Optimization of UV-A/PAR ratio for phenolic accumulation in kale using 3D radiation interception modeling

Abstract

Background

In vertical farms using light-emitting diode radiation, supplemental ultraviolet (UV)-A radiation at specific wavelengths and energy levels has been shown to enhance photosynthesis and biomass accumulation, indicating its potential as an additional light source. However, plant responses to supplemental UV-A vary depending on the intensity of photosynthetically active radiation (PAR). In this study, three-dimensional (3D) radiation interception analysis based on ray-tracing simulation was conducted to examine how the UV-A/PAR ratio affects phenolic content and antioxidant capacity of kale and to determine appropriate UV-A levels, under different photosynthetic photon flux densities (PPFD), selected based on previous studies. Kale transplants were grown for 3 weeks under 150 or 300 µmol m^− 2 s^− 1 PPFD, and supplemented with UV-A radiation (385 nm) at 15, 30, or 45 W m^− 2 for 6 days before harvest. Radiation interception was analyzed using a 3D plant model and ray-tracing simulation.

Results

Kale growth was substantially greater under 300 than under 150 µmol m^− 2 s^− 1 PPFD. Growth responses to supplemental UV-A radiation varied significantly with PPFD level. UV-A radiation increased shoot fresh weight and leaf area under 150 µmol m^− 2 s^− 1 PPFD, but not under 300 µmol m^− 2 s^− 1 PPFD at 6 days after the onset of treatment. Supplemental UV-A radiation also enhanced total phenolic content and antioxidant capacity, with the magnitude of these effects dependent on PPFD level. Under 150 µmol m^− 2 s^− 1 PPFD, the 15 W m^− 2 UV-A treatment enhanced phenolic content and antioxidant capacity, whereas no increase was observed under 300 µmol m^− 2 s^− 1 PPFD. Radiation interception analysis indicated that differences in absorbed radiation between the two PPFD conditions were primarily attributable to plant size. The correlation between UV-A radiation intensity and phenolic accumulation was stronger when based on intercepted radiation than on measured intensity. Notably, higher phenolic content and antioxidant capacity were observed at UV-A/PAR ratio ranging from 0.1 to 0.2.

Conclusions

These findings suggest that supplemental UV-A radiation should be optimized considering PPFD level and radiation interception to enhance phenolic content and antioxidant capacity in kale.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08522-w.

Background

Plants modulate their metabolism in response to environmental conditions to ensure growth, development, and survival. They respond at molecular and cellular levels, triggering physiological adjustments that influence morphology and metabolism. These processes collectively determine plant productivity, quality, and environmental adaptability [1, 2]. Abiotic stresses such as temperature, light, drought, salinity, and mineral imbalances pose significant challenges to plants and their responses vary with stress type, intensity, and duration. Moderate or tolerable levels of abiotic stress (often referred to as mild stress or eustress) can stimulate the biosynthesis of secondary metabolites without suppressing growth, thereby enhancing defense capacity [3]. Plant secondary metabolites (PSMs), including terpenoids, phenolics, and alkaloids, play critical roles in defense mechanisms, contribute to crop quality, and provide substantial benefits to human health. Notably, PSMs constitute approximately 25% of known pharmaceuticals and serve as valuable industrial materials. In Brassica vegetables such as kale contain diverse PSMs, including glucosinolates, phenolic compounds, and carotenoids, which contribute to nutritional and functional quality. Given the industrial relevance of PSMs, the strategic application of mild abiotic stress represents a promising approach for producing high-quality plants enriched in PSMs [4–6].

Ultraviolet (UV) radiation functions as an abiotic elicitor, because its high-energy photons can generate reactive oxygen species (ROS) in plants and trigger photoreceptor-mediated responses through cryptochromes, phototropins, and UV‒B resistance 8 (UVR8). These responses promote the accumulation of antioxidant phenolics as part of the plant defense mechanisms [7, 8]. Flavonoids, which constitute approximately half of all phenolics, accumulate in epidermal cells in response to UV radiation exposure, where they absorb harmful radiation and quench UV-induced oxidative stress [9]. UV radiation is categorized into UV-A (315–400 nm), UV-B (280–315 nm), and UV-C (< 280 nm) according to wavelength. Among these, UV-A has relatively lower photon energy than UV-B and UV-C, making it safer and more suitable for agricultural and industrial applications [10]. For example, in vertical farming systems using light-emitting diodes (LEDs) as primary light sources, supplemental UV-A radiation has been shown to enhance biomass and PSM accumulation [11–13], suggesting its potential as an additional light source for crop production [10, 14, 15].

However, not all studies have reported a significant increase in PSMs in response to UV radiation. Some have observed no change or even a reduction in PSM levels. For example, Bernal et al. [16] applied UV-A and UV-B radiation to woody Mediterranean species and observed no increase in carotenoid or phenolic contents, whereas Wilson et al. [17] reported decreased flavonoid contents in Brassica napus leaves in response to UV-A radiation. We hypothesized that these inconsistencies arise from two main factors. The first is variation in background photosynthetic photon flux density (PPFD) and differences in UV radiation intensity among studies. Fan et al. [18] demonstrated that leaf thickness and mesophyll development depend on light intensity. Previous studies have also shown that plants acclimate leaf structure to optimize photosynthetic efficiency under different light environments [19–21]. For example, under high PPFD conditions, plants typically develop thicker epidermal tissues to protect against excessive radiation, resulting in increased leaf thickness [22]. Such structural changes influence UV radiation absorption and transmittance, and plants grown under low PPFD tend to exhibit greater sensitivity to UV stress [23, 24]. The second factor involves methodological limitations in previous studies, where optimal UV radiation conditions were often determined based solely on incident UV radiation intensity measured on a two-dimensional (2D) surface [11, 25–27]. This approach does not account for the three-dimensional (3D) architecture of plants, including variation in UV radiation interception among leaves at different orientations.

