Use of Under-Vine Living Mulches to Control Noxious Weeds in Irrigated Mediterranean Vineyards

This article assesses the use of under-vine living mulches in Mediterranean vineyards characterized by limited water resources, one of the reasons why this agronomic practice is currently unusual in these environments. The aim of the study was to test whether the use of this alternative method in Mediterranean vineyards could suppress noxious weeds without hindering optimal vineyard development. For this purpose, four native species were selected as living mulches: Festuca ovina, Pilosella officinarum, Plantago coronopus, and Plantago lanceolata. The variables measured during three years in two different experimental farms were: (a) living mulch cover, as a possible predictor of weed suppression success; (b) weed density and weed biomass, with special attention to noxious weed species; and (c) pruning weights, measured in the last year to analyze the cumulative effect of the treatments on the grapevine vegetative growth. Our results revealed that living mulches with high cover rates (average over 70%) also showed weed suppression of up to 95%, significantly controlling the occurrence of noxious weeds such as Erigeron canadensis. No significant effect of the different treatments on vine vegetative growth was found, although further studies would be necessary. Based on these findings, it can be concluded that under-vine living mulches could be an efficient and environmentally friendly method for weed control in Mediterranean vineyards where irrigation is available.


Introduction
Vineyards are one of the essential crops of the Mediterranean basin, where they have occupied a central role in the landscape, culture, and economy since ancient times [1]. Indeed, Spain (961 kha), France (797 kha), and Italy (719 kha) account for 34% of the world's vineyard area [2]. Although traditionally, vineyards have been cultivated under rainfed conditions, in recent decades, the use of irrigation has become widespread in Mediterranean vineyards [3]. In the context of climate change, given the predictions of reduced rainfall in the Mediterranean basin [4], the area occupied by irrigated vineyards is expected to increase. Although this could prevent the detrimental effects of water shortage on vineyard yields [5], it could also have other undesirable effects, such as an increase in the presence of competitive and noxious weed species [6,7]. Winegrowers have tried to eradicate weeds in vineyards through conventional management based on the recurrent application of mechanical (tillage) and/or chemical (herbicides) control methods. However, numerous studies have highlighted in recent years critical problems associated with these weed management practices, such as the emergence of herbicide-resistant species [8], soil erosion [9], or biodiversity decline [7]. This has led to a search for alternative weed management in line with the European Green Deal (EU Farm to Fork Strategy), being Table 1. Mean temperature (T mean ) and cumulative rainfall (R) on the two experimental farms (La Poveda, farm A; El Socorro, farm B) during the study period, grouped in two intervals: the summer season (June to September) and the rest of the year (October to May). The percentage variation of accumulated precipitation (∆R) in each time interval with respect to the average precipitation value (data for the period 2001-2021) for the same interval is also indicated.

Plant Species Selected for Living Mulches
The key premise was to seek species that could permanently cover the soil without the need for reseeding. Thus, after an initial screening (apart from the four species chosen for this study, Astragalus hamosus L., Lotus corniculatus L., Medicago polymorpha L., and Sulla coronaria (L.) B.H.Choi & H.Ohashi were assessed), four indigenous species that are common in the Mediterranean flora were selected: Festuca ovina L., Pilosella officinarum Vaill., Plantago coronopus L., and Plantago lanceolata L. These species were chosen on the basis of their different functional traits to assess their ability to compete with other species and form monospecific covers ( Table 2). Festuca ovina is a tuft-forming perennial grass previously used as under-vine plant cover [36], which forms a dense grass cover that could intercept weed seed rain, as described by Doisy et al. [44]. Pilosella officinarum is a prostrate growing plant with sexual, apomictic, and vegetative (by stolons) reproduction [45,46], which has a high colonizing potential as it is able to rapidly form monospecific patches and exclude other species [47], being considered a potentially invasive species in grasslands outside its natural range [48,49]. Plantago coronopus is a biennial (rarely perennial) rosette plant, while Plantago lanceolata is a perennial rosette species that has already been used in cover crop mixtures in vineyards [50]. Both Plantago species could control weed emergence by competing effectively for light and nutrients [20,23], with this competition expected to be greater in Plantago lanceolata, given its higher biomass. Nonetheless, their competitive ability may be conditioned by their different life cycles, i.e., perennials vs. biennials. Table 2. Traits of the plant species used as living mulches. Trait data have been compiled from FAWC Database [6]. Biomass estimation is described in Appendix A. RLF, Raunkiaer life form; PHV, plant height (vegetative); LA, leaf area; LM, leaf mass; LDMC, leaf dry matter content, SLA, specific leaf area; CSR, plant CSR strategy according to Hodgson et al. [51].

