Weed and Grassland Community Structure, Biomass and Forage Value Across Crop Types and Light Conditions in an Organic Agrivoltaic System

Abstract

Agrivoltaics represents a crucial technology and an innovative solution to promote sustainability. After a cropping season in an agrivoltaic system in Northern Italy, this study investigated the floristic composition and biomass of weed communities across three crops, evaluating their variation under shaded and full light conditions. In addition, the research assessed the role of uncultivated grassland areas in agrivoltaic-shaded conditions by quantifying their biomass and evaluating their potential feed value. Weed floristic diversity and biomass were surveyed at three different times. Soil and canopy parameters were analyzed in relation to photosynthetically active radiation (PAR). Grassland biomass was assessed after four different cuts and its suitability as a feed source was evaluated by the pastoral value and near infrared (NIR) spectroscopic analysis. Results showed that tomato had the lowest weed presence, and Setaria italica and Sorghum halepense were predominant in rice, while in durum wheat, higher nutrient availability favored Echinochloa crus-galli and Cirsium arvense. In weed community composition and biomass, no significant differences were observed for the effect of different light conditions (sun/shadow), and this may be attributed to their high environmental plasticity. PAR was strongly correlated with both soil and canopy temperatures. The analysis of floristic composition, biomass yield, pastoral value and nutritional quality of grassland vegetation indicated that spring cuts can be effectively used as forage, including for grazing. These findings suggest that integrating livestock activities could offer a win–win strategy for managing uncultivated areas within agrivoltaic systems, thereby enhancing their sustainability under organic farming practices.

1. Introduction

Increased global demand for food and energy due to growing populations and economic growth implies higher competition for agricultural land [1]. To overcome this competition and support sustainability, farming and energy production could be combined by pairing photovoltaics (PV) with open-field crop production in the same area [2,3]. This concept, known as agrivoltaics, is one of the new agricultural techniques under development, where research has increased significantly in recent years [4]. An agrivoltaic system (AS) can increase land productivity and play a role in the expansion of renewable energy production [5]. There exist several types of AS with various module layouts and associated crops. Open agrivoltaic systems can be broken down into ground-level, interspace PV and stilted, overhead PV. PV modules in overhead systems are mounted at least 2.1 m above the ground [1]. Beneath PV modules that are spaced, there can be sufficient sunlight to grow certain crops. Furthermore, crops in between PV module rows can utilize uncaptured solar irradiation. Specifically, with overhead systems, the land under the PV modules is used for farming, whereas with interspace systems, it is usually the land between PV modules that is farmed [5].

AS helps to improve the output of farms, as there is a 30% increase in economic value for those that combine shade-tolerant crop production and solar-generated electricity [6]. AS can provide synergistic benefits by functioning as a protection for plants from extreme weather conditions such as hail, frost, snow, and sunburn and by improving water availability and water use efficiency [7,8]. For livestock farming, a significant increase in late-season biomass of forages can be gained and solar panels can establish a cool microclimate for grazing, promoting animal welfare by providing shelter from sun, wind, and predators [9]. Nevertheless, more efforts should be dedicated to overcoming barriers for enhancing agrivoltaics sustainability, such as examining the social, economic and ecological risk factors [7]. From an agricultural point of view, AS shading reduces the light availability for crop photosynthesis and consequently biomass production and yield [5]. There is still little information available about plants’ ability to tolerate and compensate for low light levels in temperate latitudes [10] and many studies are still ongoing on crop yield and quality of crops beneath solar panels. Weed management remains a major challenge in agrivoltaic systems. By outcompeting crops for nutrients, light and water, weeds can significantly reduce both yield and quality. Their dynamics are further influenced by environmental factors such as temperature, light availability and climate variability, which affect seed dormancy, growth patterns and crop-weed interactions. Moreover, fertilization practices, especially nitrogen application, play a pivotal role in shaping weed populations and enhancing their competitive behavior.

Crop and weed management are continuous and adaptive processes. Weeds are one of the most important bioaggressors competing for minerals, water, and nutrients, and thus result in a potentially relevant reduction in crop yield [11]. Over multiple seasons, uncontrolled weeds contribute seeds or other propagative structures to the soil weed seed bank, allowing them to sprout later in the season, the following year or even decades into the future. This problem under agrivoltaic conditions, where mechanization is not possible in the whole area, is particularly critical. In addition, weeds tend to tolerate larger variations in environmental conditions compared to crops since they are not selected to perform best under specific local conditions, unlike cultivated varieties [12]. Regarding the effect of light conditions on spontaneous vegetation/weeds and their management in organic AS, there is a paucity of research works.

In AS, area losses (uncultivated area due to the mounting structures) are estimated to be around 10–25% [10,13]. This is a parameter that could be optimized and from an economic point of view, this makes a strong case for manual or other cultivation practices that allow losses to be limited to the built-up area. Ecologically, though, these spared areas of land pave the way for increases in biodiversity as the natural vegetation remains rather unaffected within this area [7]. This special type of spontaneous vegetation/grassland could serve as an important reservoir of biodiversity inside the AS but also provides a wide range of material and non-material benefits, i.e., feed production and several ecosystem services [14,15,16]. As a management practice, livestock, in particular sheep, can be used in targeted grazing areas (i.e., uncultivated areas) to control the excess understory biomass to reduce the cost of labor and eventually herbicide applications [17].