Therefore, in this study, we investigated the effects of supplemental UV-A radiation under different PPFD levels on growth characteristics, phenolic accumulation, and antioxidant capacity of kale (Brassica oleracea L. var. acephala). We quantified photosynthetically active radiation (PAR) and UV radiation interception at the whole-plant level using 3D structural modeling and ray-tracing simulation (Fig. 1), compared these values with conventional 2D radiation intensity measurements, and analyzed how radiation interception influences the phenolic accumulation. We hypothesized that the magnitude of UV-A-induced phenolic accumulation is fundamentally contingent upon background PPFD, which drives the requisite structural and physiological acclimation of the leaves. Furthermore, we posited that actual radiation interception, quantified through 3D structural modeling, effectively accounts for the spatial complexities of plant architecture, thereby providing a significantly more robust predictive model for phenolic biosynthesis than conventional 2D incident radiation metrics.

Fig. 1.

Workflow of radiation interception analysis based on 3D scanned models for kale. Kale plants grown in a controlled environment module were 3D scanned to generate mesh representations, which were subsequently segmented and reconstructed into detailed surface models. These models were integrated into a virtual growth chamber, where ray-tracing simulations were performed to calculate light energy distribution and quantify radiation interception across five kale specimens per treatment

Results

Growth characteristics

Growth of kale transplants were markedly greater under 300 µmol m^− 2 s^− 1 PPFD than under 150 µmol m^− 2 s^− 1 PPFD (Fig. 2A, B). The growth response to supplemental UV-A radiation varied significantly with PPFD level. At 6 days after the onset of UV-A treatment, UV-A radiation increased shoot fresh weight (FW) under 150 µmol m^− 2 s^− 1 PPFD, but no such increase was observed under 300 µmol m^− 2 s^− 1 PPFD (Fig. 2A). Under 150 µmol m^− 2 s^− 1 PPFD, UV-A treatments at 15, 30, and 45 W m⁻² increased shoot fresh weight by 28, 49, and 16%, respectively, compared with the control. In contrast, under 300 µmol m^− 2 s^− 1 PPFD, the control exhibited the greatest shoot FW, whereas all UV-A treatments showed 4‒27% lower values.

Fig. 2.

Shoot fresh weight (A), leaf area (B), and representative images (C) of kale plants grown under different PPFD and UV-A radiation levels. Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Treatments are denoted as PPFD–UV-A intensity (e.g., 150–0, 150–15, 300–0, etc.), with 0 representing the control without UV-A radiation. Within each PPFD level, different lowercase letters (a-c) indicate significant differences among UV-A intensity levels, whereas within each UV-A intensity level, different uppercase letters (A-B) indicate significant differences between PPFD levels (150 vs. 300) (one-way ANOVA followed by LSD test, P < 0.05; n = 5)

A similar trend was observed for leaf area measured at 6 days after the onset of UV-A treatment (Fig. 2B). Under 150 µmol m^− 2 s^− 1 PPFD, all UV-A treatments significantly increased leaf area relative to the control. However, under 300 µmol m^− 2 s^− 1 PPFD, the control had the largest leaf area, and leaf area decreased with increasing UV-A radiation intensity, showing the lowest value with 45 W m⁻² UV-A radiation. These results were further supported by the representative images shown in Fig. 2C.

Maximum photochemical efficiency of photosystem II (PSII)

The maximum photochemical efficiency of PSⅡ [variable fluorescence (Fv)/maximum fluorescence (Fm)] was highly sensitive to supplemental UV-A radiation (Fig. 3). In both control groups under 150 and 300 µmol m^− 2 s^− 1 PPFD, Fv/Fm values remained above 0.8 throughout the 6-day UV-A treatment period. However, under 150 µmol m^− 2 s^− 1 PPFD, Fv/Fm values dropped below 0.8 at 1 day after the onset of the 30 and 45 W m⁻² UV-A treatments. By 2 days after the onset of all UV-A treatments, Fv/Fm values declined below 0.8, regardless of PPFD level. The decline in Fv/Fm values became more pronounced with increasing UV-A intensity, showing the lowest value with 45 W m⁻² UV-A treatments at 4 and 6 days after the onset of UV-A treatments under both PPFD levels.

Fig. 3.

Maximum photochemical efficiency of photosystem II (Fv/Fm) in kale plants grown under different PPFD and UV-A radiation levels. Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Treatments are denoted as PPFD–UV-A intensity (e.g., 150–0, 150–15, 300–0, etc.), with 0 representing the control without UV-A radiation (n = 5)

Total phenolic content

Supplemental UV-A radiation tended to increase the total phenolic content per gram of leaf dry weight (DW) under both PPFD levels (Fig. 4A). Total phenolic content was significantly higher in the 30 and 45 W m⁻² UV-A treatments than in the control under both PPFD levels. Under 150 µmol m^− 2 s^− 1 PPFD, even the 15 W m⁻² UV-A treatment significantly increased total phenolic content, whereas under 300 µmol m^− 2 s^− 1 PPFD, a significant difference was observed only at 4 days after the onset of UV-A treatment.

Fig. 4.