Experimental Design
The experimental design consisted of randomized blocks with four replications and six treatments: (i) an untreated control (UC); (ii) a weed-free control by manual weeding (WC) to estimate the vineyard growth development in the absence of weeds; and four under-vine living mulches, (iii) Festuca ovina (FO), (iv) Pilosella officinarum (PO), (v) Plantago coronopus (PC), and (vi) Plantago lanceolata (PL). Each treatment was distributed along eight inter-vine spacing, i.e., 9.6 m in La Poveda (8 × 1.2 m) and 8.8 m in El Socorro (8 × 1.1 m). The four selected species were planted in autumn 2018, in a strip 0.5 m wide under the vineyard row. With the exception of Pilosella officinarum, which was transplanted (density of 12 plants regularly distributed in the inter-vine spacing) from vegetative material collected from a nearby population, the rest of the species were sown from seed (seeding rates following the recommendations of the seed supplier: Festuca ovina, 9.5 g m −2 ; Plantago coronopus, 1 g m −2 ; Plantago lanceolata, 2 g m −2 ).

Field Data Collection
After the year of establishment, in the following three years (i.e., from 2019 to 2021) plant surveys were conducted at two different times, coinciding with the early and late crop growing season (May and September, respectively), by placing three sampling frames (0.33 × 0.33 m) in each of the 40 plots (2 fields × 4 replications × 5 treatments, excluding the weed-free control). At each sampling frame, the percentage cover of target species used as living mulch was estimated [52].
Previous studies in non-permanent cover crops have commonly used cover crop biomass as a predictor of weed suppression effect [22,53,54]. In our case, being a permanent cover crop, a non-destructive method was chosen in order not to affect plant viability, so the percentage cover of living mulches was measured as a possible predictor of weed suppression. Additionally, a rough estimation of living mulches biomass was made, as explained in Appendix A, in order to compare our results with previous studies.
Weed species (in a broad sense, all spontaneous flora) were identified from Castroviejo [55], updating the nomenclature according to International Plant Names Index [56], and weed density (number of individuals per m 2 ) was recorded. All weeds at each sampling frame were collected in paper bags, dried in the laboratory (oven at 105 • C for 24 h) and weighed on a precision balance to obtain the weed biomass data.
Finally, the cumulative effect of the three years of treatments on grapevine vegetative growth was estimated. For this purpose, the pruning weights of the six central vines in each plot (i.e., excluding border vines) in each of the 48 plots (2 fields × 4 replications × 6 treatments, including the weed-free control) were measured in the last year (i.e., 2021).

Statistical Analysis
The effect of the treatments on the response variables was measured separately for each experimental farm using independent models because of contrasting differences (i.e., soil type, weather, etc.) between the two farms. On the one hand, the degree of establishment of different species used as under-vine living mulches was analyzed, using the vegetation cover as the response variable, treatment (FO, PO, PC, PL), and year as the fixed factors. Year was used as a fixed effect since the temporal evolution of living mulches could follow a non-random pattern specific to each species. Living mulches vegetation cover was expressed as a decimal (i.e., range 0 to 1) to fit a generalized linear model with beta distribution [57].
On the other hand, weed data were pooled from the three sampling frames per plot, thereby obtaining a matrix that included data per plot for two sampling times and for three years (40 plots × 2 sampling times × 3 years; n = 240). Weed density models were constructed for overall species and, individually, for those species with a significant Plants 2022, 11, 1921 5 of 21 presence. To determine which species had significant presence, the absolute weed density data were transformed into relative density: Relative density = number of individuals of a species on a plot x total number of individuals on plot x × 100 A significant presence was thus established in this work for those species with relative density >3%. The effect of the different treatments (UC, FO, PO, PC, PL) on weed density and weed biomass was estimated using generalized linear mixed models, fitted with a negative binomial distribution for weed density and a gamma distribution for weed biomass. In total weed density and weed biomass models, the year was used as a fixed effect, considering that both total density and biomass of species depend on living mulches cover, which in turn depends on the year factor. However, in the individual species density models, the year was considered a random effect since the higher or lower density of a species can be driven by stochastic processes. In general, block was used as a random effect. The relationship of weed density and weed biomass (response variables) with living mulches cover was explored using linear regression models and Spearman's correlation test. In the same way, the relationship between living mulch cover and the ratio weed biomass/density was analyzed. The lower this ratio, the lower the average biomass of weeds present in a plot. This correlation could provide an analysis of whether weed biomass per plant unit increases or decreases as a function of living mulch cover.
Additionally, a linear mixed model was constructed to measure the effect of the different treatments (WC, UC, FO, PO, PC, PL) on grapevine vegetative growth, including block as a random effect. A logarithmic transformation (log[x + 1]) was performed to fit the response variable to a normal distribution. The relationship between grapevine vegetative growth and weed biomass was measured using a Spearman correlation test.
All statistical analyses were performed in R 4.1.0. Beta regression models were fitted with the betareg package in R [58], estimating its fit by pseudo R 2 . Generalized linear mixed models were fitted with the GLMMtmb package [59], computing marginal R 2 (fixed effects variance) and conditional R 2 (full model variance) using the performance package [60]. The corrected Akaike Information Criterion (AICc) was applied to screen the best models, selecting those with the lowest AICc score. For all models, the statistical significance of the fixed effects was estimated by a type III ANOVA test. To compare the mean of the levels of one factor, the Student's t-test was performed, adjusting the p-values with the Bonferroni correction.

