Organic and Conventional Management Practices That Improve Soil Quality and the Yield of Theobroma cacao in the Upper Huallaga Valley (Peru)

Abstract

Cocoa accounts for 5.20% of Peru’s cultivated land and is growing at a rate of 8.80% per year; however, yields remain low due to deficiencies in crop management. Therefore, this study used a multiple linear regression (MLR) model to evaluate effects of an organic agroforestry system (OAF) and conventional monocultures (CMs) on soil and production in high-yielding T. cacao plantations in the Upper Huallaga Valley, Peru. Four plantations were evaluated: organic agroforestry (Pa) and conventional monoculture (LE, Sa, and Sh). Soil physicochemical variables and cocoa production were assessed. The MLR analysis revealed that in OAF systems with mature trees, there will be slight losses of clay and silt fractions; the latter can be offset by high planting density. The OAF system showed a significant positive effect on pH. However, the CM system showed significant decreases in pH, CEC, Ca²⁺, and Mg²⁺. The interaction between OAF and CM optimized production, increasing the weight of dry beans. Planting density is associated with improvements in pH and bases, as well as fruit index and the weight of dry beans. MLR modeling suggests that integrating OAF systems with conventional management practices in high-density plantations offers valuable alternatives for the design of local agricultural policies and producer support programs, by identifying the factors that link management systems to soil quality and sustainable cocoa productivity in this valley.

1. Introduction

In the 2023/2024 season, global cocoa bean production reached 4,449,000 tons, with Ivory Coast (40.50%), Ghana (13.00%), Ecuador (9.70%), and Cameroon (6.70%) being the leading producers. Peru ranks eighth among producing countries with 160,000 tons of cocoa, making it the sixth most important crop, representing 5.20% of the country’s harvested area [1]. Furthermore, the National Institute of Statistics and Informatics [2] indicates that between 2014 and 2023, cocoa production grew by 8.8% annually, and the Huánuco region, home to the Upper Huallaga Valley, ranks fourth, accounting for 10.00% of national production. Peruvian cocoa production has increased at an annual rate of 12.60% over the last decade, with a cultivated area of 172,078 ha, based on monoculture systems with conventional management (56%) and the adoption of CCN-51 as the main variety cultivated [3]. In contrast, 44% of producers apply agroforestry systems and organic management practices, positioning Peru as the third largest exporter of organic cocoa in the world after Mexico and Indonesia [4]. However, compared to the countries with the highest yields in the world, Thailand (3000 kg ha⁻¹) and Guatemala (2600 kg ha⁻¹) [5,6], Peru achieves an average of 827 kg ha⁻¹ [4,6], highlighting a significant yield gap.

Organic management is an alternative that comprehensively improves soil properties [6,7,8] and has demonstrated positive benefits when applied in agroforestry systems (AFSs), as multifunctional forms of agriculture [9] that provide additional benefits to cocoa production. An AFS includes a set of land management techniques to combine crops and forest species [10] that contribute to the system’s sustainability [11,12], serving as a mitigation [13] and resilience strategy, thus controlling the negative effects of climate change [14,15] and promoting food security [16]. The benefits of organic management and agroforestry systems have been studied by researchers who have shown that these systems improve yields in the medium and long term [14] and produce physical and chemical changes that improve the soil through greater biomass incorporation, recycling, nutrient retention, soil structuring, and erosion control [6,7,9,17]. In contrast, a significant segment of scientists currently asserts that monoculture with conventional agricultural practices and the use of agrochemicals degrades the soil [18] and reduces its productive potential [19,20]. Furthermore, this production model is attributed to environmental problems such as climate change, biodiversity loss, ecosystem fragmentation, soil degradation, water regulation problems, and social disruptions [9,21,22].

Regarding the literature, there are reports that support chemical fertilization, among which the following stand out: Ballesteros et al. [23] in Nariño, Colombia, reported higher yields with chemical fertilization (2250 kg ha⁻¹) compared to organic fertilization (2007.80 kg ha⁻¹); also, Armijos et al. [24] in the province of El Oro in Colombia reported higher dry weight of 100 seeds with chemical fertilization compared to organic fertilization in CCN-51 cacao; and Tuesta-Pinedo et al. [25] in San Martin, Peru, reported higher yields with inorganic fertilizer (2502.4) compared to organic fertilizer (1560.6). Conversely, there are favorable reports for organic fertilization, among them, Florida et al. [6]. In the Upper Huallaga Valley, a production model integrating an agroforestry system (T. cacao and G. crinita) with organic management demonstrated yields ranging from 2700 to 3500 kg ha⁻¹, representing the highest yield reported in Peru. Finally, in the Ucayali region of Peru, Florida et al. [5] reported higher yields with organic fertilization (1125 kg ha⁻¹) compared to chemical fertilization, which reached 1080 kg ha⁻¹. These backgrounds highlight that there are divergent results, with limited analytical tools such as MLR that allow evaluations integrating production models and new variables (such as age and planting density) to find causal relationships and more suitable production models that represent an alternative for the 56% of producers who apply conventional monoculture in Peru, for agroforestry and organic management [26,27], a sustainable and productive way to guarantee food security [15,28].

In this context, Peruvian cacao exhibits low yields, 56% of farms practice conventional monoculture, and there is a limited availability of outstanding studies that highlight the benefits of different production models. Therefore, there is a gap in comprehensive research that quantifies the benefits of agroforestry systems, organic management, and the interaction between these two compared to conventional monoculture. Such research should identify changes in soil quality and yields in a specific local context, such as the Upper Huallaga Valley. Given this, this study hypothesizes that agroforestry systems and organic management, compared to monoculture systems with conventional management, have significant impacts on soil quality and production, achieving high yields in cacao plantations developed in the Upper Huallaga Valley, Huánuco region.

Therefore, the overall objective is to evaluate the effects of organically managed agroforestry systems and conventionally managed monoculture systems, as well as the integration of both through sustainable land-use planning (SLP), on soil quality and production in cacao plantations in the Upper Huallaga Valley of Peru. The aim is to generate information to guide concrete actions in decision making regarding cacao production systems that improve soils, yields, and the living conditions of farmers in the tropical Amazon.

2. Materials and Methods

2.1. Study Area

The research was conducted in the Huánuco region (Figure 1), involving four high-yield cacao plantations located in the villages of La Esperanza and Shiringal, in the Pueblo Nuevo district, and the village of Papayal, in the Castillo Grande district of Leoncio Prado province. Also included was the village of Sachavaca, in the Monzón district of Huamalíes province. All these areas are located in the Upper Huallaga Valley in the Huánuco region.