As an emerging agricultural technology, organic agrivoltaic systems are of great significance in solving the land use competition problem brought about by the growth of global food and energy demand. In this research framework, the aim of the study was twofold:

• to describe how weeds/spontaneous vegetation communities respond to different light conditions in terms of floristic composition and biomass;
• to evaluate the potential function of uncultivated areas (grassland) by assessing their quantity in terms of total biomass (yield) and their quality as feed through the determination of the pastoral value.

The study examined different crops (durum wheat, rice and tomato) cultivated through a whole year (autumn-spring and spring-summer cycles) inside an AS in the Po Valley, North of Italy (Figure 1). The methodological approach could be considered as a snapshot point of view, in which weed and grassland floristic composition and biomass were evaluated in a time span of one year, considering the AS typical setting—namely, PV module configuration, crop type and agronomic management. After this first step evaluation, a possible win-win strategy for spontaneous vegetation (grassland) in uncultivated areas was elaborated: controlling this type of vegetation while simultaneously valorizing it as a feed resource as an alternative strategy to the current site management, where uncultivated areas are periodically mowed by operators using small manual equipment.

2. Materials and Methods

2.1. Study Site

Field trials were established in an AS installed in 2011 in the North of Italy (Borgo Virgilio, Mantua, Lombardia, 45°05′50″ N–10°47′30″ E). A suitable classification of the agrivoltaic plant under study was proposed. This type of AS falls halfway between two “pure” classifications (grassland/arable farming), potentially leveraging the grassland areas for further agricultural activities/ecosystem services (Figure S1). It has a total size of 11.42 ha and a capacity of 2150 kWp. Seven hundred sixty-eight trackers (7680 solar panels) were installed with bi-axial technology (Poly 280 WP, Bisol Group, d.o.o., Latkova, Prebold, Slovenia). Photovoltaic modules were installed above the crops by mounting them on an open-air structure (i.e., stilt-mounted arrays). The height of the poles and panels from the ground is 4.5 m, the inter-row distance between each line of trackers allows a width clearance of up to 12.5 m, and the angle of inclination is ±50° on the main axis and ±40°on the secondary. The solar panels were installed dynamically by modifying their inclination according to the sunshine and the management of the crops. The total panel module area is 1.43 ha, so the Ground Coverage Ratio (GCR), defined as the area of solar panels/area of the land used for the AS, is 13%. As shown by Figure 2, elaborated with QGIS, release Firenze 3.28 [18], the AS includes areas where crop cultivation is prevented by support structures (e.g., tie-rods, see also Figure 1). These inner uncultivated areas (five strips), which are not mechanizable, are interspersed with the tracker’s rows and have a total surface of 1.66 ha. In addition, close to plant boundaries, outer uncultivated areas are present (buffer zone), which have a width of about 2 m and result in a small surface of 0.34 ha. Thus, areas potentially classified as grassland, where crops are not present and the species pool is maintained by natural processes, have an overall surface of 2.00 ha.

The research was performed from September 2023 to August 2024; weed samples (3 replicates) were picked on durum wheat, tomato and rice, both in full light and partially shaded zones and in areas where environmental conditions and the species pool are maintained by natural processes (grassland-shaded conditions). The crops were selected from those typical of the region (the Po Valley) and were chosen to cover an entire agricultural year: durum wheat as the autumn–winter cereal, rice as the spring–summer cereal and tomato as the spring–summer horticultural crop. The study area was selected to be representative of the Po Valley region, considering its agrometeorological characteristics and the relevance of the chosen crops, which are key species in this agricultural context in terms of cultivated surface and farm income. Marginal areas that received repeated ecological disturbance (i.e., mowing), such as field borders and strips inside the trackers, can be ecologically equated to grassland.

Figure 3 shows, monthly, the main agrometeorological measurements retrieved by the on-site weather station. During the experimental period (September 2023–August 2024), the annual mean air temperature was 15.1 °C and the total rainfall was 815 mm (Figure 3a), while the mean air relative humidity and the total solar radiation were 83.3% and 1618 W/m², respectively (Figure 3b). The field trials were established on a silty sand soil with a pH measured in water of 8.5. The sand, silt, and clay contents were 500, 400, and 100 g kg⁻¹ at 0–60 cm depth, respectively. The organic C was 1.11% and the total N was 0.18%. Assimilable P and exchangeable K were 76.42 and 810 mg kg⁻¹, respectively. The entire area was managed according to organic farming standards. Soil tillage included ripping to a depth of 40 cm and the application of the false seedbed technique to limit weed development. Durum wheat and rice were sown on 9 October 2023 and 20 April 2024, respectively, while tomato was transplanted on 17 May 2024. The preceding crops covering the soil surface in the previous growing season were tomato, before durum wheat; a cover crop, before tomato; and a forage mixture, before rice. In the AS, one slurry application per year was made in the autumn as a pre-plant soil amendment. The slurry was obtained from a bordering dairy farm equipped with a liquid-solid separator. Wheat and grassland were managed as rainfed areas, while rice and tomato were fully irrigated using the ‘Irriframe’ decision support system. Top-dressing fertilization and pest and disease control followed the organic farming standards of Lombardia Region. Harvest dates were adapted to climatic conditions and plant development stages. Grassland areas have been only periodically mowed; where possible, mowing was carried out by mechanization using a tractor-mounted mulcher, and in other areas, cutting operations were carried out with a manual mulcher.

