC in March could be related to both higher humidity and temperature than in September, and secondly, to the recent application of the fertilizers (in December and March), which could increase the available substrate. However, different fertilization schemes (chemical vs organic, and different doses applied) did not have significant influence on MBC. This suggests that substrate availability was not the main driver for the increase in biomass. Since in our study area, higher temperatures coincide with higher humidity it is difficult to elucidate the effect of both parameters. However, no correlation was observed between MBC and humidity of samples (data not shown) suggesting that, in our case, temperature could be the main driver of microbial biomass abundance. This could also be the case in the study of Srivastava et al. (2020) where higher MBC was observed in summer although soil moisture was lower. Our present results for MBC and previous MBC values obtained in 2011 were used for a preliminary diachronic study following Abreu et al. (2009). However, while some values exhibited increases and others decreases, we did not observe any significant effects of diversity levels or management practices.

MBP was also higher in the rainy, warmer season than in dry season, this coincides with the observations of Tripathi and Singh (2013). As in the work of De Sousa et al. (2021) for a Brazilian forest, no correlation between MBP and MBC were found within each season. Moreover, no effect of mineral fertilizer or compost could be observed, which is also in agreement with the results of De Sousa et al. (2021), who attributed this to rapid absorption on soil material or root uptake of the available phosphorus.

Similar to microbial biomass measurements, values obtained for biological activity (measured through CO2 efflux of soil samples) and three hydrolytic enzymes (urease, protease and β-glucosidase) were higher in the rainy, warmer season. Singh et al. (2020) also found an increase in microbial and hydrolytic enzyme activities in summer months. Lang et al. (2017) and Yu et al. (2021), through in situ measurements of CO2 efflux also found higher values in warmer months and report that temperature has more influence than humidity on seasonal variation as also seems to be our case, since no strong correlation was observed between humidity of samples and microbial activity (data not shown).

Our previous work reported that the system had a significant effect on MBC in the dry season, but it was not the case in the rainy season due to spatial variability (Alfaro-Flores et al. 2015). In the present work we also found high spatial variability for MBC, MBP and microbial activity in both sampling seasons as can be observed from standard deviation between replicate plots (Fig. 3). This spatial variability, which might have been enhanced with time, could override any potential effects of the system and management. High spatial variability in diverse soil parameters that can affect microorganisms under agroforestry has been observed by many authors (see for example Marín-Castro et al. (2016) for spatial heterogeneity of soil hydraulic conductivity, Dolácio et al. (2021) for diverse soil chemical parameters, Yu et al. (2021) for spatial variation in CO2 efflux, where variability increases in warmer months). Liu et al. (2010) also reported high spatial variation in MBC. The spatial variability affecting MBC and biological activity also extends to the respiratory coefficient and the proportion of microbial carbon. Consequently, we didn't observe any significant effects of management or diversity on these parameters.

Enzymatic activities in soil samples can be due to enzymes produced by cells at the moment of sampling, but also to enzymes that were produced previously and have been stabilized in the soil matrix (Nannipieri et al. 2018). By then, in many cases, no strong correlation of enzymatic activities with microbial activity or biomass is expected. In our study, protease, β-glucosidase and urease activities showed much higher values in rainy season as compared to dry season. Since a similar trend was observed for MBC and microbial activity, this suggests that a great portion of the enzymatic activity was linked to viable cells at the moment of sampling. The correlation that we observed between enzymatic activities and microbial biomass in dry season and between enzymatic activities and organic matter in rainy season (Figure S3, Supplementary Information) could indicate that when climatic conditions are optimal for microbial biomass, abundance of substrate would be the limiting factor for enzyme production and possible accumulation in soil. On the contrary, when climatic conditions are not optimal, microbial biomass could be the limiting factor. However, this requires further investigation.

Enzymatic activities were highly correlated between them (Figure S3) and also showed strong spatial variability. Fosfatase and β-glucosidase activities showed significantly higher values in fallow plots than under monocultures and could be used as simple and sensitive indicators of soil quality. The higher soil enzymatic activities observed in fallow plots may be attributed to the presence of varied and decomposable plant material, fewer intervention events impacting soils, and the presence of more stable microclimatic conditions, as described by Niether et al. (2018). Higher enzymatic activities in fallow plots has been reported by several authors such as Lungmuana et al. (2017) who observed that older fallows show higher soil enzymatic activities. de Medeiros et al. (2017) also report an increase in enzymatic activities over long regeneration times (more than 30 years) and in the 0–5 cm layers.

The lack of effect of management (organic vs conventional) on microbial and enzymatic properties was also observed for MBC and MBN by Alfaro-Flores et al. (2015). Organic management in the present trial includes the addition of relatively low quantities of compost in simple agroforestry systems and no external inputs in successional agroforestry (Schneider et al. 2017). Additionally, the time from the beginning of the trial to the moment of sampling (eight years) was relatively short to build up organic content and the associated microorganisms in soil.