Violin plots of total phenolic content per gram of leaf dry weight (A) and per shoot dry weight (B) of kale plants grown under different PPFD and UV-A radiation levels. Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Treatments are denoted as PPFD–UV-A intensity (e.g., 150–0, 150–15, 300–0, etc.), with 0 representing the control without UV-A radiation. Within each PPFD level, different lowercase letters (a-c) indicate significant differences among UV-A intensity levels, whereas within each UV-A intensity level, different uppercase letters (A-B) indicate significant differences between PPFD levels (150 vs. 300) (one-way ANOVA followed by LSD test, P < 0.05; n = 5)

Total phenolic content per shoot DW was higher under 300 µmol m^− 2 s^− 1 PPFD than under 150 µmol m^− 2 s^− 1 PPFD (Fig. 4B). At 6 days after the onset of UV-A treatment, the 30 W m⁻² UV-A treatment resulted in the highest values under both PPFD levels. Under 150 µmol m^− 2 s^− 1 PPFD, the 15 and 45 W m⁻² UV-A treatments also showed significantly higher values than the control, whereas under 300 µmol m^− 2 s^− 1 PPFD, no significant difference was observed between the control and the 45 W m⁻² UV-A treatment.

Antioxidant capacity

Supplemental UV-A radiation tended to increase antioxidant capacity per gram of leaf and shoot DW under both PPFD levels (Fig. 5), similar to the trend observed for total phenolic content (Fig. 4). Under 150 µmol m^− 2 s^− 1 PPFD, antioxidant capacity per gram of leaf DW was significantly higher in all UV-A treatments than in the control (Fig. 5A). The increase generally became more pronounced with higher UV-A intensity, although no significant difference was found between the 15 and 30 W m⁻² UV-A treatments. Under 300 µmol m^− 2 s^− 1 PPFD, the 30 and 45 W m⁻² UV-A treatments significantly increased antioxidant capacity at 4 and 6 days after the onset of UV-A treatment, with no significant difference between the two (Fig. 5A). In contrast, the 15 W m⁻² UV-A treatment did not increase antioxidant capacity until 6 days after the onset of UV-A treatment.

Fig. 5.

Violin plots of antioxidant capacity per gram of leaf dry weight (A) and per shoot dry weight (B) of kale plants grown under different PPFD and UV-A radiation levels. Two-week-old plants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Treatments are denoted as PPFD–UV-A intensity (e.g., 150–0, 150–15, 300–0, etc.), with 0 representing the control without UV-A radiation. Within each PPFD level, different lowercase letters (a-c) indicate significant differences among UV-A intensity levels, whereas within each UV-A intensity level, different uppercase letters (A-B) indicate significant differences between PPFD levels (150 vs. 300) (one-way ANOVA followed by LSD test, P < 0.05; n = 5)

Antioxidant capacity per shoot DW was higher under 300 µmol m^− 2 s^− 1 PPFD than under 150 µmol m^− 2 s^− 1 PPFD (Fig. 5B). Under 150 µmol m^− 2 s^− 1 PPFD, all UV-A treatments significantly increased antioxidant capacity compared with the control, with the magnitude of increase generally rising with UV-A intensity, except at 6 days after the onset of UV-A treatment. Under 300 µmol m^− 2 s^− 1 PPFD, a significant increase in antioxidant capacity was observed only at 6 days after the onset of UV-A treatment. The highest increase occurred in the 30 W m⁻² UV-A treatment, whereas antioxidant capacity tend to be lower in the 45 W m⁻² UV-A treatment than in the control.

Statistical significance of measured parameters

Two-way analysis of variance (ANOVA) revealed that PPFD, UV-A, and PPFD × UV-A interaction significantly affected shoot FW of kale (P < 0.001), except at 4 days after the onset of UV-A treatment (Table 1). UV-A treatment also had a significant effect on Fv/Fm throughout the entire treatment period (P < 0.001), whereas the effects of PPFD and PPFD × UV-A interaction were significant only at 6 days after the onset of UV-A treatment (Table 1). Similarly, total phenolic content and antioxidant capacity per gram of leaf DW were significantly affected by PPFD and UV-A, respectively, except for the PPFD effect at 6 days after the onset of UV-A treatment. Furthermore, total phenolic content and antioxidant capacity per shoot DW were significantly influenced not only by PPFD and UV-A individually, but also by their interaction throughout the UV-A treatment period (Table 1).

Table 1.

Two-way ANOVA of shoot fresh weight and dry weight, maximum photochemical efficiency of photosystem II (Fv/Fm), total phenolic content, and antioxidant capacity of kale grown under different PPFD (µmol m^− 2 s^− 1), UV-A radiation intensities (W m^− 2), and their interaction. Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days

Parameter	Treatment	Significance
		2 DAT	4 DAT	6 DAT
Shoot fresh weight	PPFD	***	***	***
	UV-A	***	ns	***
	PPFD × UV-A	***	***	***
Shoot dry weight	PPFD	***	***	***
	UV-A	***	ns	***
	PPFD × UV-A	***	***	***
Maximum photochemical efficiency of photosystem II	PPFD	ns	ns	***
	UV-A	***	***	***
	PPFD × UV-A	ns	ns	*
Total phenolic content per g leaf dry weight	PPFD	***	***	ns
	UV-A	***	***	***
	PPFD × UV-A	ns	ns	ns
Total phenolic content per shoot dry weight	PPFD	***	***	***
	UV-A	***	*	***
	PPFD × UV-A	**	***	**
Antioxidant capacity per g leaf dry weight	PPFD	***	***	ns
	UV-A	***	***	***
	PPFD × UV-A	ns	ns	ns
Antioxidant capacity per shoot dry weight	PPFD	***	***	***
	UV-A	***	*	***
	PPFD × UV-A	**	***	*

DAT Days after treatment

*, **, ***Significant at P < 0.05, 0.01, and 0.001, respectively, according to two-way ANOVA and least significant difference (LSD) test

ns Not significant

Radiation interception on leaves

PAR and UV-A radiation intensities measured at the cultivation bed surface closely matched their set values (Table 2). Ray-tracing simulation using the 3D model for kale showed that intercepted radiation decreased by an average of 24% under 150 µmol m^− 2 s^− 1 PPFD compared with the measured values. Under 300 µmol m^− 2 s^− 1 PPFD, where plants were larger than those grown under 150 µmol m^− 2 s^− 1 PPFD, intercepted radiation was on average 58% lower than the measured values. The discrepancy between measured and intercepted values was smaller for UV-A radiation than for PAR intensity, resulting in a slightly higher UV-A/PAR ratio in the intercepted values.