2.2. Characteristics of the Areas Evaluated

2.2.1. Plantation 01 (LE)

Plantation 01 is located in zone 18 L, UTM coordinates 387,198 E and 8,994,039 N, at 630 m above sea level. Politically, it is located in the town of La Esperanza (LE), Pueblo Nuevo district, Leoncio Prado province, in the Huánuco region. The plantation consists of a 12-year-old CCN-51 clone, with a density of 1250 plants per hectare, under a monoculture system with combined organic and chemical management. The plantation is pruned in August to maintain the appropriate shape and height of the plants; it also undergoes pruning to open up rows and shape the canopy, as well as ongoing phytosanitary pruning. Weed control is carried out manually two to three times a year with the aid of a brush cutter.

Regarding the plantation’s fertilization plan, as described in

Table 1. Organic–conventional fertilization plan in cacao plantations.

Input	1st Fertilization (August)			2nd Fertilization (November)			3rd Fertilization (February)
	LE	Sa	Sh	LE	Sa	Sh	LE	Sa	Sh
Dolomite (kg ha−1)							1500.00
Island guano (kg ha−1)	200.00	125.00	125.00	120.00	75.00	75.00	80.00	50.00	50.00
Diammonium phosphate (kg ha−1)	125.00	62.50	100.00	75.00	37.50	60.00	50.00	25.00	40.00
K sulphate (kg ha−1)	10.00	10.00	10.00	15.00	15.00	15.00	25.00	25.00	25.00
Urea (kg ha−1)	7.30	--	16.50	10.95	--	24.75	18.25		41.25
Magnocal (kg ha−1)	87.50	50.00	75.00	52.50	--	45.00	35.00		30.00
Triple superphosphate (kg ha−1)	--	25.00	43.75	--	15.00	26.25	--	10.00	17.50
Ulexite (kg ha−1)	97.50	87.50	97.50	58.50	52.50	58.50	39.00	35.00	39.00
Kieserite (MgSO4)	--	150.00	--	--	--	--	--	--	--
Phosphoric rock (kg ha−1)	--	75.00	--	--	45	--	--	30	--
Copper sulfate (kg ha−1)	1.50	2.50	1.13	--	--	--	--	--	--
Zinc Sulfate (kg ha−1)	3.00	4.00	3.25	--	--	--	--	--	--
Mg sulfate (kg ha−1)	1.00	--	12.50	--	--	--	--	--	--
Mn sulfate (kg ha−1)	1.00	1.00	1.00	--	--	--	--	--	--
Subtotal (kg ha−1)	533.80	592.50	485.63	331.95	240.00	304.50	247.25	175.00	242.75
Total, LE	1113.00 kg ha−1
Total, Sa	1007.50 kg ha−1
Total, Sh	1032.88 kg ha−1

LE, La Esperanza plantation; Sa, Sachavaca village plantation; Sh, Shiringal village plantation; Pa, Papayal village plantation; --, data not recorded.

, prior to the first application of fertilizers and before pruning, organic soil fertilization is applied (agos), consisting of compost (1 kg plant⁻¹ year⁻¹) and potassium humate (10 L 200 L drum⁻¹ ha⁻¹ year⁻¹). After pruning and before the start of flowering, foliar applications of calcium and boron, and phosphorus and potassium (P-K) as potassium sulfophosphate (2 L 200 L drum⁻¹) are applied every 15 days, 3 times per season. During this dormant period, copper chelate (1 L 200 L drum⁻¹) is also applied twice per season.

2.2.2. Plantation 02 (Sa)

Plantation 02 is located in zone 18 L, UTM coordinates 374,838 E and 898,0676 N at 854 m above sea level. Politically, it is located in the town of Sachavaca (Sa), Monzón district, Huamalíes province. It is a 9-year-old CCN-51 plantation with a density of 1250 plants per hectare under a monoculture system with conventional management. The owner applies the management and fertilization plan used in plantation 01, described in the previous paragraph and in Table 1. However, unlike plantation 01 (LE), the foliar applications (Ca-B), (K), and potassium sulfophosphate (3 L 200 L drum⁻¹) are applied every 15 days, but instead of three times per season, the application is done four times. Also, copper chelate 1 L 200 L drum⁻¹, is applied twice instead of once/campaign, and potassium humate 10 L 200 L drum⁻¹, every 21 days, three times/campaign, instead of two applications.

2.2.3. Plantation 03 (Sh)

Plantation 03 is located in zone 18 L, UTM coordinates 385,505 E and 8,989,495 N at 627 m above sea level. Politically, it is located in the town of Shiringal (Sh), Pueblo Nuevo district, Leoncio Prado province. It is a 12-year-old CCN-51 plantation with a density of 1500 plants per hectare under a monoculture system with conventional management. The same fertilization and management plan as plantation 01 is being applied, described in item 2.2.1 and Table 1.

Table 1 describes the amendments and fertilizers applied by the farmer to the plantations in the last 6 years, with technical advice from DEVIDA.

2.2.4. Plantation 04 (Pa)

Plantation 04 is located in zone 18 L, UTM coordinates 389,235 E and 8,976,797 N at 661 m above sea level. Politically, it is located in the town of Papayal (Pa), Castillo Grande district, Leoncio Prado province. It is a 7-year-old CMP-15 aromatic cacao plantation with a planting density of 4000 plants ha⁻¹. It applies the 100% organic technology called “MOM management”, described in Florida et al. [6], which consists of:

-

Management in an agroforestry system (T. cacao L. and G. crinita Mart.).

-

Annual drastic pruning (70% of the foliage).

-

Application of compost to the soil before drastic pruning (400 g/plant).

-

Triple foliar biofertilization (nutrients, fungicide, and insecticide), based on local inputs, 3 days after pruning and 21 days after the first application.

-

Use of the CMP-15 genotype from the Méndiz Paredes Collection.

2.3. Bioclimatic Characteristics of the Evaluated Areas

According to Holdridge’s [29] classification of the world’s life zones or vegetation formations, the districts under evaluation (Castillo Grande, Monzón, and Pueblo Nuevo) belong to a very humid subtropical forest (bmh-S). The entire province of Leoncio Prado and the district of Monzón are located in the high jungle of Peru, with varied topography including plains, slopes, and rugged hillsides. The general climatic conditions are as follows: an average annual temperature of 24.5 °C, relative humidity of 84%, and annual rainfall of 3400 mm. Therefore, according to Pulgar [30], this area belongs to the Rupa Rupa ecoregion, or high jungle.

2.4. Selection Criteria for the Evaluated Plantations

Four plantations were evaluated. They were selectively identified and chosen within the jurisdiction. Information was sought from producers using organic management practices who achieve high yields (>2000 kg ha⁻¹) and are registered with the National Program for Life Without Drugs (DEVIDA), operating in the Upper Huallaga region. Three areas registered in the program were identified and compared with a model plantation independent of the program. This model plantation (Pa) is not part of the program but has reported the highest yield in the region and in Peru [6], and was therefore used as the reference area. Once identified, the following criteria were applied:

-

Location and collection of management information, as recorded in the traceability records of the relevant plantations.