2.2. Weeds/Grassland Sampling Method and Data Collection

During the trial period, one weed sampling session was carried out for each crop (durum wheat—10 May; tomato—1 July; rice—17 July 2024). Weed biomass samplings with 3 replicates were collected both in full light zones—i.e., outside the panel trackers and inside the tracker rows where solar panels were removed—and in partially shaded zones inside the trackers, that is, under solar panels. Considering the entire growing cycle of each crop, the reduction in PAR between full sun and shaded zones was 45% in durum wheat, 22% in tomato and 24% in rice. For tomato crops, the weeds were collected from the inter-row spacing, as plastic mulch was present on the rows. For the grassland, the sampling sessions were four, with 3 replicates each, always in shaded zones inside the tracker lines (10 May, 5 June, 1 July and 6 August 2024). Surveys were carried out in areas where crop cultivation is prevented by support structures (Figure 1 and Figure 2). These inner uncultivated areas, with shaded conditions and not mechanizable, represent a real issue for spontaneous vegetation management in the AS.

For crops and grassland, plants considered included broadleaved and grass weeds cut to a height of 10 cm. As in previous research experiences [19,20] on areas representative of the site, each survey was conducted through a quadrat with a sampling area of 58 cm × 62 cm in which floristic composition was identified and recorded. For both weeds and grassland, the number of species observed was annotated together with their total density. Density was the number of individuals per species within the measured plot (quadrat) area. Plant nomenclature was done according to Pignatti [21]. Species were later grouped into main functional groups as follows: Poaceae (graminoids), Fabaceae (legumes), and other species (further herbs). For each group, after fresh weight recording, dry weight was retrieved by oven-drying the samples at 65 °C for 24 h and from this, the biomass, expressed as Mg DM ha⁻¹, was calculated for crop weeds and grassland. These sampling operations and the subsequent processing described above were used to determine the floristic composition and estimate the biomass of weeds and grassland. In addition, for each sampling point, the soil moisture expressed as volumetric water content (VWC) (%), the electrical conductivity (EC) (dS/m), soil temperature (°C), and infrared (IR) canopy temperature (°C) were collected throughout the field campaign, using a portable Fieldscout TDR 300 probe (Spectrum Technologies Inc., Aurora, IL, USA) with 12 cm-long rods. In addition, photosynthetically active radiation (PAR) was measured for all the treatments with a calibrated photosensor (RK200-02 Quantum PAR Sensor).

2.3. Near-Infrared Reflectance Spectroscopy Analysis

Regarding grassland, a set of 12 biomass samples (3 replicates × 4 sampling times) was collected to estimate the dry matter (DM) biomass. The batches were then analyzed in the laboratory using near infrared (NIR) reflectance spectroscopy, and the residual DM of the samplings was attributed. To estimate the nutritional feed value, grassland samples were ground to 2 mm and the spectra acquisition was performed using the instrument Foss NIR-System 5000 monochromator (NIR-System, Silver Spring, MD, USA) with the spinning ring cup cell as sample transport module by double scanning each sample in the 1098–2500 nm spectral region. Spectral data were processed using WinISI II V1.5 software (Infrasoft International, Port Matilda, PA, USA). NIRS calibration equations were developed for Italian hay’s predicted in vitro neutral detergent fiber parameters. Predictions were performed using the equation developed and validated by Brogna et al. [22] and Palmonari et al. [23], considering a coefficient of determination higher than 0.85 and a standard error ranging from 1.9 and 3.4 in cross-validation. Indeed, the following parameters were retrieved: dry matter (%), ash (% of DM), crude protein content (CP, % of DM), acid detergent-insoluble protein (ADIP, % of DM) and soluble protein (SolP, % of DM) according by Goering and Van Soest [24] and Licitra et al. [25]; neutral detergent fiber (NDF, % of DM) with amylase and sodium sulfite method according to Mertens et al. [26], acid detergent fiber (ADF, % of DM), acid detergent lignin (ADL, % of DM), undigested NDF after 240 h of in situ rumen incubations (uNDF, % of NDF), digestible NDF evaluated after 24 h of in situ rumen incubations (dNDF, % of NDF) as reported by Palmonari et al. [27], fat (% of DM), starch (% of DM), sugar (% of DM), metals and other elements (Ca, P, Mg, and K) as recommended by AOAC [28] and net energy for lactation (NE_L, kcal kg DM⁻¹). The data on the grassland floristic composition were analyzed by computing the Pastoral Value index (range 0–100) (Equation (1)), which synthesizes grassland biomass and nutritional parameters as follows [29,30]:

$$Pastoral Value=0.2 \sum _{i=1}^{i=n}{CS}_i \times {IS}_i$$

(1)

with CS_i, being the single species percentage contribution to vegetation composition (species relative abundance) of each plot, and IS_i, being the index of specific forage value (0–5). The indices of each species were assigned following Cavallero et al. [31], while an average value (compared to the botanical family to which it belongs) was assigned for those species not present in the list. The lab analyses and the determination of the pastoral value described in this paragraph were leveraged to determine the adequacy of grassland vegetation as a feed source.

