Effect of Different Indole Butyric Acid (IBA) Concentrations in Various Rooting Media on the Rooting Success of Loropetalum chinense var. rubrum Yieh Cuttings and Its Modeling with Artificial Neural Networks

Abstract

This study aimed to evaluate the rooting success of Loropetalum chinense var. rubrum Yieh cuttings in three different rooting media: 100% peat, 100% perlite, and a 50% peat–50% perlite mixture. Additionally, three concentrations of Indole Butyric Acid (IBA)—1000 ppm, 3000 ppm, and 6000 ppm—were tested, along with a control group consisting of non-hormone-treated cuttings. The chlorophyll content of the leaves was measured in µmol/m², and its relationship with rooting success was examined. Measurements were conducted every 15 days over a 120-day period. The collected data were analyzed using both an artificial neural network (ANN) and SPSS 29.0.2 statistical software. Results indicated that perlite medium yielded the highest rooting rate and chlorophyll concentration, whereas the peat medium performed the poorest. While 1000 ppm IBA led to the greatest improvement in rooting rate, 6000 ppm resulted in the highest chlorophyll concentration. The highest chlorophyll levels were observed during measurement periods M7, M8, and M9. Analyses of peat moisture and pH indicated that the physicochemical properties of the rooting media significantly influenced cutting development. This study aims to support the identification of optimal propagation methods for this species and to contribute to the literature by developing an ANN model based on the measured parameters.

1. Introduction

Loropetalum chinense var. rubrum Yieh is a valuable plant taxon belonging to the Hamamelidaceae family, which is increasingly being used for its high aesthetic and ecological values [1]. It is a medium-sized evergreen shrub taxon native to Hunan Province in China [2,3]. Its most striking feature is its red leaves and flowers with calligraphic value. With its calligraphic flowers and characteristic leaves, it is extensively used in recreational and residential areas [4,5]. In addition to its color characteristics, the plant habitus is characterized by a compact and richly branched structure [6,7]. The most dramatic feature of this plant is its long flowering period [5,8]. Since the vegetation period is quite long, it blooms several times a year and when it blooms, the flowers remain on the plant for a long time [9]. In this way, it offers a flowering appearance throughout the year. In addition to these features, Loropetalum chinense stands out as an excellent landscape plant due to its color, texture, and form characteristics [10,11], and also offers notable ecological contributions [2,12,13]. Loropetalum chinense provides an interesting visual effect with its dense and dense growth and provides an important and sheltered habitat for pollinators such as insects [14,15]. In addition, the roots of the plant secrete organic compounds that improve soil quality, while the leaves help to improve the air quality of the areas where it is located by trapping dust and metal accumulation [16,17]. Improving soil quality is of ecological importance [18]. Climate is a critical external factor affecting the variation in bioactivity compounds and genetic variation for plants [19,20,21]. The climatic tolerance of Loropetalum chinense is quite high [22,23,24]. In Rize province, which has a temperate Black Sea climate, the average annual total rainfall between 1991–2020 was 2309.5 (mm). In addition to this high precipitation rate, the average annual highest temperature between 1991–2020 is 19 °C and the average lowest temperature is 11.7 °C [25]. These values show that Rize province, which is the study area, has a suitable habitat structure that will allow Loropetalum chinense to remain flowering and grow healthy from spring to summer and even until the beginning of the autumn season. Many generative and vegetative methods are used for propagation of plant material. Plant propagation by cuttings is an effective vegetative method used to create plant individuals with the same genetic characteristics of the mother plant [26,27]. In this method, homogeneity and efficiency in plant production are ensured while preserving the genetic characteristics of the mother plant [28,29]. These values are important criteria in the propagation of taxa with aesthetic and ecological value. This method enables the production of high-quality individuals that are well adapted to environmental conditions, with minimal genetic variation, making it suitable for ecological restoration and landscape applications [28,30]. For the rehabilitation of forest areas or landscaping works, plant production method with cuttings provides ideal plant individuals at low cost [31,32]. As a result of the literature search, no detailed information on the nutrient medium and hormone doses for the propagation of Loropetalum taxon by cuttings was found [33,34]. In addition, there is no study examining the relationship between chlorophyll synthesis in the leaves left on the cuttings during rooting of this taxon and rooting performance.

