1 Introduction

Shade in agroforestry systems has been widely studied for perennial crops such as coffee and cacao due to shade’s potential to reduce the harmful effects of extreme climatic conditions (Niether et al. 2018). It still remains a controversial topic due to mixed results found throughout the literature. Several studies support the potential of shade to buffer extreme temperatures (Tscharntke et al. 2011; Blaser et al. 2018; Niether et al. 2020), as well as regulate the occurrence of pests and diseases in the system (Allinne et al. 2019; Bagny Beilhe et al. 2020). However, others have found that shade trees may indeed buffer temperatures, but also create stress due to resource competition (Abdulai et al. 2018). Therefore, further research is still required to fully understand the potential of shade in order to maximize its benefits in terms of climate buffering and pest and disease regulation, and reduce its negative impact in terms of production (Babin et al. 2010; Gidoin et al. 2014a).

Cacao trees characteristically have low light saturation and require a stable, warm, and humid climate, making them vulnerable to extreme climates (Suárez et al. 2021). Temperatures below 15 °C and above 34 °C have been reported to cause significant declines in photosynthesis rates and pod growth (Lahive et al. 2019). Microclimate conditions underneath shaded areas in agroforestry systems tend to meet cacao trees’ microclimatic requirements, rendering them suitable for cacao cultivation. Different cacao production systems are cultivated throughout the Peruvian Amazonia. These range from full sun systems where only cacao trees are planted to diversified agroforestry systems, where cacao trees are planted with a variety of timber and fruit trees (Tuesta Hidalgo et al. 2014). Full sun monocultures are often considered highly productive systems, though they tend to be unsustainable in the long term, whereas agroforestry systems tend to be less productive but highly sustainable, and some studies have found them to be as productive as full sun monocultures in the long term (Mortimer et al. 2018). Additionally, they provide additional benefits by creating ecosystem services and helping improve farmers’ livelihood through crop diversification (Cerda et al. 2020).

In cacao agroforestry systems, as depicted in Fig. 1, differentiation between the shade created by associated shade trees and that created by the cacao tree itself has not been studied. Shade from associated trees modifies the light (Gidoin et al. 2014a), throughfall (Niether et al. 2018), and airflow (Boudrot et al. 2016) conditions that cacao trees experience. These conditions created by the associated shade trees then further modify the pre-existing conditions below the cacao tree canopy, thus creating different microclimatic conditions than those created beneath the cacao self-shade under full sun. We hypothesize that by exploring the interaction between these two types of shade and the roles they play in modifying the microclimatic conditions underneath the cacao tree, we can comprehend how the total shade projected in the system affects the composition and frequency of pest and disease occurrence.

Fig. 1
figure 1

a Cacao full sun system with cacao trees projecting their own shade. b Cacao agroforestry system with associated shade trees projecting their shade over the cacao trees. c Diagram illustrating cacao production systems become heterogeneous when grown with associated shade trees, resulting in varying shade conditions within the same system. In full sun cacao production systems, the cacao trees themselves project shade, which can be referred to as the total shade; however, in cacao agroforestry systems, total shade encompasses the combined shade from the cacao trees’ self-shading and the shade projected by associated shade trees, referred to as the associated shade. With the presence of associated shade trees, the total shade becomes denser and alters the microclimate differently compared to cacao trees under full sun conditions. As a result, this variance in shade conditions creates different environments for the pests and diseases capable of affecting the cacao tree (depicted by the white rectangle and orange symbols) within the same production system.

Full size image

Most studies that relate shade to the microclimate tend to focus on temperature and relative humidity variables (Blaser et al. 2018; Lahive et al. 2019). This is also the case for studies that relate microclimate to the incidence of pests and diseases (Boudrot et al. 2016; Leandro Munoz et al. 2017; Fachin et al. 2019). Indicators that incorporate temperature and relative humidity have recently become more common for studying the relationship between shade and microclimate. One such indicator is the air vapor pressure deficit (VPD) (Jiménez-Pérez et al. 2019; Blaser-Hart et al. 2021). VPD provides a better understanding of the air’s capacity to retain water, which affects plant physiology. Notably, there is a negative correlation between stomatal conductance and VPD, and photosynthesis rates in cacao trees may decrease at VPD levels above 1.8 kPa (Lahive et al. 2019). At lower VPD levels, the formation of water films in the environment becomes more favorable. These water films are especially important for fungal spore colonization (Leandro Munoz et al. 2017). However, to our knowledge, VPD has rarely been linked to the incidence of pests and diseases in cacao, with only one study connecting VPD, pod wetness, and black pod rot disease back in the 1980s (Butler 1980).

Due to the physiological requirements of the cacao tree, shade plays an important role in the cacao tree yield, although the role of each type of shade has not yet been fully described. Yield is normally used as an indicator of productivity, and to establish its relationship to the types of shade, understanding the different types of yield is essential. Yield can be decomposed in potential yield, attainable yield, and actual yield. Potential yield refers to the yield amount that a cacao plant is capable of producing based on its genetic potential, without limitations imposed by water, nutrients, or biotic stress. However, it is not practical in real-world situations. Attainable yield pertains to the yield that farmers can achieve within their specific production circumstances, and it has not yet been influenced by yield-limiting factors such as water, light, and nutrient availability. The attainable yield is closely related to associated shade as it affects the light reaching the crop, thus influencing the photosynthesis rate. Actual yield refers to the yield that the farmers actually harvest after accounting for yield reducing factors (pests and diseases and climatic disasters). Actual yield is closely related to shade trees and the cacao self-shade for its effects on the development of pests and diseases (Savary et al. 2006; van Ittersum et al. 2013). The attainable yield and the actual yield provide more practical information; however, the attainable yield is rarely studied due to its measurement complexity, with no current studies conducted in cacao to date. This remains an important knowledge gap, which can be crucial for evaluating pest and disease regulation in the agroecosystem. Yield loss can be calculated by determining the difference between the attainable yield and actual yield. Yield loss has been proposed as an indicator to evaluate the pest and disease regulating service in agroecosystems, as it may serve as an output of the interaction network that reduces pest and disease occurrence in the agroecosystem (Avelino et al. 2018). Yield loss provides a potential means to assess the impact of external factors, such as shade and its influence over microclimate, over the pest and disease regulation. Understanding these diverse relationships among shade, microclimate, and pests and diseases is essential to grasping the conditions under which cacao pests and diseases are regulated. This comprehension is crucial for effectively managing these conditions to maximize yield.