2.4. Statistical Analysis

Descriptive statistics of data were elaborated with IBM SPSS software, version 29 [32]. The same statistical software was employed to accomplish the analysis of variance (ANOVA) on the collected data, once the assumption of normality (Kolmogorov–Smirnov test) and the homogeneity of variance (Levene median test) were satisfied. Then, Duncan’s range test was used for the post hoc comparison of means at the 5% significance level (p ≤ 0.05). In particular, mean values were compared to evaluate the effect of (i) crop type and shading on biomass and plants per plot in weed samplings, through a two-way ANOVA and (ii) cutting date on biomass, pastoral value, plants per plot and quality (NIRS) in grassland, through a one-way ANOVA. A correlation plot was built with the Origin software (version 2023b) [33] to calculate and visualize the pairwise correlations separated by agrometeorological measurements (VWC, EC, T soil, T canopy, PAR). Three levels of rejection of the null hypothesis (Ho) were considered and indicated by asterisks: * p-value ≤ 0.05; ** p-value ≤ 0.01; *** p-value ≤ 0.001.

3. Results and Discussions

Table 1. Floristic composition (%) of plant species and botanical families monitored in durum wheat, tomato and rice and grassland.

	Land Use
	Wheat	Tomato	Rice	Grassland	Grassland	Grassland	Grassland
Sampling Date	10 May	1 July	17 July	10 May	5 June	1 July	6 August
Species
Avena sativa L.	3.8	-	-	8.2	20.1	0.5	-
Cynodon dactylon L.	-	3.8	2.1	2.2	3.5	6.7	4.7
Digitaria sanguinalis L.	-	-	17.6	-	-	-	-
Echinochloa crus-galli (L.) P. Beauv.	19.6	-	-	0.3	0.2	-	-
Hordeum murinum L.	-	-	-	1.0	1.3	-	-
Lolium multiflorum Lam.	22.8	-	-	1.9	0.5	-	-
Lolium perenne L.	-	-	-	4.3	1.4	-	-
Setaria italica (L.) P. Beauv	5.7	-	35.6	0.3	-	-	-
Sorghum halepense L.	13.3	7.3	20.5	1.8	3.4	1.1	1.1
Poaceae	65.2	11.2	75.8	20.1	30.3	8.2	5.7
Trifolium pratense L.	4.4	-	-	0.2	-	-	-
Trifolium repens L.	-	-	0.3	0.2	-	-	-
Fabaceae	4.4	-	0.3	0.4	-	-	-
Abutilon theophrasti Medik.	3.8	-	-	-	-	-	-
Chenopodium album L.	5.1	-	-	-	-	-	-
Cichorium intybus L.	-	-	-	0.5	0.1	-	-
Cirsium arvense (L.) Scop.	13.3	-	1.1	0.4	1.9	1.8	0.6
Convolvulus arvensis L.	-	5.4	-	2.9	5.9	1.6	1.6
Crepis biennis L.	-	-	0.3	-	-	-	-
Erigeron canadensis L.	-	-	0.8	-	-	-	-
Euphorbia serpens Kunth	-	-	6.9	-	-	-	-
Fumaria officinalis L.	1.9	-	-	-	-	-	-
Plantago lanceolata L.	-	-	-	0.3	-	-	-
Portulaca oleracea L.	-	83.5	10.9	-	-	-	-
Potentilla reptans L.	-	-	3.5	4.0	6.8	2.6	-
Ranunculus bulbosus L.	-	-	0.3	-	-	-	-
Rumex spp.	0.6	-	-	0.3	-	-	-
Solanum nigrum L.	-	-	0.3	-	-	-	-
Stellaria media (L.) Vill.	5.1	-	-	2.9	0.2	-	-
Taraxacum officinale Weber	0.6	-	-	0.5	0.2	-	-
Other species	30.4	88.8	23.9	11.8	15.1	6.0	2.2
Total amount	100	100	100	32.3	45.5	14.3	7.9

Each survey was conducted in a quadrat with a sampling area of 58 cm × 62 cm, in which floristic composition was identified and recorded. Floristic composition was reported as a percentage of specimens.

shows the floristic composition of plant species, grouped by botanical families (Poaceae, Fabaceae and other species), across different land uses (durum wheat, tomato, rice, and grassland), based on data collected over multiple sampling dates throughout the cropping season. The sum of individuals is reported for the two sampling zones (full light + shadow) in durum wheat, tomato and rice. For the grassland, plant species are shown throughout the four sampling dates. Floristic composition was reported as a percentage of specimens both for crops and grassland.

From the surveys, 28 weed species were identified. The weed community was largely dominated by the Poaceae, particularly in durum wheat (65.2%) and rice (75.8%), where grass weed competition may be tough to manage. Tomato exhibited a high percentage of Portulaca oleracea (83.5%) and this reflected the significant contribution of the group called other species to the overall weed composition in this crop (88.8%). In the same production site, growing season and similar water regime, Portulaca oleracea was found to be a particularly prevalent weed [34]. Deepening the analysis for the Poaceae, Sorghum halepense was prominent in all crop types. It is a perennial and rhizomatous weed that tends to form dense stands derived from sexual reproduction but also from vegetative propagation [35]: this is the main factor explaining its success as an agricultural weed. Other graminoid species gave a spotted contribution throughout the sampling, such as Digitaria sanguinalis, Echinochloa crus-galli, Lolium multiflorum and Setaria italica. In durum wheat, nutrient availability due to soil amendment was related to the growth of Lolium multiflorum (22.8%), Echinochloa crus-galli (19.6%), and Cirsium arvense (13.3%) [36]. In fact, the application of slurry in autumn can significantly affect the floristic composition of treated areas.