The cuttings method of plant propagation is the process of growing young plant parts under favorable conditions to develop root systems. In this method, the ability of plants to form roots is enhanced by using hormones, especially indole-3-acetic acid (IBA) [35,36,37]. IBA is among the plant growth hormones that promote root development in plants and, when applied, it accelerates root formation processes and provides improved root morphology [38,39,40,41]. IBA applications significantly increase the rooting rates of plants and promote root development under stress conditions such as water stress [42,43,44,45]. Application of IBA to the lower part of the cuttings increases the survival rate of new plant individuals by enabling roots to form faster and more densely [46]. Moreover, the general effects of IBA not only on root development but also on plant growth are remarkable [47,48,49,50]. Studies have shown that IBA increases the general health and ability of plants [51,52]. The importance of IBA hormone in plant propagation by cuttings method becomes evident with factors such as providing faster and healthier development of roots, increasing resistance to stress conditions and improving plant health. Therefore, IBA is considered an indispensable component, especially in applications such as propagation by cuttings.

In most of the studies, statistical methods are used to determine the optimum production parameters. Among these methods, the Statistical Package for the Social Sciences (SPSS 29.0.2) program is one of the widely preferred tools, along with other software such as SAS and R. For example, Azad et al. (2016) analyzed the effects of different concentrations of IBA on rooting and biomass production in Santalum album cuttings [53]; Sevik and Güney (2013) analyzed the effects of IAA, IBA, NAA, and GA₃ hormones on rooting and morphological characteristics [47]; Song et al. (2023) studied the effects of different doses of IBA on root growth [54]; Ma et al. (2024) examined the effects of propagation methods of Malus ‘Baiyun’ [55]; and Zhang et al. (2023) conducted a comparative analysis of endogenous hormones and antioxidant enzymes [56], all utilizing SPSS software for statistical analysis.

Artificial neural networks (ANNs) are computational models that mimic the biological nervous system and are generally used in many different fields in terms of efficiency and prediction accuracy [57,58,59]. ANNs achieve effective results in analyzing complex data thanks to their multi-layered structure [60,61,62]. Artificial neural networks (ANNs) are a widely used prediction method for modeling complex relationships between input and output data sets in industrial applications. This method enables output predictions to be made for untested points by building a model based on experimental data and determined parameters. In particular, ANN models offer effective alternatives when it is difficult to determine process characteristics with traditional experimental equations [63]. The basic structure of ANN is based on a multilayer architecture consisting of an input layer, one or more hidden layers and an output layer. The input and output layers consist of a set of neurons representing the variables of interest, while the hidden layers perform the data processing. Although theoretically there is no limit to the number of hidden layers, in most applications one or two layers are considered sufficient [61,64,65,66,67]. The output layer processes the information from the hidden layers to generate the final output vector. Research shows that single output ANN models have higher accuracy rates compared to multiple output models. Each neuron in the network structure is connected to other neurons with weights that can be adjusted during the learning process. These weights are optimized during the training process until the error rate is minimized. The ANN can be developed to provide the best predictions for specific problems by using an appropriate training algorithm.

In light of this framework, the present study was designed with the following specific objectives:

• To evaluate the rooting success of Loropetalum chinense var. rubrum Yieh cuttings in three different rooting media: 100% peat, 100% perlite, and a 50% peat–50% perlite mixture.
• To assess the effects of different indole-3-butyric acid (IBA) concentrations (1000 ppm, 3000 ppm, and 6000 ppm) on the rooting performance of the cuttings, alongside a control group with no hormone application.
• To measure the chlorophyll content (µmol/m²) of the cuttings’ leaves and analyze the relationship between chlorophyll synthesis and rooting success.
• To determine the responses of the cuttings to the physical and chemical characteristics of the rooting media, such as pH, moisture, temperature, and electrical conductivity.
• To model the rooting and chlorophyll response using artificial neural networks (ANNs) and validate experimental results through statistical analyses (ANOVA and t-test).
• To contribute to identifying the optimal propagation conditions for Loropetalum chinense in ecological restoration, commercial nursery production, and similar applications.
• To improve understanding of how leaf chlorophyll production interacts with root development in order to clarify metabolic processes involved in the rooting phase.
• To provide guiding data for future research and practical applications in the vegetative propagation of this and other ornamental plant species.