In Latin America, three diseases are responsible for most of the production losses. The frosty pod rot (Moniliophthora roreri, hereafter FPR) can result in losses of up to 90% of the total production (Thevenin and Trocme 1996). If left unattended, black pod rot (Phytophthora palmivora, hereafter BPR) disease may also cause losses of up to 90% (Acebo-Guerrero et al. 2012). Witch’s broom (Moniliophthora perniciosa) losses range between 50 and 90% (Meinhardt et al. 2008). Moreover, a new threat has emerged in the Amazonia in the last decade, known as the American cocoa pod borer (hereafter APB), locally known as “Mazorquero” (Carmenta foraseminis (Busck) Eichlin). Although this pest directly damages the seeds to a lesser extent (between 11 and 24%), it inflicts a considerable amount of indirect damage by introducing secondary infections through the lesions caused by the nymphs emerging from the pod (Fachin et al. 2019).

This paper aims to address how the two types of shade influence the pest and disease regulation through the modification of the microclimate. We intend to answer this question by first establishing individual relationships between the two types of shade and the microclimatic, pests and diseases, and yield variables; subsequently, by employing a systemic approach to test the effect of the interaction network over the yield loss due to pests and diseases. Our objectives encompass the utilization of integrative indicators like VPD and yield loss. Furthermore, we aim to employ a specific methodology to calculate yield loss at distinct time points, thereby quantifying yield loss in terms of the number of seeds.

We hypothesize that the two proposed types of shade identified in cacao agroforestry systems in the Peruvian Amazonia have the potential to exert both positive and negative influence over the microclimatic conditions, as well as on the attainable yield and actual yield. This innovative approach, which considers the dual effects of shade on microclimatic conditions and yield, represents a novel avenue that has not been explored in previous research on the impacts of shade in agroforestry systems. Additionally, we postulate that these alterations in microclimatic conditions will alter the frequency and variability of the investigated pests and diseases. Finally, we propose that the interaction network involving shade types, microclimate, pests and diseases, and yield primarily accounts for agroecosystem yield loss, which simultaneously serves as a proxy for pest and disease regulation.

2 Materials and methods

2.1 Study area and experimental design

The study took place over the course of two periods: from October 2021 to November 2021 and from May 2022 to June 2022, within the Alto Huayabamba Valley, situated in the San Martin region of the Peruvian Amazonia (Huayabamba River: 7°15′05.67″ S, 76°55′45.68″ W) as illustrated in Fig. 2.

Fig. 2
figure 2

a Map of the study zone in the Alto Huayabamba Valley, San Martin region in Peru. Five communities distributed along the valley and eight plots distributed between the communities. Credits for San Martin region in Peru’s map (https://commons.wikimedia.org/wiki/File:San_Martin_in_Peru.svg). Alto Huayabamba Valley picture was generated in Google Earth Pro (version 7.3.6.9345 (64-bit) [year 2023], Huayabamba River: 7°15′05.67″ S, 76°55′45.68″ W, date accessed 02/08/2023). b Diagram of the spatial structure of the plots along with the distribution of the 40 monitored trees in black (A1 to A40) within the plot.

Full size image

This region features a humid tropical climate, as illustrated in Fig. 3, with a monthly average temperature of 25 °C. Rainfall occurs consistently throughout the year, peaking between February and April, tapering off in May and June, and rising again by late September. The estimated average rainfall for the years 2020 to 2022 was 1080 mm (Harris et al. 2020). The prevalent cacao variety cultivated in this zone is CCN-51, a highly productive and witch’s broom disease-resistant clone (Arévalo-Gardini et al. 2017). The landscape predominantly consists of lowlands, where small farms along the river cultivate cacao and plantain. On average, farmers tend to three hectares of CCN-51, yielding between 600 and 900 kg ha−1 (Tuesta Hidalgo et al. 2014); this figure can rise to as much as 2000 kg ha−1 in other Latin American countries (Jaimez et al. 2022). Cacao plots are largely managed using organic practices, encompassing organic fertilization, copper application, sanitary harvesting, and maintenance pruning. Pruning strategies vary according to the farmers’ resources, including cacao pruning practices, from a robust biannual pruning to maintenance pruning conducted three to four times annually. These pruning approaches hold particular significance, as changes to cacao tree branches have a direct impact on total shade measurements.

Fig. 3
figure 3

Precipitation and average temperature charts for the Alto Huayabamba Valley for years a 2020, b 2021, and c 2022 according to the calculations obtained through (Harris et al. 2020) model. SI in b and SII in c stand for monitoring seasons I and II, respectively.

Full size image

A total of 320 CCN-51 cacao trees were monitored, distributed in lots of 40 across eight plots within five communities in the Alto Huayabamba Valley. These trees were ranged between 15 and 20 years according to the owners of their respective plots. Cacao trees were planted with a 3 × 3 m planting distance. The monitored cacao trees were planted in lines of five leaving a single cacao tree between each monitored tree as a buffer in the line. There was also a line of buffer cacao trees between lines of monitored cacao trees (Fig. 2b). The aim was to gather data on microclimatic conditions of cacao trees, cacao self-shading, shade provided by associated trees, and incidence of pests and diseases. These plots varied in function of their cropping systems, ranging from cacao monocultures to agroforests with a mix of timber and fruiting tree species, albeit with low degrees of diversity. The range of cropping systems and the uniformity in cacao genetics and organic practices facilitated direct comparisons among shade conditions affecting cacao trees, while minimizing biases in yields and pest and disease incidence attributable to widely varying management scenarios.