In rice, Setaria italica (35.6%), and Sorghum halepense (20.5%) were observed in over 50% of the total observations; also, Digitaria sanguinalis was noted as an impactful weed (17.6%). Therefore, regarding weed management strategies, these species should receive more attention. Mechanical weed control methods could be a possible solution, considering the organic management of the site. However, some species, such as Cirsium arvense [37] and Sorghum halepense, have roots and rhizomes distributed to a large depth over the soil profile and the mechanical surface soil treatments carried out at the experimental site may not pose a problem for their spread. Thus, integrated weed management strategies should be specifically developed for this type of AS, such as the adoption of long crop rotations—including, for example, the introduction of a perennial species such as alfalfa—and the application of deep tillage during the summer.

The data in Table 1 indicate a notable variation in species composition between the crops and the grassland. The grassland showed a higher diversity of species over multiple sampling dates, suggesting a more diversified and stable ecosystem compared to the monoculture of durum wheat, tomato, and rice. Avena sativa, Cynodon dactylon, Convolvulus arvensis and Potentilla reptans are the most representative species. Nevertheless, species biodiversity in grassland decreased after repeated mowing, especially in summer—only five species were recorded on the final sampling date, contributing just 7.9% to the overall grassland community. Although elevated end-of-season temperatures (in July and August, see Figure 3) were observed, this factor appeared to have no significant effect on plant biodiversity [38,39]. In contrast, as a documented explanation, frequent mowing was demonstrated to be the primary driver, as it increased total productivity but reduced other ecosystem services, such as species richness [40,41].

Dry matter aboveground biomass and floristic distribution patterns of weeds were assessed in the AS in shady and full light conditions on the three crops (

Table 2. Biomass and plants per plot of weed species in full light and shaded conditions.

Item		Wheat	Tomato	Rice	Mean	Crop p-Value	Shading p-Value
Biomass (Mg of DM ha−1)	Sun	0.71	0.42	1.06	0.73	0.010	n.s.
	Shadow	0.47	0.12	1.39	0.66
	Mean	0.59 ab	0.28 a	1.23 b	0.70
Plants per plot (number)	Sun	7.33	50.67	60.30	39.43	0.001	n.s.
	Shadow	10.25	23.59	65.05	32.96
	Mean	8.78 a	37.14 b	62.67 c	36.20
Poaceae (plants per plot)	Sun	3.17	3.00	43.00	10.36	-	n.s.
	Shadow	7.00	5.00	49.75	15.15
Fabaceae (plants per plot)	Sun	-	-	0.50	0.09	-	n.s.
	Shadow	0.58	-	-	0.35
Other species (plants per plot)	Sun	3.67	49.67	16.00	18.45	-	n.s.
	Shadow	2.17	20.50	14.50	8.30

Each survey was conducted in a quadrat with a sampling area of 58 cm × 62 cm, in which floristic composition was identified and recorded. Letters indicated statistical grouping. Different letters within each row indicate significant differences (within a variable) between treatments (crop and shading) to Duncan’s range test (p ≤ 0.05). n.s. = not significant.

). The data showed a significant impact of the weed groups in terms of biomass. The data were collected from plots in both shaded and full light conditions, and the overall average of the observations was processed.

Among the evaluated crops, tomato showed the lowest weed biomass (0.28 Mg DM ha⁻¹), whereas durum wheat (0.59 Mg DM ha⁻¹) and rice (1.23 Mg DM ha⁻¹) showed approximately twofold and fourfold higher weed biomass, respectively. The number of plants per crop was lower in durum wheat (8.78); in tomato, an average of 37.14 plants was counted, while in rice, the average number was 62.67. Biomass and plants per plot did not differ significantly, considering the light condition treatments. The interaction between the two main factors (crop and shading) was also not significant. As indicated by Table 2, weed biomass had higher values in full light treatments in durum wheat and tomato with 0.71 vs. 0.47 Mg DM ha⁻¹ and 0.42 vs. 0.12 Mg DM ha⁻¹, respectively. In rice, the number of other species counted varied between the full light (18.45) and shadow (8.30) treatments, but without a statistically significant difference. In rice, the number of other species markedly differed between sun (18.45) and shade (8.30) treatments, but the difference was not statistically significant.

The results showed a significant impact of the main botanical weed communities typical of the forage system in northern Italy, which provides for the traditional rotation between maize and Lolium spp. The use of organic manure (especially slurry) and the non-use of herbicides were also documented as factors affecting weed populations [42,43]. Ingraffia et al. [44] reported the results of an organic field experiment in which the responses of durum wheat and weeds to no-tillage were evaluated compared with conventional tillage. The weed biomass in conventional tillage retrieved by the authors (0.63 Mg ha⁻¹) was higher than our observation in shadow conditions (0.47 Mg ha⁻¹) and lower than what was observed in the full light condition (0.71 Mg ha⁻¹). However, Ingraffia et al. [44] measured the biomass of the weed species at the end of the cycle, while in our study, it was assessed on 10 May. The above suggests a higher production of biomass at the end of the cycle under the AS. In addition, many studies have shown that soil inversion by moldboard plowing distributes weed seeds along the soil profile, most of them deep in the soil, thereby reducing the likelihood of their emergence [45,46]. In our experimental site, plowing was replaced by ripping, thus lessening the potential effect of limiting weed germination. Among crops, weed biomass and plants per plot differed significantly, mainly due to the different climatic conditions that occurred in the different growing seasons. Climatic growing parameters influence plant distribution, growth, development, and phenology, with specific species thriving within, for example, temperature [47] or solar radiation [48] ranges. In organic AS, weeds are one of the greatest concerns for crop production, considering that their environmental plasticity hinders successful control. This may partly explain the absence of significant differences in weed biomass and plant density between the full light and shade treatments. In addition to the high plasticity of weeds, which enables rapid adaptation to environmental fluctuations, this response can also be attributed to the fact that weeds develop in the interference of crop canopies. As such, shaded conditions do not represent a substantial constraint to their growth and development [49] and weeds tend to compensate for the total biomass, for example by accumulating more carbon in other organs, such as the stem [50]. Studies on adaptation to light gradients of weed species demonstrate that populations exposed to shade evolve traits like increased chlorophyll b and larger leaves to enhance light capture [51,52]. This could inform management strategies: establishing dense crop canopies or deploying timely shading interventions, such as cover crops, can significantly impede weed establishment. Weed response to soil heterogeneity—such as varying pH, fertility, or texture—is also an important issue. While generalist annual weeds tolerate broad soil conditions, other species show localized distribution linked to specific properties like organic matter content [20,53]. This suggests the value of site-specific strategies; targeted amendments or tailored crop rotations, which alternate soil nutrient exploitation to disrupt weed adaptation cycles and shift weeds from preferred niches, could be implemented. In specific cropping conditions like those of organic AS, it could also be reasonable to consider localized soil amendments or micro-environmental alterations (e.g., pH modification) to undermine weed establishment.

