Introduction

1. Grapevine yields

Grapevine yield has historically received less attention compared to other crops, largely due to an assumed inverse relationship between grape yield and wine quality (Bisson et al., 2002; Jackson & Lombard, 1993; Poni et al., 2018). European viticulture is often grown under the supervision of wine geographical labels, where maximum yield thresholds are set to maintain wine quality (Stranieri & Tedeschi, 2019). Despite these regulations, many vineyards do not reach the maximum permissible yields established by these quality standards. In France, research by Schauberger et al. (2018) indicates a plateau in grapevine yields since the 1980s, leading some researchers to suggest a potential “vineyard decline”, which may be attributed to environmental and management factors (Riou et al., 2016). Furthermore, Mediterranean regions, already known for their warm climates and low precipitation, are expected to experience negative impacts on grape yields due to climate change, manifesting through extreme temperatures, drought, and intense rainfall (Bernardo et al., 2018; Droulia & Charalampopoulos, 2021; Simonneau et al., 2017; Touzard et al., 2017). This underscores the urgent need to deepen our understanding of the factors affecting grapevine yield to investigate possible management practices to sustain or increase it.

In Languedoc-Roussillon, vineyard area declined from 250,000 to 180,000 hectares between 2000 and 2020 (DRAAF Occitanie, 2024). This one-third reduction in viticultural activity highlights the fragility of the sector and the profitability risks faced by wine producers in the region. Given that increasing the added value of wines is not always a viable strategy for these companies, we hypothesise that reducing grapevine yield gaps could contribute to sustaining viticultural activity in the region. Grapevine cropping systems present a high diversity of farming management practices performed by winegrowers according to their objectives for wine yield, wine quality and vineyard sustainability. While pedoclimate conditions (Fernandez-Mena et al., 2023) and vineyard design at planting (Fernandez-Mena et al., 2025) have an impact on yield, this is also the case for annual farming practices in vineyards that ensure adaptation to daily weather and sanitary conditions. By irrigation, soil and canopy management, winegrowers deal with water availability, the main driver for grapevine yield in the Mediterranean wine regions (Mirás-Avalos & Araujo, 2021), and they conduct pest management to preserve grapevine health. Therefore, the implementation of key management practices is crucial for addressing both the annual variability of climate conditions and long-term climate change impacts on grapevine yield (Gutiérrez-Gamboa et al., 2021; Naulleau et al., 2021), providing ecosystem services in vineyards (Giffard et al., 2022) and obtaining a high-quality wine (Reynolds, 2022).

2. Annual practices in vineyards that may be involved in yield losses

Difficult to predict and manage by winegrowers, some climate events have always had a great impact on the grapevine yield gaps. The climate risks include several phenomena, such as hail, extreme dryness, spring frost and heatwaves. Exacerbated by climate change, their frequency has increased and will long-term severely affect grapevine yield for most future climate projections (Fraga et al., 2015), although linked to high uncertainty (Bois et al., 2023). In addition, pest hazards are exceptional pest attacks, often linked to fungi’s fast spread, such as downy or powdery mildew, that can significantly impact grapevine yield (Kassemeyer, 2017). Unfortunately, it is difficult for winegrowers to predict these attacks and integrate an adaptive management during the year, because of the uncertainty of local weather conditions and the intensity of regional pest pressure (Fermaud et al., 2016).

Apart from the exceptional pest events, winegrowers can protect their vineyard against regular pest attacks by implementing decision rules in an integrated protection strategy. Integrated pest management strategies in vineyards include several annual practices, such as the optimisation and timing for the pesticide application, the reduction of the number of treatments and the substitution of chemical pesticides with biocontrol agents (Pertot et al., 2017). There is a wide range of pest management options depending on winegrowers’ choices (Hossard et al., 2022), which are influenced by technical advisers (Hillis et al., 2016) and decision support tools (Carlos & do Carmo, 2022).

Another key component modulating grapevine yields is the soil management of the vineyard. The soil management strategy includes a set of techniques that may help grapevine development and health, as well as other ecosystem services (Giffard et al., 2022). Tillage, herbicide spraying and cover cropping mowing serve to manage the spontaneous or planted vegetation, since some of these plants may compete with the grapevine for nutrient and water resources, depending on particular periods, spatial coverage and growth levels (Fernandez-Mena et al., 2021; Ripoche et al., 2010). In addition, organic amendments and mineral fertilisation can help soil structure and nutrient availability in vineyards, enhancing grapevine growth and yield (Arrobas et al., 2014).