2.2 Microclimate data collection and VPD estimation

Microclimate data were collected over the course of approximately 30 days during each of the two seasons: season I, spanning from October to November 2021 (corresponding to the end of the dry season), and season II, spanning from May to June 2022 (corresponding to the end of the rainy season). Measurements were conducted using calibrated mini data loggers (iButtons, Model DS1923, by Measurement Systems Ltd.), which utilize a digital thermometer and a hygrometer to capture air temperature and relative humidity values, respectively. Calibration was performed using data collected from a Campbell Weather Station (CR1000) under similar conditions, ensuring the reliability of the iButton data. Within each plot, a total of 32 and 28 trees were monitored by iButtons for seasons I and II, respectively, out of the 40 trees surveyed. As the aim was to compare cacao trees within a plot over the same period, all iButtons were simultaneously placed within each plot for a span of 48 h (excepting in two plots where data collection was reduced to 24-h periods). Temperature (in °C) and relative humidity readings were taken every 30 min. Each iButton was attached to a cacao tree branch at 1.70 m from the ground, avoiding branches exposed to direct sunlight, into a handcrafted shield that protected them from sun radiation and rainfall.

For each cacao tree, VPD was calculated through the difference in saturated vapor pressure and actual vapor pressure. These two were calculated through formulae that used the temperature and relative humidity that was collected over the 48-h periods (as depicted in Table 1), along with VPDmax, VPDmin, and VPDmean. VPD serves as an indicator of the air-drying capacity, indicating that at high VPD levels, the air is drier causing water stress to plants. Conversely, lower VPD levels suggest water saturated air, alleviating plant water stress. We selected VPDmax as an indicator as it better represents the extreme values at which plants experience more stress.

Table 1 Collected variables by category along with the collection and transformation methods.
Full size table

2.3 Shade characterization

Shade cover was assessed using hemispherical photography. Photos were captured using a GoPro HERO6 Black camera equipped with a 190° fish-eye lens. The camera was mounted on a gimbal stabilizer (Zhiyun Crane M2) with its fish-eye lens pointing upward. This setup enabled us to capture photos directly beneath the cacao tree canopies to obtain total shade measurements. Additionally, the camera and gimbal were attached to an extendable carbon fiber rod for capturing photos directly above the cacao tree canopy and beneath the canopies of associated trees canopies, allowing us to obtain associated shade measurements. To mitigate excessive sunlight glare in the photos, image acquisition took place during periods of diminished sunlight, such as before dawn, dusk, or under cloud cover. Photo analysis involved two distinct software applications: ImageJ, utilizing the Hemispherical 2.0 plugin (Beckschäfer 2015), and Gap Light Analyzer (Frazer et al. 1999). Both programs estimated canopy openness, from which we derived shade percentage values by using the inverse of the canopy openness percentage. The photos were analyzed using both programs, and a correlation coefficient of R = 0.994 was obtained when both results were compared. As both programs provided nearly identical results, most photos were analyzed with ImageJ, due to its automation capabilities. Gap Light Analyzer was employed as a supplementary tool to scrutinize images that ImageJ could not process with complete automation, necessitating some manual inputs. To prevent introducing bias through manual software inputs, the same individual operated both programs during the analysis.

The resulting shade percentages were further divided into intervals to assess their effects on our microclimate, yield, and pest and disease variables. For associated shade, four categories were set: “full sun” for shade percentages less than 1%; (1–29) for percentages greater than or equal to 1% but less than 30%; (30–59) for percentages between 30 and less than 60%; and (≥60), encompassing percentages of 60% or higher. These categories were chosen based on recommendations from the literature (Beer et al. 1998; Somarriba et al. 2018; Blaser-Hart et al. 2021). For total shade, we defined three categories: (30–59) for percentages between 30 and less than 60%; (60–80) for percentages between 60 and 80%; and (>80) for percentages above 80%. These categories were designed to encompass varied total shade conditions, allowing us to compare them with the associated shade categories and evaluate their impact on our selected variables.

2.4 Pest and disease characterization and yield loss calculation

Pests and diseases were monitored from 2019 to 2022 for each of the surveyed cacao trees; however, datasets closest to the months coinciding with our study period were extracted for subsequent data analysis (September 2021 and June 2022). Each extracted dataset contained data of approximately 1 month of pest and disease monitoring. The monitored diseases were black pod rot and frosty pod rot, and the pest was the American cocoa pod borer. Considering that these three pests and diseases are the most relevant in the area, the types of lesions and plant tissues they affect (all in the cacao pod) lead us to design the following monitoring strategy: (i) Cacao pod quantities per tree were accounted in terms of the pod’s phenology, considering four stages: cherelle, young pod, adult pod, and mature pod; (ii) Observations were performed every 15 days to inspect for diseased pods. Any discovered pods with disease symptoms were removed from the tree and tallied as affected pods within the corresponding phenological stage in which they were detected.

Incidence and yield loss were calculated for each disease for seasons I and II. At the tree level, incidence refers to the number of infected pods attributed to a specific pest or disease within the total pod count of a particular tree during the monitored timeframe. Yield loss pertains to the quantity of compromised and/or destroyed seeds per tree resulting from the presence of each specific pest, disease, or disease combination. To determine yield loss per tree, we established injury profiles (IPs) for each tree. These IPs denote specific combinations of pests and diseases that may manifest in a given cocoa tree. Seven profiles were formulated in accordance with the three monitored pests and diseases: IP1 = black pod rot; IP2 = frosty pod rot; IP3 = American cocoa pod borer; IP4 = black pod rot + American cocoa pod borer; IP5 = black pod rot + frosty pod rot; IP6 = frosty pod rot + American cocoa pod borer; and IP7 = black pod rot + frosty pod rot + American cocoa pod borer. Additionally, IP0 was introduced to denote trees unaffected by pests and diseases.