-

The technical criterion applied was that the plantation must be in the productive stage, with active management and traceability in place at the time of the evaluation.

-

The management practices during the production cycle were characterized and described, including the timing and quantity of amendments, biofertilizers, fertilizers, etc., applied.

-

A sampling sub-area (1000 m² of cultivated area) was delimited to randomly obtain soil and cacao production samples.

2.5. Determination of Physicochemical Indicators of the Soil

Soil samples were analyzed from T. cacao plantations under organic and organic–chemical management. A 1000 m² sub-area was selected, and sampling was carried out at 10 random points within each subplot. These points coincided with 10 T. cacao plants selected for evaluation. Four soil subsamples were extracted from each plant to obtain a composite sample. These samples were taken at the canopy height of each selected plant, according to the Soil Taxonomy methodology, considering a stratum of 0.00–0.20 m depth. Physical and chemical soil indicators were evaluated according to Bazán [31].

2.6. Determination of Cacao Production Indicators

Ten cocoa plants were randomly selected in the 1000 m² sampling sub-area, in which the methodological criteria proposed by Dos Santos & Midlej [32] and Orozco-Aguilar et al. [33] were used to determine performance parameters through the number of fruits (a) and the pod index (b).

(a). 

Number of fruits or ears of corn

This was obtained by counting the total number of fruits or ears of corn with dimensions ≥ 10.00 cm. The result was expressed in ears/plant.

(b). 

Ear index

These are the pods needed to obtain one kilogram of dry cocoa and are expressed with the following formula:

Im = #M/kg dry cocoa

where:

• Im = pod index # pods/kg dry cacao;
• #M = number of pods.

(c). 

Yield

The calculation according to the references indicated is based on a simple method that combines the ear count and the ear index, and only ears larger than 10 cm are counted as survival discount factors, since once this minimum size is reached, the fruit’s survival probability is high. The yield calculation is as follows:

R = #m*Im

where:

R = yield in kg/plant;

#m = number of ears per plant;

Im = ear index.

2.7. Design and Statistical Analysis

This research is framed within a non-experimental, cross-sectional, retrospective design with an explanatory scope. The collected data was subjected to statistical analysis to infer the significance of the parameters of the proposed models. Subsequently, an econometric analysis was performed through the estimation and interpretation of multiple linear regression models, as well as various robustness tests that support the validity of these models. Stata IC 15 software was used for this purpose.

2.8. Multiple Linear Regression (MLR) Model

The Multiple Linear Regression model is represented vectorially in Equation (1), based on Greene [34]. This methodology allows quantifying the effect of the independent variables (X) on the dependent variable (Y). The estimation of the parameters (β) was performed using the Ordinary Least Squares (OLS) method:

Y = Xβ_OLS + μ

(1)

where the error term (μ) must follow a normal distribution with zero mean (E(μ) = 0) and constant variance (σ²_μ).

To validate the model, the Jarque–Bera test [35] was used to verify whether μ follows a normal distribution. White’s test [36] was also applied to verify whether σ²_μ is homoscedastic; if not, the parameters were estimated using generalized least squares (GLS). Furthermore, to evaluate the relationship between the independent variables, the variance inflation factor (VIF) was considered to detect potential multicollinearity problems, following O’Brien’s proposal [37], with the expectation that the VIF would be less than 10. Additionally, the coefficient of determination (R²) [31] was used to observe the extent to which the variability of the independent variables explains the variability of the dependent variable. The global relevance test was also performed using Fisher’s F-test [31], and the individual relevance test using Student’s t-test [38] on the parameters, to confirm their statistical significance. Finally, the root mean square error (RMSE) was used as a measure of the predictive accuracy of the model [39].

3. Results

3.1. Soil and Production Indicators

Table 2. Descriptive statistics on the physical and chemical properties, production, and yield in cocoa plantations managed using organic agroforestry (Pa) and in organic–conventional monoculture (Le, Sa, and Sh).

Indicator	Treatment				Statistics
	LE	Sa	Sh	Pa	$\stackrel{-}{\mathbf{X}}$	Sig.
Sand (%)	48.60 ± 2.63 b	20.40 ± 2.84 a	22.00 ± 6.48 a	45.40 ± 16.80 b	34.10 ± 15.83	<0.01 **
Clay (%)	21.40 ± 2.99 b	45.20 ± 2.15 c	20.00 ± 2.49 b	9.00 ± 1.70 a	23.90 ± 13.56	<0.01 **
Silt (%)	30.00 ± 2.54 a	34.40 ± 2.12 a	58.00 ± 5.10 c	45.60 ± 15.41 b	42.00 ± 13.56	<0.01 **
Texture	loamy texture	Clayey	silty loam	sandy loam	--	--
pH	4.36 ± 0.26 b	4.12 ± 0.17 a	7.49 ± 0.10 c	7.40 ± 0.18 c	5.84 ± 1.64	<0.01 **
MO (%)	2.12 ± 0.55 a	2.51 ± 0.72 a	2.22 ± 0.71 a	1.93 ± 1.15 a	2.20 ± 0.81	0.472 ns
P (ppm)	3.92 ± 1.61 a	7.77 ± 2.15 b	3.33 ± 1.38 a	7.36 ± 2.81 b	5.59 ± 2.82	<0.01 **
K (cmol+ kg−1)	140.43 ± 34.29 a	141.73 ± 52.86 a	136.06 ± 31.36 a	171.54 ± 17.05 a	147.44 ± 37.59	0.129 ns
Ca (cmol+ kg−1)	3.91 ± 1.61 a	2.64 ± 0.55 a	13.65 ± 2.21 c	10.31 ± 1.88 b	7.63 ± 4.86	<0.01 **
Mg (cmol+ kg−1)	0.47 ± 0.25 a	0.30 ± 0.11 a	2.24 ± 0.54 b	1.86 ± 0.73 b	1.22 ± 0.97	<0.01 **
Al (cmol+ kg−1)	0.82 ± 0.44 b	1.69 ± 0.59 c	0.00 ± 0.00 a	0.00 ± 0.00 a	0.63 ± 0.79	<0.01 **
CEC (cmol+ kg−1)	5.95 ± 1.67 a	5.72 ± 0.59 a	16.57 ± 2.30 c	12.95 ± 2.18 b	10.30 ± 5.01	<0.01 **
Fruits (N°)	38.70 ± 4.22 c	36.50 ± 4.90 c	28.80 ± 5.71 b	20.60 ± 1.89 a	31.40 ± 8.30	<0.01 **
FI (F kg−1)	12.77	13.12	14.12	20.51	15.13 ± 3.19	--
Yield (kg ha−1)	4667.03 ± 508.63 b	4406.71 ± 592.15 ab	3900.17 ± 747.52 a	4639.27 ± 427.3 b	4403.29 ± 639.32	0.02 *
Plant age (years)	12	9	12	7	10.25 ± 2.07	--
PD (plant ha−1)	1 250	1 250	1 500	4 000	2400 ± 1300	--

LE, plantation in the town of La Esperanza; Sa, plantation in the town of Sachavaca; Sh, plantation in the town of Shiringal; Pa, plantation in the town of Papayal; FI, Fruit Index; PD, plant density; different letters (^a, ^b and ^c) in the same row indicate differences between the evaluated areas; Sig.: significance; **: highly significant (p < 0.01); *: significant (p < 0.05); ns, not significant; --: data not reported; n = 40; data matrix = 600.

shows the mean values and standard deviations of the individual values for each variable across the four plantations, as well as the overall mean, which allows us to determine the characteristics and levels of each variable evaluated in these plantations. Regarding soil physical properties, silt had the highest mean content (42.00%) and the lowest variability (SD = 13.56%), compared to sand (34.10% ± 15.83%) and clay (23.90% ± 13.56%).