3. Gap of knowledge and purpose of the study

To date, no comprehensive study has investigated the impact on grapevine yield of management practices in vineyards made by a large sample of winegrowers within a specific viticultural region. In addition to accomplishing a whole grapevine yield gap analysis of Languedoc-Roussillon, annual farming practices are an important component of the grapevine yield equation that completes the environmental and the vineyard design factors already analysed in previous studies. These previous studies evidenced that environmental factors explained more than half of the yield variability at the municipality scale (Fernandez-Mena et al., 2023), whereas vineyard design factors, plant material choices and related label options explained an important part of the yield variability at the plot scale (Fernandez-Mena et al., 2025). Therefore, leaving an important unknown part of the grapevine yield gap that may correspond to annual adverse events and farming practices in Languedoc-Roussillon vineyards.

In this study, we examined how management practices and climate and pest hazards could be involved in grapevine yield gaps by analysing a large dataset derived from surveys conducted with randomly selected winegrowers and vineyard plots in the Languedoc-Roussillon region, in southern France. First, we investigated the discrepancy between the target yield set by winegrowers, the wine label yield threshold and the actual yield achieved to better understand yield gaps. Second, we analysed the impact of exceptional climate and health hazards on these yield gaps. Third, we assessed how pest management practices employed by the winegrowers in our dataset contributed to yield losses. Finally, we explored the effect of soil management, namely cover cropping and fertilisation, on possible yield gaps.

Materials and methods

1. Languedoc-Roussillon viticultural region

The Languedoc-Roussillon viticultural region, formerly an administrative region in France until 2015 and now part of the Occitanie administrative region, is situated on the Mediterranean coast and shares a southern border with Spain (Figure 2). According to national statistics by FranceAgriMer (2020), during the years under study, Languedoc-Roussillon accounted for an area of around 240,000 hectares of planted vineyards, the largest in France and the second-largest wine-growing region in Europe. Fernandez-Mena et al. (2023) provided detailed descriptions of the climate and soil properties in the wine-growing areas in Languedoc-Roussillon. The authors divided the region into seven zones. Areas with higher soil water-holding capacity are generally located near the coast, while shallower soils are found inland and along the southern coast. Soils are predominantly basic, although acidic soils are present in mountainous areas. The region is characterised by a dry Mediterranean climate, except for the western area near Carcassonne, where a more humid climate begins to appear. Heat waves are most frequently observed in the areas surrounding Nîmes.

Vineyards in this region adhere to the regulations of two primary types of wine labels, each with specific constraints concerning maximum yield. These labels are PDO (Protected Designation of Origin), covering 86,000 hectares and PGI (Protected Geographical Indication), covering 140,000 hectares. While numerous labels exist for both types, the largest in terms of both area and number of winegrowers are the PDO label AOC Languedoc and the PGI label IGP Pays d’Oc. Typically, PDO label declarations limit cultivation to a small number of grapevine varieties, emphasising the preservation of local wine typicity and imposing more stringent requirements compared to PGI labels. For instance, PDO labels in Languedoc-Roussillon generally set maximum thresholds for grapevine yield at around 50 to 60 hl·ha^–1·year^–1 of wine, whereas PGI labels allow higher thresholds, reaching 90 to 100 hl·ha^–1·year^–1 (INAO, 2010).

2. Data sources and dataset

We used data from surveys conducted by Agreste, the agricultural statistical service under the French ministry of Agriculture. These surveys were conducted by public agents who collected information on vineyard characteristics and farming practices from randomly selected plots engaged in full wine production in the specified year and aged over three years. Our dataset comprises information from 3,507 vineyard plots situated in the Languedoc-Roussillon region, obtained from four surveys conducted in the following years: 742 plots in 2016 (Agreste, 2018); 846 plots in 2013 (Agreste, 2015); 940 plots in 2011 (Agreste, 2012); and 979 plots in 2006 (Agreste, 2008). Access to the surveys involved confidential data handling, and thus, we accessed and processed the information within a secure environment provided by CASD—Centre d’accès sécurisé aux données (Fernandez-Mena et al., 2020). The exact location of the plots inside the Languedoc-Roussillon region was confidential and could not be extracted.