2. Materials and Methods

2.1. Material

The main material of this study was Loropetalum chinense (Recep Tayyip Erdoğan University Zihni Derin campus in Fener Neighborhood, Rize province, Türkiye (41°2′13.58″ North Latitude, 40°29′36.28″ East Longitude)) cuttings collected from healthy, mature mother plants located at Recep Tayyip Erdoğan University. These plants were not grafted, and the term “mother plant” is used here to refer to the original source of the cuttings. The cuttings were taken from new shoots growing in May, all collected from the same aspect to ensure uniformity.

Indole-3-butyric acid (IBA) was used as the rooting hormone in three different concentrations: 1000 ppm, 3000 ppm, and 6000 ppm. Sterile pruning shears were used during cutting collection.

In this study, the rooting media were coded as follows: T1 (100% peat), T2 (100% perlite), and T3 (50% peat–50% perlite mixture). Similarly, the hormone treatments were coded as H0 (control, no hormone application), H1 (1000 ppm IBA), H2 (3000 ppm IBA), and H3 (6000 ppm IBA). These labels are consistently used throughout the manuscript to describe the different treatment groups.

2.2. Data Sets

All measurements of plant individuals in the experimental design were repeated every 15 days. A total of nine measurements were carried out between May and September 2024 at approximately 15-day intervals.

The 15-day measurement interval was selected to systematically monitor the progressive changes in rooting and physiological responses without imposing excessive mechanical stress on the cuttings. This time frame allowed for capturing dynamic developments at regular periods, balancing the need for detailed data with the necessity to minimize potential damage caused by frequent handling.

These measurements are labeled as M1 to M9, corresponding to the following dates:M1—6 May 2024; M2—21 May 2024; M3—5 June 2024; M4—20 June 2024; M5—5 July 2024; M6—20 July 2024; M7—4 August 2024; M8—19 August 2024; and M9—3 September 2024. According to this table, the chlorophyll content of the cuttings, and the temperature, pH value, electrical conductivity (EC), and humidity of the rooting medium were measured on the measurement day in each experimental environment. Chlorophyll values were measured with the Apogee MC-100 chlorophyll concentration meter, which is in the inventory of Recep Tayyip Erdoğan University. This device can measure three different parameters, SPAD, CCI and μmol m⁻². In this study, μmol m⁻² parameter was used as the measurement unit of Loropetalum chinense var. rubrum Yieh. In order to obtain accurate results, measurements were made from the leaves at the same angle in all plants. After the leaves were placed in the device so that the upper part of the leaves was placed where the device sent the beam, the measurements were made from the centre of the leaf, avoiding the leaf veins. Measurements were made between 14:00 and 16:00.

Rooting medium temperature and electrical conductivity (EC) were determined as the average of the data taken from 3 points in all pots. Celsius (°C) scale was used for temperature values. Electrical conductivity (EC), also known as electrical conductivity, refers to the capacity of water to conduct electric current in units of ‘mS/cm’ (milliSiemens/cm) due to the dissolved substances in the water [21]. Rooting medium pH and moisture (%) values were measured with Yieryi brand device in all environments. At the end of 120 days, the number of rooted individuals, root development parameters and the effectiveness of rooting media were evaluated. In the study, precise measurements were made for steel diameters using digital callipers.

Although rooting success and survival rates were considered the primary indicators of propagation success, additional physiological measurements such as chlorophyll content were included to provide complementary information. Maintaining photosynthetic activity during the rooting process is essential for the survival and development of cuttings, and chlorophyll measurements served as supportive data to assess the physiological status and vigor of the plant material.

2.3. Application Studies

The study was carried out in the Plant Laboratory of the Department of Landscape Architecture, Recep Tayyip Erdoğan University, Rize. Cuttings of Loropetalum chinense var. rubrum Yieh were collected from individuals at previously determined locations. The cuttings were cut with the help of sterilized pruning shears at a length of 30 cm, wrapped in damp cloths and placed in plastic containers containing ice molds and water. The cuttings brought to the laboratory were trimmed to 15 cm in length, each containing at least two vegetative buds and two leaves. The cuttings were kept in a damp cloth until planting. The cuttings were randomly selected and divided into three groups and three different rooting media were prepared: 100% perlite, 100% peat and peat–perlite mixture (50–50%). After the greenhouse pads were divided with wooden materials, these areas were filled with rooting media and organized as plant pads.

The compositions of the rooting media were selected based on their proven effectiveness in optimizing aeration, drainage, and moisture retention. Peat and perlite mixtures in different ratios were chosen to create varying physical environments, aiming to investigate their influence on root formation and development.