We estimated yield loss per IP through Eq. 1.

$$YL IPi=\sum\nolimits_{i=1}^{n}\left(APi\times 46\times {SDR}_{i}\right)$$
(1)

Equation 1 defines APi as affected pods of disease i; 46 represents the average number of seeds per pod; and SDRi signifies the seed damage ratio for disease i. Equation 1 illustrates the process of obtaining yield loss through the summation of the product of affected pods, average number of seeds, and the corresponding seed damage ratio associated to a specific injury profile. This calculation provides the yield loss specific to the injury profile affecting the tree. Seed damage ratios denote the proportion of affected and/or destroyed seeds found upon opening a pod afflicted by a specific pest or disease. These seed damage ratios serve as a link between the visible incidence and the real loss incurred due to pests and diseases. To obtain these ratios, 30 pods were collected and opened for each specific pest and disease. Pod selection adhered to two criteria: (i) presence of initial pest and disease symptoms and (ii) proximity of cacao pods to maturation (approximately 1 month before maturation). Pods were gathered during the farmers’ sanitary harvest practice, coinciding with the moment when farmers open pods and salvage viable seeds—thus accurately representing the actual yield. Seeds were categorized as healthy, affected, and destroyed. For each category, the seed count was divided by the total number of seeds within the opened pod. The coefficients derived from each opened pod were aggregated and then divided by the number of opened pods. Two sets of damage ratios per pest and disease were obtained: one for affected seeds and another for destroyed seeds, referred to as affected seed ratio and destroyed seed ratio, respectively. The seed damage ratios for each specific pest and disease are then derived by aggregating the sets of affected seed ratios and destroyed seed ratios corresponding to their specific pest and disease. The average number of seeds per pod was estimated to be 46 (45.39 ± 8.73) through the average of seed counts from 90 pods (pods infected by FPR were excluded from evaluation due to the fused seeds often caused by the disease, which hinders accurate seed counting.). The seed damage ratios were found to be 64.7% for the black pod, 91.54% for the frosty pod rot, and 50.43% for the American cocoa pod borer. These seed damage ratios and seed average per pod are highly specific to the region where we conducted the study and to the CCN-51 cacao variety in the zone. In addition, they apply under conditions of sanitary harvest for cocoa pods in their adult and mature stages. Having calculated the yield loss per tree and the average number of seeds per pod allowed us to calculate the attainable yield and actual yield.

2.5 Data analysis

For all statistical analyses, we included only those trees with available microclimate data and which had produced at least three cacao pods during either season I or II. This pod quantity equates to approximately 10% of the average cacao pod production per tree within our datasets. To understand the effects of the associated shade over the pest and disease frequency and variability across the seasons, we calculated frequency percentage of the injury profiles (as seen in Table 1). We estimated this percentage of injury profiles by season and by the previously described associated shade conditions. This frequency percentage provides information on the presence of a specific injury profile among the cacao trees in its associated shade condition for each season.

To examine the relationship between our response variable, VPDmax, and the predictor variables—season, associated shade conditions, and the injury profiles—we performed an analysis of variance (ANOVA) on generalized linear models (GLM). Moreover, to explore significant differences detected in the ANOVA for the significant factors, we conducted a post hoc Tukey HSD (honestly significant difference) test. ANOVA tests and GLMs were conducted using the base R package (version 4.1.0), while Tukey HSD tests were carried out with the “multcomp” R package (version 1.4-25) (Hothorn et al. 2008).

To assess the effects of the predictor variables, namely, associated and total shade conditions, on our response variables of yield (attainable yield, actual yield, and yield loss), we employed Kruskal-Wallis tests. In cases where the Kruskal-Wallis tests yielded significant results, we conducted post hoc Dunn’s multiple comparison tests under the Bonferroni family adjustment. Additionally, we performed Wilcoxon signed ranked tests between the attainable yield and actual yield for each associated and total shade condition. Both Kruskal-Wallis tests and Wilcoxon signed ranked tests were executed using the base R package (version 4.1.0), while the Dunn’s multiple comparison test was conducted using the “dunn.test” R package (version 1.3.5) (Dinno 2014).

We performed a structural equation model (SEM) using the “lavaan” R package (version 0.6.15) with the mixed linear model (MLM) estimator. The overall model to explain yield loss, along with its sub-models, was constructed as outlined in Table 2. These models were built based on preliminary findings and supported by relevant bibliographic references that elucidate the intricate relationship between yield loss concerning pests and diseases and the available resources, microclimate, and associated shade. Notably, actual yield was omitted from the model due to its computation as the difference between the attainable yield and the yield loss, which could introduce multicollinearity and redundancy into the global model. The entire calculation and data analysis were carried out using RStudio software (version 2022.12.0+353).