Table 2.

Set, measured, and intercepted values of PAR (µmol m^− 2 s^− 1), UV-A radiation (W m^− 2), and UV-A/PAR ratio in the controlled environment module used for kale cultivation

Set value			Measured value			Intercepted value
PAR	UV-A	UV-A/PAR	PAR	UV-A	UV-A/PAR	PAR	UV-A	UV-A/PAR
150	0	0	150.2 ± 12.5	0	0	135.9 ± 4.7	0	0
150	15	0.1	152.1 ± 11.3	15.8 ± 3.4	0.10 ± 0.022	104.3 ± 10.1	11.0 ± 1.1	0.11 ± 0.002
150	30	0.2	148.6 ± 14.4	32.2 ± 2.6	0.23 ± 0.042	110.9 ± 60.3	24.3 ± 2.6	0.22 ± 0.008
150	45	0.3	150.6 ± 11.1	41.8 ± 7.0	0.30 ± 0.066	105.7 ± 11.4	34.6 ± 3.7	0.33 ± 0.008
300	0	0	301.9 ± 27.4	0	0	120.6 ± 24.7	0	0
300	15	0.05	301.7 ± 30.5	15.7 ± 2.3	0.05 ± 0.009	120.6 ± 22.7	7.2 ± 1.1	0.06 ± 0.002
300	30	0.1	299.9 ± 25.1	30.0 ± 1.7	0.10 ± 0.009	120.6 ± 9.8	14.3 ± 1.1	0.12 ± 0.003
300	45	0.15	297.7 ± 33.1	45.9 ± 5.0	0.16 ± 0.024	138.7 ± 15.7	24.4 ± 3.7	0.18 ± 0.006

Measured value represents the averages obtained at the cultivation bed surface, 30 cm below the light source, whereas intercepted values correspond to radiation interception outputs generated from the 3D model for kale

The spatial distribution of PAR and UV-A radiation interception was visualized in the 3D model at 6 days after the onset of UV-A treatment (Fig. 6). PAR interception showed that leaves near the growing point, closest to the LED light source, received the highest PPFD, whereas lower, shaded leaves received substantially less PPFD (Fig. 6A). Under 300 µmol m^− 2 s^− 1 PPFD, both the spatial variation and the magnitude of intercepted radiation were greater than under 150 µmol m^− 2 s^− 1 PPFD. The upper leaf surface also received more UV-A radiation in the 350–400 nm range (Fig. 6B). As expected, no UV-A radiation interception was observed in the controls without UV-A radiation exposure under either PPFD level. Overall, UV-A radiation interception increased with UV-A intensity, indicating greater absorption of UV-A energy by the leaves at higher UV-A radiation intensities.

Fig. 6.

Representative radiation interception distributions of PPFD (400–700 nm) (A) and UV-A energy (350–400 nm) (B) on 3D scanned models of kale plants grown under different PPFD and UV-A radiation levels. Two-week-old plants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Treatments are denoted as PPFD–UV-A intensity (e.g., 150–0, 150–15, 300–0, etc.), with 0 representing the control without UV-A radiation

Regression analysis of phenolic content and antioxidant capacity according to UV-A/PAR

Nonlinear regression analysis of total phenolic content (Fig. 7) and antioxidant capacity (Fig. 8) per unit gram of leaf DW was performed at 6 days after the onset of UV-A treatment. Regression results for UV-A/PAR based on measured 2D values were compared with those based on intercepted radiation obtained from 3D simulations. For total phenolic content, the coefficient of determination (R²) values based on measured UV-A/PAR were 0.56 and 0.48 under 150 and 300 µmol m^− 2 s^− 1 PPFD, respectively, and 0.49 when data from both PPFD levels were combined (Fig. 7A). The corresponding R² values based on intercepted UV-A/PAR (0.63, 0.58, and 0.54, respectively) were higher than those based on measured UV-A/PAR (Fig. 7B). For antioxidant capacity, the R² value based on intercepted UV-A/PAR was 0.59 under 150 µmol m^− 2 s^− 1 PPFD, higher than that based on measured UV-A/PAR (0.52) (Fig. 8A, B). Under 300 µmol m^− 2 s^− 1 PPFD, however, a higher R² (0.64) based on measured UV-A/PAR was obtained than that based on intercepted UV-A/PAR (0.61). For the combined data under both PPFD levels, R² values were identical at 0.44 (Fig. 8A, B).

Fig. 7.

Nonlinear regression of total phenolic content of kale according to UV-A/PAR ratio based on measured (A) and intercepted radiation values (B). Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Symbols indicate biological replicates (blue dots, 150 PPFD; red dots, 300 PPFD; n = 20 per PPFD). Solid lines represent PPFD-specific nonlinear regression fits (blue line, 150 PPFD; red line, 300 PPFD), and the green dashed line represents the pooled (combined) fitted curve using data from both PPFD levels (n = 40). R², coefficient of determination. ***Significant at P < 0.001

Fig. 8.