Key variables of the raw survey used in this study are available in an anonymised version in Fernandez-Mena (2024a) and Fernandez-Mena (2024b). The variables used in this study are presented in Table 1. Not all variables were available for all the years, reducing the number of plot data points for these variables.

3. Grapevine yield gap analysis for annual farming practices in vineyards

We presented the yield distribution of the survey sample, as well as their wine label and other management characteristics, to assess the sample representativeness of the grapevine cropping systems in Languedoc-Roussillon. Then, we analysed the possible yield gaps as described by winegrowers and four annual factors that may cause yield gaps, explained as follows.

Firstly, particularly important for yield gap analysis in grapevine cropping systems, we estimated the existing gap between the target yield established by the winegrowers themselves and the actual yield they obtained. Differently to other crops, the target yield in grapevine is often different to the potential yield. It is usually limited by a wine label threshold to preserve wine quality, but it can also be constrained by winegrowers, seeking a specific quality or lowering their expectations because of environmental limitations. Therefore, we compared the target yield declared by winegrowers in the surveys with the threshold generally established by wine labels (PDO and PGI) in Languedoc-Roussillon, to observe variations from that threshold due to environmental limitations or target wine quality. Then, we assessed the degree of achievement of the winegrower’s target yield, comparing it with the wine label threshold yield and with the actual yield obtained. In Languedoc, there are around 50 different wine labels, either PDO or PGI. In this survey, only the type of wine label was recorded: PDO, PGI or no label, but the specific name of the label was not specified. Therefore, we calculated the yield threshold for each plot by using the threshold from the most probable wine label, based on statistical data (DRAAF Occitanie, 2024).

Secondly, we analysed how exceptional climate and pest hazards affected grapevine yield gaps in our dataset. Climate hazards included several phenomena such as hail, frost or heatwaves and pest hazards included any exceptional pest attack. Climate hazards were not measured but described by the winegrowers, who noted if the surveyed plot suffered from one of these hazards during the corresponding year, which could importantly affect plot yields. Therefore, we compared yields obtained from the plots suffering hazards with the plots not suffering any hazards.

Thirdly, we assessed how yield gaps could arise based on the pest management practices employed by the winegrowers in our dataset. We started quantifying the pest pressure declared by winegrowers from the four main pests in the region: powdery mildew (Erysiphe necator), downy mildew (Plasmopara viticola), botrytis bunch rot or grey mould (Botrytis cynerea) and the larvae of the European grapevine moth or eudemis (Lobesia botrana) that could eventually also be associated with larvae of cochylis (Eupoecilia ambiguella). We then analysed how grapevine yield was affected by the level of pest pressure for each of these pests, according to different types of pest management. Pest management was measured by the Treatment Frequency Index (TFI), as described in Mailly et al. (2017). The total TFI includes all pesticides, including fungicides, insecticides and herbicides. TFI is calculated as the number of full treatments per year, considering as full treatments those using the recommended dose of active ingredients per application. In addition, organic and wine quality labels were considered as differentiated factors for pest management.

Fourthly, we analysed how cover cropping affected grapevine yield gaps. In the surveys, winegrowers described their row and inter-row management with three categories for each one: bare soil, temporary, and permanent cover cropping. We used the Coverage Index (CI) proposed by Fernandez-Mena et al. (2021) to classify the cover cropping practices. We considered a bare soil row or inter-row as CI = 0; a temporary cover cropped row or inter-row as 0.5; and a permanent cover cropped one as 1. Then, we calculated the total CI using the standard row and inter-row width in Languedoc-Roussillon vineyards, i.e., inter-rows equivalent to two-thirds of the total surface area and rows to one-third of it. The CI helped us to analyse the degree of soil coverage and bare soil that could be linked to yield gaps.

