Contribution of Artificial Neural Networks (ANNs) in Analyzing and Modeling Phenological Synchronization of Fig and Caprifig in Northern Morocco

Abstract

The Mediterranean fig (Ficus carica L.) is a dioecious fruit tree of high nutritional and economic value in the Mediterranean basin. In northern Morocco, phenological desynchronization between male and female fig trees limits pollination and production. This study aimed to characterize the phenological stages of indigenous fig and caprifig varieties using the BBCH scale and to evaluate the predictive capacity of artificial neural networks (ANNs). This study was conducted in the Bni Ahmed region over two consecutive years (2021 and 2022) at two sites. At each site, a total of 80 female fig trees were selected. Caprifig trees were selected in accordance with their availability (37 trees/site 1; 24 trees/site 2). Local meteorological data were incorporated into the analysis to evaluate the influence of climatic conditions on phenological stages. Our results revealed significant effects of temperature, humidity, and rainfall on phenological dynamics, along with a clear inter-varietal variability and pronounced desynchronization between male and female fig trees. Early-ripening caprifig varieties showed limited pollination efficiency, whereas late-ripening varieties were better synchronized with the longer receptivity period of female fig trees. Importantly, the ANN model demonstrated exceptional predictive performance (R² up to 0.985, RMSE < 1 day), serving as a robust and practical tool for forecasting key phenological stages and minimizing potential yield losses. These findings demonstrate the value of combining phenological monitoring with AI-based modeling to improve adaptive management of fig orchards under Mediterranean climate change. This is the first study in Morocco to implement such an integrated approach to fig and caprifig trees.

1. Introduction

Morocco exhibits exceptional agro-biodiversity and notable intra-specific genetic diversity [1,2,3,4], shaped by contrasting environments from coastal plains to mountain areas. However, this richness is increasingly threatened by water scarcity and climate change, which accelerate genetic erosion and reduce the resilience of agroecosystems [5,6,7]. Perennial fruit crops are particularly affected, as climatic variability impacts their phenology, physiology, yield, and quality [8,9,10]. In this context, the common fig (Ficus carica L.) is no exception. It is highly vulnerable to climate change, which alters its cultivation and physiological processes, reducing agronomic performance [11]. Globally, these upheavals manifest as significant alterations in phenological cycles, impacting a succession of key seasonal events including flowering, fruit maturation, leaf senescence, and abscission [12].

In this respect, analyzing the phenological patterns of crops within well-defined environments is crucial for optimizing yield and selecting effective agronomic practices [13]. The phenology of fruit trees refers to the sequential changes that occur throughout the annual growth cycle. Systematic studies in perennial plants improve understanding of these stages and their relevance for crop production. This information supports planning of agricultural practices, including irrigation, fertilization, and pest and disease management. Such knowledge enhances input efficiency, stabilizes yields, and ensures optimal fruit quality [14]. Furthermore, detailed phenological data allow for assessment of the biological and economic value of each stage and serve as a key tool for identifying relevant phenophases and characterizing genotypes [15]. Various phenological scales have been devised to ensure accurate and standardized descriptions of plant development, including the BBCH scale (Biologische Bundesanstalt, Bundessortenamt, und Chemische Industrie). This scale offers a standardized coding system for both monocotyledonous and dicotyledonous plants, with adaptability to each species. The BBCH scale has successfully been implemented a diverse array of fruit crops, including cardamom (Elettaria cardamomum) [16], jamun (Syzygium cumini) [17], mangosteen (Garcinia mangostana) [18], cashew (Anacardium occidentale) [15], jackfruit (Artocarpus heterophyllus) [19], Indian gooseberry (Phyllanthus emblica L.) [20,21], jujube (Ziziphus jujuba) [22], custard apple (Annona squamosa L.) [23], sweet cherry (Prunus avium) [24], mango (Mangifera indica L.) [25], and kiwi (Actinidia deliciosa ‘Hayward’) [26]. Recently, this scale has also been used to accurately delineate the developmental stages of fig (Ficus carica L.), particularly in studies addressing its phenological behavior across varying agroclimatic settings [27,28]. The BBCH scale offers a clear advantage for fig trees by providing uniform codes for each developmental stage [29], facilitating comparisons across varieties and agroclimatic conditions [30,31,32], and improving the accuracy of phenological monitoring and modeling [33,34,35].

Concurrently, crop models rely on mathematical algorithms to simulate and quantify crop responses to environmental conditions, particularly climate [36]. These tools provide valuable knowledge for understanding and optimizing plant growth processes. Among the most promising approaches, artificial neural networks (ANNs) stand out for their capacity to model intricate, non-linear relationships, without requiring prior assumptions regarding the links between independent and dependent variables [37]. As a machine learning method, ANNs perform as well as conventional statistical models or traditional regressions. Their effectiveness has been validated across numerous fields, including agriculture, medical sciences, education, finance, engineering, security, and trade [38,39,40,41,42,43,44,45,46,47]. In agriculture, the modeling approach proves especially beneficial for evaluating the impacts of fluctuating environmental conditions on phenological patterns and plant physiological responses, especially for vulnerable species such as fig (Ficus carica L.). Thus, crop models enable the accurate simulation of phenological dynamics, growth, development, and agronomic performance as a function of different climatic variables [48,49,50]. However, most traditional models require prior assumptions regarding these relationships, whereas ANNs excel at capturing non-linear interactions without such constraints [51,52,53], which reinforces their relevance in our study. In recent years, ANNs have been increasingly applied to phenological prediction across diverse crops [54,55,56]. For example, they have been used to model budburst, flowering, and ripening dynamics in grapevine [57,58]; to predict anthesis and grain-filling stages in cereals such as wheat and maize under temperature and water stress [59]; and to simulate flowering and fruit development in perennial fruit trees such as avocado, dragon fruit, and mango [60]. These studies demonstrate the robustness of ANN-based models in forecasting crop phenology and highlight their capacity to capture complex climate–plant interactions [61]. Building on this evidence, the present work applies ANN to the fig–caprifig system, where phenological synchronization is crucial for ensuring pollination success and sustainable yields.

The common fig (F. carica L., 2n = 26) is an emblematic species of Mediterranean agroecosystems, playing a central role in traditional agricultural practices [62,63]. Belonging to the Moraceae family, the species was among the earliest domesticated fruit crops [64,65,66]. F. carica L. is a gynodioecious species comprising male trees (caprifigs) and female trees, with the presence of hermaphroditic flowers in male trees justifying its gynodioecious classification [67,68]. Four main types of figs are recognized according to their pollination requirements: the common fig (parthenocarpic), Smyrna and San Pedro types (requiring pollination), and the caprifig, which provides pollen [69,70,71]. Caprifig trees produce three annual fruit cycles (profichi, mammoni, mamme), with profichi being the most important for caprification, a process mediated by Blastophaga psenes L. and essential for achieving optimum yields in Smyrna and San Pedro types [72,73,74].

In Morocco, more specifically in the Rif mountains, fig trees represent a crucial source of income for local farmers [75]. Despite their essential role, caprifig trees remain insufficiently studied [74,76,77]. In this context, the village of Talandaoued (Bni Ahmed Charqia commune, Chefchaouen province) is a representative example of agroecosystems where fig and olive trees fulfill vital economic and food functions [78]. This locality is distinguished by notable varietal diversity, featuring over 15 fig varieties and four types of caprifig used for pollination [74,77,79]. Caprification, a key practice in production areas, is traditionally implemented to ensure fruiting and enhance the quality of edible figs [80,81,82]. Nevertheless, despite the good pollination capacity exhibited by specific varieties such as L’morr or L’louizi [77], farmers are facing a significant decline in production, mainly due to a desynchronization between the female fig trees’ receptivity period and the emergence of the pollinator (B. psenes) [74,77]. This context highlights the urgent need for strategies to preserve the reproductive balance of fig trees. Accordingly, the present study pursues two main objectives: (1) to characterize the phenological stages of male and female fig trees in a semi-arid region of northern Morocco using the BBCH scale, and (2) to assess the predictive capacity of artificial neural networks (ANNs) for forecasting these stages.