Nonlinear regression of antioxidant capacity of kale according to UV-A/PAR ratio based on measured (A) and intercepted radiation values (B). Two-week-old transplants were grown under 150 or 300 µmol m⁻² s⁻¹ PPFD for 3 weeks and then exposed to supplemental UV-A radiation at 0, 15, 30, or 45 W m⁻² for 6 days. Symbols indicate biological replicates (blue dots, 150 PPFD; red dots, 300 PPFD; n = 20 per PPFD). Solid lines represent PPFD-specific nonlinear regression fits (blue line, 150 PPFD; red line, 300 PPFD), and the green dashed line represents the pooled (combined) fitted curve using data from both PPFD levels (n = 40). R², coefficient of determination. ***Significant at P < 0.001

Regression curves based on intercepted radiation for both total phenolic content and antioxidant capacity generally increased with the UV-A/PAR ratio up to approximately 0.1–0.2, after which they reached a plateau (Figs. 7B and 8B). This pattern was apparent when analyzing combined data under both PPFDs. In contrast, under 300 µmol m^− 2 s^− 1 PPFD alone, regression based on intercepted radiation exhibited a linear relationship between UV-A/PAR and both parameters.

A multiple regression model using Gaussian equations was developed to predict total phenolic content and antioxidant capacity per shoot DW as a function of total leaf area and UV-A/PAR ratio (Fig. 9). The model demonstrated high accuracy, with R² = 0.95 for both parameters. To evaluate potential overfitting, the dataset (n = 40) was randomly split into training (80%) and test (20%) subsets. Similar performance was observed between the training and test sets (total phenolic content: R² = 0.95 vs. 0.91; antioxidant capacity: R² = 0.95 vs. 0.90), indicating adequate generalization and limited overfitting. In addition, both total phenolic content and antioxidant capacity tended to increase with increasing total leaf area.

Fig. 9.

Response surface multiple nonlinear regression of total phenolic content (A) and antioxidant capacity (B) per shoot dry weight of kale according to the simulated UV-A/PAR ratio and total leaf area at 6 days after the onset of UV-A treatment (n = 40). A second-order polynomial model was fitted by ordinary least squares, and the fitted 3D response surface is shown together with the observed data points (gray). R², the coefficient of determination. ***Significant at P < 0.001

Discussion

All UV-A treatments (15, 30, and 45 W m^{− 2}) under 150 µmol m^{− 2} s^{− 1} PPFD increased the shoot FW of kale from 2 to 6 days after the onset of UV-A treatment, whereas no biomass enhancement was observed under 300 µmol m^{− 2} s^{− 1} PPFD (Fig. 2B). These findings indicate that the background PAR intensity is a critical factor contributing to the inconsistent growth responses to supplemental UV-A radiation. Under sub-saturating PPFD levels, supplemental UV-A radiation can enhance photosynthetic activity and thereby promote biomass accumulation [8]. Under such conditions, UV-A radiation is absorbed by photosynthetic pigments, such as chlorophylls and carotenoids [28]. In addition, UV-A-excited phenolics re-emit blue-green fluorescence, which provides additional excitation energy to photosynthetic pigments [29, 30]. Cryptochromes activated by this blue fluorescence also facilitates stomatal opening, improving CO₂ uptake and photosynthesis [31]. Conversely, under sufficiently high PPFD levels, excessive UV-A radiation has been hypothesized, based on previous studies, to damage the PSII protein complex by inducing dissociation of the D1 and D2 proteins from the PSII reaction center, thereby reducing the maximum photochemical efficiency and electron transport rate [32, 33].

Although both PPFD levels used in this study (150 and 300 µmol m^{− 2} s^{− 1}) were below the reported light saturation point of kale (750–1,000 µmol m^{− 2} s^{− 1} PPFD) [34], supplemental UV-A radiation under 150 µmol m^{− 2} s^{− 1} PPFD may enhance biomass accumulation by stimulating photosynthesis rather than inducing UV radiation stress. The use of 385-nm UV-A radiation, which lies close to the PAR wavelength range, might further contribute to this positive response. In contrast, under 300 µmol m^{− 2} s^{− 1} PPFD, supplemental UV-A radiation suppressed growth relative to the control (Fig. 2A), indicating that UV-A radiation induced stress during rapid growth phase. As a result, carbon assimilates were likely diverted toward defense mechanisms rather than leaf expansion, which may account for the reduction in leaf area (Fig. 2B).

Maximum photochemical efficiency of PSⅡ (Fv/Fm) has been used as an indirect indicator of photosynthetic capacity. Typically, plant stress can be assessed based on a reference Fv/Fm range of 0.80–0.83. In this study, the control groups maintained Fv/Fm values above 0.80, whereas all UV-A-treated plants exhibited the values below 0.80 at 2 days after the onset of UV-A treatment (Fig. 3). This reduction indicates that supplemental UV-A treatment induced photoinhibition, thereby reducing photochemical efficiency. Stronger UV-A radiation caused a proportionally greater decline in Fv/Fm. Although supplemental UV-A radiation promoted growth under 150 µmol m^− 2 s^− 1 PPFD, but inhibited growth under 300 µmol m^− 2 s^− 1 PPFD, reductions in photochemical efficiency were consistently proportional to UV-A radiation intensity under both PPFD levels. This suggests that supplemental UV-A radiation activates photoprotective mechanisms independently of background PPFD level, which in turn affects photochemical efficiency.

Supplemental UV-A radiation can affect the accumulation of bioactive compounds in kale leaves through two primary mechanisms. First, UV-A radiation energy can generate ROS, which induce oxidative stress and subsequently stimulate antioxidant synthesis to quench ROS [35–37]. Second, UV-A radiation activates biosynthetic pathways mediated with photoreceptors, such as cryptochrome, phototropin, and UVR8, leading to phenolic accumulation in the epidermal tissues [8, 38]. These phenolics function both as UV-absorbing molecules and as antioxidants [39].