Finally, we studied how fertilisation influenced grapevine yield gaps. We analysed the yields of plots fertilised in the surveyed year and fertilised in the last five years, with both organic and mineral product applications. We also assessed the influence of the nitrogen (N) balance by calculating the N exported in the harvested grapes and contrasting it to the N applied to the soils. We finally estimated, using linear regressions, the yield gap produced by the N, P and K rates in kg per ha.

Generally, we applied data visualisation and statistical analyses to see the influence on yield gaps of the different annual factors presented in our database (Table 1). Since we were dealing with complex grapevine systems with multiple variables in interaction, the complete isolation of these factors is not possible from rest of the factors creating yield gaps, hence, we plotted yield results by adjusting the levels of some of the most important management categories, such as the wine label or the organic management. For each factor analysed here, we used yield mean comparisons that collaborated to identify yield gaps associated with those factors. For categorical factors, we assessed normality of residuals and conducted Levene’s test, failing to apply parametric tests. Therefore, we used non-parametric tests: the Kruskal–Wallis test for ranked data and the Wilcoxon signed-rank test for post-hoc mean comparisons. In our study, the basic unit for the statistical analysis is the plot. All plots count the same in the analysis, regardless of the area.

Data mining, data transformation, data visualisation and statistical analyses were performed using the R software (R Core Team, 2014) and the following R packages: ‘tidyverse’ (Wickham et al., 2019a); ‘dplyr’ (Wickham et al., 2019b); ‘tidyr’ (Wickham et al., 2019c); ‘ggplot2’ (Wickham, 2016); ‘ggExtra’ (Attali & Baker, 2023); ‘sjPlot’ (Lüdecke, 2021), ‘FactoMineR’ (Husson et al., 2017), ‘ranger’ (Wright et al., 2019), ‘lmtest’ (Hothorn et al., 2015).

Results

1. Sample and yield distribution

According to the survey sample, PDO wine plots represented 42 % of the data, and PGI wine plots 49 % of the data. The wine label plot proportions corresponded very well with the actual proportions in the official census by FranceAgriMer (2020). Only unlabelled wine plots, i.e., plots actually managed without any associated wine label, were slightly over-represented in the surveyed sample (9 % of the plot data) compared to the Languedoc-Roussillon census (5 % of the viticultural area).

The average yield of all plots was 54.95 wine hl·ha^–1·year^–1. The global yield distribution of the surveyed plots presented two peaks, a first peak around 40 wine hl·ha^–1·year^–1 and a smaller second peak around 75 wine hl·ha^–1·year^–1 (Figure 2A). The first peak of the whole yield distribution involved numerous PDO wine plots, whereas the second peak corresponded to PGI wine and unlabelled wine plots. Logically, PDO wine plots exhibited lower yields compared to PGI wine plots and unlabelled wine plots, with no observed differences between the latter two, as determined by the Wilcoxon signed-rank test (Figure 2B). Across the four surveyed years, average grapevine yields were comparable. However, yields in 2006 stood out significantly, demonstrating 6 % higher values than the average for both PDO and PGI wine plots (Figure S1).

2. Target yield and yield achievement

We estimated the yield threshold required by the type of wine label for each of the surveyed plots. This value can actually differ from the target yield of the winegrower. In 63 % of plots, target yield and yield threshold were equivalent, whereas in 36 % of plots the target yield was lower (Figure 3).

The actual yield met or exceeded the target yield set by the winegrower on 42 % of the plots, whereas in the remaining 58 %, the target yield was not achieved (Figure 4A). Regarding the percentage of the target yield achieved for each plot, we observed that plots with no wine label achieved their target yield better than PDO wine plots (Figure 4B). When selecting only plots with target yield not achieved, the percentage of achievement was around 63.2 % of the target yield.

3. Climate and pest hazards associated with grapevine yield gaps

In total, 33 % of the plots suffered a climate or a pest hazard declared by the winegrower. Out of these plots, 71 % were only affected by climate hazards (hail, frost or heatwaves), 17 % only by a pest hazard, and 14 % were affected by a combination of both types of hazards. The majority of plots, 67 % did not suffer any hazard for the year surveyed.

According to the Kruskal–Wallis test, these hazards presented a significant effect on grapevine yield (Figure 5). Climate and pest presented a similar level of yield gaps of around 12-15 wine hl·ha^–1·year^–1. The combination of both hazards presented the highest yield losses, with a gap of almost 25 wine hl·ha^–1·year^–1 compared to non-damaged plots (Figure 5).