2. Materials and Methods

2.1. Test Site

This study was conducted over two consecutive seasons (2021–2022) at two locations: Site 1 (Anassel) and Site 2 (L’marj L’bardi), situated southwest of the village of Bni Ahmed, in the province of Chefchaouen. The two sites are separated from each other by a hill at an altitude of 668 m. Site 1 is 455 m above sea level, while site 2 is 521 m above sea level (Figure 1). During the observation period, the mean annual temperature was recorded at 18.8 °C, whereas the precipitation ranged from 742 to 755 mm. The region exhibits high thermal amplitude, with recorded temperatures reaching a maximum of 47.6 °C and a minimum of 1.1 °C. The orchard soils, composed mainly of clay and shale, have an aqueous pH of 8.6.

2.2. Microclimatic Data Collection

The region’s climate is classified as Mediterranean according to the Köppen–Geiger system [83]. During the years 2021 and 2022, microclimatic data were collected in situ at both experimental sites (Anassel and L’marj L’bardi) using DALLAS iButton DS1990A-F5 sensors (Analog Devices Inc. (Wilmington, NC, USA)), with three stations per site. These sensors enabled continuous monitoring of air temperature and relative humidity at an hourly resolution throughout the study period (Figure 2). In addition, precipitation data were obtained from the nearest official meteorological station of the Moroccan National Meteorology Directorate, located in Chefchaouen, which provides daily rainfall records. This combination of local sensor data and official precipitation records ensured an accurate characterization of the climatic conditions in the study area.

2.3. Phenological Observations

This study involved eight receptive female fig varieties, namely, L’h’archi, L’messari, Aïcha, L’hamri, L’qellal, L’mdar, L’qouti, and L’âassal, as well as three caprifig (male fig) varieties, comprising L’hlou, L’morr, and L’louizi. At each site, a total of 80 female trees (10 trees per variety) were selected for consistent monitoring. The selection of caprifig trees was chosen based on their availability at both sites, totaling 37 trees at the Anassel site, distributed as follows: 19 for L’morr, 10 for L’hlou, and 8 for L’louizi. In the L’marj L’bardi site, 24 trees were selected, with a distribution of 14 for L’morr, 7 for L’hlou, and 3 for L’louizi. To ensure a homogeneous sample, the number of monitored trees was determined by both practical and methodological considerations, and trees were selected according to age and vigor.

For each site, five branches were marked on each tree, positioned according to the four cardinal points (north, south, east, west) and a central branch. Observations on the marked branches were carried out weekly and biweekly during winter, over two consecutive years (2021 and 2022) at both sites. Phenological stages were recorded using the BBCH scale. These stages were assessed through flowering participatory observations involving all studied varieties, with particular emphasis on the leafing, flowering, fruiting, and senescence stages of both fig and caprifig trees. The duration of each phenological stage was recorded in days. The observed stages and their corresponding BBCH codes are summarized in

Table 1. Correspondence between observed phenological stages and BBCH codes for fig and caprifig.

BBCH Code	Description (Based on BBCH Scale for F. carica)
01	Beginning of bud swelling (leaf buds)
11	Beginning of leaf development
15	Mid leaf set (leaves ≈ 50% of final size)
51	Emergence of reproductive bud (fig flower bud; birth of profichi, mamme, mammoni for caprifig)
53	Beginning of syconium elongation (autumn fig bud)
65	Peak flowering (receptive period for autumn figs; start of caprification period)
87	Fig fruit ripe for picking (autumn figs ripe for eating)
89	End of mammoni and mamme (over-mature, senescent)
91	Beginning of senescence (first leaves change color)
93	End of senescence (≥90% leaves fallen)

.

2.4. Phenological Model

In order to predict the phenological stages of fig trees, an artificial neural network (ANN) modeling approach was adopted. A multilayer feedforward neural network (MLP) was implemented and trained using R software (version 4.3.1). Before training, the dataset was carefully preprocessed: the variables were screened for outliers using exploratory data analysis. The dataset was then randomly partitioned into three subsets, 70% for training, 15% for testing, and 15% for validation, using all available observations across both study years and sites. This random allocation ensured that each subset contained a representative distribution of data, thereby avoiding bias due to temporal (year) or spatial (site) separation. To ensure stable convergence of the network, all input variables (mean temperature, relative humidity, and precipitation) and the output variable (phenological stage duration in days) were normalized to a [0–1] scale using min–max transformation. Several ANN architectures were tested in preliminary runs, with hidden neuron numbers ranging from 3 to 10. The final chosen structure (4–7–1) (Figure 3) minimized both the root mean square error (RMSE) and the mean absolute percentage error (MAPE) on the test and validation sets, and thus provided the best trade-off between predictive accuracy and generalization. All factors were considered fixed because they represent specific experimental conditions rather than random samples. To prevent overfitting, the model was evaluated systematically across the three subsets, and early stopping criteria were applied by monitoring validation error during training. Model performance was assessed using statistical measures such as coefficient of determination (R²), root mean square error (RMSE), and mean absolute percentage error (MAPE), calculated according to established equations [84,85,86,87,88]. This approach made it possible to quantify the model’s accuracy and robustness in predicting phenological stages.

$$\mathrm{MAPE}={\displaystyle \frac{1}{n}}\times {\sum }_{i=1}^n({\displaystyle \frac{\left|(P_i-A_i)\right|}{A_i}}\times 100)$$

(1)

$$\mathrm{RMSE}=\sqrt{{\displaystyle \frac{1}{n}}{\sum }_i^n{\left(P_i-A_i\right)}^2}$$

(2)

$${\mathrm{R}}^2=1-\sqrt{{\displaystyle \frac{\sum _{i=1}^n{(P_i-A_i)}^2}{\sum _{i=1}^n{A_i}^2}}}$$

(3)

where P_i and A_i denote the simulated (predicted) and observed values respectively, while n corresponds to the total number of training vectors for the i-th training vector.

2.5. Statistical Analysis

Statistical analysis was performed to assess the impact of climatic changes (temperature, relative humidity, precipitation, and orientation) and study years (2021 and 2022) on the phenological stages of fig and caprifig. A three-factor ANOVA model was applied with variety, year, and branch orientation as fixed factors, including their interactions. The assumptions of the ANOVA were checked before analysis: residual normality was verified using the Shapiro–Wilk test, and homogeneity of variances was assessed using Levene’s test. Only when these assumptions were met were the results interpreted at a significance threshold of p < 0.05. Simple non-parametric correlations were executed to analyze relationships between phenological stages and environmental factors, for both fig tree receptivity and profichi ripening periods. Furthermore, Gantt charts were generated to visualize the seasonality of phenological stages and assess their synchronization between fig and caprifig. All statistical analyses were performed in R (version 4.3.1). The artificial neural network model was implemented using the “neuralnet” package (version 1.44.2). Data manipulation and preprocessing were supported by the packages “plyr” (version 1.8.9) and “MASS” (version 7.3-65), while graphical visualizations were produced using “ggplot2” (version 3.5.2) and “ggpmisc” (version 0.6.2).

3. Results

3.1. Microclimatic Data

Analysis of microclimatic data collected over two consecutive years (2021 and 2022) indicates notable discrepancies between the two sites (Figure 4), mainly in terms of temperature and humidity. The L’marj L’bardi site recorded consistently higher mean annual temperatures, with deviations of 1.0 °C in 2021 and 1.5 °C in 2022 during peak months. Episodes of temperatures exceeding 30 °C and 40 °C were more prevalent in L’marj L’bardi, particularly in July, with a 13% rise in intervals exceeding 40 °C recorded in 2021. Relative humidity trends were similar across both sites, though L’marj L’bardi displayed marginally higher levels, especially in winter and spring. Conversely, precipitation patterns were comparable at both locations, showing no significant discrepancies. The observed inter-site microclimatic variations over the two years underlined the necessity of considering local conditions in the adaptation of farming practices, particularly regarding temperature and humidity variations.