Table 2 Variables and sub-models utilized to construct to the global model. AttY, attainable yield; AS, associated shade percentages; TS, total shade percentages; S, seasons; VPDmax, maximum vapor pressure deficit; YL, yield loss.
Full size table

3 Results

3.1 Injury profile frequency and variability in function of the seasons and associated shade conditions

The frequency percentage and variability of injury profiles exhibit changes across seasons and in accordance with various associated shade conditions as depicted in Fig. 4. The frequency percentage values for IP0 generally showed higher values under all associated shade conditions during season I (average frequency of 57.46%) compared to season II (average frequency of 46.32%). For both seasons, the lower IP0 frequency percentage values were observed under 30% associated shade conditions (44.66% for season I and 42.10% for season II). Among trees exhibiting injury profiles related to pest and disease infections, the dominant injury profiles during season I varied with changes in associated shade conditions. Specifically, IP4 (black pod + American cocoa pod borer) displayed the highest frequency percentage value when associated shade conditions were lower than 30% (28.15%) and also under full sun conditions (14.58%). Conversely, when associated shade conditions ranged between 30 and less than 60%, IP2 (frosty pod rot) exhibited the highest frequency percentage value (20%). Moreover, when associated shade conditions reached 60% or more, IP3 (American cocoa pod borer) had the highest frequency percentage (28.57%). In contrast, during season II, IP2 consistently predominated, regardless of the associated shade conditions. As for the variability of injury profiles, during season I, the variability of injury profiles decreased as associated shade conditions increased. Notably, all injury profiles were present under full sun conditions, as opposed to only IP3’s presence when associated shade conditions were 60% or greater. In contrast, during season II, all injury profiles were observed under full sun conditions and when associated shade was less than 30%. The variability began to decrease beyond 30% associated shade conditions.

Fig. 4
figure 4

Injury profile frequency percentages are represented under the associated shade conditions examined. Injury profile frequency was examined by season, where a represents season I and b season II. IP0 in bluish green represents trees with no disease; IP1 to IP3 represent infected trees with only one pest or disease (IP1 = black pod rot (orange); IP2 = frosty pod rot (sky blue); IP3 = American cocoa pod borer (yellow)); IP4 to IP6 represent infected trees with a combination of two pests and diseases (IP4 = black pod rot + American cocoa pod borer (blue); IP5 = black pod rot + frosty pod rot (vermillion); IP6 = frosty pod rot + American cocoa pod borer (reddish purple)); and IP7 represents trees that are affected with all pests and diseases (IP7 = black pod rot + frosty pod rot + American cocoa pod borer (black)). Abbreviations for associated shade conditions: less than 1% associated shade or full sun (FS); more than 1% and less than 30% associated shade (1–29); between 30 and less than 60% associated shade (30–59); and 60% or more associated shade (≥60). Colors were selected according to Wong (2011) to provide accessibility for colorblind readers.

Full size image

3.2 Microclimate in function of season, associated shade conditions, and injury profiles

VPDmax is significantly influenced by the seasons (p < 0.001), with VPDmax being significantly higher during season I compared to season II, as depicted in Fig. 5a. Associated shade conditions also influence VPDmax (p < 0.001), since VPDmax was significantly higher under full sun conditions and 1–29 compared to 30–59 (Tukey HSD, p < 0.05) and ≥60 (Tukey HSD, p < 0.05) conditions. Similarly, VPDmax was significantly higher at 1–29 conditions compared to 30–59 (Tukey HSD, p < 0.001) and ≥60 (Tukey HSD, p < 0.05) conditions (Fig. 5b). Injury profiles appear to have no significant influence over the VPDmax (Fig. 5c).

Fig. 5
figure 5

Average VPDmax (mean ± standard error) per season (a), per associated shade conditions (b), and per injury profiles (c). The p values shown in each graph correspond to ANOVA tests, where a and b yielded significant results (p < 0.001). Letters above the graphs a and b indicate significant differences between averages (Tukey HSD, p < 0.001), no letters were added to graph c due to a lack of significance in the ANOVA and Tukey tests.

Full size image

3.3 Attainable, actual yield and yield loss in function of associated shade and total shade

Significant differences (Kruskal-Wallis, p < 0.001) were observed for attainable yield and actual yield across various associated shade conditions. However, no significant differences were identified for yield loss. Associated shade conditions exhibited a significant diminishing effect on both attainable yield and actual yield (Fig. 6a). Moreover, the attainable yield proved significantly higher (WMW, p = 0.041) than the actual yield under 1–29 associated shade conditions.

Fig. 6
figure 6

Bars represent the attainable yield (bluish green), actual yield (blue), and yield loss (vermillion) in number of seeds per associated shade conditions (a) and total shade conditions (b); the p values correspond to Kruskal-Wallis test for attainable yield (bluish green), actual yield (blue), and yield loss (vermillion). Letters above the bars in graph a represent significant differences in yield between the associated shade conditions (Dunn test). No letters were added above the bars in a and b when no significant differences were found in yields between the different associated or total shade conditions. Asterisk above letters in a indicate significant differences between the attainable yield and actual yield (Wilcoxon signed ranked test, p < 0.05).

Full size image

Attainable yield and actual yield exhibited a slight increase with ascending total shade conditions, up to 80% total shade conditions. Beyond this point, they experienced a decrease when exceeding 80% total shade conditions. However, despite these trends, no statistically significant differences were ascertained. Similarly, yield loss showed a slight increase as total shade conditions increased (Fig. 6b). Notably, akin to attainable yield and actual yield, no statistical significance was detected.

Figure 7a illustrates that under full sun conditions of associated shade, total shade conditions exert no significant influence on either attainable yield or actual yield. However, yield loss experiences a significant increase with escalating total shade conditions (Kruskal-Wallis, p < 0.05). At 1–29 associated shade conditions (Fig. 7b), attainable and actual yield significantly increase when total shade conditions increase. This is specifically evident between total shade conditions 30–59 and 60–80 (Dunn test, p < 0.001). At 30–59 associated shade conditions (Fig. 7c), there is no significant influence of total shade conditions on the number of seeds. Nonetheless, a tendency is observed for attainable yield, actual yield, and yield loss to rise as total shade conditions increase. Finally, at ≥60 associated shade conditions (Fig. 7d), no significant influences were found attainable yield, actual yield, and yield loss, though we see tendencies of the attainable yield and actual yield to decrease above 80% total shade conditions. We see the opposite with yield loss, as it tends to increase above 80% total shade. No significant differences were found either in between the attainable yield and actual yield under any combination of associated and total shade conditions.