4. Yield gaps associated with pest and disease pressure and crop protection

Powdery mildew is the most common pest in the Languedoc-Roussillon region, evidenced by the high percentage of winegrowers declaring medium and high pressure (Figure S2). Downy mildew and European grapevine moth were also notable pests that affected around 30-40 % of the plots for some years with medium and high pressure. Botrytis bunch rot or grey mould was less frequent in the region and presented the highest pest pressure episode in 2013, with 18 % of plots under medium or high pressure. The highest overall pest pressure was observed in the 2013 vintage, regardless of the pest (Figure S2), probably linked to pest pressure and climate conditions, since 2013 was the most humid year in the region (as calculated in Table S1). TFI data were only gathered for 2011 and 2016, two years that presented an average pest pressure.

Over the two years of pest management data available, the average total TFI of all plots was 11.9, and the standard deviation was 4.2. Organic plots exhibited a significantly lower TFI (9.8), with two points less than the conventional plot that averaged 12.1. Around 80 % of the TFI corresponded to fungicides, which were the most used products for grapevine protection in our dataset. Interestingly, whereas a decrease is observed in organic winegrowers, we did not observe any significant link between fungicide TFI and pest pressure felt by conventional winegrowers for the main fungi diseases, powdery mildew, downy mildew and botrytis bunch rot or grey mould (Figure 6). On the contrary, we evidenced significantly lower insecticide TFI for plots with low pressure felt by winegrowers of European grapevine moth, by two or three treatments less than in the high-pressure plots (Figure 6).

Regarding the TFI of the main wine labels, the average TFI of PDO wine plots was 10.8, two TFI points lower than PGI wine plots and plots without wine label, with an average TFI of 12.7 and 12.2, respectively. Inside each wine label category, there was no significant increase in wine yield with higher TFI (Figure 7). Nevertheless, some conventional plots with very few treatments, i.e., TFI lower than six, displayed yield gaps of around 10-20 % losses compared to the rest. Additionally, yield gaps were not appreciated in plots with high pest pressure (Figure S3). These results evidenced that increasing pesticide treatments over 12 points did not serve to increase yields for all levels of pest pressure.

5. Yield gaps associated with cover cropping management

The spatial and temporal distribution of cover cropping in the row and inter-row of the surveyed plots led to a large range of CI (coverage index), from 0 to 1 (Figure S4). The average CI was 0.147, quite low, since the most common soil management was bare soil (CI equal to 0), performed either by tillage or herbicide application and adopted by 71 % of winegrowers. The herbicide use in the region was low, with only a 0.5 TFI for herbicides per year on average and 0.55 for plots with bare soil. Across years, the average CI was higher in 2006 and lower in 2013 (Figure S5). For PDO wine plots, the average CI was 0.154, higher than PGI wine plots with a CI equal to 0.133. Moreover, organic wine plots exhibited a CI of 0.189, significantly higher than the CI of the conventional plots, which was 0.141. The wine yield of the different cover crop management strategies was comparable (Figure 8), both for PGI and PDO wine plots (Figure S6). The plots leaving bare soil the whole year (CI = 0) presented higher yields, but comparable to other CI levels with full coverage (CI = 1) or high (CI = 0.5-1).

6. Yield gaps associated with fertilisation

In 54 % of the plots, fertilisation was applied during the surveyed year, either with organic or mineral products. In the last five years, 82 % of the plots were fertilised at least once. Mineral fertilisation was used in most cases, i.e., in 91 % of all the plots, whereas organic fertilisation was applied only to 24 % of the fertilised plots, most of them combining both types of fertilisation. A similar proportion of plots were fertilised for both PDO and PGI wine labels (Figure 9). Fertilisation increased yield in PGI wine plots, especially when applied in mineral form. Unfertilised plots exhibited a 10 % yield gap compared to fertilised plots in PGI plots, whereas there was no such yield gap in PDO plots.