Regarding chemical properties, higher pH (7.49 ± 0.10) and concentrations of Ca, Mg, and CEC were observed in Sh (7.49 ± 0.10, 13.65 ± 2.21, 2.24 ± 0.54 b, and 16.57 ± 2. 30 cmol⁺ kg⁻¹, respectively), OM, P, and Al in Sa (2.51 ± 0.72%, 1.69 ± 0.59 ppm, and 7.77 ± 2.15 cmol⁺ kg⁻¹, respectively), and K in Pa (171.54 ± 17.05 cmol⁺ kg⁻¹).

Regarding production indicators, although the LE plantation did not have higher values for the different nutrients, it had the highest number of fruits and yield (38.70 ± 4.22 and 4667.03 ± 508.63 kg ha⁻¹, respectively), and the Pa plantation had the highest fruit index (20.51 F kg⁻¹). Regarding the characteristics of the plantation, the average age of the plants was 10.25 ± 2.07 years (median: 11 years; range: 7–12 years), and the planting density was 2400 ± 1300 plants/ha (median: 1.72; range: 1.54–4.62). Finally, Table 2 shows significant differences for most indicators, except for the means of organic matter (OM) and potassium (K⁺).

3.2. Modeling

3.2.1. Physical Fractions of Soil

The results of the soil physical properties modeling (sand, clay, and silt) are presented in

Table 3. Multiple linear regression model for soil quality based on physical properties (Dimension 1) in cocoa crops.

Independent Variable	Dimension 1: Physical Properties
	Sand (%)	Clay (%)	Silt (%)
Intercept	−0.8105	1.8086 ***	−1.154 ***
Independent variable 1: High-performance organic management systems
Dimension 1: Cultivation System
Agroforestry and monocultura (kg/plant)	0.0316	−0.0145	0.0142
Dimension 2: Crop Management
Organic and Conventional (organic = 1 and conventional = 0)	11.7049 ***	−0.5151 ***	−1.9715 ***
Interaction
Agroforestry×Organic	3.5056 **	---	−0.7635 **
Independent variable 2: Control variables
Planting density (Plants/ha)	−4.0306 ***	−0.0755 *	0.9384 ***
Age (Years of the plant)	0.5713 ***	−0.1195 ***	−0.0025
Validation test
RMSE	0.4390	0.0237	0.0776
Coefficient of determination (R2)	0.7041	0.9726	0.7144
Test of global relevance: F-Fisher statistic (df1 = k − 1 and df2 = n − k)	16.18 ***	310.60 ***	17.01 ***
Jarque–Bera test: Chi-square statistic (df = 2)	5.707 *	4.444	5.071 *
White’s test: Chi-square statistic (df = 9)	11.3242	10.4102	15.4901 *
Multicollinearity test: Average VIF	127.08	112.79	127.08

*** p < 0.01, ** p < 0.05, * p < 0.10, df represents the degrees of freedom, df₁ represents the degrees of freedom of the numerator, df₂ represents the degrees of freedom of the denominator, k represents the number of parameters, and n represents the sample size.

. The high coefficients of determination (R² = 70.41%, 97.26%, and 71.44%, respectively) indicate that the independent variables explain a substantial proportion of the variability in these properties. The overall F-test confirms that all models are statistically significant (p < 0.01). The residuals of the three models do not show significant deviations from normality (Jarque–Bera test, p > 0.05) and are homoscedastic (White’s test, p > 0.05). Although multicollinearity was detected (average VIF > 10), this is attributable to the natural interdependence among the predictor variables, a phenomenon documented in the literature (e.g., Florida et al., [12]). The low RMSE values (0.44, 0.02, 0.08) support the predictive accuracy of the models.

The results of the multiple linear regression analysis show that dimension 1—the farming system (agroforestry or monoculture)—does not have a significant causal relationship with the sand, clay, and silt fractions of the soil. Regarding dimension 2, crop management (organic and conventional) exhibits a positive and significant causal relationship (p < 0.01) with sand content (11.70%) and a negative and significant causal relationship (p < 0.01) with clay and silt content (−0.515% and −1.97%, respectively) compared to conventional management.

The interaction between “agroforestry systems and organic management” was significantly enhanced, which was positively associated (p < 0.05) with the sand fraction (3.51%) and negatively associated with the silt fraction (−0.76%). In contrast, planting density showed a negative relationship (p < 0.01) with sand and clay (−4.03% and −0.075%, respectively), and a positive relationship with silt (0.938%); however, plantation age had a significant positive relationship (p < 0.01) with sand (0.57%) and a negative relationship with the clay fraction (−0.119%).

3.2.2. Chemical Variables of the Soil

The models for the chemical properties (

Table 4. Multiple linear regression model for soil quality based on chemical properties (Dimension 2) in cocoa crops.

Independent Variable	Dimension 2: Chemical Properties
	pH	OM (%)	P (ppm)	K+ (cmol+ kg−1)	CEC (cmol+ kg−1)	Ca2+ (cmol+ kg−1)	Mg2+ (cmol+ kg−1)
Intercept	−23.2257 ***	1.4070	22.093 **	0.8410 **	−42.248 ***	−59.0933 ***	−77.661 ***
Independent variable 1: High-performance organic management systems
Dimension 1: Cultivation System
Agroforestry and monoculture (kg/plant)	0.0333 **	0.4360	−0.1614	−0.0619	−0.0200	0.0069	0.0298
Dimension 2: Crop Management
Organic and Conventional (organic = 1 and conventional = 0)	−2.2902 ***	−4.2086	−0.2376	0.3466	−4.9558 ***	−5.2995 ***	−6.8837 ***
Independent variable 2: Control variables
Planting density (Plants/ha)	3.0521 ***	1.2348	−1.4219	−0.1357	5.6768 ***	7.2233 ***	9.2920 ***
Age (Years of the plant)	0.8897 ***	−0.1994	−3.990 ***	−0.0102	0.9617 **	2.9569 ***	3.4150 ***
Validation test
RMSE	0.1660	0.8150	0.3626	0.0795	0.1791	0.2625	0.4572
Coefficient of determination (R2)	0.7429	0.0926	0.5631	0.2393	0.8884	0.8868	0.8094
Test of global relevance: F-Fisher statistic (df1 = k − 1 and df2 = n − k)	1082.92 ***	0.89	11.28 ***	2.75 **	69.65 ***	68.54 ***	37.16 ***
Jarque–Bera test: Chi-square statistic (df = 2)	0.0682	2.896	5.609 *	2.344	5.849 *	5.434 *	5.888 *
White’s test: Chi-square statistic (df = 8)	12.4772	11.1874	2.5988	14.6056 *	10.1767	13.4124 *	7.9672
Multicollinearity test: Average VIF	636.16	112.79	44.38	112.79	44.38	44.38	44.38