Fig. 7
figure 7

Bars represent the attainable yield (bluish green), actual yield (blue), and yield loss (vermillion) in number of seeds by total shade conditions under different associated shade conditions. From left to right, associated shade conditions of full sun (a), 1–29 (b), 30–59 (c), and ≥60 (d) are represented. The p values in a and b correspond to Kruskal-Wallis test for attainable yield (bluish green), actual yield (blue), and yield loss (vermillion). Letters above the bars in graph a represent significant differences in yield between the associated shade conditions (Dunn test). No letters above the bars in a, c, and d due to no significant differences found in the Dunn test.

Full size image

3.4 Influence of season, microclimate, and types of shade over the pest and disease related yield loss

The SEM (p < 0.005) parameters of CFI = 0.977; TLI = 0.0909; SRMR = 0.031; and RMSEA = 0.102 collectively show that the proposed model is well suited to represent and explain the complex relationships within the dataset.

As depicted in Fig. 8, yield loss exhibited positive correlations with attainable yield (path coeff. = 0.58, p < 0.001) and associated shade (path coeff. = 0.13, p = 0.005); conversely, it displayed a negative correlation with VPDmax (path coeff. = −0.10, p < 0.05). Attainable yield was negatively correlated with associated shade (path coeff. = −0.38, p < 0.001) and positively correlated with total shade (path coeff. = 0.14, p < 0.05); no significant correlation was found with the seasons. VPDmax demonstrated a negative correlation with the seasons (path coeff. = −0.43, p < 0.001), closely followed by a negative correlation with total shade (path coeff. = −0.40, p < 0.001), and lastly, a negative correlation with associated shade (path coeff. = −0.22, p < 0.001).

Fig. 8
figure 8

Diagram representing all the significant interactions and the influence they exert over the VPDmax, attainable yield, and yield loss. Bluish green and vermillion arrows indicate positive and negative relationships, respectively. Numbers on the arrows indicate the path coefficient for each interaction.

Full size image

4 Discussion

As we intended to address how the two types of shade influenced the pest and disease regulation through the modification of the microclimate. As well as to quantify the yield loss in terms of seed numbers, we believe that our investigative approach enabled us to obtain specific results that addressed our hypotheses. These results clearly demonstrate the distinct positive and negative impacts of shade on all the microclimatic, yield, and pest and disease variables we examined. In the following sections, we will delve into the most pertinent findings.

4.1 Injury profile’s variability and frequency vary in function of the seasons and the associated shade

As hypothesized, the frequency and variability of the injury profiles for the different pests and diseases changed across the different seasons and associated shade conditions. We observed that under all associated shade conditions, the frequency of injury profiles indicating pests and diseases is lower during season I. Season I is characterized by lower humidity and higher temperatures, compared to season II, which presents higher humidity and lower temperatures. Our results also reveal that during season I, when humidity acts as a limiting factor, associated shade conditions significantly contribute to the variability of injury profiles indicating diseases. Distinct injury profiles become more prevalent under different shade conditions. For instance, under full sun conditions, all injury profiles are present, while at ≥60 associated shade conditions, only IP3 is observed. Throughout this season, the emergence of detrimental injury profiles seems to be constrained as shade intensifies, possibly due to the influence of shade on the microclimate (Blaser-Hart et al. 2021). This tendency, though present in season II as well, appears to be less pronounced. While we still observe lower injury profile variability with increasing shade conditions, the number of injury profiles per associated shade condition is greater compared to the same conditions during season I. Factors such as the diversity and density of associated shade trees may indeed play a pivotal role in this scenario, as studies have indicated that an augmented density of shade trees contributes to reduced airflow within the system (Niether et al. 2018), thus constraining the dispersal of fungal spores (Gidoin et al. 2014a). Additionally, evidence suggests that enhancing the diversity of shade trees introduces a broad spectrum of microbial diversity, including fungal endophytes, that act as antagonists to cacao fungal pathogens (Arnold et al. 2003; Mejía et al. 2008; Tscharntke et al. 2011).

Moreover, the physiological requisites of cacao pests and diseases add another layer of complexity. Both black pod rot and frosty pod rot, fungal diseases, thrive under conditions of high humidity and low temperature (Fulton 1989; Thevenin and Trocme 1996; Guest 2007; Meinhardt et al. 2008; Marelli et al. 2019). On the other hand, the American cocoa pod borer benefits from elevated temperatures and dry conditions (Fachin et al. 2019). Our findings also mirror this association between drier conditions and the APB, especially during season I. Remarkably, the presence of the APB, either as IP3 or IP4, persisted regardless of the associated shade conditions in season I. While not definitively proven, it seems that the presence of the APB might have inadvertently favored the emergence of BPR. IP4 was notably prevalent at lower associated shade conditions (full sun, 1–29, and 30–59), despite microclimatic conditions perhaps not being optimal for BPR dispersal and colonization during season I. A potential link between lesions induced by the APB and BPR colonization has been proposed by Fachin et al. (2019) and Alomía et al. (2021), providing a potential explanation for this outcome. In contrast during season II, fungal diseases become predominant as humidity is no longer a limiting factor, thus having presence of IP1 under all shade conditions, and IP2 becoming predominant at all levels.