Among the fertilised plots in the last year, 85 % received nitrogen (N), 79 % phosphorus (P), and 73 % potassium (K). The three elements were applied in 77 % of the cases. From both organic and mineral fertilisation, N and P applications ranged from 15 to 80 kg·ha^–1·year^–1. An estimation of the exported N ratio by grape harvest compared to the N applied by fertilisation showed a high ratio of likely over-fertilised plots (Figure S7). When representing the influence of N and P applications on yield (Figure S8), we observed a high dispersion (R² < 0.01) that did not yield meaningful results. In these relationships, there is a small positive influence of N and P applications on yield, where a 50 kg N application could potentially increase yields by almost 5 % in PGI wine plots. Finally, no observable link was found between K fertilisation and grapevine yield.

Discussion

1. Particularities of grapevine yield gap analysis

The originality of this work lies in its foundation, not on experimental results but on observations gathered through a large-scale survey of the decisions, practices, adverse events and yields declared by numerous winegrowers across a broader range of conditions than experimental settings typically permit. This extensive approach, commonly applied to annual crops in yield gap research (van Ittersum et al., 2013), represents a significant advancement in documenting the impact of various farming management practices on grapevine yield gaps.

Specifically for yield gap analysis in viticulture, two yield references should be considered: the wine label yield threshold and the winegrower’s target yield. The first informs about the maximum yield that can be produced to reach the quality standard of the label, whereas the second is the winegrower’s production expectation for each plot (Merot et al., 2022). Target yields can be lower than the label yield threshold in cases when winegrowers target a specific wine quantity or when environmental or pest constraints limit their yield expectations. In the surveys that we used, target yields were reported after harvest. They may integrate some of the constraints experienced during the season (pest attacks, droughts, etc.) compared to the target yields surveyed at the beginning of the season. In our database, the winegrowers’ target yields were below the wine label thresholds in 36 % of the cases. Furthermore, our findings revealed that 58 % of the plots did not achieve the target yield. For these plots, only 68 % of the total target yield was reached on average, involving important yield gaps that may cause a great lack of profitability.

Average annual yields for PDO, PGI and unlabelled wines were of the same order of magnitude as those published in official statistics for the same region, yet the latter do not show a higher yield in 2006 (FranceAgriMer, 2013; FranceAgriMer, 2020). Unsurprisingly, climate and pest hazards varied during the surveyed years and within the region, certainly contributing to important yield gaps across a large range of management options. For each of the four years, the weather chronicles from MeteoFrance (2024) report climatic events which may have been very localised (hail, frost) or have affected larger areas of the region (drought, heat waves). Even if climatic events were noted by the subjective perception of the winegrowers, the regional impact on grapevine yield of several key climatic indicators for the whole Languedoc-Roussillon region was quantified by Fernandez-Mena et al. (2023). For instance, the 2019 heatwave was assessed in Languedoc-Roussillon by Lopez-Fornieles et al. (2022), evidencing yield losses from 10 to > 40 % depending on the area. Climate change predictions for heatwaves impact established up to 35 % yield losses in southern Europe (Fraga et al., 2020). Concerning droughts, Yang et al. (2022) estimated yield losses from 5-15 % due to water stress in Languedoc-Roussillon, in future scenarios when shifting the flowering-veraison 25 % earlier.

Powdery mildew was the disease with the highest pressure, consequently capable of creating higher yield gaps than other pests and diseases. An important trade-off has been demonstrated between maintaining or increasing grape yield and reducing grapevine susceptibility to powdery mildew (Guilpart et al., 2017). Hence, vigorous vineyards in plots targeting high yields could have been more affected by powdery mildew attacks. Growing resistant grapevine varieties will reduce this source of yield losses (Merdinoglu et al., 2018), estimated to save up to $48 million per year in California (Fuller et al., 2014).

Overall, the two types of hazards lead to significant reductions in yield, but our data do not make it possible to establish in what way. In addition, Guilpart et al. (2014) have observed that environmental conditions during the post-flowering period in one year have an impact on yield the following year. Consequently, inter-annual variations of climatic and pest conditions hinder correlations between them and the yield of the surveyed year and also of the next year.