Establishment of Under-Vine Living Mulches
The percentage cover observed in living mulches varied according to the species used, the year and the farm studied ( Figure 1). Living mulches of Festuca ovina and Plantago lanceolata (hereinafter, FO living mulch and PL living mulch) were successfully established on the two farms studied, showing an average cover (over the three years) of about 80% on both farms. In the case of FO living mulch, although in the first year the percentage of cover barely exceeded 40%, from the second year onwards, it reached cover close to 100%. Regarding PL living mulch, no significant inter-annual variations were found, although a decrease in the percentage cover was observed in the third year. The living mulch of Plantago coronopus (hereinafter, PC living mulch) was successfully established in the first year but showed a drastic decrease in the percentage cover from the second year onwards, thus registering in the whole three years and the two farms with the lowest establishment values compared to the other living mulches. In the case of living mulch of Pilosella officinarum (hereinafter, PO living mulch), the observed response was clearly different between farms. On farm "La Poveda" (hereinafter, farm A), although in the first year the percentage cover did not differ significantly from that observed for the other species, from the second year onwards, a decline began, which was significantly marked in the third year, when in some plots the percentage cover dropped below 10%. In contrast, the percentage cover of PO living mulch on farm "El Socorro" (hereinafter, farm B) increased over the years, from 68% to 95% average cover from 2019 to 2021, respectively.
with the lowest establishment values compared to the other living mulches. In the case of living mulch of Pilosella officinarum (hereinafter, PO living mulch), the observed response was clearly different between farms. On farm "La Poveda" (hereinafter, farm A), although in the first year the percentage cover did not differ significantly from that observed for the other species, from the second year onwards, a decline began, which was significantly marked in the third year, when in some plots the percentage cover dropped below 10%. In contrast, the percentage cover of PO living mulch on farm "El Socorro" (hereinafter, farm B) increased over the years, from 68% to 95% average cover from 2019 to 2021, respectively. . Significance levels (p-values) of the factor "treatment" and of the interaction between "treatment" and "year" are indicated in the text in red and black ink, respectively. For the "treatment" level, means were significantly different if red capital letters next to the dashed lines are different. For the "treatment: year" level, boxes with different lower-case black letters had significantly different means. In the upper left corner, pseudo R 2 indicates the variance explained by the beta regression model. Significance levels (p-values) of the factor "treatment" and of the interaction between "treatment" and "year" are indicated in the text in red and black ink, respectively. For the "treatment" level, means were significantly different if red capital letters next to the dashed lines are different. For the "treatment: year" level, boxes with different lower-case black letters had significantly different means. In the upper left corner, pseudo R 2 indicates the variance explained by the beta regression model.

Effect of Under-Vine Living Mulches on Weed Density
A total of 59 plant species were identified during the three years of weed sampling, 35 on farm A and 42 on farm B, of which 18 species were present on both farms (Table 3). On farm A, the most abundant species were Erigeron canadensis (relative density of 39%), Bromus madritensis (13%), Cyperus rotundus (12%), and Lolium arundinaceum (12%). Weed community composition was different on farm B, with Euphorbia prostrata (52%), Convolvulus arvensis (13%), B. madritensis (7%), and Sonchus asper (6%) as the most abundant species. Table 3. List of weeds present on the two study farms during the two sampling dates, May and September (average of the three years 2019 to 2021), ranked in order of highest to lowest relative density. Those species classified as noxious weeds are indicated with an asterisk next to the EPPO code. The + and -signs indicate that the relative density of the species was less than 0.01% (+) or that the species was not found (−).