In response to our first hypothesis, our results showed a positive effect of plant biodiversity on the enzymatic activities of phosphatase, β-glucosidase, and urease, mostly during the rainy season, with fallow plots exhibiting significantly higher values than monocultures. However, there was no effect on microbiological parameters. This leads us to interpret that plant diversity can have a positive effect on soil quality but that further research is needed. It is noteworthy that organic carbon, measured as a complementary parameter, also showed higher values under agroforestry systems compared to monocultures, indicating encouraging results for soil quality under agroforestry. For future research it is necessary to take into account the spatial variation that was observed in our results. We strongly recommend that further studies collect a much larger number of soil subsamples and thoroughly homogenize the resulting mixture to obtain more representative samples of the plots. Additionally, since biological processes mostly occur in superficial soil layers (Piotrowska-Długosz et al. 2022), it is recommended that sampling be conducted at a depth of 0–10 cm to better observe changes in microbiological and enzymatic parameters. On the other hand, we can reject the second hypothesis because, under our conditions, no effect of management (organic versus conventional) was observed for any parameter in any of the systems or seasons.

Plant inputs

Except for fallow plots, the highest quantities of plant material inputs to soil were due to pruning (more than 60%), as was also reported by Schneidewind et al. (2019).The most abundant pruning residues were leaves, which degrade faster than stems or branches. In monocultures, the amount of cacao pruning residues was higher than in simple or succesional agroforestry, but, in the agroforestry systems, additional pruning residues were introduced to the system from the shade trees. However, it is noteworthy that residues from each tree were primarily deposited at the base of the respective tree. Specifically, around the trunks of the cacao trees, where our samples were extracted, the predominant residues were those of cacao. The estimated contribution of plant material due to litterfall or herbaceous strata did not compensate the lack of pruning residues in the fallow plots. Hence, the fallow plots received less plant inputs than the other plots. However, they showed higher enzymatic activities, suggesting the lack of a direct relationship between quantities of plant inputs and enzymatic activities of soil. These results are similar to those reported by Feng et al. (2019), who found no significant relationship between soil enzymatic activities and litterfall. Partey et al. (2019) report that quality (for example lignin and nutrient contents) of plant inputs to soil could be more important than the quantity for determining the effects on biological parameters. Further investigation into the diversity of plant inputs and their decomposability across the different cultivation systems is necessary to establish a relationship between the characteristics of these inputs and the enhanced presence of soil enzymes. Regarding our third hypothesis, our results suggest its rejection, as we could not detect any effect of the quantity of plant residues on the measured parameters.

Conclusions

Eight years after the establishment of the trial, no significant effect of the plant diversity in the different systems (monocultures, simple agroforestry, successional agroforestry, fallows) on microbiological properties was observed. However, total organic carbon, phosphatase, β-glucosidase and urease activities of soil samples from 0 to 20 cm depth showed lower values in the monocultures compared to the fallows. Our findings indicate that the enzymatic parameters that we examined hold promise for evaluating the effect of different cacao production systems on soil. Increased values for enzymatic activities in the warmer, rainy season suggest that this could be a better sampling time, in which these parameters can show more sensitive response to cultivation systems.

The higher values observed for total organic carbon and enzymatic activities in fallow plots was not directly related to the quantity of plant inputs to soils, however they might be related to higher diversity and/or decomposability of these residues. Thus, the effect of the quality (cellulose, lignin, nutrient content) of these inputs on microbiological and biochemical parameters should be further explored. Additionally, higher values in fallow plots could reflect the fact that these plots are less affected by perturbation due to cultivation activities and that they present more microclimatic stability. No effect of management (organic, conventional) was observed on any of the measured parameters, possibly because addition of compost in the organic monocultures and simple agroforestry systems was relatively low and no external inputs were added to successional agroforestry.

Moderate to strong spatial variability was observed for all the studied parameters. In case of microbial biomass, microbial activity and enzymatic activities, there was also a strong seasonal variation, with higher values in the rainy season. This soil spatial and seasonal variability can make it challenging to discern the effect of the cultivation system on certain soil properties. It is then crucial to consider these sources of variation in the sampling design, for example, increasing the number of samples or using geostatistical methods.

Regarding cacao cultivation, further studies on the decomposability of specific plant inputs and the possibilities of increasing diversity in agroforestry systems must be performed, this could allow recommendations for introducing particular plant species that could contribute to soil quality. Additionally, the feasibility of increasing quantities of compost for organic management practices should be studied, since increased organic matter inputs may be necessary to achieve significant improvements in soil characteristics.