4.2 Microclimate is altered in function of the seasons and the associated shade

A recent study, using machine learning to predict global warming trends, has estimated that we are rapidly approaching the critical threshold of 1.5 °C set by the Paris Agreement. Furthermore, it predicts that we could reach 2 °C within the next three decades (Diffenbaugh and Barnes 2023). Consequences of the increase in temperature could be severe, as shown by a study with predictions on VPD thresholds in coffee (Kath et al. 2022). Agroforestry has been strongly suggested as a potential mitigation strategy against these changes. Our findings align with previous research that highlights the capacity of shade to modify microclimates (Lin 2007; Blaser et al. 2018; Blaser-Hart et al. 2021; Merle et al. 2022), to the extent of manifesting the capacity of shade to buffer extreme climates (Niether et al. 2018, 2020; Lahive et al. 2019). While many studies focus on temperature and humidity measurements to test their hypotheses, only a few delve into utilizing VPD. Our results reveal the impact of seasonality in the region on VPDmax. This is obvious, as VPD integrates temperature and relative humidity measurements for its calculation, and the difference in terms of these measurements is well established between the two seasons (dry and rainy seasons). Furthermore, we found that VPDmax decreased as associated shade conditions increased. This could be explained through factors that are modified with the introduction of associated shade trees, such as a reduction in the light radiation that reaches the cacao trees, resulting in lower temperatures (Beer et al. 1998; Lin 2007; Tscharntke et al. 2011; Merle et al. 2022); diminished wind flow that hinders the dissipation of moisture, leading to elevated relative humidity over prolonged periods (Somarriba et al. 2018; Niether et al. 2018, 2020; Merle et al. 2022); and while not measured in our study, alterations at the soil level such as soil moisture and competition for water and nutrients with cacao trees. These alterations can subsequently affect cacao tree evapotranspiration (Abdulai et al. 2018; Blaser-Hart et al. 2021).

We observed no significant differences in VPDmax between full sun conditions and less than 30% associated shade conditions, which may be related to the spatial distribution of shade trees. For instance, shade trees could be clustered and unevenly distributed across the plot, leading to localized shade spots with lower VPDmax, and unshaded zones exhibiting higher VPDmax. When these values are averaged, they may not yield a significant effect. In contrast, having a homogeneous distribution may reduce VPDmax across the entire plot (Somarriba et al. 2018), possibly creating a noticeable difference if compared with full sun conditions when averaged. This may also help explain the high variability of injury profiles in both full sun and less than 30% associated shade conditions, as having conditions of both high VPDmax and lower VPDmax, provide conditions that are ideal for the cacao fungal diseases to develop different stages of their life cycle (Leandro Munoz et al. 2017; Abad-Sánchez et al. 2018; Fachin et al. 2019; Marelli et al. 2019). In contrast, we found significant differences in the VPDmax under 30–59 and ≥60 associated shade conditions when compared to full sun and less than 30% associated shade conditions. This outcome can likely be attributed to the factors previously discussed, such as reduced light radiation and decreased airflow. However, it could also explain the observed decrease in injury profile variability seen under higher associated shade conditions. As VPDmax remains lower and more stable due to increased shade, along with fewer unshaded spots for completing their life cycle and reduced airflow for spore transportation, the decrease in injury profile variability becomes evident. Furthermore, while there is a minor decline in VPDmax from 30–59 to ≥60 associated shade conditions, this change lacks significance. This observation suggests that it is after surpassing 30% associated shade conditions that the microclimate experiences significant modifications. Future research efforts should focus on this shade range to better comprehend the factors that become more influential to microclimate modification. Lastly, while we can see slight variations in VPDmax between the different injury profiles, we found no significant differences between them. This outcome indicates that VPDmax alone may not be the primary driver for the appearance and variability of injury profiles.

4.3 Shade management contributes to comparable attainable and actual yields

The presence of shade in cacao and coffee agroecosystems has undergone extensive investigation concerning its effects on yield. Results have been diverse, ranging from shade having a detrimental impact on yield to shaded systems demonstrating the potential to rival full sun systems in terms of yield at long terms (Beer et al. 1998; Somarriba et al. 2018; Mortimer et al. 2018; Cerda et al. 2020; Wainaina et al. 2021; Yamoah et al. 2021; Cocoletzi Vásquez et al. 2022). To date, no studies in cacao agroforestry systems have comprehensively examined the effects of associated and total shade on cacao yield, nor have they explored the resulting impact on yield loss due to pests and diseases. Our results align with studies that have shown a negative impact of shade over the attainable yield, and consequently on actual yield, as both decrease as associated shade conditions augment. However, we have also uncovered that attainable yield and actual yield do not significantly differ when associated shade conditions remain under 30%. When this threshold is exceeded, then the attainable yield and actual yield become significantly impacted. This similarity in attainable yield and actual yield between full sun and less than 30% associated shade conditions reinforces the notion that agroforestry system yields can indeed rival those of full sun systems when associated shade is judiciously managed. Moreover, we observed that yield loss due to pests and diseases is not significantly influenced by the associated shade conditions, which suggest that further factors need to be considered. In contrast, if we examine these same variables through the total shade, we find that attainable yield and actual yield maintain a notable consistency across all conditions. However, as total shade conditions escalate, we identify a significant augmentation in yield loss. This trend implies that it is the interplay between the inherent self-shading of cacao and the introduced associated shade that synergistically amplifies the yield penalties stemming from pests and diseases. Finally, by separating the cacao trees by their total shade conditions and, furthermore, by their associated shade conditions, we found that the proportion of associated shade within the total shade plays a role in increasing or decreasing the yield variables as well as the yield loss. For instance, in scenarios where total shade conditions increase under full sun conditions, yield loss experiences a noteworthy increase. This scenario reinforces the significance of total shade as an indicator of the cacao self-shading, signifying that inadequately pruned cacao trees that create higher total shade conditions (typically exceeding 60%) are more susceptible to disease. Under these conditions, the attainable yield and actual yield are not significantly impacted by the total shade conditions. In contrast, at less than 30% associated shade conditions, attainable yield and actual yield significantly increase as total shade conditions increase, while the yield loss remains similar across the different total shade conditions. This has two implications: first, due to the increased associated shade conditions, cacao trees can maintain more leaves, therefore increasing the total shade, as they experience less stress due to decreased sunlight (Mensah et al. 2023); and second, the increase in associated shade conditions allow for a more stable environment for the pests and diseases keeping the yield loss similar across all total shade conditions. Going further, at 30–59 associated shade conditions, we can see a considerable reduction in attainable yield and actual yield when total shade conditions are below 60–80 if compared to the previous associated shade conditions, while the yield loss remains stable as at less than 30% associated shade conditions. At this point, the cacao trees have an abundance of leaves and their attainable yield is still lower than under similar total shade conditions at full sun or less than 30% associated shade conditions. Finally, at ≥60 associated shade conditions, attainable yield and actual yield sharply decrease due to the excessive amount of associated shade and small amount of sunlight the cacao trees receive.