*** p < 0.01, ** p < 0.05, * p < 0.10.

) showed variable fit. The models for pH, CEC, Ca²⁺, and Mg²⁺ had high R² values (74.29% to 88.84%) and were globally significant (p < 0.01). The models for P (R² = 56.31%) and K⁺ (R² = 23.93%) were also significant (p < 0.01 and p < 0.05, respectively), while the model for OM (R² = 9.26%) was not significant. The residuals of all models met the assumptions of normality and homoscedasticity. Multicollinearity was detected, which is justified by the nature of the variables.

The multiple linear regression analysis (Table 4) shows that Dimension 1, the farming system (agroforestry or monoculture), exhibits a positive and significant causal relationship (p < 0.05) only with pH levels (0.03%). Regarding dimension 2, crop management (organic and conventional) exhibits a negative and significant causal relationship (p < 0.01) with pH, CEC, Ca²⁺, and Mg²⁺ (−2.29, −4.955, −5.299, and −6.88%, respectively).

Cocoa planting density showed a significant positive causal relationship (p < 0.01) with pH, CEC, Ca²⁺, and Mg²⁺ values (3.05%, 5.68%, 7.22%, and 9.29%, respectively). The age of cocoa plantations showed a significant positive causal relationship (p < 0.01) with pH, CEC, Ca²⁺, and Mg²⁺ (0.889, 0.96, 2.96, and 3.42%, respectively), and a negative relationship (p < 0.01) with phosphorus (−3.99%), whereas no significant causal relationships were observed for OM and K⁺.

3.2.3. Cocoa Production Variables

The production and yield models (

Table 5. Multiple linear regression model for cacao production based on fruit production (Dimension 1) and yield (Dimension 2).

Independent Variable	Dimension 1: Fruit Production		Dimension 2: Performance
	Number of Fruits (F Plant−1)	Fruit Index (F kg−1)	Dried Almonds (kg ha−1)
Intercept	2.7462 ***	−0.9721 ***	4.4462 ***
Independent variable 1: High-performance organic management systems
Dimension 1: Cultivation System
Agroforestry and monoculture (kg/plant)	0.4231 ***	−0.0026	0.4110 ***
Dimension 2: Crop Management
Organic and Conventional (organic = 1 and conventional = 0)	0.1349 *	-	−0.2774 **
Interaction
Agroforestry×Organic	-	−0.0924 ***	0.6186 ***
Independent variable 2: Control variables
Planting density (Plants/ha)	−0.0250	0.4911 ***	0.3784 ***
Age (Years of the plant)	−0.0327	−0.0293 **	0.0347
Validation test
RMSE	0.0357	0.0058	0.0225
Coefficient of determination (R2)	0.9856	0.9992	0.9812
Test of global relevance: F-Fisher statistic (df1 = k − 1 and df2 = n − k)	351.85 ***	6740.79 ***	1411.61 ***
Jarque–Bera test: Chi-square statistic (df = 2)	2.098	5.308 *	0.7671
White’s test: Chi-square statistic (df = 9)	-	-	-
Multicollinearity test: Average VIF	8771.05	29.08	74.13

F fruits, *** p < 0.01, ** p < 0.05, * p < 0.10.

) exhibited excellent fit (R² > 98%) and overall significance (p < 0.01). Residuals were normally distributed, but heteroscedasticity was detected, requiring the use of GLS for estimation. Multicollinearity was present but considered inherent to the design.

The multiple linear regression analysis (Table 5) shows that Dimension 1, the farming system (agroforestry or monoculture), has a positive and significant causal relationship (p < 0.01) with the number of fruits and the weight of dried cocoa beans (0.42 and 0.41%, respectively). Regarding dimension 2, crop management (organic and conventional) shows a positive and significant causal relationship (p < 0.10) with a slight increase in the number of fruits (0.13%), but with a significant (p < 0.05) decrease in dry beans (−0.28%). The interaction between the agroforestry system and organic management shows a highly significant causal correlation (p < 0.01), reducing the fruit index (−0.09%) and substantially increasing dry beans (0.62%), a very positive aspect for cocoa production.

With regard to the control variables, planting density showed a significant positive causal relationship (p < 0.01) with the fruit index and the weight of dry almonds (0.49% and 0.38%, respectively). Planting age, on the other hand, had a significant negative causal relationship (p < 0.05) with the fruit index (−0.03%), without showing significant effects on the other production variables.

4. Discussion

4.1. About Modeling

Multiple linear regression (MLR) analysis, which examines the causal relationship between the organically managed agroforestry system (OAF) and conventional monoculture (CM) and the physical and chemical properties of the soil as well as cocoa production, shows statistically robust fits. The models for the physical fractions of sand, silt, and clay in the soil (Table 3) exhibit high coefficients of determination (R² = 70.41, 97.26, and 71.44%, respectively), indicating that the independent variables explain a substantial proportion of the variability in these properties. The overall F-test confirms that all models are statistically significant. The model residuals do not show significant deviations from normality (Jarque–Bera test) and are homoscedastic (White’s test). The low RMSE values (0.44, 0.02, and 0.08) support the predictive accuracy of the models.

The models for the chemical properties (Table 4) showed variable fit. The models for pH, CEC, Ca²⁺, and Mg²⁺ had high R² values (74.29 to 88.84%) and were globally significant. The models for P (R² = 56.31%) and K⁺ (R² = 23.93%) were also significant, while the model for OM (R² = 9.26%) was not significant. The residuals of all models met the assumptions of normality and homoscedasticity. Multicollinearity was detected, which is justified by the nature of the variables. Likewise, the production and yield models (Table 5) exhibited excellent fit (R² > 98%) and overall significance. The residuals were normally distributed, but heteroscedasticity was detected, requiring the use of LCM for estimation. Multicollinearity was present but was considered inherent to the design.