Figure 4 shows the raw (not model-adjusted) Pearson linear correlations between the main soil parameters (VWC, EC, soil and canopy temperature) and PAR. The figure aims to provide a descriptive overview of the main soil parameters and PAR links in the AS during the experiment, which was conducted with numerous variables: crop, season, light condition and irrigation level (rainfed/irrigated). The correlation plot was built with all the aggregated data of the full light and shadow measurements of the crops and grassland throughout the experimental period. Figures S2 and S3 show the correlation trend separating the shading effect for full light and shadow, respectively.

Overall correlations were moderate between VWC with EC (0.46), soil temperature (−0.47) and canopy temperature (−0.63). Correlations were higher in soil temperature with canopy temperature (0.92) and PAR (0.81). Finally, the correlation between canopy temperature and PAR was 0.70. As is known, the implementation of AS is expected to affect plant growth due to changes in microclimatic conditions [5]. Specifically, light and temperature, together with water availability, are key indicators of plant success in a specific growing environment [54]. The shaded area under the panels affects the availability of PAR over the plant canopy with a reduction of PAR affected by AS setup and, on the contrary, a sharp increase in the PAR level, which may occur passing from shading to open full light conditions [55]. In this study, PAR showed a positive correlation with canopy and soil temperature; thus, the microclimate established under the AS reduces the exposure of plants to excessive sunlight and moderates the heat stress, particularly during hot summer months [56]. Naturally, as C3 plants (like those in this study) tend to saturate at a significantly lower PAR, they are more adaptable under an agrivoltaic array [57]. Moreover, the soil temperature is closely associated with plant growth, plant stress (particularly at the initial root zone) and soil microbial diversity [54]. According to other studies [58,59], where temperatures are lower, VWC showed higher values (negative correlation), due to diminishing water evaporation caused also by the AS wind-blocking effect [60,61], suggesting potential mitigation of soil water deficit [62]. Adeh et al. [56] investigated the effect of periodic AS shading on grasslands under drought stress in Oregon (USA). Low-mounted photovoltaic panels caused higher soil moisture content and increased grass biomass late in the season. The results of this study demonstrated that EC increased where VWC increased. In zones with higher soil moisture, EC may also be influenced by agricultural activities, which can increase leaching and, in turn, nutrients dissolved in soil water [61]. Microclimatic conditions under the AS are hard to find in open fields; as mentioned, the lower canopy and soil temperature reduced heat stress for the plants, resulting in higher yields compared to the open system. Hence, AS, also accounting for a trade-off between reduced PAR and enhanced water availability, could be beneficial not only to crops but also to weed growth. In this sense, attention must be paid to weed control by developing an ad hoc management strategy that considers microclimatic conditions and typical AS constraints (i.e., spatial limitation for machinery).

Results about dry matter biomass, pastoral value and floristic diversity in grassland are represented in

Table 3. Biomass, Pastoral Value and plants per plot for grassland.

Item	Cut	Grassland 1	Grassland 2	Grassland 3	Mean ± s.d.	Cut p-Value
Biomass (Mg of DM ha−1)	I	0.94	0.84	0.64	0.81 ± 0.52	n.s.
	II	1.50	1.85	1.48	1.61 ± 0.69
	III	0.34	0.36	2.40	1.03 ± 0.87
	IV	0.49	0.38	0.53	0.47 ± 0.78
Pastoral Value	I	36.25	46.25	37.14	39.88 b ± 5.53	0.001
	II	36.36	32.00	38.00	35.45 b ± 3.10
	III	20.00	25.00	20.00	21.66 a ± 2.87
	IV	20.00	20.00	20.00	20.00 a ± 0
Plants per plot (number)	I	101.00	122.00	83.00	102 bc ± 19.52	0.010
	II	131.00	204.00	95.00	143.33 c ± 55.54
	III	44.00	33.00	58.00	45.00 ab ± 12.53
	IV	21.00	26.00	28.00	25.00 a ± 3.61
Poaceae (plants per plot)	I	52.00	77.00	61	63.33 bc ± 12.66	0.010
	II	75.00	145.00	67	95.67 c ± 42.91
	III	15.00	28.00	35	26.00 ab ± 10.15
	IV	14.00	21.00	19	18.00 a ± 3.61
Fabaceae (plants per plot)	I	-	4.00	-	1.33 ± 2.31	n.s.
	II	-	-	-	-
	III	-	-	-	-
	IV	-	-	-	-
Other species (plants per plot)	I	49.00	41.00	22.00	37.33 b ± 13.37	0.022
	II	56.00	59.00	28.00	47.67 c ± 17.10
	III	29.00	5.00	23.00	19.00 ab ± 12.49
	IV	7.00	5.00	9.00	7.00 a ± 2.00