3.2. Phenological Stages

3.2.1. Phenology of Fig Tree Varieties

Phenology of Leafing out and Senescence

An analysis of phenological data pertaining to the leafing phase across both study sites demonstrated substantial inter-annual fluctuations (Figure 5). In 2021, the onset of budburst among the various fig varieties occurred between late February and early March, with onset ranging from 28 February (L’qouti) to 10 March (L’mdar). At the Anassel site, the duration of the foliage stage varied among varieties, from 227 days (L’mdar) to 244 days (L’h’archi), with an average of 234.88 ± 5.69 days. Defoliation began on 17 October for L’qouti and extended to 5 November for L’hamri. In 2022, the leafing period ranged from 244 days (L’qouti) to 258 days (L’âassal), with a mean of 249.63 ± 4.96 days. Onset occurred between 4 February (L’âassal) and 13 February (L’hamri), while the end of the cycle was recorded between 10 October (L’qouti) and 23 October (L’hamri).

The L’marj L’Bardi site displayed analogous dynamics, with slightly shorter leafing periods. In 2021, the duration ranged from 217 days (L’mdar) to 230 days (L’h’archi), averaging 222.75 ± 4.46 days. Initial leaf emergence was noted on 24 February for L’qouti and on 4 March for L’hamri. Defoliation occurred between 6 October (L’qouti) and 22 October (L’hamri). In 2022, the leafing period ranged from 241 days (L’qouti) to 256 days (L’âassal), averaging 247.50 ± 5.26 days. Budburst began on 2 February (L’âassal) and 10 February (L’hamri). The phase ended between 2 October (L’qouti) and 18 October (L’hamri).

An examination of the senescence stage showed differences between sites and years (Figure 5). In 2021 at Anassel, senescence duration ranged from 43 days (L’qellal) to 54 days (L’qouti), with a mean of 48.25 ± 4.46 days. The stage began between 17 October (L’qouti) and 5 November (L’hamri) and ended between 10 December (L’qouti) and 20 December (L’hamri). In 2022, the average duration was 36.88 ± 6.27 days, ranging from 24 days (L’âassal) to 43 days (L’hamri). Onset dates varied between 7 October (L’qouti) and 23 October (L’hamri), with end dates between 13 November (L’âassal) and 5 December (L’hamri).

At L’marj L’Bardi in 2021, the mean senescence duration was 49.13 ± 3.27 days, ranging from 44 days (L’âassal) to 53 days (L’qouti). Senescence started between 6 October (L’qouti) and 22 October (L’hamri) and ended between 28 November (L’qouti) and 8 December (L’hamri). In 2022, the average duration was 32.25 ± 5.31 days, with the shortest period observed in L’âassal (21 days) and the longest in L’qellal (38 days). Onset occurred between 2 October (L’qouti) and 18 October (L’hamri), with end dates between 7 November (L’qouti) and 23 November (L’hamri).

Fig Flower Phenology

The fruiting phase of fig flowers on female fig trees at the two observation sites exhibited inter-annual fluctuations between 2021 and 2022 (Figure 6). In 2021, the mean duration of this phase was shorter, at (58.57 ± 6.90 days) in Anassel and (60 ± 6.06 days) in L’marj L’bardi. Within Anassel, the L’âassal variety showed the shortest durations at (48 days), whereas the L’hamri variety showed the longest durations at (69 days). In 2022, a notable prolongation of the fig-flower phase was recorded at both sites, with an average duration of (104.29 ± 5.74 days) at Anassel and (100.57 ± 2.70 days) at L’marj L’bardi. This increase was evident in minimum periods of 95 days (Anassel) for the variety L’messari and 97 days (L’marj L’bardi) for L’messari and L’mdar, while maximum periods peaked at (111 days) and (104 days), respectively, for L’h’archi and L’hamri. Among the examined varieties, L’h’archi exhibited the longest fig-flowering period in 2022 at Anassel (111 days), closely trailed by L’âassal (109 days), while L’hamri and L’qouti shared an identical duration of (107 days). At L’marj L’bardi, L’hamri presented the longest duration of (104 days), succeeded by L’qouti with (103 days).

Autumn Fig Phenology

As expected, following the analysis, observations of the autumn fig stage across various female fig varieties at the two study sites elucidated pronounced inter-annual variations between 2021 and 2022, as well as between varieties (Figure 7). In 2021, the mean duration of this stage was longer, averaging of (93.38 ± 4.10 days) at Anassel and (89.25 ± 6.54 days) at L’marj L’bardi. At Anassel, the shortest durations were noted for the L’qouti variety (86 days), whereas L’messari recorded the longest at (99 days). In L’marj L’bardi, L’qouti revealed the shortest duration (80 days), whereas L’messari extended to 98 days. In 2022, a reduction in cycle length was observed at both sites, averaging (70.25 ± 5.34 days) at Anassel and (68.50 ± 5.53 days) at L’marj L’bardi. This was reflected in minimum durations of 63 days (Anassel) and 61 days (L’marj L’bardi), while maximum durations reached (79 days) and (78 days), respectively. Among the varieties investigated, L’hamri exhibited the longest duration in 2022 at Anassel (79 days), followed by L’h’archi and L’âassal (74 days each), while at L’marj L’bardi, L’hamri also achieved the longest duration (78 days), succeeded by L’h’archi (73 days).

Analysis of the harvesting period for female fig varieties at the two study sites unveiled considerable inter-annual (2021 and 2022) and inter-site (Anassel and L’marj L’bardi) variability. In 2021, the average harvest period was shorter, averaging (27.63 ± 2.77 days) at Anassel and (26.25 ± 1.67 days) at L’marj L’bardi. Within Anassel, the L’qellal variety had the shortest harvest duration, at (23 days), whereas varieties such as L’mdar, L’h’archi, L’hamri, and Aïcha extended to 30 days each. Conversely, in L’marj L’bardi, the variety L’qellal had the shortest duration (23 days), while the varieties L’messari, L’mdar, L’hamri, and Aïcha reached longer harvest durations of up to 28 days. In 2022 the harvest period significantly lengthened at both sites, averaging (32.88 ± 4.52 days) at Anassel and (32.50 ± 6.05 days) at L’marj L’bardi. At Anassel in 2022, Aïcha recorded the most extended harvest period (40 days), followed by L’qouti and L’messari (37 days each). Conversely, at L’marj L’bardi, L’messari had the longest period (41 days), followed by Aïcha (39 days).

3.2.2. Phenology of Caprifig Varieties

Phenology of Leafing out and Senescence

Analysis of the foliation dynamics of caprifig varieties at the Anassel and L’marj L’bardi sites uncovered marked inter-annual variability between 2021 and 2022 (Figure 8). In 2021, the mean foliation duration was shorter, measuring (239.00 ± 2.00 days) at Anassel and (227.33 ± 1.53 days) at L’marj L’bardi. The L’hlou variety exhibited the shortest durations (237 days) at Anassel and (226 days) at L’marj L’bardi, while L’morr displayed the longest durations (241 days) at Anassel and (229 days) at L’marj L’bardi. In 2022, a pronounced extension of the foliation period was recorded at both sites, averaging (252.00 ± 2.65 days) at Anassel and (248.67 ± 2.08 days) at L’marj L’bardi. The shortest foliation periods were observed in the L’louizi variety—250 days at Anassel and 247 days at L’marj L’bardi—whereas L’morr had the longest durations, reaching 255 days and 251 days, respectively.