2. Low pesticide use is not necessarily associated with yield gaps

Concerning pest management, our results highlighted that, unexpectedly, high TFI was not necessarily associated with high pest pressure (as experienced by winegrowers) in conventional plots, which reveals that organic winegrowers were better at adjusting fungicide use to pest pressure. Moreover, higher levels of TFI did not demonstrate superior yield performance across a large range of management options and levels of pest pressure, meaning that even low and medium levels of pesticide use were enough to protect the grapevine from pests and diseases. As an exception, only some conventional plots with very low TFI levels (< 6) showed yield gaps of around 10 % reduction on average. Therefore, a substantial reduction of pesticide use seems to be possible, i.e., lowering TFI to less than 12 in Languedoc-Roussillon, without compromising yields. Better adjusting the treatments with the pest pressure and timing can be one of the keys to providing avenues for pesticide reduction (Pertot et al., 2017). Although TFI was available for only two years, the strength of the study lies in the large dataset, comprising 1,772 plots. Regarding the inter-annual variability of pest pressure, reductions in pesticide use observed in certain years may not be feasible in years with severe pest outbreaks.

Only a few studies analysed large samples of winegrowers’ strategies to perform pest and disease control. Namely, Fouillet et al. (2022) with 161 grapevine cropping systems sampled in France, evidenced a significant average yield reduction of 19 % after reducing TFI by 33 % on average over 10 years. However, their sample for Languedoc-Roussillon did not demonstrate any significant yield change. In addition, Fouillet et al. (2024) complementary statistical analysis did not show any significant correlation between pesticide reduction and yield. These studies confirm our findings, evidencing that there is room for reducing pesticide use without generating yield gaps in vineyards of the Mediterranean region.

Similarly to other studies examining pesticide use, we only analysed TFI data in the present study. TFI reflects the annual quantity of pesticides applied, but not the protection strategy or the relevance of the applications. Therefore, for a given TFI, the dates and doses of application may not be properly selected to enable good control of pests. By analysing surveys on pesticide use in 11 French wine-growing regions, Mailly et al. (2017) observed that delaying the first fungicide application was the most common way of reducing TFI. Yet they did not analyse the impact of such a change on the quality of disease control, nor on the resulting yield.

3. Yield gaps appeared only with full soil coverage and a lack of fertilisation

Cover cropping with different CI levels did not result in significant yield gaps. Changes of CI may be partly explained by a strategy of flexible cover crop management that takes into account water stress during the vegetative season to adapt the control or destruction of the cover crop (Ripoche et al., 2010). This flexible cover crop management could be done at the spatial scale, depending on the plot location and its features. However, we did not find temporal adaptation to seasonal changes in water availability according to the climatic data that we observed (Table S1). The average CI was lower in 2013 compared to 2016, although the water balance in the season from October 2012 to September 2013 (P-PET = –620 mm) was not dryer than in the 2015-2016 season (P-PET = –856 mm), where P is the total precipitation and PET the total potential evapotranspiration, calculated over the main weather stations across the Languedoc Roussillon (Table S1). These results aligned with Fernandez-Mena et al. (2021) that analysed the soil management practices of 334 winegrowers in Languedoc-Roussillon, evidencing that the cover cropping strategy chosen by the winegrowers was not particularly linked to the soil and climate characteristics of the vineyards, but it was rather associated to other factors, such as the expected yield, the wine label, or the organic management. Our results also evidenced that more work needs to be done for increasing CI while adapting the service crop termination strategy that increases soil ecosystem functions and limits competition with the grapevine (Garcia et al., 2024).

Fertilisation was not performed every year, and only half of the vineyard plots were fertilised the previous year. However, more than 80 % of plots were fertilised at least once in the last five years. In our dataset, a 10 % reduction in yield gap was observed in PGI wine plots that were not fertilised either in the last year or in the last five years. Nevertheless, PDO wine plots did not showcase differences between fertilised and non-fertilised plots, probably due to a lower level of nutrient extraction or a nutrient balance fulfilled by crop residue return. Some authors argued that N fertilisation needs in grapevine greatly depend on nutrient recycling from crop residues returned to soils. Arrobas et al. (2014) and Celette et al. (2009) estimated that grape clusters removed about 20-30 kg N ·ha^–1, although the cropping system was able to recycle almost 50 kg N ·ha^–1 by returning the leaves and shoots into the soil. Unfertilised PDO wine plots in our database could probably fulfil their N extraction with crop residues returned to the soil. However, information about the crop residues returned to soil, and soil N organic and mineral content was not available in this survey.