In general, multicollinearity (average VIF > 10) was detected in the soil physical fractions model (Table 3), which is due to the natural interdependence among the predictor variables, a phenomenon documented in the literature (e.g., Florida et al., [12]). Furthermore, the coefficients of determination obtained in all models indicate that the variability of the dependent variables is explained by the independent variables, demonstrating a remarkable explanatory power [34,38]. Regarding the fundamental assumptions for model validity, homoscedasticity was verified using White’s test, suggesting the absence of heteroscedasticity in the residuals [36]. The normality of the errors was confirmed by the Jarque–Bera test, which validates the normal distribution of the stochastic component in all models [35]. The convergence of these global significance criteria, the fulfillment of the assumptions (homoscedasticity and normality) and the high explanatory power (R²) support the suitability of the proposed model to analyze the influence of the selected variables on the physicochemical properties of the soil and cocoa production.

In this context, MLR modeling allows for the establishment of positive or negative causal relationships between variables, and research in similar areas has demonstrated its usefulness in the Peruvian tropics for explaining the impact of rotational grazing on soil physical and chemical parameters [40], the impact of permanent Harton banana plantations on soil quality [41], and the impact of agroforestry–organic management in cacao plantations on soil properties and production [6]. The results, compared with these references, show similar statistical fit in terms of meeting assumptions, validity, and high predictive power.

4.2. Physical Indicators of the Soil

Soil texture, expressed as the proportion of sand, silt, and clay fractions, is a physical property directly or indirectly related to other physical properties that define air and water flow in the soil [42]. These aspects are important for plant development and its variations and are partly conditioned by management practices and by spatial and soil formation factors [6,43,44]. The proportion of these physical fractions was evaluated based on the requirements of T. cacao, which needs deep, permeable, loamy alluvial soils with good drainage, free from flood risk, and that allow easy penetration of the taproot. In general, balanced proportions of sand, silt, and clay are needed to allow for good root system development [6,45]. According to the FAO [46], the soils evaluated correspond to sandy loam textures for LE, clay for Sa, silty loam for Sh, and sandy loam for Pa. Therefore, the area with the least physical suitability for cacao cultivation would be Sa, due to its high average clay levels (45.20%), although this does not appear to affect production, since it is the third area evaluated with the highest yield (4406.71 kg ha⁻¹), well above the Peruvian average of 840 kg ha⁻¹ [47].

The results of the multiple linear regression on the fractions (Table 3) that determine texture show:

(a).

The cropping system (agroforestry and monoculture) shows a non-significant causal relationship with the sand, silt, and clay fractions. This indicates that the agroforestry system maintains soil particles in equilibrium.

(b).

Farming management (organic and conventional) shows a significant negative causal relationship with clay and silt.

(c).

The interaction between the agroforestry system and organic management is negatively associated with the silt fraction.

(d).

Planting density has a significant negative relationship with clay and a positive relationship with silt.

(e).

Plantation age has a significant negative causal relationship with the clay fraction.

These results suggest that in a cacao plantation where an agroforestry system is implemented, with organic management and older trees, there will be slight losses of clay and silt fractions. This result can be explained by the high rainfall in the study area. According to SENAMHI [48], the area ranges from 400 to 1500 m above sea level, with temperatures between 17 and 29 °C and annual rainfall ranging from 2848 to 3500 mm, with maximum rainfall intensities of up to 160 mm in 24 h; therefore, the study area is classified as very rainy. In addition to rainfall, the evaluated plantations were selected for their high yield within the Upper Huallaga Valley; consequently, they are geographically dispersed, with distances ranging from 5 to 20 km between them, resulting in spatial variability. Finally, an important consideration is the proportionality of the three fractions (which add up to 100%); the reduction in silt and clay shown in the regression clearly implies an increase in sand, due to the simple proportionality resulting from the loss of the finer fractions.

However, the modeling also shows that the losses of silt and clay can be compensated for by a high planting density. This behavior is largely due to the actions of physical weathering by abrasion [49] and chemical weathering, which determine the size of the fine particles [50] and are strongly influenced by climate and tree roots [51]. These conditions are present in an agroforestry system with organic management and high planting density in a highly rainy area such as the Upper Huallaga Valley.

In this context of high rainfall and spatial variability, the slight losses of the finer fractions (silt and clay) as evidenced by the multiple linear regression is a reasonable finding and consistent with the findings of Sainju et al. [52], who note that the relationship among physical properties including the different fractions of sand, silt, and clay in the soil is influenced by soil conditions, climate, and management practices. Furthermore, according to Sharma & Kumar [53], particle size varies significantly depending on land use and is correlated with vegetation types, vegetation cover, and crop management; these factors can promote soil erosion and/or cause the loss of primarily silt and clay in the surface layer. Finally, De Wang et al. [54] state that areas with high precipitation experience high erosion rates and, consequently, the loss of fine particles, which is largely determined by land use, as evidenced by the results of this study.

The sandy fraction is chemically inert and lacks colloidal properties [17,46] and does not pose a threat to crops; however, proportionally, its increase is due to the loss of silt and clay; these fractions are indeed important and are likely related to the soil’s low organic matter content [55] and to the aforementioned factors (climate, plant age, vegetation cover, and spatial variability). Silt and clay are the active fractions of the soil [17,19,56], and their loss from the soil, in this case, is attributable to water erosion; according to the regression analysis, this can be controlled by high planting density, which generates greater shade and root density; this can lead to differences in soil fertility levels [57], since greater shading helps reduce soil erosion and, consequently, the loss of colloidal fractions (silt and clay), which control nutrient availability.

Overall, these results suggest a slight reduction in the percentage of clay and silt attributable to management practices, as the soil tends to lose these fractions due to high rainfall (3400 mm) in the valley [6,12], in addition to the factors mentioned above. This is likely naturally associated with the evolution of the soil profile, which exhibits high spatial variability and strong spatial dependence due to variations in topographic position [58], the lithology of the parent material [55,59], and soil formation processes [43,44]. Finally, the results do not align with those of Florida et al. [6], who evaluated the Pa area with different ages, spacings, and applications of agroforestry–organic management; they concluded that no significant changes had occurred in soil fractions over 13 years. In this study, the agroforestry system with organic management is 7 years old, and the system is not yet established; therefore, the regression did not find a significant causal relationship, as these changes become more pronounced over the long term and are not visible in the short and medium term [11,19,57]. However, the regression suggests that the combination of an agroforestry system with organic management and high planting density would be the ideal planting system to maintain a balance against the loss of fine particles, attributed mainly to the high rainfall in the study area.