Analysis of the senescence stage across caprifig varieties at both sites highlighted inter-annual variations between 2021 and 2022, along with site-specific variability. In 2021, the average duration of senescence was (55.67 ± 5.13 days) at Anassel and (50.00 ± 4.58 days) at L’marj L’bardi. L’morr demonstrated the shortest duration at Anassel (50 days) and L’marj L’bardi (45 days), whereas L’hlou recorded the longest duration at Anassel (60 days) and L’marj L’bardi (54 days). In 2022, the senescence period showed a marginal extension at both sites, with an average duration of (59.33 ± 8.08 days) at Anassel and (52.67 ± 5.86 days) at L’marj L’bardi. The L’morr variety consistently exhibited the shortest senescence period at L’marj L’bardi (46 days), while L’hlou displayed the maximum prolonged duration reaching 64 days at Anassel and 57 days at L’marj L’bardi.

Fruiting Phenology

Analysis of the profichi period across the two studied sites, Anassel and L’marj L’bardi, indicated a general extension in 2022 relative to 2021 (Figure 9). In 2021, the mean profichi period was shorter, averaging (100.67 ± 8.74 days) at Anassel and (86.00 ± 4.36 days) at L’marj L’bardi. Within Anassel, the shortest profichi duration was noted in the L’hlou variety (91 days), while L’morr exhibited the longest (108 days). Likewise, in L’marj L’bardi, L’hlou showed the shortest period (83 days), and L’morr displayed the longest (91 days). In 2022, the profichi phase extended substantially at both sites, with an average reaching (137.33 ± 0.58 days) at Anassel and (136.00 ± 1.00 days) at L’marj L’bardi. Among the examined varieties, L’hlou exhibited the longest duration at Anassel (138 days). Conversely, L’morr and L’louizi recorded equivalent durations (137 days). In L’marj L’bardi, durations were sequentially distributed across varieties: L’morr (135 days), L’hlou (136 days), and L’louizi (137 days).

Analysis of the mammoni period across the two examined sites, Anassel and L’marj L’bardi, revealed variations between the years 2021 and 2022 (Figure 9). In 2021, the mammoni period averaged longer durations, with means of (142.00 ± 10.58 days) at Anassel and (132.00 ± 11.36 days) at L’marj L’bardi. At Anassel, L’hlou experienced the longest duration (154 days), while L’louizi recorded the shortest (134 days). Similarly, in L’marj L’bardi, L’hlou persisted as the variety with the longest period (145 days), and L’louizi maintained the shortest period (124 days). In 2022, a marginal reduction in mammoni duration was observed at both sites, with means of (132.00 ± 9.64 days) at Anassel and (127.33 ± 4.93 days) at L’marj L’bardi. Among the three studied varieties, L’hlou recorded the longest duration at Anassel (143 days), while L’louizi recorded the shortest (125 days). At L’marj L’bardi, L’hlou also showed the longest period (133 days), and L’louizi the shortest (124 days).

Regarding the analysis of the mamme period, substantial variations were identified between 2021 and 2022 (Figure 9). In 2021, at the Anassel site, the average duration of mamme across the studied varieties was (169.33 ± 1.53 days), with a minimum duration of 168 days (L’morr) and a maximum duration of 171 days (L’hlou). In 2022, this average duration extended to (184.33 ± 6.43 days), with minimum and maximum durations recorded at 177 days (L’hlou) and 189 days (L’louizi), respectively. This upward shift reflects a noticeable extension of the mamme phase in 2022 relative to 2021. In 2021, the mamme period at L’marj L’bardi was markedly prolonged, with an average of (209.67 ± 1.15 days), with minimum durations of 209 days (L’morr and L’louizi) and a maximum of 211 days (L’hlou). By 2022, this average duration declined to (170.33 ± 5.69 days), with minimum and maximum durations of 164 days (L’hlou) and 175 days (L’louizi), respectively.

3.2.3. Female Fig Tree Receptivity Period and Profichi Maturation

The analysis of caprification and receptivity periods at Anassel and L’marj L’bardi revealed significant inter-annual and inter-varietal variations (Figure 10 and Figure 11,

Table 2. Comparative phenological characteristics and synchronization efficiency of fig and caprifig varieties.

Variety	Type	Mean Receptivity Duration (Days)	Receptivity Period (Observed Range)	Main Synchronized Caprifig(s)	Synchronization Efficiency
L’hlou	Caprifig	14–16	20 May–23 June	L’qouti, L’messari	Good in 2021, partial in 2022 (shift)
L’morr	Caprifig	12–25	28 May–28 June	L’mdar, L’qellal, L’âassal, Aïcha	Good in 2021, partial in 2022
L’louizi	Caprifig	12–21	6 June–July 2	L’âassal, L’qellal, L’hamri, Aïcha	Good in 2021, partial in 2022
L’qouti	Female	15–28	15 May–14 June	L’hlou	Strong overlap in 2021, partial in 2022
L’messari	Female	19–25	20 May–17 June	L’hlou	Good in 2021, partial in 2022
Aïcha	Female	18–40	13 May–28 June	L’hlou, L’morr, L’louizi	Variable: good with L’louizi, partial with L’hlou and L’morr
L’âassal	Female	19–34	15 May–23 June	L’morr, L’louizi	Good in 2021, partial in 2022
L’qellal	Female	19–35	14 May–24 June	L’morr, L’louizi (secondary)	Moderate synchronization
L’mdar	Female	18–33	16 May–21 June	L’morr (main)	Satisfactory synchronization
L’h’archi	Female	18–36	15 May–26 June	Broad overlap	Good but variable
L’hamri	Female	18–33	18 May–28 June	L’louizi, L’morr (late)	Good in 2021, partial in 2022

). In 2021, the mean caprification duration reached (20.7 ± 4.5 days) at Anassel and (17.3 ± 4.5 days) at L’marj L’bardi, ranging from 13 to 25 days depending on the variety. In 2022, this duration declined markedly to (13.7 ± 1.2 and 12.3 ± 0.6 days), respectively. Female receptivity periods followed the same trend: in 2021 they lasted 31.4 ± 3.6 days (25–35) at Anassel and 29.5 ± 6.5 days (19–40) at L’marj L’bardi, but in 2022 decreased to 19.1 ± 1.3 and 18.0 ± 1.3 days.

Synchronization efficiency varied according to variety and year. In 2021, early female figs such as L’qouti and L’messari showed strong overlap with L’hlou caprifig, while mid- and late-season varieties (Aïcha, L’hamri) aligned better with L’morr and L’louizi. In 2022, the shortening of caprification phases led to partial overlaps with early figs, whereas late varieties maintained or improved synchronization due to their extended receptivity. Overall, results highlight a progressive desynchronization between caprifig pollen release and female receptivity, reducing pollination efficiency under shorter phenological phases. The comparative patterns are synthesized in Table 2, while Figure 10 and Figure 11 visually illustrate the contraction of overlap windows between 2021 and 2022

3.3. Correlation Analysis

Analysis of the correlation coefficients highlighted the influence of climatic factors on the various phenological phases of both fig and caprifig trees. Within fig (Figure 12B), foliation exhibited a strong negative correlation with relative humidity (r = −0.614) and temperature (r = −0.782), while it showed a positive association with rainfall (r = 0.814), indicating that higher rainfall levels promote foliage development. Flowering figs showed a weak correlation with humidity (r = 0.160) and a negative correlation with temperature (r = −0.441), while being markedly affected by rainfall (r = 0.979). The appearance of autumn figs correlated positively with both humidity (r = 0.772) and precipitation (r = 0.868), while displaying a negative association with temperature (r = −0.559). Likewise, the receptivity period of female fig trees showed a positive correlation with humidity (r = 0.534) and rainfall (r = 0.901), while being strongly negatively correlated with temperature (r = −0.803).