EPPO
Amaranthus albus L. The use of living mulches had a significant effect on weed density, although the effect varied depending on the mulching species used and the weed species to be controlled. On farm A, FO and PL living mulches drastically reduced total weed density compared to the control, and a smaller but significant reduction in PC living mulch was observed ( Figure 2). On farm B, only the two Plantago species showed a significant reduction in total weed density, with the suppression by PL living mulch being particularly marked. Beyond the effect on noxious weeds, FO and PL living mulches significantly reduced the density of B. madritensis (BROMA) on both farms. In addition, PL living mulch significantly reduced the occurrence of the most abundant species on farm B, Euphorbia prostrata (EPHCH). Significance levels (p-values) of the factor "treatment" and of the interaction between "treatment" and "year" are indicated text in red and black ink, respectively. For "treatment" level, means were significantly different if red capital letters next to the dashed lines are different. For "treatment: year" level, boxes with different lower-case black letters had significantly different means. In the upper left corner, R 2 m indicates the variance explained by the fixed effects of the mixed model. Outliers are indicated with an asterisk (*). Table 4. Effect of treatments on the density of weeds with a significant occurrence (relative density > 3%). Those species classified as noxious weeds are indicated with an asterisk next to the EPPO code. The percentage of variation in weed density that is explained by the fixed effects of the model is indicated (R 2 m). A type III ANOVA was used to calculate the significance level of the effect (p-values: *** p < 0.001, ** p < 0.01, * p < 0.05, ns not significant), followed by a t-test adjusted with a Bonferroni correction for pairwise comparison. For each farm, treatment means sharing the same letter within each row are not significantly different. When the year factor was considered (i.e., treatment: year), different patterns were shown according to the living mulch species (Figure 2). For example, a regular behavior in all three years, either because showing significant differences with respect to untreated control (hereinafter UC), e.g., in PL living mulch on both farms and FO living mulch on farm A, or because showing no significant differences with respect to UC, e.g., in PO living mulch on farm A. However, other remarkable patterns were found in species such as PC Plants 2022, 11, 1921 9 of 21 living mulch, which showed a significant reduction in weed density only during the first two years, then losing this weed-suppressive capacity in the third year on both farms. An opposite pattern was observed in FO living mulch on farm B, since during the first year had no effect on weeds (UC = 279 plants m −2 ; FO = 308 plants m −2 ), but from the second year onwards, drastically reduced the total weed density.
The effect of living mulches on the density of weed species with a significant presence (i.e., relative density > 3%) is shown in Table 4. On farm A, the four living mulches significantly reduced the presence of the noxious weed E. canadensis (ERICA), especially FO and PL living mulches, with weed suppression of 92% and 97%, respectively. The density of S. asper (SONAS) was significantly reduced by PL living mulch on both farms, by FO living mulch on farm A, and by PO and PC living mulches on farm B. Noxious weed C. arvensis (CONAR) was also significantly suppressed by PL living mulch on farm B. Table 4. Effect of treatments on the density of weeds with a significant occurrence (relative density > 3%). Those species classified as noxious weeds are indicated with an asterisk next to the EPPO code. The percentage of variation in weed density that is explained by the fixed effects of the model is indicated (R 2 m ). A type III ANOVA was used to calculate the significance level of the effect (p-values: *** p < 0.001, ** p < 0.01, * p < 0.05, ns not significant), followed by a t-test adjusted with a Bonferroni correction for pairwise comparison. For each farm, treatment means sharing the same letter within each row are not significantly different.   Beyond the effect on noxious weeds, FO and PL living mulches significantly reduced the density of B. madritensis (BROMA) on both farms. In addition, PL living mulch significantly reduced the occurrence of the most abundant species on farm B, Euphorbia prostrata (EPHCH).

Effect of Under-Vine Living Mulches on Weed Biomass
Similar to that observed for weed density, the results of this study have shown how the use of different living mulches largely explains the observed variance in weed biomass in both farm A (R 2 m = 0.84) and farm B (R 2 m = 0.75) (Figure 3). Considering the three-year average, FO and Plantago (i.e., both PC and PL) living mulches significantly reduced weed biomass on farm A compared to the control, with an average reduction of 95% for PL living mulch (UC = 248.64 g m −2 ; PL = 12.25 g m −2 ) and 93% for FO (UC = 248.64 g m −2 ; FO = 18.07 g m −2 ). On farm B, a significant reduction was observed for all four mulching species, particularly in PL living mulch, which showed a 93% reduction in weed biomass compared to the control (UC = 204.59 g m −2 ; PL = 13.73 g m −2 ).
The analysis accounting for the year factor (i.e., treatment: year) showed that weed biomass reduction by FO living mulch on both farms was not statistically significant in the first year but was significant in the following two years. Regarding PC living mulch, weed biomass reduction was only significant in the first two years on farm B. Regular patterns throughout the years were found for the other two species used as living mulches, with significant differences with respect to UC in PL on both farms and PO on farm B, while no significant differences in PO on farm A.

Relationships between Living Mulch Cover, Weed Density and Weed Biomass
Overall, negative correlations were observed between living mulch cover and weed density and weed biomass, being clearly stronger on farm A than farm B and more robust when weed biomass was considered as a response variable (Table 5). These relationships were consistent as a similar response was observed on both farms. Indeed, a strong negative correlation was observed between living mulch cover and weed biomass in both farm A (r = −0.79) and farm B (r = −0.57), which exhibited a better fit using a quadratic regression models, where the slope of the curve flattens out at cover percentages above 75% ( Figure 4A). However, a different response depending on the type of living mulch was found when studying the relationship between weed biomass and living mulch cover ( Figure 4B).