4.3. Chemical Indicators of the Soil

The results for soil chemical variables (Table 2), corresponding to plantations with organic agroforestry management (Pa) and conventional monoculture (Sa, LE, and Sh), show, according to FAO criteria [46], that the pH varies between 4.12 and 7.49. Based on these values, the LE and Sa areas are classified as strongly acidic (pH < 5.00), while the Sh and Pa areas are classified as moderately alkaline (pH < 8.5). Regarding organic matter content, it ranges from 1.93 to 2.51%, corresponding to a medium level (1.50–3.50%) in all evaluated areas. Soil phosphorus (P) ranges from 3.33 to 7.77 ppm, corresponding to low levels (p < 5.5 ppm) in the LE and Sh areas, and to medium levels (p < 11 ppm) in the Sa and Pa areas. Regarding potassium (K⁺), it ranges from 136.06 to 171.54 ppm, corresponding to medium levels (120–240 ppm) in all areas. Calcium (Ca²⁺) ranges from 2.64 to 13.65 cmol⁺ kg⁻¹, corresponding to a low level (LE and Sa) and a high level (2.00–5.00 and >10.00 cmol⁺ kg⁻¹) for Sh and Pa, respectively. Magnesium (Mg²⁺) ranges from 0.30 to 2.24 cmol⁺ kg⁻¹, corresponding to a very low level (below 0.50 cmol⁺ kg⁻¹) for areas LE and Sa, and a medium level (below 3.00 cmol⁺ kg⁻¹) for areas Sh and Pa. As for aluminum (Al) levels, only the Sa area (1.69 cmol⁺ kg⁻¹) shows toxic levels (above 1.00 cmol⁺ kg⁻¹).

The results (Table 2) show two soil types. The first corresponds to plantations LE and Sa, with a conventionally managed monoculture system. Although a combination of organic (biofertilizers and amendments) and chemical (formulated fertilizers) practices are applied, it is still considered conventional. These soils have a strongly acidic pH, low phosphorus, calcium, and magnesium content, and high aluminum levels. This corresponds to the typical acidic residual soils of tropical Peru, classified by the Soil Taxonomy as the order Inceptisols, subgroup Typic Dystrudep [60]. The second corresponds to plantations Sh with conventional monoculture (CM) and Pa agroforestry with organic management (OAF). These soils are moderately alkaline, with medium phosphorus levels, high calcium levels, medium magnesium levels, and no aluminum. These are recent alluvial soils that correspond to the order Entisols, subgroup Typic Udifluvent [41,60].

The MLR model shows a significant positive causal relationship between organic and conventional management (OAF) and pH compared to monoculture. Organic management, compared to conventional management, shows a significant negative causal relationship with pH, Ca²⁺, Mg²⁺, and CEC levels. Conversely, the control variables of planting density and age show significant positive causal relationships with pH, Ca²⁺, Mg²⁺, and CEC levels. Furthermore, to understand the effects shown by the RML, consider that the Upper Huallaga Valley has high rainfall [48], with annual precipitation ranging from 2848 to 3500 mm, with maximum rainfall intensities of up to 160 mm in 24 h, and the distance between plantations naturally generates spatial variability in its different indicators.

In this context, the positive causal relationship shown by OAF on pH has an R² = 0.74; that is, the difference (36%) could be influenced by variables not evaluated such as precipitation and spatial variability, although the latter would apparently be contradictory to the statements of Chinea and Rodríguez [43] and Inocencio and Florida [44], that the chemical properties of the soil present low variability and spatial dependence in specific contiguous areas; on the contrary, this research evaluated four plantations within a valley and, contrasting with the results (Table 2), it was evidently observed that variability occurs between the individual pH values among the evaluated areas. Furthermore, we must understand that MLR modeling is a statistical tool that allows us to collect this variability presented by the data of the different plantations and finds causal relationships on them [31,34], with the R² value (74%) being fundamental; consequently, the observed effect is closely related to the characteristics of the agroforestry system, in this case, made up of the species Guazuma crinita and Calycophyllum spruceanum that provide multiple benefits such as cover, prevention of the direct impact of raindrops, and the contribution of biomass that improves recycling, OM content and soil cover [13,14,17,22]. All these benefits in a context of high rainfall reduce runoff and loss of fine particles [50,51], determinant in the retention of Ca²⁺ and Mg²⁺ that balance or improve the pH of the soil [56,61]. Therefore, the positive effect shown by OAF on pH, compared to monoculture, is reasonable and supported by references, including Florida et al. [6] and Watemberg et al. [61], among others.

The MLR analysis also found that organic management, compared to conventional management, shows a significant negative causal relationship with pH, Ca²⁺, Mg²⁺, and CEC levels. This result can be explained by the fact that organic management only includes biocides (antifungals), biofertilizers (foliar nutrients), and compost. These biofertilizers have low nutrient concentrations and are released more slowly compared to chemical fertilizers (Florida-Rofner et al. [62]; Ríos-Velasquez et al. [63]; Aguirre-Illatopa et al. [64]). Additionally, high rainfall and high intensity lead to losses of fine particles (silt and clay), as described in Section 4.2. Combined with the weak nutrient incorporation by the biofertilizers, the result is very evident when compared to conventional management, which in this case applies biofertilizers plus chemical fertilizers (Table 1) at different stages of the cocoa production process. Therefore, the results suggest the adoption of conventional management, which in this case combines biofertilizers and formulated fertilizers. This combination (organic–chemical fertilization), according to Tripathi et al. [65] and Ning et al. [66], reduces toxic effects and improves pH levels over long periods. This latter assertion is supported by Andoh-Mensah et al. [55], who suggest combined organic and chemical fertilization as an integrated soil fertility management and nutrient transfer strategy can significantly improve soil chemical properties and cocoa production.

MLR modeling also suggests older plantations with higher planting densities, as they show significant positive causal relationships with pH, Ca²⁺, Mg²⁺, and CEC levels. The plantations are between 7 and 12 years old and have a planting density of 1250 to 4000 plants ha⁻¹. This result is reasonable because, with increasing age, the plantation not only reaches its maximum productive capacity but also stabilizes as an agroecosystem and incorporates more dry leaves from the forest species included in the system [6,9,67], as well as contributing leaves and other parts of the cacao plants. This is due to the greater number of plants, increased natural defoliation, and the resulting increase in biomass from pruning [68] and leaf fall, which contributes between 5 and 10 t ha⁻¹ year⁻¹ of biomass [69]. In both cases, light regulation and biomass incorporation occur [11,13,70,71], which directly influences recycling and nutrient levels. In the long term, they can increase soil carbon by up to 6% and nitrogen by up to 4% [61]. In addition, carbon incorporated through amendments based on production waste [72,73,74] provides nitrogen (N), phosphorus (P), and sulfur (S), as well as smaller amounts of calcium (Ca) and magnesium (Mg) [6,54,75]. These bases enrich the exchange complex [76] and, in acidic soils, displace Al³⁺. This neutralizes hydrogen ions, raises the pH, and increases nutrient availability [56,60,77], which is further enhanced by the combined application of chemical fertilizers and biofertilizers.