2.4. Experimental Design

The collected Loropetalum chinense var. rubrum Yieh plants were planted in 3 different rooting media. The first rooting medium was peat, the second was perlite and the third was a mixture of peat and perlite in equal volumetric amounts. In addition, indole butyric acid (IBA) rooting hormone was applied at 3 different doses (1000 ppm, 3000 ppm and 6000 ppm) in each rooting medium except the control group. Each experimental medium was designed with 3 replications. There are 10 cuttings in each replicate. In total, 360 cuttings were used in the study. Plant cuttings of the same size were planted in each experimental medium.

2.5. ANOVA and T Test

In the analysis of the research, ANOVA (analysis of variance) and t-tests, which allow comparing the averages of different groups that are not homogeneous, were used. Whether the differences in the data obtained as a result of the measurements were statistically significant or not and their trends were determined by this method. Square root transformation was used in data sets that did not show homogeneous distribution.

2.6. Artificial Neural Network

ANN modeling for chlorophyll (C) and rooting (R) obtained from the measurements of plants grown in 100% perlite, 100% peat and 50% peat + 50% perlite environments was carried out by using a sub-program in MATLAB 2020a software. The modeling process with ANN consists of two stages by using training and test data. In the training process, the input and output values given to the network were checked to minimize the margin of error, and in the test phase, the results were predicted according to the input values without changing the weight values. As a result of the experiments, C and R values were analyzed separately using a generalized feed forward network structure. Environment (O), hormone amount (H), Ph (P), EC (E), temperature (T), and humidity (M) variables were defined as input parameters and C and R were defined as output parameters. For each input parameter, a total of 25 experiments were performed according to five levels. Out of 3240 data, 2592 were categorized for training and 648 for testing. Input and output data were transferred to the software and the optimum network structures were determined by performing different experiments with these data (Figure 1).

3. Result

3.1. Statistical Anaylses (ANOVA)

As a result of the experimental study, the rooting numbers of plants grown on different media (peat, perlite, mixture) and hormone treatments (control, 1000 ppm, 3000 ppm, 6000 ppm), the amount of chlorophyll and other factors affecting them (pH, EC, humidity, temperature) were analyzed.

Figure 2 presents the mean pH values measured in three different rooting media (peat, perlite, mixture), with statistically significant differences identified by ANOVA (F = 1436.74, p < 0.01). According to the Dunnett T3 multiple comparison test, the perlite medium exhibited the highest pH value (6.14), while the peat medium had the lowest (5.41). The groupings based on different letters represent statistically distinct categories. The regression curve and R² value indicate a non-linear relationship between rooting media and pH levels.

Figure 3 displays the average pH values measured across nine different measurement periods (M1–M9) specifically for the peat medium. The analysis revealed significant temporal variation in pH (F = 4.46, p < 0.01). The highest values were recorded during M2 and M3 (5.60–5.65), whereas the lowest values occurred in M7 and M8 (5.10), suggesting an increase in medium acidity during mid-summer.

Figure 4 illustrates the temporal pH changes in the perlite medium. Statistically significant differences were observed across measurement periods (F = 6.93, p < 0.01). The highest pH was measured in M3 (6.88), and the lowest in M9 (5.78). The regression curve and R² value (0.6658) highlight a moderate non-linear association between sampling periods and pH in this medium.

Figure 5 shows the same analysis for the mixture medium. According to the ANOVA results, significant differences were observed among measurement periods (F = 9.25, p < 0.01). The highest pH value (6.03) was recorded in M3, while the lowest (5.32) was observed in M8. A moderate non-linear trend (R² = 0.4057) was identified between measurement periods and pH in the mixture medium.

Electrical conductivity (EC) values varied significantly among the different rooting media (peat, perlite, and mixture). According to ANOVA results, the mean EC values differed statistically (F = 1401.12, p < 0.01), with the peat medium exhibiting the highest EC value (1.10 µS/cm) and the perlite medium showing the lowest (0.11 µS/cm). The Dunnett T3 test grouped the media into two statistically distinct categories (Figure 6). The regression curve and R² value (0.5121) indicated a moderate non-linear relationship between EC and rooting media.

In addition, EC values measured over nine different time points (M1–M9) revealed dynamic trends depending on the rooting medium. For the peat medium, EC values showed statistically significant variation across periods (F = 2.60, p < 0.01). The highest value was observed in M2 (1.53 µS/cm), while the lowest occurred in M9 (0.57 µS/cm), suggesting a gradual decline over time (Figure 7). The regression curve and R² value (0.4198) supported a non-linear decreasing pattern.