Relationships between Living Mulch Cover, Weed Density and Weed Biomass
Overall, negative correlations were observed between living mulch cover and weed density and weed biomass, being clearly stronger on farm A than farm B and more robust when weed biomass was considered as a response variable (Table 5). These relationships were consistent as a similar response was observed on both farms. Indeed, a strong negative correlation was observed between living mulch cover and weed biomass in both farm A (r = −0.79) and farm B (r = −0.57), which exhibited a better fit using a quadratic regression models, where the slope of the curve flattens out at cover percentages above 75% ( Figure 4A). However, a different response depending on the type of living mulch was found when studying the relationship between weed biomass and living mulch cover ( Figure 4B). Table 5. Correlations between living mulch cover and weed density and weed biomass, calculated for the two farms (overall) and for each farm separately. The correlation value (r) was estimated using a Spearman correlation test. The significance level of the correlations is also indicated (*** p < 0.001).  Table 5. Correlations between living mulch cover and weed density and weed biomass, calculated for the two farms (overall) and for each farm separately. The correlation value (r) was estimated using a Spearman correlation test. The significance level of the correlations is also indicated (*** p < 0.001).

Weed density
Weed biomass Living mulch cover −0.60 *** −0.69 *** Farm A Weed density Weed biomass Living mulch cover −0.72 *** −0.79 *** Farm B Weed density Weed biomass Living mulch cover −0.45 *** −0.57 *** While FO and PO living mulch covers were strongly correlated with weed biomass, this relationship was weaker in the two Plantago species, especially in PL living mulch, where the observed correlation was not even significant. These differentiated responses therefore conditioned the average weed suppression reached by each living mulch. For example, PC living mulch resulted in high weed suppression (71%) with intermediate levels of cover (54%) on farm B, while PO living mulch required much higher cover (83%) to reach a similar level of suppression (68%) ( Table 6). When the correlation between living mulch cover and the ratio weed biomass/weed density was analyzed, contrasting responses were also observed according to the type of living mulch ( Figure 5). In this case, it was the FO and PL living mulches that showed a weaker correlation, exhibiting, on average, a markedly lower ratio than that observed for PO and PC living mulches ( Table 6).  While FO and PO living mulch covers were strongly correlated with weed biomass, this relationship was weaker in the two Plantago species, especially in PL living mulch, where the observed correlation was not even significant. These differentiated responses therefore conditioned the average weed suppression reached by each living mulch. For example, PC living mulch resulted in high weed suppression (71%) with intermediate levels of cover (54%) on farm B, while PO living mulch required much higher cover (83%) to reach a similar level of suppression (68%) ( Table 6). When the correlation between living mulch cover and the ratio weed biomass/weed density was analyzed, contrasting responses were also observed according to the type of living mulch ( Figure 5). In this case, it was the FO and PL living mulches that showed a weaker correlation, exhibiting, on average, a markedly lower ratio than that observed for PO and PC living mulches (Table 6).

Effect of Under-Vine Living Mulches on Grapevine Vegetative Growth
The different treatments applied explained part of the variance observed in grapevine vegetative growth (based on pruning weights), both on farm A (R 2 m = 0.36) and farm B (R 2 m = 0.24) ( Figure 6). However, although a slight reduction in pruning weight was observed in living mulches compared to weed-free control (WC) on both farms there were no significant differences between treatments. An analysis of the relationship between grapevine vegetative growth and weed biomass showed a negative correlation between the two variables, with pruning weight being lower the higher the weed biomass ( Figure 7). However, this correlation was only significant on farm A (r = −0.51*) while on farm B, it was clearly weaker (r = −0.10).

Effect of Under-Vine Living Mulches on Grapevine Vegetative Growth
The different treatments applied explained part of the variance observed in grapevine vegetative growth (based on pruning weights), both on farm A (R 2 m = 0.36) and farm B (R 2 m = 0.24) ( Figure 6). However, although a slight reduction in pruning weights was observed in living mulches compared to weed-free control (WC) on both farms, there were no significant differences between treatments. An analysis of the relationship between grapevine vegetative growth and weed biomass showed a negative correlation between the two variables, with pruning weight being lower the higher the weed biomass ( Figure 7). However, this correlation was only significant on farm A (r = −0.51*), while on farm B, it was clearly weaker (r = −0.10).