4.4. Cocoa Production

The results show the average yields (Figure 1) of the conventional (organic–chemical) monoculture plantations LE with 4667.03 kg ha⁻¹, Sa with 4406.71 kg ha⁻¹, and Sh with 3900.17 kg ha⁻¹, and the agroforestry–organic plantation Pa with 4639.27 kg ha⁻¹. The numbers of fruits and the fruit indexes are respectively 38.70 and 12.77 for LE, 36.50 and 13.12 for Sa, 28.80 and 14.12 for Sh, and 20.60 and 20.50 for Pa. The almond yield results are in the following order: LE > Pa > Sa > Sh, the number of fruits LE > Sa > Sh > Pa, and the fruit index Pa > Sh > Sa > LE. Additionally, it is necessary to point out that the planting density follows the order Pa > Sh> Sa and LE (4000, 1500, and 1250 plants ha⁻¹, respectively); all areas have an average higher than the Peruvian average of 820 kg ha⁻¹ and the averages reported for countries with the highest yields in the world (2700 and 3000 kg ha⁻¹) for Thailand and Guatemala, respectively [1,2,3,5], and are much higher than the world average, which in the last decade remained at 450 kg ha⁻¹ year⁻¹ [78].

From these overall results, we can highlight that the Pa (OAF) plantation has the highest planting density but the lowest number of fruits and the highest fruit index. The latter indicates smaller fruits, which influence yield [5,6], but this effect is offset by its high density. Conversely, LE (CM) shows the highest yields, the greatest number of fruits, and the lowest fruit index, indicating that it produces larger fruits. Although it has a lower density than the other areas, its yield was higher. This behavior—a greater number of fruits and a higher fruit index, resulting in smaller ears—was reported by Florida et al. [5] when applying NPK fertilizer. Therefore, the yields achieved in these plantations through the application of combined organic and chemical fertilization are an alternative that, according to current literature, confirms significant improvements in cocoa yields [55,61,76,79].

The yield results (4667.03 kg ha⁻¹) for LE exceed the average values (3500 kg ha⁻¹) for CMP-15 of 2.5 × 1 m of 4 years) reported by Florida et al. [6]. This result could be related to the higher proportion of amendments (island guano) and fertilizers (diammonium phosphate, magnocal, and ulexite) applied in the first application during the dormancy and flowering phases [6], compared to the other areas. This would have generated an adequate supply of nutrients, particularly nitrogen, phosphorus, and potassium [80,81], which are the most limiting elements and have a significant impact on cacao productivity [55]. Furthermore, the application of compost in Pa and of island guano in Sa, LE and Sh contributes positively to the incorporation of these key nutrients, which allow high levels of production to be achieved [82]; this also takes into account that the areas are between 7 and 12 years old, which is the age of maximum production, and that their productive useful life varies between 20 and 30 years [82,83].

Multiple linear regression modeling revealed four aspects that deserve to be highlighted, among them:

(a). 

The agroforestry system had a positive and significant effect on the number of fruits and dried almonds.

(b). 

Organic management alone was associated with a small increase in the number of fruits but with a slight decrease in dried almonds.

(c). 

The interaction between agroforestry and organic management was highly significant, reducing the fruit index but substantially increasing dried almonds.

(d). 

Planting density showed a positive effect on the fruit index and dried almonds.

The model’s findings suggest the development of cacao plantations in agroforestry systems with organic management and high planting density. This would result in a greater number of fruits, a lower fruit index, and a higher yield of dried beans, leading to high-yield plantations with a system and management practices that protect the agroecosystem.

In this context, at the local level, Florida et al. [5] evaluated the effects of compost, NPK fertilizer, and the combination of both, and found that purely organic management reduces the pod index, meaning larger pods and, therefore, higher yield. Conversely, chemical treatment produces a greater number of pods per plant, but smaller ones. Florida et al. [6] also studied the effect of agroforestry systems (OAF) on the production of CMP-15 cacao in the Pa area under the MOM management system and found yields of 2700 kg ha⁻¹ for CMP-15 planted at 2 × 2 m spacing and 4 years old, 3500 kg ha⁻¹ for CMP-15 planted at 2 × 1 m spacing and 5 years old, and 3300 kg ha⁻¹ for CMP-15 planted at 3 × 2 m spacing and 13 years old. Therefore, it is quite clear that increasing planting density is a viable alternative in cacao cultivation, since, according to its requirements, T. cacao is a species with slow nutrient absorption (Puentes-Páramo et al. [84]) and does not respond to excessive fertilization. If this limit is exceeded, it can respond negatively due to nutritional imbalances caused by stress [5,84,85], conditions that can arise in a chemical management system.

Finally, the average global cocoa yield is 485 kg/ha, while Peru reached an average of 840 kg ha⁻¹ in 2020 [47], and the Alto Huallaga Valley in the Huánuco region has an average yield of less than 412 kg ha⁻¹ [45,79]. Therefore, the results of Florida et al. [6] on organic management under an agroforestry system with yields between 2700 and 3500 kg ha⁻¹ for CMP-15 represented the first report on a high-yield management alternative. However, this research identified three new areas with similar management and production characteristics, corresponding to monoculture areas with conventional management, and numerically surpassed the report by Florida et al. [6]. Yields of 4667.03 kg ha⁻¹ were obtained for Sa (CCN-51 plantation, 12 years old, 2.70 × 2.70 m) and 4639.27 kg ha⁻¹ for Pa (CMP-15 plantation, 7 years old, 2.50 × 1.00 m). These results are significantly higher than the regional averages and exceed the average yield recorded in Guatemala of 3000 kg ha⁻¹ [86]. They are also much higher than those obtained with different fertilization schemes, including Álvarez-Carrillo et al. [87], who found averages of 2730 kg ha⁻¹ for conventional agriculture and 2521.43 kg ha⁻¹ for conventional agriculture. None of the references achieved the yields recorded in this study. Therefore, based on MLR modeling, it is recommended to adopt high-density agroforestry organic systems as a comprehensive alternative that optimizes land use, improves soil quality, increases crop yields, and promotes responsible production.

5. Conclusions

The evaluated soils exhibit physical and chemical qualities suitable for the requirements of T. cacao production and correspond to soils of the order Inceptisols, subgroup Typic Dystrudep for the LE and Sa plantations, and of the order Entisols, subgroup Typic Udifluvent for the Sh and Pa plantations.

The multiple linear regression (MLR) analysis found causal relationships that suggest the adoption of a production model based on the integration of agroforestry systems with organic–conventional management (combining biofertilizers and chemical fertilizers) in plantations with high planting densities (between 1250 and 4000 plants ha⁻¹). This model shows a positive effect on pH, Ca²⁺, Mg²⁺, and CEC levels; it prevents the loss of fine particles (silt and clay), decreases the fruit index, and improves dry bean yields.

The evaluated plantations show yields that far exceed the averages of the references consulted; therefore, they are production alternatives that through MLR analysis identifies causal relationships that respond to the information gap regarding comprehensive studies that quantify the benefits of different cocoa production models in the Alto Huallaga valley in Peru.