Each survey was conducted in a quadrat with a sampling area of 58 cm × 62 cm, in which floristic composition was identified and recorded. Letters indicated statistical grouping. Different letters indicate significant differences (within a variable) between treatments (cut) according to Duncan’s range test (p ≤ 0.05). n.s. = not significant.

. Although biomass did not differ significantly across cuts, a decreasing trend was observed toward cut IV, accompanied by substantial variability among replicates. For the pastoral value, the highest mean value was recorded in the first cut, 39.88 ± 5.53. This parameter varied significantly between the first two cuts (I = 39.88 ± 5.53 and II = 35.45 ± 3.10) and the last two cuts (III = 21.66 ± 2.87 and IV = 20.00). The highest number of plants per plot was recorded in the II cut (143.33 ± 55.54) and then in the I cut (102 ± 19.52); subsequently, a more limited presence of species was observed, but with significant differences between the III (45.00 ± 12.53) and IV cut (25.00 ± 3.61). The highest number of Poaceae (plants per plot) was recorded in the II cut (95.67 ± 42.91) and then in the I cut (63.33 ± 12.66); in summer cuts, the number of Poaceae was reduced at the III cut (26.00 ± 10.15), and then further at the IV (18.00 ± 3.61). The same trend was observed for other species (plants per plot), where the highest number was recorded in the II cut (47.67 ± 17.10), and then in the I cut (37.33 ± 13.37). Biomass yield in uncultivated areas should be considered both in terms of the intrinsic value of the product, estimated at 100 € per ton on the local market [63], and as a cost saving in labor, which would otherwise be employed for the manual cutting of unused biomass.

Although the plant species composition of grassland is well documented, an important research effort has still to be devoted to understanding their dynamics and growth under AS. The relationship between species richness and aboveground net primary productivity is still a central and much-debated issue in community ecology [64]. Future research in the AS sector should explore additional topics such as grassland functional traits and their biodiversity that are directly linked to ecosystem services, their trade-offs and synergies with crop and energy production.

The nutritional value of grassland biomass was investigated with NIRS analyses (

Table 4. Statistical analysis of grassland biomass and Near-Infrared Reflectance Spectroscopy (NIR) data.

Item	Unit	Sampling Dates				p-Value
		10 May	5 June	1 July	6 August
Biomass	t DM ha−1	0.81	1.61	1.03	0.47	n.s.
DM	%	90.55	91.30	91.37	91.35	n.s.
Ash	% of DM	11.00	9.32	10.63	10.61	n.s.
CP	% of DM	14.83	8.79	13.09	13.08	n.s.
Fat	%	2.90	2.70	3.08	3.06	n.s.
ADIP	% of DM	1.14	1.54	1.74	1.72	n.s.
SolP	% of DM	5.22 a	2.74 b	2.83 b	2.81 b	<0.001
NDF	% of DM	52.85	62.25	55.29	55.27	n.s.
ADF	% of DM	33.06	42.98 a	40.07	40.05	n.s.
ADL	% of DM	3.91 b	6.53 a	6.69 a	6.67 a	0.009
dNDF	% of NDF	54.75 a	41.43 b	41.24 b	41.22 b	0.025
uNDF	% of NDF	7.26 b	21.02 a	22.66 a	22.64 a	<0.001
Starch	% of DM	2.94	3.72	2.40	2.38	n.s.
Sugar	% of DM	9.82 a	6.98 b	4.89 c	4.87 c	<0.001
Ca	% of DM	0.82	0.65	1.04	1.02	n.s.
P	% of DM	0.44	0.35	0.38	0.36	n.s.
Mg	% of DM	0.21	0.10	0.21	0.19	n.s.
K	% of DM	3.23	2.38	2.84	2.82	n.s.
NEL	kcal kg DM−1	1353.49 a	1114.62 b	1166.63 b	1166.61 b	0.041

DM, dry matter; CP, crude protein content; ADIP, acid detergent-insoluble protein; SolP, soluble protein; NDF, neutral detergent fiber with α-amylase, sodium sulfite and correcting for ash contamination; ADF, acid detergent fiber; ADL, acid detergent lignin; dNDF, digestible NDF evaluated after 24 h of in situ rumen incubations; uNDF, undigested NDF after 240 h of in situ rumen incubations; NEL, net energy for lactation. Different letters within each row indicate significant differences between treatments (sampling dates) according to Duncan’s range test (p ≤ 0.05). n.s. = not significant.