Finally, fig tree senescence appears to be significantly influenced by humidity (r = 0.859) and rainfall (r = 0.873), with high temperatures exerting an inhibitory effect, evidenced by its negative correlation with temperature (r = −0.803). In caprifig (Figure 12A), foliation mirrored those observed in fig, showing strong negative correlation with humidity (r = −0.819) and temperature (r = −0.742), alongside a positive association with rainfall (r = 0.831). The emergence of profichi fruit showed a weak negative correlation with humidity (r = −0.257) and temperature (r = −0.397), while maintaining a moderate correlation with rainfall (r = 0.689), suggesting a relatively minor climatic influence on this phenological stage. In contrast, the caprification period was strongly impacted by these factors, displaying a very strong positive correlation with humidity (r = 0.949) and a pronounced negative correlation with temperature (r = −0.913), indicating that high temperatures can adversely affect this stage. Mammoni and mamme production exhibited comparable patterns, with moderate to strong correlations with humidity and precipitation, alongside a notable negative correlation with temperature (r = −0.729 for mammoni and r = −0.985 for mamme). Finally, caprifig senescence was moderately affected by these climatic factors, displaying positive correlations with humidity (r = 0.350) and rainfall (r = 0.727), and a strong negative correlation with temperature (r = −0.900), confirming that high temperatures expedite the defoliation process of caprifig trees.

3.4. Analysis of Variance

Analysis of variance (ANOVA) was conducted to evaluate the effects of variety, year, orientation, and their interactions on the phenological stages of fig and caprifig trees. The results, summarized in

Table 3. Results of ANOVA testing the effects of variety, year, orientation, and their interactions on phenological stages of fig and caprifig trees.

Phenological Stage	Factor	F	Pr > F	η2
Fig Trees
Foliation	Variety	322.592	<2 × 10−16 ***	0.89
	Year	12,141.367	<2 × 10−16 ***	0.98
	Orientation	1721.818	<2 × 10−16 ***	0.86
	Variety × Year	80.040	<2 × 10−16 ***	0.66
	Variety × Orientation	1.678	0.1141	0.04
	Year × Orientation	735.957	<2 × 10−16 ***	0.72
	Variety × Year × Orientation	2.601	0.0129 *	0.06
Fig Flowers	Variety	403.89	<2 × 10−16 ***	0.91
	Year	82,454.65	<2 × 10−16 ***	0.99
	Orientation	103.51	<2 × 10−16 ***	0.29
	Variety × Year	317.86	<2 × 10−16 ***	0.88
	Variety × Orientation	24.95	<2 × 10−16 ***	0.37
	Year × Orientation	264.36	<2 × 10−16 ***	0.51
	Variety × Year × Orientation	41.98	<2 × 10−16 ***	0.50
Fig Autumn	Variety	161.763	<2 × 10−16 ***	0.80
	Year	14,248.381	<2 × 10−16 ***	0.98
	Orientation	289.129	<2 × 10−16 ***	0.50
	Variety × Year	259.439	<2 × 10−16 ***	0.86
	Variety × Orientation	13.024	1.46 × 10−14 ***	0.24
	Year × Orientation	35.824	6.41 × 10−9 ***	0.11
	Variety × Year × Orientation	5.731	3.19 × 10−6 ***	0.12
Receptivity	Variety	109.157	<2 × 10−16 ***	0.73
	Year	4585.097	<2 × 10−16 ***	0.94
	Orientation	128.439	<2 × 10−16 ***	0.31
	Variety × Year	95.149	<2 × 10−16 ***	0.70
	Variety × Orientation	17.679	<2 × 10−16 ***	0.30
	Year × Orientation	4.273	0.0396 *	0.01
	Variety × Year × Orientation	13.377	6.02 × 10−15 ***	0.25
Senescence	Variety	188.62	<2 × 10−16 ***	0.82
	Year	5424.64	<2 × 10−16 ***	0.95
	Orientation	124.22	<2 × 10−16 ***	0.30
	Variety × Year	119.86	<2 × 10−16 ***	0.74
	Variety × Orientation	13.35	6.42 × 10−15 ***	0.25
	Year × Orientation	192.74	<2 × 10−16 ***	0.40
	Variety × Year × Orientation	12.09	1.56 × 10−13 ***	0.23
Caprifig Trees
Foliation	Variety	25.487	3.98 × 10−10 ***	0.27
	Year	2469.686	<2 × 10−16 ***	0.95
	Orientation	778.421	<2 × 10−16 ***	0.85
	Variety × Year	4.526	0.01251 *	0.06
	Variety × Orientation	10.350	6.55 × 10−5 ***	0.13
	Year × Orientation	74.728	1.32 × 10−14 ***	0.35
	Variety × Year × Orientation	6.910	0.00139 **	0.09
Mamme	Variety	11.98	1.61 × 10−5 ***	0.15
	Year	2421.99	<2 × 10−16 ***	0.95
	Orientation	1595.73	<2 × 10−16 ***	0.92
	Variety × Year	50.58	<2 × 10−16 ***	0.43
	Variety × Orientation	63.64	<2 × 10−16 ***	0.48
	Year × Orientation	11,212.71	<2 × 10−16 ***	0.99
	Variety × Year × Orientation	77.18	<2 × 10−16 ***	0.53
Mammoni	Variety	441.788	<2 × 10−16 ***	0.87
	Year	255.836	<2 × 10−16 ***	0.65
	Orientation	286.878	<2 × 10−16 ***	0.68
	Variety × Year	16.470	3.94 × 10−7 ***	0.19
	Variety × Orientation	5.416	0.005456 **	0.07
	Year × Orientation	35.756	1.86 × 10−8 ***	0.21
	Variety × Year × Orientation	7.947	0.000545 ***	0.10
Profichi	Variety	218.10	<2 × 10−16 ***	0.76
	Year	10,338.01	<2 × 10−16 ***	0.99
	Orientation	11,099.69	<2 × 10−16 ***	0.99
	Variety × Year	266.55	<2 × 10−16 ***	0.80
	Variety × Orientation	48.90	<2 × 10−16 ***	0.42
	Year × Orientation	1683.23	<2 × 10−16 ***	0.93
	Variety × Year × Orientation	43.04	3.3 × 10−15 ***	0.39
Caprification	Variety	115.187	1.09 × 10−14 ***	0.63
	Year	1043.214	<2 × 10−16 ***	0.88
	Orientation	115.059	<2 × 10−16 ***	0.46
	Variety × Year	173.994	<2 × 10−16 ***	0.72
	Variety × Orientation	1.308	0.273850	0.02
	Year × Orientation	15.469	0.000133 ***	0.10
	Variety × Year × Orientation	6.020	0.003124 **	0.08
Senescence	Variety	539.560	<2 × 10−16 ***	0.89
	Year	59.317	2.49 × 10−12 ***	0.30
	Orientation	374.931	<2 × 10−16 ***	0.73
	Variety × Year	22.076	4.98 × 10−9 ***	0.25
	Variety × Orientation	7.543	0.000782 ***	0.10
	Year × Orientation	1.380	0.242111	0.01
	Variety × Year × Orientation	3.698	0.027307 *	0.05

* p < 0.05; ** p < 0.01; *** p < 0.001.

, demonstrate highly significant effects (p < 0.001) of these factors on all phenological stages, including foliation, flowering, autumn fruiting, receptivity, and senescence for fig trees, and foliation, mammoni, mamme, profichi, caprification, and senescence for caprifig trees. The year factor exhibited the strongest influence (p < 2 × 10⁻¹⁶, η² up to 0.99 for fig flowers), highlighting the sensitivity of phenological stages to inter-annual climatic fluctuations. Orientation significantly shaped phenological dynamics (p < 2 × 10⁻¹⁶, η² up to 0.99 for caprifig profichi), underscoring the role of local microclimatic conditions. Variety also had a significant impact (p < 2 × 10⁻¹⁶, η² up to 0.89 for caprifig senescence), reflecting genetic diversity’s role in phenological responses. Interactions between variety, year, and orientation were generally significant (p < 0.001), except for specific cases such as variety × orientation for fig foliation (p = 0.114) and caprifig caprification (p = 0.274), indicating nuanced adaptations to environmental conditions. The tripartite interaction (variety × year × orientation) further illustrated the complex interplay of these factors, with significant effects across most stages (p < 0.05). Detailed F-values, p-values, and η² effect sizes are presented in Table 3, providing a comprehensive overview of the factors influencing phenological dynamics in fig and caprifig trees.