In the perlite medium, although significant variation was found (F = 9.92, p < 0.01), no statistical differences were observed among periods based on multiple comparisons (all groups labeled “a”). EC values remained consistently low, ranging from 0.030 to 0.010 µS/cm (Figure 8). The regression curve and R² value (0.7928) nonetheless revealed a moderate non-linear trend.

For the mixture medium, EC values also differed significantly between measurement periods (F = 4.23, p < 0.01). The highest values were detected in M1 (1.50 µS/cm) and M2 (1.79 µS/cm), with a subsequent decline observed in later periods (Figure 9). However, most time points belonged to the same statistical group, indicating moderate variation. The regression curve and R² value (0.7112) confirmed a non-linear decreasing relationship.

As illustrated in Figure 10, the mean temperature values measured across different rooting media (peat, perlite, mixture) exhibited statistically significant differences (F = 12.78, p < 0.01). According to the T3 test results, two statistically distinct groups were identified. Among the tested media, the peat medium recorded the highest average temperature at 28.48 °C, while the perlite medium showed the lowest value at 27.72 °C. The regression curve and R² value (0.8777) indicated a strong non-linear relationship between rooting media and temperature distribution.

Figure 11 presents the variation in temperature values across the nine measurement periods (M1–M9) specifically for the peat medium. Statistically significant differences were detected between periods (F = 109.29, p < 0.01), with the highest temperature recorded in M7 (32.08 °C) and the lowest in M1 (21.64 °C). The regression trend and a high R² value (0.9422) revealed a pronounced parabolic relationship, demonstrating seasonal influence on temperature dynamics in the peat environment.

In Figure 12, temperature fluctuations observed in the perlite medium across the same measurement periods showed significant variation (F = 132.8, p < 0.01). The highest temperature was noted in M7 (32.68 °C), while M1 showed the lowest value (21.16 °C). The regression curve and R² value (0.855) confirmed a strong non-linear relationship between measurement periods and temperature behavior in the perlite medium.

Lastly, Figure 13 displays the temporal temperature changes in the mixture medium. ANOVA results indicated significant variation between periods (F = 6.48, p < 0.01), with the highest average temperature observed in M4 (31.92 °C) and the lowest in M1 (21.62 °C). The regression curve and R² value (0.8707) again confirmed a strong curvilinear pattern, consistent with environmental temperature trends across the study period.

The graph given in Figure 14 shows that the rooting medium type creates significant differences in the humidity ratio (F: 1634.33, p < 0.01). While peat (%59.84) has the highest value in terms of humidity ratio, perlite (%38.66) has the lowest value. Rooting mediums exhibited statistically significant differences from each other according to humidity values. In Figure 15, the analysis between humidity values and measurement periods are examined. Although humidity values showed significant differences (F: 52.21, p < 0.01) when examined according to different measurement periods, the R² (0.0264) value indicated that the change did not show a significant trend between the periods.

The rooting rates of the plant cuttings taken according to the rooting medium are given in Figure 16 and the rooting amounts according to the applied rooting IBA hormone concentrations (ppm) are given in Figure 17. When these two graphs are examined, it is seen that the rooting mediums and the applied IBA concentrations create significant differences. It was determined that the best medium among the rooting mediums was perlite and the best IBA dose was 1000 ppm.

The change in the amount of chlorophyll in the cutting leaves according to the rooting environment is given in Figure 18. The values for chlorophyll changes according to IBA hormone concentrations are given in Figure 19 and the change in the amount of chlorophyll in the leaves according to the measurement periods are given in Figure 20. When these figures are examined, it is determined that there is a significant strong relationship between the rooting environment and the change in the amount of chlorophyll (R²: 0.9017), and similarly, there is a moderately significant relationship between the change in the amount of chlorophyll with different measurement periods (R²: 0.5237). When hormone application and chlorophyll changes were examined, a significant relationship was observed, but it was determined that this relationship was weak. It was understood that the best rooting medium in terms of chlorophyll synthesis was perlite.