In this study, total phenolic content and antioxidant capacity significantly increased in most UV-A treatments under both PPFD levels (Figs. 4 and 5). Overall, both total phenolic content and antioxidant capacity were higher under 300 µmol m^{− 2} s^{− 1} PPFD than under 150 µmol m^{− 2} s^{− 1} PPFD. However, the UV-A radiation intensity required to induce these increases depended on the PPFD level. For example, under 150 µmol m^{− 2} s^{− 1} PPFD, the 15 W m^{− 2} UV-A treatment increased total phenolic content by approximately 11% on a leaf DW basis and 9% on a shoot DW basis compared with their respective controls (Fig. 4). In contrast, under 300 m^{− 2} PPFD, the 15 W m^{− 2} UV-A treatment did not produce a significant increase. Cen and Bornman [24] reported that beans (Phaseolus vulgaris) grown under low PPFD exhibited greater sensitivity to UV stress due to PPFD-dependent changes in leaf structure. Under low PPFD, leaves tend to have thinner and less dense epidermal tissues, allowing deeper penetration of UV radiation into mesophyll tissues [23]. Conversely, high PPFD increases leaf thickness, reducing UV sensitivity [18, 40, 41]. Consistent with these observations, the specific leaf weight measured before UV-A treatment was approximately 1.3-fold higher under 300 µmol m^{− 2} s^{− 1} PPFD than under 150 µmol m^{− 2} s^{− 1} PPFD (data not shown), suggesting that PPFD-induced increases in leaf thickness contribute to reduced UV-A sensitivity.

Most studies on plant responses to the light environment have measured radiation intensity on a 2D surface, either at the cultivation bed or at canopy height relative to the light sources. However, in real cultivation systems, preset radiation intensities do not necessarily represent the actual radiation intercepted by leaves, because light is distributed within a 3D space surrounding the canopy [42]. Whole-plant radiation interception is strongly influenced by 3D architecture and morphology of the plant [43, 44]. Therefore, quantifying the radiation that plants actually intercept is crucial for accurately evaluating their responses to light environments.

In this study, kale plants grown under 300 µmol m^{− 2} s^{− 1} PPFD developed larger canopies and generated more self-shading than those grown under 150 µmol m^{− 2} s^{− 1} PPFD (Fig. 2). As a result, the vertical distance to the light source decreased and the irradiation angle narrowed, causing considerable variation in radiation interception among leaves depending on their spatial position (Fig. 6). Consequently, radiation interception per unit leaf area decreased by approximately 28% more under 300 µmol m^{− 2} s^{− 1} PPFD than under 150 µmol m^{− 2} s^{− 1} PPFD (Table 2). Radiation interception also varies with wavelength due to differences in leaf reflectance and transmittance properties [45]. UV-A radiation generally penetrates more deeply into the canopy due to higher leaf transmittance. Yoon et al. [46] reported a smaller vertical gradient for UV radiation interception compared with PAR. Consistent with this, PAR decreased by approximately 9% more than UV-A radiation when comparing measured and intercepted values in this study (Table 2).

Previous studies using 3D plant models and ray-tracing simulations have primarily examined PAR interception [42, 47, 48]. Yoon et al. [49] extended this approach to UV-B radiation and demonstrated a relationship between leaf-position-dependent UV-B radiation interception and phenolic distribution in kale. In this study, we established the relationships between UV-A/PAR ratio and both phenolic content and antioxidant capacity (Fig. 9), and demonstrated the potential of 3D radiation interception analysis as a fundamental tool for optimizing plant quality through supplemental UV-A radiation.

Since it is not feasible to scan every plant, only a limited number of scanned individuals can be used as representative models for simulation. This may not fully capture the morphological variation present in a population. Since plant morphology varies with planting density, growth stage, and environmental conditions, it is important to include a sufficiently diverse range of scanned individuals. Despite these limitations, this approach has strong potential for improving production strategies in controlled-environment agriculture and vertical farming systems, provided that future research addresses these challenges.

Regression analysis of total phenolic content per gram of leaf DW at 6 days after the onset of UV-A treatment showed a higher R² value when based on intercepted UV-A/PAR than on measured UV-A/PAR (Fig. 7). In contrast, this trend was less apparent for antioxidant capacity (Fig. 8). This may be attributed to the wider distribution range of antioxidant capacity values and greater variability associated with UV-A radiation and PAR intensities, which limited improvements in R² even when using intercepted values. Nevertheless, under 150 µmol m^− 2 s^− 1 PPFD, the R² value based on intercepted UV-A/PAR (0.60) remained higher than that based on the measured UV-A/PAR (0.52), indicating that intercepted radiation may serve as a more effective parameter than conventional radiation measurements for analyzing plant light responses.

Under 300 µmol m^− 2 s^− 1 PPFD, the relationships between the intercepted UV-A/PAR ratio and both phenolic content and antioxidant capacity were linear rather than a saturating (Figs. 7 and 8). This suggests that photoprotective regulation, resistance to photooxidative stress, and adjustment in photosynthetic efficiency differ depending on PPFD level. Therefore, the increases in phenolic content and antioxidant capacity appear to be driven not only by the relative proportion of UV-A radiation but also by the absolute light environment and the associated photophysiological responses of plants.

Regression analysis suggested a favorable UV-A/PAR ratio of approximately 0.1–0.2 for maximizing bioactive compound accumulation per unit of light energy. Experimentally, the 15 and 30 W m^− 2 UV-A treatments under 150 µmol m^− 2 s^− 1 PPFD, and the 30 and 45 W m^− 2 UV-A treatments under 300 µmol m^− 2 s^− 1 PPFD fell within this optimal range and significantly increased total phenolic content and antioxidant capacity per gram of leaf DW compared with their respective controls (Figs. 4A and 5A). Furthermore, a multiple regression model incorporating leaf area and UV-A/PAR ratio accurately predicted total phenolic content and antioxidant capacity per shoot DW (Fig. 9), demonstrating very high predictive performance.