Seeing all the different trends exhibited by the attainable yield, actual yield, and yield loss across the different variations in associated shade and total shade suggest that understanding this dynamic is of key importance to develop productive shaded agroforestry systems. For instance, we found that under our research conditions, the cacao trees under total shade conditions 60–80 at full sun and less than 30% associated shade conditions provided the best conditions for cacao trees. Developing tools to properly measure the self-shading of the cacao trees may also prove useful in this endeavor.

4.4 Yield losses due to pests and diseases are not influenced only by the types of shade

Introducing shade into agroecosystems is often a controversial topic because of the duality in terms of benefits/drawbacks that it may cause (Guest 2007; Babin et al. 2010; Tscharntke et al. 2011; Ratnadass et al. 2012; Avelino et al. 2011; Blaser-Hart et al. 2021). Throughout this paper, we have explored the different effects shade has over the pests and diseases, microclimate, and yield separately. However, our model aimed to untangle the connections among the variables we observed and their collective influence on yield loss caused by pests and diseases. Our hypotheses held true, as we found that yield loss is influenced by these three variables, albeit with varying intensities. In addition, each of these variables interacts among themselves to result in positive or negative effects. Yield loss is affected mostly by the attainable yield. This suggests that the most important factor for pests and diseases, and consequently the yield loss, is related to the resource availability (Ratnadass et al. 2012; Gidoin et al. 2014b), and the fact that the pests and diseases attack and develop themselves in the cacao pods. The higher the attainable yield, the higher the yield loss due to pests and diseases. Additionally, the associated shade affects the yield loss through direct and indirect pathways. The direct relationship is positive, meaning that the yield loss increases as associated shade increases. This corresponds with our results, as at higher associated shade conditions, the injury profiles with frosty pod rot are more common and known to cause higher amounts of yield loss. Conversely, the indirect connections are negative, primarily channeled through attainable yield. This observation is in line with our results—higher levels of associated shade tend to reduce attainable yield. The second route is through the microclimate, measured through VPDmax. Notably, our model aligns with our findings, showcasing that higher associated shade levels correspond to reduced VPDmax values. Importantly, VPDmax wields the least impact on yield loss among the variables, exerting a negative influence—meaning that as VPDmax decreases, yield loss increases. This contrasts with our previous results, where VPDmax had no influence over the injury profiles. However, this might be explained by the fact that our model considered VPDmax as part of a dynamic interaction network. This emphasizes the significance of considering VPDmax when scrutinizing yield loss. VPDmax, in turn, is subject to the influence of various factors beyond associated shade, including seasons and total shade. Seasons exert substantial influence over VPDmax, with a marked decrease from season I to season II. Total shade also emerges as a pivotal regulator of VPDmax, mirroring associated shade patterns—higher total shade leads to lower VPDmax. This particular interaction between associated and total shades over VPD and its consequent relationship with yield loss has not been previously studied and it may have interesting ramifications in how associated and total shades affect yield loss. VPDmax holds particular importance in the physiology of mature cacao trees, negatively affecting their evapotranspiration and photosynthesis rates when excessively high (Suárez et al. 2021; Cocoletzi Vásquez et al. 2022). Optimal shading levels can mitigate VPDmax, maintaining it within tolerable ranges for cacao trees. Moreover, the VPD also governs the formation of moisture or dew on cacao pods and leaves (Butler 1980), which are crucial for spore colonization. Therefore, having excessive levels of shade may allow for VPD to decrease considerably and allow the formation of these water films, resulting in detrimental conditions for the cacao yields as it will stimulate the appearance of fungal diseases.

5 Conclusion

We showed that when researching the effects of shade on cacao growth, development, and production, it is important to consider and distinguish between cacao self-shade and shade cast on the cacao trees from associated shade. We found that each type of shade is linked to a specific trend, for example, associated shade had a negative influence over attainable yield, while total shade was positively correlated with attainable yield. This indicates that having a proper management of both types of shade to keep a balance is key to maximize the attainable yield and consequently the actual yield. By adding seasonality, the associated shade became a mitigating factor for the extreme vapor pressure deficit values during the end of the dry season. In addition, during this season, the associated shade limited the variety and frequency of the specific injury profiles for pests and diseases. These results corroborate our first two hypotheses by clearly establishing effects of the different types of shade over the microclimate and yield variables, as well as over the pest and disease regulation. Furthermore, it provides insights into how the different types of shade need to be managed differently across seasons. Finally, our most important result is that yield loss due to pests and diseases is not driven only by shade, but by a complex interaction network, in which shade is only one of the variables. This result corroborates our third hypothesis as while it was possible to explain the yield loss due to pests and diseases by examining each tested variable individually, it was only by examining them as a web through a systemic approach that we were able to better explain the drivers of yield loss. Having a better comprehension of these dynamics between shade, microclimate, yield, and pests and diseases is key to better understand the pest and disease regulation in the system and develop improved strategies for integrated pest management. We also believe that further work should focus on developing tools to properly measure the self-shade of the cacao trees that is contained in the total shade measurement. In fact, delving deeper into this shade dynamic may provide insights as to how properly balance their densities and maximizing the attainable yield while keeping the pests and diseases at manageable levels.