3.2. ANN

In ANN, logsig–purelin activation functions and Scaled Conjugate Gradient Backpropagation Algorithm (trainscg) learning algorithm were used for chlorophyll. The weight and threshold values obtained for K according to the activation functions and learning algorithms are given in

Table 1. Weight and threshold values obtained for K.

i	W1	W2	W3	W4	W5	W6	θi	Lw	θ11
1	−8.8872	−1.0549	−47.336	−37.8254	84.0242	3.0384	1.2183	47.6539	217.1088
2	8.1794	1.0004	−152.61	−133.963	−40.0058	−53.3014	−31.7692	45.2017
3	−13.0838	−0.0832	136.53	54.0415	25.812	−5.2988	−3.9067	62.9046
4	−14.029	−6.0072	61.144	186.788	97.2744	−9.0497	5.1317	−20.7534
5	82.012	−17.422	15.732	−65.9827	363.886	532.292	6.1269	−49.1541
6	−0.481	1.7411	−17.012	2.7393	−86.6964	−116.583	−13.1457	−18.978
7	−27.592	0.0778	−157.65	−23.1222	14.9759	19.6133	−45.9056	−51.7034
8	45.8897	26.604	100.19	39.1366	−58.38	−54.3333	−16.812	−29.8601
9	1.9125	−0.7573	−42.293	−79.6553	1.0574	−0.9486	−24.224	−6.6785
10	43.7316	0.5859	28.071	5.5977	−92.6572	−42.025	11.6064	62.5415

. The mathematical models obtained for K according to weight and threshold values are given in Equations (1)–(3). The most appropriate network structure for ANN modeling was determined as 6-10-1.

Equation (1), the logsig between the input and hidden layer for the amount of chlorophyll (K) produced as a result of the growth of the plant produced in 100% perlite, 100% peat, and 50% peat + 50% perlite environments, Equations (2) and (3) are the models determined according to the purelin activation functions used between the hidden layer and the output layer.

$$F_i={\displaystyle \frac{2}{1+e^{-[O.W_{OFi}+H.W_{HFi}+P.W_{PFi}+E.W_{EFi}+T.W_{TFi}+M.W_{MFi}+\theta i]}}}$$

(1)

$$C=\sum _{i=1}^{10}{F}_i.W_i+{\theta }_j$$

(2)

$$C=\sum _{i=1}^{10}\left({\displaystyle \frac{2}{1+e^{-[O.W_{OFi}+W_{HFi}+P.W_{PFi}+E.W_{EFi}+T.W_{TFi}+M.W_{MFi}+\theta i]}}}\right).W_i+{\theta }_j$$

(3)

In the ANN, tansig–purelin activation functions and Scaled Conjugate Gradient Backpropagation Algorithm (trainscg) learning algorithm were used for the root. The weight and threshold values obtained for R according to the activation functions and learning algorithms are given in

Table 2. Weight and threshold values obtained for R.

i	W1	W2	W3	W4	W5	W6	θi	Lw	θ11
1	1.7403	−0.028	4.1406	−1.0125	−0.2986	0.6381	8.4076	5	9.9929
2	5.7677	0.0099	0.2313	1.327	−1.4458	−1.8879	−14.3251	−1.4187
3	2.252	0.0401	−10.3116	−1.152	−0.323	−0.8172	−22.7695	2
4	1.4877	−0.0742	1.6346	1.3536	−0.5111	−0.2602	−13.177	7
5	−0.2448	−7.9738 × 10−11	−1.41 × 10−6	−5.91 × 10−7	1.22 × 10−7	1.13 × 10−7	13.731	−20.8338
6	9.3115	−0.3131	−0.4306	−0.4948	5.0959	10.7002	−2.0956	−5
7	0.6823	0.0345	−6.137	−2.2646	−0.8864	−1.7388	−17.1508	−1
8	1.2562	0.0072	1.9285	−0.1133	−1.034	1.8899	−17.6737	−5.4262
9	−3.1606	−0.0019	−2.856	1.9153	−1.2155	1.4472	14	3.4146
10	−0.5993	0.1124	0.799	−0.0701	0.0176	0.1962	−0.706	3

. The mathematical models obtained for R according to weight and threshold values are given in Equations (4)–(6). The most suitable network structure for ANN modeling was determined as 6-10-1.