3.5. Evaluation and Analysis of the ANN Model

Examination of the scatter plots comparing observed and predicted data highlighted the robustness of the ANN model in predicting the phenological stages of fig and caprifig (Figure 13 and Figure 14). For fig foliation, the model exhibited an accurate fit, with low RMSE values of 3.678, 3.300, and 2.587 for the test, validation, and training phases, respectively, and a coefficient of determination (R²) consistently exceeding 0.95. The mean absolute percentage error (MAPE), ranging from 4.528% to 7.067%, further supported the model’s stability (Figure 13A,

Table 4. Evaluation of model performance for phenological stage prediction in test, validation, and training phases. Metrics used include mean absolute percentage error (MAPE), root mean square error (RMSE), and coefficient of determination (R²).

	Test			Validation			Training
	MAPE	RMSE	R2	MAPE	RMSE	R2	MAPE	RMSE	R2
Foliation of Fig	4.528	3.678	0.200	7.067	3.300	0.969	4.992	2.587	0.956
Fig flowers	0.970	1.516	0.997	1.214	1.728	0.974	0.744	1.266	0.981
Fig autumn	0.677	2.147	0.755	0.822	2.114	0.938	0.310	0.749	0.989
Receptivity	1.522	2.550	0.907	1.101	2.018	0.991	1.117	1.946	0.979
Senescence	0.562	2.189	0.914	0.424	1.356	0.967	0.345	1.219	0.968
Foliation of Caprifig	7.065	2.700	0.986	3.930	1.777	0.985	6.088	2.373	0.968
Caprification	0.388	0.648	0.993	0.676	0.999	0.991	0.262	0.480	0.995
Mammoni	0.827	1.137	0.900	0.369	0.515	0.999	0.585	0.879	0.979
Mamme	0.943	2.783	0.884	1.352	4.369	0.782	0.477	1.525	0.954
Senescence	1.249	3.800	0.853	0.726	2.683	0.999	0.358	1.427	0.967

). For fig flowering, the model achieved an R² surpassing 0.97 across all phases. RMSE fluctuated between 1.266 and 1.728, while MAPE remained below 1.3% (Figure 13B, Table 4). Simulations for the autumn fig stage showed an R² of 0.989 during training but decreased to 0.755 in the testing phase. RMSE values ranged from 0.749 to 2.147, while MAPE remained below 1% (Figure 13C, Table 4). Scatter plot analysis for the receptivity period yielded an R² of 0.98 and an RMSE of 2.55, with consistent results across training, validation, and testing sets (Table 4, Figure 13D). Observed versus predicted scatter plots for the senescence stage indicated a good fit (R² = 0.96) with RMSE values of 1.219 and 1.356 for training and validation, and 2.189 for testing (Figure 13E, Table 4). For the caprifig foliation stage, observed and predicted dates showed a strong correlation (R² = 0.97), with RMSE values between 1.777 and 2.7 depending on the dataset (Figure 14A, Table 4). For caprification, the ANN model yielded high accuracy with an R² of 0.99 and a RMSE of 0.65 in the test phase. Training and validation results were similarly strong, with R² values of 0.995 and 0.991 and RMSE values of 0.480 and 0.999, respectively, (Figure 14B, Table 4). Regression plot analysis for the mammoni stage revealed R² values ranging from 0.900 (test) to 0.999 (validation), with RMSE values between 0.515 and 1.137 (Figure 14C, Table 4). The mamme stage showed moderate performance compared to others. While training yielded an R² of 0.954 and RMSE of 1.525, validation dropped to R² = 0.782 with RMSE = 4.369. Testing remained acceptable (R² = 0.884, RMSE = 2.783) (Figure 14D, Table 4). For caprifig senescence, the ANN achieved R² values between 0.853 and 0.999 across datasets, with RMSE values between 1.427 and 3.800 (Figure 14E, Table 4). Overall, the ANN model yielded high R² values, low RMSE and MAPE scores, and limited error dispersion across most phenological stages. Nonetheless, reduced accuracy was observed for the mamme and senescence stages, suggesting higher variability in these events.

4. Discussion

The identification of phenological stages is essential for optimizing crop management, guiding breeding programs, and assessing climate change impacts on productivity [15,20]. In order to effectively monitor the phenological development and ensure the comparability of observations, precise identification of species-specific growth stages using a standardized coding system, such as the BBCH scale, is imperative [89]. Although the BBCH scale has been applied to several fruit species worldwide, its application to fig trees (F. carica L.) in Morocco is reported here for the first time. This regional novelty enhances the originality of this study and provides a standardized framework for monitoring fig phenology under local agroecological conditions. In this study, the different phenological stages of male and female fig trees were identified and described using the BBCH scale as a standardized reference, offering valuable insights into eco-physiological processes.

In a context where optimizing crop management assumes increasing importance amid the challenges posed by climate change, mathematical modelling has emerged as a strategic approach for forecasting key phenological and agronomic traits, including the duration of developmental stages (budding, flowering, ripening), yield, and production quality [90,91]. For the fig tree (F. carica L.), an emblematic species of Mediterranean agroecosystems, such modelling approaches provide a robust scientific basis for varietal enhancement and the sustainable management of its cultivation amid changing climatic conditions. In this work, we developed a predictive model for the critical phenological stages of male and female fig trees cultivated in northern Morocco, based on local meteorological data and phenological observations gathered across diverse agroecological settings. Our modeling approaches integrate advanced tools such as automated machine learning (AutoML) and artificial neural networks (ANNs).

Artificial neural networks offer significant predictive advantages in identifying key phenological stages and in addressing synchronization challenges [56,92]. Compared with more conventional approaches, such as regression models or the Growing Degree Days (GDD) method, ANNs provide additional flexibility [93,94]. Regression models and GDD approaches are useful for capturing linear relationships and temperature-based thresholds [95,96,97], yet they often underestimate the influence of multiple interacting factors such as humidity, precipitation, and varietal differences [98,99]. In contrast, ANNs can incorporate these non-linear and multidimensional interactions, thereby achieving higher predictive accuracy and robustness under heterogeneous environmental conditions [100,101]. This methodological advantage highlights the specific contribution of ANNs in phenological modeling and underscores their potential for application to other fruit species in variable Mediterranean environments.

This investigation revealed several major findings regarding the impact of climatic factors on the phenological stages of male and female fig trees grown in northern Morocco. Correlation analysis revealed that phenological stages are significantly influenced by temperature, relative humidity, and precipitation. Overall, stages showed significant negative correlations with ambient temperature, suggesting that higher temperatures accelerate phenological progression. Mechanistically, this negative correlation can be explained by the acceleration of metabolic and enzymatic processes in fig and caprifig trees at elevated temperatures [102,103,104]. Higher temperatures enhance cell division, elongation, and differentiation [105,106,107], which in turn promote faster bud development, leaf expansion, and flowering [108,109,110]. Consequently, the duration of certain phenological stages is shortened. Conversely, relative humidity and precipitation modulate plant water status and turgor pressure [111,112,113]. These observations align with previous research on other fruit species, including Malus domestica [114,115] and Castanea mollissima and Melia azedarach [116]. Furthermore, relative humidity and precipitation play a decisive role in modulating the phenological stages of fig and caprifig, although their effects vary according to phase. In female figs, foliation and flowering are markedly enhanced by precipitation. Likewise, in caprifig, the caprification period exhibits particular sensitivity to humidity and precipitation. These observations confirm that water availability and humidity prolong certain phenological phases, corroborating findings from other fruit species [115,116,117]. Moreover, within the same species, distinct phenological stages respond differently to climatic variables. For instance, in the female fig, foliation is stimulated by precipitation but suppressed by temperature, whereas flowering is less influenced by humidity yet strongly dependent on rainfall. In caprifig, the caprification period is also highly sensitive to climatic conditions, showing positive association with humidity and a negative response to temperature, emphasizing these factors’ critical roles in reproductive success. These results concur with previous findings, which showed that early and late stages within the same species may exhibit divergent responses to climatic variability [12,116].