Extend of Establishment of Under-Vine Living Mulches
Our results revealed that the living mulches with the most successful establishment were PL and FO living mulches on farm A, and PL and PO living mulches on farm B. In general, similar patterns of establishment were found in both farms for FO, PC, and PL living mulches, which suggests that they were more influenced by intrinsic traits of each species than by differential environmental factors of each farm. Indeed, FO living mulch had a limited establishment in the first year (50.9% on farm A, 41.1% on farm B), but in  Correlation between weed biomass (log weed biomass) and grapevine vegetative growth (log Pruning weight) according to the factor "farm". Spearman's correlation coefficient (r) and significance level (no asterisks, non-significant; * p < 0.05) are indicated for each of the estimated correlations. Confidence intervals (95%) are shown with light-grey shaded areas.

Extend of Establishment of Under-Vine Living Mulches
Our results revealed that the living mulches with the most successful establishment were PL and FO living mulches on farm A, and PL and PO living mulches on farm B. In general, similar patterns of establishment were found in both farms for FO, PC, and PL living mulches, which suggests that they were more influenced by intrinsic traits of each species than by differential environmental factors of each farm. Indeed, FO living mulch had a limited establishment in the first year (50.9% on farm A, 41.1% on farm B), but in Figure 7. Correlation between weed biomass (log weed biomass) and grapevine vegetative growth (log Pruning weight) according to the factor "farm". Spearman's correlation coefficient (r) and significance level (no asterisks, non-significant; * p < 0.05) are indicated for each of the estimated correlations. Confidence intervals (95%) are shown with light-grey shaded areas.

Extend of Establishment of Under-Vine Living Mulches
Our results revealed that the living mulches with the most successful establishment were PL and FO living mulches on farm A, and PL and PO living mulches on farm B. In general, similar patterns of establishment were found in both farms for FO, PC, and PL living mulches, which suggests that they were more influenced by intrinsic traits of each species than by differential environmental factors of each farm. Indeed, FO living mulch had a limited establishment in the first year (50.9% on farm A, 41.1% on farm B), but in the second and third year they reached coverages above 80%, regardless of the farm. This is consistent with previous studies showing that Festuca ovina is a slow-growing species with slow establishment [61] and with a lower emergence rate than Plantago species used as living mulches [62]. On the other hand, the decrease in PC living mulch cover from the second year onwards probably occurred because Plantago coronopus is a biennial species; this study further found that spontaneous regeneration from self-seeding was very limited. Seedling recruitment of Plantago coronopus is strongly conditioned by rainfall [63,64], so low rainfall during the October 2020-May 2021 period (see Table 1) may have limited the recruitment of new seedlings on both farms. A lower recruitment was observed on farm A (21.4% less cover of PC living mulches than on farm B), which could be due to the lower rainfall received and a lower water retention capacity of the sandy loam soil in this farm. In contrast, the perennial species of the same genus Plantago lanceolata, showed a high cover on both farms during the three years of observations. On this basis, selecting perennial species as living mulch would seem more appropriate since they are able to achieve successful establishment without reliance on the success of seeding and recruitment, as these processes, in turn, depend on environmental conditions that are difficult to manage.
With regard to PO living mulches, contrasting differences were observed in the percentage of cover recorded on both farms (a 77.8% higher cover on farm B than on farm A in the last year). Pilosella officinarum is a drought-tolerant species that, however, does not seem to tolerate high temperatures [45]. In fact, in the Iberian Peninsula it is part of dry grasslands of cool environments, being mainly found in mountain areas [55]. The presence of a natural population of P. officinarum on farm B suggested that its establishment could be successful, at least on this farm, which was later confirmed in the results. Conversely, our hypothesis for farm A is that the higher temperatures reached in the summer period, more than 2 • C higher than on farm B (Table 1), could have limited their survival. Accordingly, under a global warming scenario, the use of PO living mulches in our latitudes should be limited to areas with cooler temperatures where the survival of P. officinarum is assured.