Equation (4), the tansig between the input and the hidden layer for the amount of rooting (R) that occurs as a result of the growth of the plant produced in 100% perlite, 100% peat, and 50% peat + 50% perlite environments, Equations (5) and (6) are the models determined according to the purelin activation functions used between the hidden layer and the output layer.

$$F_i={\displaystyle \frac{2}{1+e^{-2[O.W_{OFi}+H.W_{HFi}+P.W_{PFi}+E.W_{EFi}+T.W_{TFi}+M.W_{MFi}+\theta i]}}}-1$$

(4)

$$R=\sum _{i=1}^{10}{F}_i.W_i+{\theta }_j$$

(5)

$$R=\sum _{i=1}^{10}\left({\displaystyle \frac{2}{1+e^{-2[O.W_{OFi}+H.W_{HFi}+P.W_{PFi}+E.W_{EFi}+T.W_{TFi}+M.W_{MFi}+\theta i]}}}-1\right).W_i+{\theta }_j$$

(6)

4. Discussion

This study comprehensively investigated the effects of different rooting media, indole butyric acid (IBA) concentrations, and measurement periods on the chlorophyll concentration and rooting rate of Loropetalum chinense var. rubrum Yieh cuttings, as well as the peat moisture, pH, and electrical conductivity of the rooting media. The findings indicate that rooting media and hormone dosages play important roles in the physiological development of cuttings.

Firstly, perlite medium was found to have the highest positive effect on both chlorophyll concentration and rooting rate compared to other media. Previous studies have determined that perlite media are balanced in terms of water holding capacity and air permeability, thereby enhancing rooting rates [68,69,70]. Similarly, in our study, the balanced water holding capacity of perlite may have provided a more favorable microclimate in the rooting process of the cuttings due to its ability to retain the required amount of water while draining excess water and balancing oxygen permeability. Nutrient uptake has been reported to be affected by the pH level in perlite media [71,72]. In addition, it has been found that the low electrical conductivity (EC) and optimal pH level of perlite allow for more efficient nutrient uptake, suggesting that these conditions may facilitate improved nutrient absorption [73,74,75,76]. The results are consistent with previous studies on perlite media and explain the high rooting success.

The lowest rooting rate and chlorophyll concentration were observed in peat medium. Studies have shown that the high water retention capacity of peat substrates may create unfavorable conditions for root growth by reducing air permeability [77,78,79]. It has been observed that rooting success and chlorophyll concentration are lower in environments with high peat content [80,81]. High electrical conductivity and low air permeability in peat media have been reported to limit or even inhibit root development and plant growth [82,83]. Based on these data found as a result of the literature research, the low rooting percentage in the peat medium used in the study can be explained by the high water holding capacity, low air permeability and high electrical conductivity values. These properties may have caused oxygen deficiency in the root zone and negatively affected root development. In addition, the high pH variance observed in the peat medium may have caused irregularities in nutrient uptake and negatively affected chlorophyll synthesis based on the literature information given above.

It has been reported that low doses of IBA (500–2000 ppm) can protect chlorophyll synthesis, whereas high doses may inhibit root development by increasing oxidative stress levels [84]. Similarly, the highest rooting percentages were obtained in low-concentration IBA media under conditions of oxidative stress [85,86,87]. High IBA concentrations have been reported to increase oxidative stress symptoms and suppress rooting by causing excessive cell growth [88]. In addition, different doses of IBA can enhance the synthesis of photosynthesis-related pigments [42,89]. Furthermore, it has been observed that 2000 ppm IBA application leads to the highest chlorophyll concentration [90]. When the effects of IBA concentrations were examined in the study, it was observed that 1000 ppm dosage provided the highest rooting rate. This may be attributed to the stimulating effect of IBA at low concentrations on the division and differentiation of root cells. In contrast, the lowest rooting rate was observed at 3000 ppm IBA treatment. As mentioned in the above references, the negative effects of high hormone concentrations may suppress the rooting process by inducing overgrowth and oxidative stress in the cells. Interestingly, 6000 ppm IBA concentration increased chlorophyll concentration to the highest level. This result indicates that high IBA concentration can induce chloroplast development and photosynthetic pigment synthesis.

A review of the literature on the variation in chlorophyll concentration according to environmental conditions indicates that chlorophyll concentration in plants reaches optimal values with increasing daylight hours (approximately 13–14 h per day) and air temperatures around 26 °C during summer [91,92,93,94].

When the measurement periods were analyzed, it was found that the chlorophyll concentration was the highest during the M7, M8, and M9 periods. These periods coincided with the summer months in Rize (specifically in the Fener district), during which extended daylight hours and optimal temperature conditions may have contributed to enhanced photosynthetic activity. This finding is consistent with the literature, supporting the direct influence of seasonal environmental changes on photosynthesis and chlorophyll production.