ANOVA results revealed that the factors of variety, year, and orientation significantly affect key phenological stages of both fig and caprifig trees, particularly budbreak, flowering, autumn fruiting, receptivity, and senescence. The predominant effect of year underlines the sensitivity of phenological stages to interannual climatic fluctuations, consistent with previous studies emphasizing the strong relationship between fig phenology and environmental variables such as temperature and humidity [28,114,118]. Moreover, [119] showed that varietal diversity significantly influences fig tree productivity and resilience, while [120] highlighted that genomic structuring of fig populations is strongly linked to environmental conditions. Similarly, ref. [121] demonstrated that genetic variability in fruit trees growing under high-altitude stress is crucial for shaping phenological performance. These findings confirm that the genetic resources of varieties play a pivotal role in determining phenological responses. The significant impact of orientation on all phenological stages across both male and female varieties confirmed the structuring role of local microclimatic conditions, aligning with studies emphasizing the role of these abiotic factors in modulating tree phenology [122]. Interactions between factors, notably variety × year and year × orientation, reflected differentiated phenological responses according to year, variety, and local exposure, indicating complex adaptations of fig and caprifig trees to climatic and microclimatic variations. The three-way interaction (variety × year × orientation) particularly illustrated the multifactorial nature of phenological mechanisms, reflecting a dynamic interplay between genetic diversity, annual climatic fluctuations, and local conditions. For instance, ref. [120] reported that genetic differentiation in fig populations is tightly associated with local environmental heterogeneity, while [123] observed contrasting reproductive strategies among related Ficus species, driven by microclimatic and climatic variability. Together, these studies reinforce the idea that phenological dynamics in figs result from the combined effects of genetic background and site-specific environmental constraints. A recent genomic study of Ficus species confirmed that genetic structure and diversity are strongly shaped by environmental heterogeneity and geographical barriers. For example, ref. [120] demonstrated that local ecological gradients generate significant genetic differentiation, which in turn supports adaptive phenological responses in diverse environments.

Climate change exerts considerable pressure on Mediterranean agroecosystems [124,125,126], and the Rif region of northern Morocco is no exception. Recent studies indicate a steady increase in average annual temperatures, more frequent and prolonged droughts, and irregular rainfall patterns, which together exacerbate water scarcity and soil degradation [127,128,129]. In the Rif, projections suggest a temperature rise of 1.5–2.5 °C by mid-century, accompanied by a 10–20% decline in precipitation [130]. These changes are expected to directly affect the phenological cycles of fig trees by advancing budburst, shortening flowering duration, and altering fruit ripening synchronization, with potential negative consequences for yields and fruit quality [131,132]. Beyond productivity, such shifts also threaten the sustainability of traditional cultivation systems, which rely on rainfed agriculture and limited water resources [133,134,135]. Addressing these challenges requires adaptive strategies, including the selection of resilient varieties, improved water management, and the preservation of local ecological knowledge to ensure the long-term viability of fig-based agroecosystems in the Rif.

In this respect, under the climatic conditions prevailing in northern Morocco, particularly within the study area, the period from February to July was the main phase in the vegetative and reproductive development of fig trees. During this interval, the development of shoots and leaves was critical, as it promoted the formation of reproductive buds, which were decisive for fruit production [28]. Moderate spring temperatures, which are conducive to optimal photosynthetic activity, are essential for both vegetative and reproductive growth, as has been reported [136]. In contrast, summer conditions marked by high temperatures and high light intensity impose significant stress on fig varieties, which manifest as decreased stomatal conductance, reduced photosynthetic efficiency, and lower photochemical performance [137]. These negative effects indicated the sensitivity of fig trees to summer abiotic stress and highlighted the importance of spring climatic conditions in sustaining their productive potential, also noted in [138].

In fig and caprifig (F. carica L.), the major growth stages were characterized by overlapping phenological phases, where vegetative growth preceded and paralleled the reproductive stages [28]. Budbreak, in particular, signified the end of dormancy and the onset of the active growth phase [139]. This stage holds significant agronomic importance, particularly for scheduling cultivation practices such as determining the optimal timing for winter pruning. In the Bni Ahmed region, dormancy typically spanned from late December to February, a pattern also reported in previous studies [27,140,141], where dormancy occurred between January and February.

The results of this study show that the dynamics of leafing and senescence in fig and caprifig trees are strongly shaped by seasonal temperature and precipitation. Phenological studies on Ficus confirm that these climatic factors drive shifts in leaf and fruit development cycles [117,142]. Budburst in female fig trees occurred from 4 to 28 February at Anassel and 2 to 24 February at L’marj L’bardi, while in caprifig trees it was recorded from 27 January to 4 March at Anassel and 25 January to 21 February at L’marj L’bardi. These findings agree with Moroccan data showing fig budburst in early March after a short, temperature-sensitive dormancy [143] and with broader evidence that warmer winters advance Mediterranean budbreak [144]. Microclimatic differences between Anassel and L’marj L’bardi, notably slightly higher temperatures at the latter, explain site-specific variations, similar to patterns reported for Ficus populations in Taiwan [145]. Comparable findings were also reported in other studies [91,146,147,148]. Regarding foliage, marked differences were observed between female fig and caprifig trees at the two study sites, Anassel and L’marj L’bardi, and between the two observational years. In 2021, leafing durations for female fig trees ranged from 227 to 244 days at Anassel, and from 217 to 230 days at L’marj L’bardi. In 2022, a significant extension was observed, spanning 244–258 days at Anassel to 241–256 days at L’marj L’bardi. This prolongation likely reflects milder climate conditions, which favor extended periods of vegetative activity. A similar phenomenon was documented in Ficus populations in Kerala, India, where atypical rainfall patterns and higher temperatures contributed to a prolonged vegetative phase [142]. These observations are consistent with results indicating leafing durations ranging from 256 to 275 days in comparable Mediterranean climates [27,141]. Caprifig trees showed comparable inter-annual variability, with longer leafing durations in 2022, indicating higher tolerance to climatic fluctuations—an adaptive trait highlighted in studies on Ficus resilience under climate change [142,145]. In terms of senescence, female fig trees displayed earlier and faster defoliation in 2022, with durations decreasing from 48.25 to 36.88 days at Anassel, and a comparable reduction at L’marj L’bardi, consistent with the accelerating effect of warmer autumns [149,150]. Caprifig trees, more tolerant overall, exhibited a slight extension in senescence, reflecting their sensitivity to autumn climate. These results confirm that phenological responses are both species- and site-specific, but consistently modulated by temperature and precipitation variability [142,143,145,151].

Analysis of the flowering stages and harvesting periods of female fig trees at Anassel and L’marj L’bardi revealed pronounced inter-annual variability in duration and phenological synchronization. In 2021, the fig-flower fruiting stage was relatively short, averaging 58.57 ± 6.90 days at Anassel and 60.00 ± 6.06 days at L’marj L’bardi. In 2022, this period was markedly prolonged, reaching 104.29 ± 5.74 and 100.57 ± 2.70 days, with observed ranges from 95 to 111 days. These results are consistent with earlier research [152,153], which reported syconium ripening durations between 84 and 112 days depending on environmental conditions. The extension observed in 2022 likely reflects higher spring temperatures, a factor known to delay senescence and extend phenological phases in Mediterranean climates [141,154]. Furthermore, the variability observed among different varieties confirmed that phenological responses are shaped not solely by climatic factors but also by genetic traits [121]. Recent genomic and phenological studies of Ficus species have underscored the pivotal role of intra-cultivar genetic diversity in shaping flowering and fruiting behavior and in enhancing adaptability to environmental fluctuations [103,120]. In terms of harvesting periods, a substantial contraction was recorded in 2022, with harvest windows narrowing to 10–14 days compared to 15–20 days in 2021. This reduction may reflect a strategic adaptation by farmers, seeking to synchronize harvests with market demand while mitigating escalating phytosanitary challenges. At both Anassel and L’marj L’bardi, many producers have adopted the immediate removal of ripe fruit post-harvest to mitigate phytosanitary risks, consistent with proposed guidelines [155].