Weed Suppression by Under-Vine Living Mulches
This three-year study has shown that, as expected, successfully established living mulches (i.e., high percentage cover) generally showed high weed suppression, in accordance with previous studies on permanent living mulches [24]. The living mulch of Plantago lanceolata showed the highest suppression capacity, higher than in the grass living mulch Festuca ovina in both farms, which challenges the previous assumption that grass cover crops provide more suppression than broadleaf species [21,22]. This prior assumption may be true for broadleaf species commonly used as cover crops (i.e., brassicas, legumes) but should not be generalized. In addition, our results revealed that the extent of living mulch cover was a good predictor of weed suppression, being inversely related to weed density and biomass, regardless of the farm. This is in line with Steinmaus et al. [65], who found a similar relationship between living mulch cover and weed suppression. In light of this, our assumption is that the higher the percentage of soil covered by living mulches, the less surface area available for weeds to colonize, germinate, and establish. Nevertheless, the relationship between living mulch cover and weed biomass varied depending on the species used as living mulch ( Figure 4B). Thus, the weed suppression success of PO and FO living mulches was highly dependent on the coverage achieved, but living mulches of Plantago were less cover-dependent, especially PL living mulch, for which no significant correlation was observed. As discussed below, these differences could be due to the different suppression mechanisms of each species. In the case of PO living mulch, the small growth habit of Pilosella officinarum may limit its ability to compete for light and generate a shading effect, a key feature for weed suppression [20,66,67]. Therefore, the success of this species in suppressing weeds depends to a large extent on achieving high percentages of cover, as any gap in this living mulch provides a good chance for weeds to establish and thrive. The living mulch of Festuca ovina was also strongly cover-dependent, showing high weed suppression at high cover levels. However, its slow establishment generating only moderate cover during the first year may compromise weed suppression in this year of implantation. Once established and due to the dense tufted cover that this species forms, we consider that it can be very effective in intercepting seed rain [44] and suppressing weeds by shading, thus limiting weed seed germination and weed seedling growth. In fact, the low ratio observed between weed biomass and density in FO living mulch (Table 6) may be indicative of this suppressive effect on seedling growth. Similarly, a very low ratio was also found in PL living mulch. Both species (i.e., Festuca ovina and Plantago lanceolata) showed similar biomass and height, clearly higher than the other two species used (see Table 2). These traits could be considered a differential factor in terms of increased competition for light (the aforementioned "shading effect") and nutrients against weeds, in line with previous work indicating that cover crop biomass is a key driver of weed suppression [22,53,54]. However, all the above does not explain why PL living mulch, with a biomass similar to that of FO living mulch, showed a higher weed suppression capacity compared to other living mulches even at intermediate cover levels (Figure 4B). A plausible explanation could lie in a physical mechanism associated with leaf traits, whereby the leaves of Plantago lanceolata, with a significantly larger leaf area than the others (Table 2), could shade a greater soil surface, therefore, a medium cover would be enough to exert a competitive shading. On the other hand, the presence of phytotoxic compounds has been cited in Plantago lanceolata [68] and other species of the genus [69], so that weed suppression could be partly related to allelopathic effects, although this could not be demonstrated in our study. In this sense, a lower relationship between weed biomass and leaf mass cover was observed in PC living mulch, achieving higher weed suppression than FO and PO living mulches at cover levels below 70% ( Figure 4B). Given its growth habit and leaf size, we consider it unlikely that this is due to an increased shading effect, but possibly by an allelopathic effect.

Control of Noxious Weeds by Under-Vine Living Mulches
Our results indicated that the use of under-vine living mulches can drastically limit the occurrence of E. canadensis, as well as other problematic weeds in vineyards, such as Convolvulus arvensis or S. asper (Table 2). Erigeron canadensis is a highly competitive weed that poses a serious threat, given its resistance to glyphosate [70,71] and its depressive effect on grapevine vegetative growth [72]. This species was significantly controlled by all living mulches, especially by FO and PL living mulches (92% and 97% suppression), similar to the suppression levels found by Pittman et al. [73] in cover crops mixtures. Considering that germination of E. canadensis is highly light-dependent [74,75], our assumption is that Festuca ovina and Plantago lanceolata, with the highest shading capacity, were also the most effective in suppressing this noxious weed. This is in line with Steinmaus et al. [65], who observed that the use of living mulches in a Californian vineyard achieved high weed suppression due to light interception. In addition, Doisy et al. [44] have reported that grass covers (Festuca arundinacea Schreb and Dactylis glomerata L.) intercept 56% of seed rain of E. canadensis, which could likewise occur in FO living mulch. Similar effects of grass cover were reported on S. asper with 66% interception of seed rain [44], a species that also has a highly dependent germination on light received [74]. Therefore, as discussed for E. canadensis, we consider that these mechanisms could explain to a great extent the high suppression of S. asper by FO living mulch on farm A and by PL living mulch on both farms. The low suppression of S. asper by FO living mulch on farm B could be due, as discussed above, to the low establishment of F. ovina during the first year. This fact might also have limited the suppressive capacity of FO living mulch on C. arvensis, which was only significantly controlled by PL living mulch on farm B, with a 72% reduction with respect to control. This result is particularly relevant given that C. arvensis is a perennial geophyte, persistent and hard to manage, which can cause significant crop losses [76]. Our results contrast with the findings of Steinmaus et al. [65], as the living mulches used by them (Avena sativa L. and Vicia benghalensis L.) not only failed to control C. arvensis, but increased its dominance. Nonetheless, our results do not clarify the mechanism involved in the suppression of C. arvensis by PL living mulch, which could be due to a combination of all the mechanisms that have been discussed above.
Likewise, it should be noted that the two most abundant species in the vineyards studied were two invasive alien species: the aforementioned E. canadensis and Euphorbia prostrata, although the latter does not a priori pose any threat to grapevines. The application of irrigation in Mediterranean woody crops during the summer months can lead to the proliferation of alien weeds [77,78]. In the current context, in which the surface area occupied by irrigated vineyards in Mediterranean countries such as Spain is increasing exponentially [79], irrigated vineyard rows could be a potential reservoir of invasive alien weeds, some of which, such as E. canadensis or other species of the genus Erigeron, could be noxious grapevine weeds. Based on our results, the use of under-vine living mulches could be a successful option to avoid the proliferation of these species.