The study comprehensively measured and analyzed the peat pH, moisture content, temperature, and electrical conductivity of the rooting media across the experimental period. These physical and chemical properties were found to influence both chlorophyll concentration and rooting success. The positive effect of perlite and the negative effect of peat media on root and shoot development were explained based on their respective environmental characteristics, such as pH stability, optimal EC levels, and better air–water balance.

In addition to the experimental findings, this study incorporated predictive modeling using artificial neural networks (ANNs) to provide deeper insights into the complex relationships between environmental conditions, hormone treatments, and physiological responses of Loropetalum chinense var. rubrum cuttings. The ANN models were developed to predict chlorophyll concentration (C) and rooting success (R) using six input variables: rooting media type, pH, electrical conductivity (EC), temperature, humidity, and IBA concentration. A 6-10-1 network structure was found to be optimal for both outputs, with logsig–purelin and tansig–purelin activation functions for chlorophyll and rooting models, respectively.

Notably, the ANN model outputs were highly consistent with experimental results, validating its predictive performance. The model identified perlite medium and 1000 ppm IBA as optimal conditions for maximizing rooting success, aligning with statistical findings. Similarly, 6000 ppm IBA was found to enhance chlorophyll concentration, which was also reflected in the model predictions. This consistency between modeled and observed outcomes demonstrates the reliability of the ANN approach in capturing non-linear biological responses that might be obscured in traditional statistical methods.

Compared to conventional analyses such as ANOVA, which assess group-level mean differences, ANN models can reveal underlying interactions and predict outcomes under untested combinations of variables. This is particularly valuable in vegetative propagation, where success depends on the simultaneous effects of multiple, often interacting factors. The model also offers the advantage of continuous, real-time estimation of key physiological parameters, thus acting not only as a diagnostic tool but also as a potential decision-support system for nursery managers and restoration practitioners.

While the current model is species-specific, its architecture and parameterization are broadly adaptable. By retraining with appropriate datasets, the same ANN framework can be applied to other ornamental or woody species, particularly those propagated via cuttings. This opens pathways for generalizing the model into a scalable tool for diverse horticultural applications. Moreover, integrating larger datasets and additional variables such as nutrient availability, light intensity, or endogenous hormone levels could further enhance the model’s robustness and scope.

Additionally, although the study included graphical representations of the trends, tabular data presentation will be expanded in future supplementary files to improve transparency and allow reproducibility of the ANN training process. This will address the need for greater visibility of the underlying data volume that supports the modeling outcomes.

In conclusion, combining empirical measurements with ANN modeling enriched the understanding of root formation and photosynthetic activity in Loropetalum chinense var. rubrum propagation. The results emphasize the potential of ANN-based systems to complement experimental work, enhance predictive accuracy, and guide optimized propagation strategies under varying environmental and hormonal regimes.

5. Conclusions

The results obtained reveal the effects of physical and chemical properties of rooting media, IBA concentrations and seasonal variations on the physiological responses of Loropetalum chinense var. rubrum Yieh cuttings. In particular, perlite medium and 1000 ppm IBA concentration optimized rooting rates, while 6000 ppm IBA concentration was effective in increasing chlorophyll concentration. In future studies, it is recommended to investigate the effects of different rooting media and various hormone combinations to develop strategies to make the rooting process more efficient.

This study revealed that perlite was the most suitable medium for rooting and chlorophyll concentration. Optimizing the pH value of the rooting medium is a critical factor for enhancing both the rooting success and chlorophyll content of cuttings. Minimization of high humidity and EC is recommended to increase the nutrient uptake of cuttings. Increasing hormone concentration applied to Loropetalum chinense cuttings had an irregular effect on rooting number, but had positive effects on chlorophyll content. In practice, the optimum hormone concentration for rooting is recommended as 1000 ppm. This study has shown that media such as perlite can be effectively used for landscape architecture and plant propagation applications and 1000 ppm can be effectively used in appropriate hormone applications. The most appropriate network structure for ANN modeling was determined as 6-10-1. For chlorophyll content, logsig activation was determined between the input and hidden layers and pureline activation between the hidden and output layers. For rooting amount, tansig activation was determined between input and hidden layers and pureline activation between hidden and output layers.