An examination of autumn fig ripening at Anassel and L’marj L’bardi revealed notable inter-annual variability between 2021 and 2022, alongside notable differences among the various cultivated varieties. In 2021, ripening was relatively long, averaging (93.38 ± 4.10 days) at Anassel and (89.25 ± 6.54 days) at L’marj L’bardi, with an average interval from pollination to full purple fruit ripeness being 36 days at Anassel, and 35 days at L’marj L’bardi. These results are consistent with previous findings [131,156], where ripening intervals of approximately 36 to 38 days between pollination and harvesting fully ripe fruit were reported. Furthermore, this variability in ripening time among fig varieties has been well documented in studies of morphological diversity and phenological development, indicating that genotype and environmental factors significantly influence fruit development timelines [28,157]. In contrast, 2022 showed a significant decline in ripening duration, averaging (70.25 ± 5.34 days) at Anassel and (68.50 ± 5.53 days) at L’marj L’bardi. The post-pollination intervals shortened to (29 days) at Anassel and (25 days) at L’marj L’bardi. This reduction could be attributed to more stressful climatic conditions, notably high summer temperatures, known to accelerate the ripening process in fig trees. Studies have shown that increased temperatures expedite the phenological phases of fig trees, leading to earlier fruit ripening and harvesting, as confirmed in previous research conducted in arid and semi-arid climates [28,157,158]. The inter-varietal and inter-annual differences in ripening times underscore the influence of genetic factors, even under comparable climatic conditions [119]. Certain fig species exhibit delayed ripening, whereas others attain full ripeness rapidly, depending on environmental conditions [117]. Additionally, fig ripening dates vary geographically and among variety. In this study, harvest typically occurred between July and August, consistent with previous reports [159]. For instance, the Bursa Siyahı variety reaches full ripeness between 15 and 25 August in Yalova, Turkey [160], while in Hatay province, ripening occurs between 22 July and 2 August [141]. Fig ripening generally initiates in July and can extend until cooler temperatures prevail from October to December, highlighting the importance of climatic conditions on harvest scheduling [161,162].

Similarly, to female fig varieties, the results of this study revealed significant differences in the duration and progress of the phenological stages of caprifig (F. carica L.) at the Anassel and L’marj L’bardi sites. These disparities highlight significant inter-annual variations between 2021 and 2022, in the principal reproductive phases (profichi, mammoni and mamme), reflecting a high sensitivity of caprifig trees to annual climatic conditions [28,123]. In 2021, the profichi stage occurred from March to July, 5 days earlier in L’marj L’bardi than in Anassel. Conversely, in 2022, this stage commenced earlier, spanning from February through June. These observations are compatible with the data reported by previous studies, which placed this stage between May and July [163], In contrast, another study reported a slightly earlier period, from mid-March to mid-June [164]. The mammoni and mamme stages also exhibited pronounced inter-annual and inter-site variability. Within this study, the mammoni stage, occurring from May to September/October, aligns with the period detailed in previous research, which identified it as extending from April or May through September [163]. Likewise, the mamme stage was observed from September to February/March, consistent with previous findings that placed this stage between September and April/May [163]. These results highlight the pronounced sensitivity of caprifig trees to annual climatic fluctuations and reinforce the importance of varietal selection to enhance phenological adaptation, agreeing with findings of previous research [165]. The adjustment of phenological cycles to environmental conditions has played a key role in optimizing the production and resilience of fig trees under changing climatic scenarios.

Data analysis revealed that climatic variability affects synchronization between male and female fig varieties, which is crucial for pollination efficiency and reproductive success. Premature varieties such as L’hlou, with their rapid ripening, can partially pollinate early-season female varieties (L’qouti, L’mdar, L’qellal), but their limited pollination capacity constrains effectiveness [77]. Farmers in Bni Ahmed often procure profichi from regions where this phase occurs earlier, reflecting adaptive management strategies to overcome synchronization gaps [74]. Varieties with extended receptivity, such as L’morr, provide better overlap with female fig trees, highlighting the importance of diversifying caprification periods to secure pollination over time [166]. In contrast, mismatches between late male varieties (e.g., L’louizi) and early female figs (e.g., L’qouti, L’messari) reduce pollination success [167]. These observations emphasize that optimizing the timing of caprification and selecting complementary varieties are key strategies for maintaining reproductive success and stable yields under climatic variability.

Analysis of the artificial neural network (ANN) model’s performance in predicting phenological stages of fig and caprifig revealed excellent simulation capability. Overall, high values of the coefficient of determination (R²) and low root mean square errors (RMSEs) and mean absolute percentage errors (MAPEs) confirm the model’s accuracy and robustness, consistent with studies applying ANNs to crops such as wheat and beans [168,169]. Fig foliation and flowering stages were simulated with exceptional performance, comparable or superior to results reported for other fruit species such as beans, cherries, and grapevines [170,171,172]. Autumn figs showed greater variability, reflecting the challenge of simulating reproductive stages at scale [173], yet receptivity stages were accurately captured, consistent with systematic reviews highlighting ANN proficiency in managing environmental variability [174].

Simulation of senescence and foliation stages of caprifig also demonstrated strong robustness, in line with previous reports for deciduous species [175,176]. Minor declines in performance were noted for certain stages, notably the mamme stage, likely due to local environmental heterogeneity, a challenge commonly observed in phenological modeling of other crops [168,169]. Concerning caprification, our ANN model achieved very high accuracy, exceeding performances reported for olive blossom and other woody species, including deep neural network approaches [177,178,179]. These results confirm the strong capacity of the ANN model for learning and generalization in predicting phenological stages. By capturing complex, non-linear responses to environmental conditions, our findings highlight the potential for integrating such models into agricultural management systems, optimizing fig and caprifig cropping strategies under climate variability [173,174]. Moreover, the practical applications of the ANN model in the field can directly support farmers and technicians. For example, it can predict optimal caprification dates, issue alerts for critical phenological stages, and guide decisions regarding irrigation and harvest scheduling. These applications demonstrate how AI-based phenological modeling can improve on-the-ground management of fig orchards and help mitigate potential yield losses under variable climatic conditions. Importantly, the combined approach of BBCH phenological monitoring and ANN modeling is highly transferable to other Mediterranean fruit trees. By adapting the input data to local conditions and specific species, this framework can support predictive phenology, enhance crop management, and integrate local ecological knowledge across diverse agroecosystems, thereby broadening the applicability and impact of this methodology beyond fig cultivation.

5. Conclusions

This study presents, for the first time in Morocco, the combined use of the BBCH scale and artificial neural networks (ANNs) to synchronize and predict the phenology of fig and caprifig (F. carica L.), offering an innovative methodological framework tailored to Mediterranean orchards. By calibrating the principal phenological stages of both female and male fig trees according to the BBCH scale, we provide a standardized reference for monitoring and comparing varietal cycles. Climatic factors such as temperature, relative humidity, and precipitation were shown to exert predominant influences across all stages, while also revealing varietal and site-specific responses. The application of ANNs yielded highly accurate predictive models (R² > 0.95, RMSE < a few days, MAPE < 5%), with caprification simulated with exceptional precision (R² = 0.99; RMSE < 1 day). These results highlight the potential of ANN-based tools to forecast critical phenological dates, optimize pollination timing, and minimize yield losses in Mediterranean fig orchards. Looking forward, future research should not only integrate finer microclimatic (soil temperature, radiation, soil moisture), agronomic, and genomic variables, but also expand to larger numbers of trees and multiple agroclimatic zones. This will strengthen the external validation of models and broaden their relevance for climate adaptation strategies. In summary, this integrative approach provides operational forecasting and management tools specifically adapted to Mediterranean fig and caprifig cultivation, thereby contributing to the resilience and long-term sustainability of Mediterranean agriculture under climate change scenarios.