Effect of Under-Vine Living Mulches on Grapevine Vegetative Growth
Our results revealed no significant effect of the different treatments on grapevine vegetative growth, which contrasts with previous studies that observed a significant effect of under-vine living mulches on vegetative growth [37,38]. However, in these studies focusing on under-vine living mulches, the treatment is the same in both vineyard rows and inter-rows, similar to what Giese et al. [36] and Vanden Heuvel and Centinari [80] correctly referred to as "complete vineyard floor cover crops". However, in other studies in which, like us, they applied the same management in vineyard inter-rows [39,40,81], they found no significant effect of the under-vine living mulches on vine vegetative growth. Notably, when under-vine living mulches are used in newly planted vineyards, a negative effect on vegetative growth can occur [82,83], even if the same management was applied in all vineyard inter-rows. Presumably, young vines with a less developed root system seem to be more susceptible to competition from plant cover [84]. Conversely, in studies contrasting different treatments in the vineyard rows (e.g., herbicides, tillage vs. cover crops) while keeping the same management on vineyard rows, a significant effect on pruning weights was found [42,43,85].
Considering the above, our assumption is that management of vineyard inter-rows (in our case, up to five times more area than the rows) could have a significant effect on grapevine vegetative growth, proportionally greater than the effect of management on the vineyard rows. Therefore, when management in the vineyard inter-rows is the same, homogeneous results of vegetative growth in mature vineyards are to be expected, regardless of the treatment applied in the vineyard rows. In this regard, although no significant differences were observed between treatments, farm A (with tilled inter-rows) showed, on average, greater pruning weights than farm B (with mown inter-rows). This is fully consistent with the results observed in a recent study on farm B [7], in which the authors found that the percentage of bare soil (considering both vineyard rows and inter-rows) is inversely related to vineyard yield, but also to grapevine vegetative growth (unpublished data). Thus, as in the present study, management with tilled inter-rows (higher percentage of bare soil) showed significantly higher pruning weights than management with mown inter-rows (lower percentage of bare soil).
On the other hand, our results also showed a slight negative correlation between weed biomass and grapevine vegetative growth, although it was only significant on farm A. For the above reasons, we assume that the effect of management on vineyard inter-rows attenuated any possible relationships between these two variables. Moreover, considering the higher weed biomass on farm A, associated with a high presence of the noxious weed E. canadensis, could justify the stronger correlation observed on farm A with respect to farm B, the latter with a clearly lower impact of weeds on the vegetative growth of the vineyard. Even so, the observed correlation supports the premise that control of weed biomass under vines may play a relative role in regulating grapevine vegetative growth.

Implications for Vineyard Weed Management
To our knowledge, under-vine living mulches have so far been a very unusual management in Mediterranean vineyards, being limited to experimental studies [40]. However, our study suggests that permanent under-vine living mulches can be successfully established in a semi-arid Mediterranean climate, suppressing noxious weeds without jeopardizing the grapevine vegetative growth. In conditions of water scarcity, it seems daring to consider keeping the whole vineyard surface (i.e., row and inter-row) with a permanent vegetation cover. Since the competition of plant cover with the vineyard will largely depend on the percentage of bare soil [7], and this percentage can be adapted according to soil properties to avoid competition for water [86], therefore a reasonable option would be to use permanent living mulches in irrigated vineyard rows, where there could be a higher occurrence of noxious grapevine weeds [7], and keeping free of vegetation the inter-row space through soil management that prevents competition with the vineyard as much as possible.
Finally, under-vine living mulches have a low establishment cost (see Appendix B) and, as has been demonstrated here, can suppress weeds for three years without the need for mechanical or chemical operations, reducing the economic and environmental costs as well as damage to vines associated with conventional weed management. It would be advisable to study their evolution on a longer time scale, comparing unmanaged living mulches and living mulches with minimal management (e.g., annual mowing, basic fertilizer needs, etc.), which could contribute to the maintenance of the living mulch and, consequently, weed suppression over time. For a large-scale extension to commercial vineyards, it would also be necessary to develop more efficient sowing methods based on improved vineyard seeders that can adapt to both small herbaceous seeds and the reduced dimensions of the inter-vine spacing.  FO living mulches, 0.28 g m −2 of PO living mulches, 0.53 g m −2 of PC living mulches, and 1.20 g m −2 of PL living mulches.