Conclusions

This study demonstrated that responses of kale to different UV-A radiation intensities, particularly changes in phenolic content and antioxidant capacity, varied depending on background PPFD level. Furthermore, clearer relationships between UV-A/PAR ratio and both phenolic content and antioxidant capacity were observed when radiation interception was quantified using a 3D plant model, highlighting the value of 3D radiation interception simulation in controlled-environment light research. A UV-A/PAR ratio of approximately 0.1–0.2 appeared to be associated with enhanced bioactive compound accumulation. Future work should focus on developing more efficient methods for constructing 3D plant models. Additionally, if real-time leaf area imaging (RGB imaging or depth sensing) or tracking becomes feasible, bioactive compound accumulation could be predicted using leaf area information in combination with UV-A/PAR ratio.

Materials and methods

Plant materials and growth conditions

Kale seeds (Asia Seed Co., Ltd., Seoul, Republic of Korea) were germinated on urethane cubes (25 × 25 × 30 mm; L × W × H) in trays under white LED radiation (D&W, Gwangmyeong, Republic of Korea) at 200 µmol m^{− 2} s^{− 1} PPFD. At 2 weeks after germination, uniform seedlings were transplanted into a controlled environment module maintained at 20℃, 60% relative humidity and 550 µmol mol^{− 1} CO₂ concentration. Transplants were grown under white LED radiation at 150 or 300 µmol m^{− 2} s^{− 1} PPFD with a 12/12 h (light/dark) photoperiod for 3 weeks. They were supplied with Hoagland nutrient solution [50] at an electrical conductivity (EC) of 1.2 dS m^{− 1} and a pH of 6.0, adjusted every 3 days using a multiparameter meter (Multi 3430, WTW, Weilheim, Germany).

UV-A treatments

At 3 weeks after transplanting, kale transplants grown under 150 or 300 µmol m^{− 2} s^{− 1} PPFD were treated with supplemental UV-A radiation at 15, 30, and 45 W m⁻² for 6 days, following the method described in our previous report [11]. UV-A LEDs (Bissol LED, Seoul, Republic of Korea) with a peak wavelength of 385 nm were used (Fig. S1), while control plants received only white LED radiation. UV-A radiation intensity for each treatment was measured at nine positions on the cultivation bed surface, adjusted to the plant canopy level (30 cm below the light source), using a spectroradiometer (Jaz-EL 200, Ocean Optics, Dunedin, FL, USA), and the mean value was applied for analysis. In some cases, tissue samples collected after each treatment were immediately frozen in liquid nitrogen and stored in a deep freezer at ‒70 °C.

Measurement of growth parameters

Shoots (excluding petioles) were collected at 0, 2, 4, and 6 days after the onset of supplementary UV-A radiation treatments. Shoot and leaf FWs were measured using an electronic scale (SI-234, Denver Instruments, Denver, CO, USA). Samples were then freeze-dried for 3 days using a freeze dryer (Alpha 1‒4 LSC Plus, Martin Christ Co., Osterode, Germany) to determine their DWs. Leaf area was measured with a leaf area meter (LI-3100, LI-COR Inc., Lincoln, NE, USA).

Assessment of photosynthetic efficiency

Photosynthetic efficiency was assessed by measuring chlorophyll a fluorescence on fully expanded leaves using a PAM chlorophyll fluorometer (Fluorpen FP 110, PSI, Drasov, Czech Republic) following a 30-min dark adaptation. Minimum fluorescence (Fo) was recorded following excitation with a low-intensity measuring beam and Fm was determined by applying a saturating light pulse of 1,500 µmol m^− 2 s^− 1 to close all PSII reaction centers. The maximum photochemical efficiency of PSII (Fv/Fm) was calculated as (Fm ‒ Fo)/Fm.

Determination of total phenolic content and antioxidant capacity

For biochemical assays (total phenolic content and antioxidant capacity), biological replicates (n = 5) were independent seedlings. Each biological replicate was extracted and assayed once (no technical replicates). For each biological replicate, frozen tissue samples were uniformly powdered using a milling tube (MMT 40.1, IKA-Werke GmbH & Co. KG, Staufen, Germany). Approximately 40 mg of powdered tissue from each treatment was mixed with 4 mL of 80% acetone and extracted for 15 min in a sonicator (SK5210HP, KUDOS, Shanghai, China). The resulting extract was incubated at 4 and ‒20°C for 12 h for the determination of phenolic content and antioxidant capacity, respectively. Total phenolic content and antioxidant capacity were determined using the Folin-Ciocalteu reagent method [51] and the ABTS (2,2’-azino-bis-3-ethylbenzothiazoline-6-sulfonic acid) assay [52], respectively, with slight modifications. Absorbance of the final reaction mixtures was measured at 765 and 730 nm using a spectrophotometer (Ultraviolet-1800, Shimadzu, Kyoto, Japan) [53]. Total phenolic content was expressed as mg gallic acid equivalent (GAE) per gram DW using gallic acid (Acros Organics, Geel, Belgium) as a standard. Antioxidant capacity was expressed as mM Trolox (6-hydroxy-2,5,7,8-tetramethylchroman-2-carboxyl acid) equivalent antioxidant capacity (TEAC) per gram DW using Trolox (Sigma-Aldrich Co., LLC, St. Louis, MO, USA) as a standard. Total phenolic content and antioxidant capacity were expressed per g DW to indicate metabolite concentration and per shoot DW to indicate whole-plant accumulation considering treatment-induced biomass changes.