). The results showed a significant decrease in the nutritional value from the first cut (May 10), passing to the next three (June 5, June 1, and August 6). This was also confirmed by six strategic parameters for dairy cows’ (and ruminants’) feeding and nutrition. First, SolP decreased passing from the first cut (5.22) to the subsequent ones (2.74, 2.83 and 2.81), and the same happened for dNDF, sugar and NE_L. In contrast, the parameters that worsen the quality of biomass (ADL and uNDF) registered an opposite trend, increasing from the first cut to the summer ones. The observed results demonstrate how the nutritional value of these biomasses is well-suited for ruminant feeding. The first cut carried out on May 10 is comparable with values reported in other studies [65,66], dealing with the nutritional characterization of forages produced in lands specifically designed to produce fodder for dairy cows. Compared to the studies mentioned above, the first cut of the grasslands under study obtained a CP of 14.83%, which is on average 4 percentage points higher. In addition, SolP (%) was higher compared to the studies mentioned (5.22 vs. 4.34 and 4.25). In the study by Andrew et al. [9], they investigated the yields and quality of forage in an agrivoltaic environment grazed by lambs. Similar to our study, the average CP content of pastures reduced sharply (p < 0.01) from 21.6% on April 19 to 15.8% on May 1 and remained relatively stable until the end of the grazing season. On the other hand, dNDF was on average about 7 percentage points lower compared to matrices that were classified as similar. These two contrasting aspects can be attributed to the significant presence of weed species, which can increase the protein content of the biomass [67,68] while, however, depressing its digestibility. In particular, Sorghum halepense and Cynodon dactylon (see also Table 1) play an important role in reducing nutritional parameters of interest to ruminants. Although Sorghum halepense is palatable to ruminants and has a similar nutritive value to Cynodon dactylon, their impact in terms of dry matter produced on the sampled plots was the main reason for the decrease in nutritional value over time. Moreover, the dry season favored the rapid development of the maturity stages of the two weed species recorded at the time of biomass sampling (June 5, June 1 and August 6).

Considering areas where crops are not present (Figure 2), the grassland has an overall surface of 2.00 ha, which could correspond to an overall potential forage biomass of 3.92 ± 0.70 Mg ha⁻¹, as the sum of the four cuts (Table 4). The management of these marginal areas could also be carried out through grazing by ruminants. The biomass produced from the first cut could be utilized by heifers and dry cows, especially in areas like the one in our study, which borders a dairy farm. Subsequent cuts, on the other hand, are more suitable for being utilized by small ruminant species, such as sheep. Comparable spring lamb growth and live weight production per hectare from open and solar pastures demonstrate that AS would not decrease the production value and potential of the land [9]. This was also confirmed by Sturchio et al. [69], who reported an increase in forage quality later into the growing season with grazing. Combining energy and pasture production appears to be greatly advantageous and could be a strategy for controlling the spread of weed species [70,71]. The separation of areas designated for grazing can be achieved using electrified fences or, more simply, with the new Nofence [72,73]. An additional possible solution is to use the entire area of the agrivoltaic site during the period between the crops (tomato, sorghum, potato) and the sowing of winter cereals. Therefore, these could represent win-win strategies for managing spontaneous vegetation in AS, which may offer a viable solution to simultaneously control vegetation in uncultivated areas and provide a valuable input for another agricultural activity, such as livestock farming. Finally, land use efficiency in agrivoltaics, particularly where energy and animal production are combined on the same land, can potentially be further enhanced through optimized system design, the selection of shade-tolerant pasture species (i.e., Dactylis glomerata and Lolium perenne) and the implementation of sustainable livestock management practices, as described above. The overall system can be identified as a win-win option among AS management strategies and a basis for promoting sustainable agriculture within organic production regulations.

4. Conclusions

This study offered valuable insights into the floristic composition and biomass of both weeds and spontaneous vegetation within an AS in the Po Valley (Northern Italy). The work is pioneering and among the first of its kind, aiming to provide an initial reference point for broader analyses under diverse environmental and agronomic conditions. Throughout an entire year, the analysis covered weeds in three cultivated crops—durum wheat, rice, and tomato—under different light conditions (full light and shade), as well as spontaneous vegetation in shaded grassland.

Regarding floristic composition, the weed community was largely dominated by the Poaceae, particularly in durum wheat and rice, where grass weed competition may be tough to manage. The crop with the least presence of weed species was tomato, while durum wheat and rice showed a higher number of weeds. Tomato exhibited a high percentage of Portulaca oleracea. In rice, Setaria italica and Sorghum halepense were observed in over 50% of the total observations; in durum wheat, higher nutrient availability was related to the growth of Lolium multiflorum, Echinochloa crus-galli, and Cirsium arvense. These indications could be useful for identifying the most frequent weed species in this specific context (crop type combined with agrivoltaic system) and for selecting the most appropriate control techniques. As an outstanding result, in weed community composition and biomass, no significant difference was registered between light condition treatments (sun/shadow). This could be due to the plasticity of the weed and to the dynamic shadow projected to the ground by PV, which may not significantly limit weed growth. To enhance agrivoltaic sustainability in terms of weed management (no use of mechanical control nor mowing operations), this study also evaluated the potential function of uncultivated areas (grassland) by assessing their total biomass and feed value. To this end, through the analysis of biomass, floristic composition and pastoral value, the research work demonstrated that the quantity and quality of grassland spring cuts in the AS might be usefully utilized for forage, also through grazing. This represents a possible sustainable and win-win strategy for grassland management in an organic AS: control spontaneous vegetation while simultaneously valorizing it as a feed resource.

Current research on weed biodiversity and forage opportunities in an organic agrivoltaic system is insufficient; further studies on the composition of vegetation communities and their control and/or management are needed and should provide practical insights, empowering farmers operating in AS and enhancing the sustainability of the organic agricultural system.