4. Limitations of the approach and perspectives for improvement

Grapevine yield results from a combination of factors, including environmental conditions, plant material, plantation patterns, wine label and management practices. Given that our analysis was based on survey data encompassing a wide variety of complex viticultural systems across the Languedoc-Roussillon region, it hindered the isolation of each factor on the final grapevine yield. To address this complexity, we prioritised cross-variable plotting and employed statistical analyses to explore the influence of farming practices on grapevine yield. Due to privacy regulations, the exact locations of the surveyed plots within the region could not be disclosed. However, with the necessary privacy permission, data on farming practices could be integrated with the climate and soil agroecological zones of Languedoc-Roussillon, as described by Fernandez-Mena et al. (2023). In future studies, the strategic choices (plant material, wine label, technological and vineyard design options), described in Fernandez-Mena et al. (2025), should be combined with farming practices and soil and climate zoning to provide a whole grapevine yield gap analysis that unravels the contribution of the different factors involved, i.e., environment, strategic choices and annual farming practices.

In this study, the basic unit used was the plot, regardless of the plot area. When available, plot area data could be used to weight the relative importance of each plot in the statistical analysis. Furthermore, the lack of economic data limited the assessment of actual profitability risks associated with yield gaps. In future research, incorporating economic information alongside farming practices could enhance the understanding of whether yield gaps can be offset by cost reductions or by increasing the added value of the resulting wine. Given that grapevines form a perennial cropping system, the yield is determined considering multiple years (Guilpart et al., 2014), whereas in the current dataset, only four isolated years were available, weakening the conclusions extracted from annual practices. More data from consecutive years could help to study inter-annual interactions on wine yield.

5. Aligning yield with environmentally friendly farming practices in vineyards

Recent changes in the wine market have led Languedoc-Roussillon vineyards, which were not under an organic label, to obtain a new label called High Environmental Value (HEV) (Vitisphere, 2023). Both organic and HEV labels drive farming practices by reducing chemical inputs in vineyards and applying sustainable soil management by minimising bare soil during the rainy season.

In our study, we verified that important yield gaps were only produced when drastically reducing inputs, both for pesticides and fertilisers. We evidenced the feasibility of a substantial pesticide reduction without compromising yields for high and medium levels of pesticide use. Organic fertilisation appeared as an effective alternative to chemical fertilisers for maintaining yield levels. Winter cover cropping and early spring cover cropping are practices that can increase the provision of ecosystem services without compromising yields.

Overall, improvements on how winegrowers take into account pest pressure and water stress during grapevine development will help to reduce input use and provide essential ecosystem services in vineyards without involving yield gaps.

Conclusion

This study highlights the specific challenges and insights of grapevine yield gap analysis, distinguishing it from traditional approaches focused on annual crops. By using a large-scale survey of winegrower practices and outcomes, our findings reveal that while 36 % of winegrowers target yields below the wine label threshold, a significant portion of their plots (58 %) did not even achieve these targets, underscoring profitability concerns in the sector. We have identified key factors affecting yield gaps, such as climatic and pest events, very low pesticide use and lack of fertilisation. Notably, pesticide reduction in conventional plots did not consistently lead to significant yield gaps until a reasonable limit (TFI > 6), suggesting that there is room for reducing pesticide use without compromising yield. Future research should integrate farming annual practices with plant material, vineyard design, wine label and soil, and climate context to better understand the full scope of yield gap drivers and define the conditions for better control of grape yields. Additionally, the evolving framework of wine labelling, particularly discussions around environmental standards, in a context of climate change, represents a critical area for aligning yield performance with sustainability goals. Addressing yield gap challenges will be key for maintaining both the economic viability and the environmental stewardship of viticulture.

Acknowledgements

We thank all the Pays d’Oc PGI collaborators in this project, who provided funding, and shared data and expertise, in particular Laure Lacombe, Florence Barthès, Olivier Simonou, Jacques Gravegeal, Damien Onorré and Gérard Bancillon. We highly appreciate the support of the Wine and Vine chair of the University of Montpellier, INRAE and Institut Agro for facilitating our partnership.