3D analysis of radiation interception

Three plants per treatment were randomly selected at 6 days after the onset of supplemental UV-A treatment and scanned using a high-resolution portable 3D scanner (GO! SCAN50™, Creaform, Lévis, Canada) with a resolution of 2 mm to create 3D models. Scan data were exported as 3D mesh files using the accompanying software (VXelements, Creaform), corrected for imperfections such as holes and noise using Geomagic Design X software (3D Systems, Rock Hill, SC, USA), and reconstructed into surface models for ray-tracing simulations. Leaf and material transmittance and reflectance were measured using a spectrophotometer (IC-2, StellarNet, Inc., Tampa, FL, USA) equipped with an integrating sphere over the 250–700 nm wavelength range. The virtual cultivation bed and LED bar were reconstructed in 3D using Solidworks (Dassault Systèmes, Vélizy-Villacoublay, France) to match the actual dimensions and layout of the plant production module. The scanned plant surface models were then positioned within the virtual cultivation space according to the actual planting density.

Ray-tracing simulations were performed using Optiworks (OTIS Inc., La Farlède, France) with five randomly selected plants per treatment. All leaf surfaces were defined as detectors, and radiation interception per unit leaf area was averaged. Results were expressed as average radiation interceptions in the PAR (400–700 nm) and UV-A (350–400 nm) ranges. The workflow is summarized in Fig. 1.

Statistical analysis

All treatments for measurements were conducted with five biological replicates per treatment (n = 5). Shoot fresh weight and dry weight, leaf area, Fv/Fm, total phenolic content and antioxidant capacity of kale grown under different PPFD levels (150 or 300 µmol m^− 2 s^− 1) and four UV-A radiation intensities (0, 15, 30, and 45 W m^− 2) were analyzed using two-way ANOVA to test the main effects of PPFD and UV-A radiation intensity and their interaction (presented in Table 1). In addition, within each PPFD level, UV-A radiation intensity effects were evaluated using one-way ANOVA followed by LSD test (P < 0.05) for presentation in Figs. 2 and 4, and 5. All statistical analyses were performed using SAS software (version 9.4, SAS Inst. Inc., Cary, NC, USA).

Using 40 biological replicates obtained from a 2 × 4 factorial design (2 PPFD levels × 4 UV-A radiation intensities), we described the relationship between the UV-A/PAR ratio (independent variable, x) and either total phenolic content or antioxidant capacity (dependent variable, y) using a saturating nonlinear model, y = d + 𝑎 × x / (k + x) (Figs. 7 and 8). Nonlinear regression was performed separately for the datasets at 150 PPFD and 300 PPFD to estimate PPFD-specific fitted curves. In addition, the same model was fitted to the pooled dataset combining observations from both PPFD levels to obtain an overall fitted curve. The parameters a, k, and d were estimated by minimizing the mean squared error between observed and predicted values using the Adam optimization algorithm with analytical gradients (8,000 iterations, learning rate = 0.05). Model fit was quantified using the coefficient of determination (R²) for each PPFD-specific model and for the pooled (combined) model, where R² was calculated from the residual and total sums of squares. The results were visualized as scatter plots for each PPFD level together with the corresponding PPFD-specific fitted curves and the combined-data fitted curve.

In addition, we included total leaf area (Y) as an additional predictor alongside the UV-A/PAR ratio (X) and analyzed its relationship with either total phenolic content or antioxidant capacity (Z) using a second-order multiple nonlinear regression (response-surface) (Fig. 9). The model was specified as Z = β₀ + β₁X + β₂Y + β₃X² + β₄Y² + β₅XY + ε, thereby incorporating the main effects of X and Y, quadratic curvature terms (X² and Y²), and the interaction term (XY). Regression coefficients were estimated by ordinary least squares. Model fit was quantified using the coefficient of determination (R²), computed from the residual and total sums of squares. Based on the fitted model, predicted values were generated and the response surface (predicted regression surface) was visualized together with a three-dimensional scatter plot of the observed data (Fig. 9). Additionally, to assess potential overfitting, the dataset (n = 40) was randomly split into a training set (80%) and an independent test set (20%). Model parameters were estimated using the training set, and predictive performance was then evaluated on both the training and test sets using the R², calculated from the residual and total sums of squares. Similar R² values between the training and test sets were interpreted as evidence of limited overfitting. All graphing and regression analyses were performed in Python (version 3.13.7, Python Software Foundation, Wilmington, DE, USA) using NumPy (version 2.4.0) and Matplotlib (version 3.10.8).

Abbreviations

ABTS

2,2’-Azino-bis-3-ethylbenzothiazoline-6-sulfonic acid

ANOVA

Analysis of variance

DAT

Days after treatment

DW

Dry weight

2D

Two-dimensional

3D

Three-dimensional

EC

Electrical conductivity

FW

Fresh weight

GAE

Gallic acid equivalent

LED

Light-emitting diode

LSD

Least significant difference

Fo

Minimum fluorescence

Fm

Maximum fluorescence

Fv

Variable fluorescence

PAR

Photosynthetically active radiation

PPFD

Photosynthetic photon flux density

PS II

Photosystem II

PSM

Plant secondary metabolite

ROS

Reactive oxygen species

TEAC

Trolox equivalent antioxidant capacity

Trolox

6-Hydroxy-2,5,7,8-tetramethylchroman-2-carboxyl acid

UV

Ultraviolet

Authors’ contributions

MJK performed the experiment, collected, and analyzed the data. MSY developed the concept, wrote the manuscript and revised it. MMO supervised, developed the concept, acquired the funding, and revised the manuscript. All authors read and approved the manuscript.

Funding

This study was supported by Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, and Forestry (IPET) through K-Vertical Globalization Project, funded by Ministry of Agriculture, Food, and Rural Affairs (MAFRA) (RS-2025-02283050).

Data availability

The datasets supporting the conclusions of this article are included within the article or are available from corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.