Plant phenology is defined as the timing of recurring life events (Lieth 1974), and involves reproductive phenomena such as flowering, fruiting, or seed germination, as well as vegetative processes like leaf flushing and shedding (Morellato et al. 2010; Wolkovich and Cleland 2011). Phenology is determined by a balance between intrinsic (e.g. constrained by phylogeny) and extrinsic factors (e.g. temperature, photoperiod; Bisigato et al. 2013). Biotic pressures, such as seasonal presence of predators, pollinators, and seed dispersers, as well as the presence of sympatric species sharing pollinators or predators may also shape phenological patterns (Rathcke 1983; Fenner 1998). Notably, plant-pollinators interactions may operate as a habitat filter shaping the phenological composition of the community (e.g. the absence of a particular pollinator prevents the establishment of a plant species in a community; Sargent and Ackerly 2008). Flowering time is particularly sensitive to environmental factors such as temperature, moisture, and photoperiod (Rathcke and Lacey 1985; Price and Waser 1998; Sherry et al. 2007; Bisigato et al. 2013). Phenology has thereby received increased attention over the last decades since it has been considered as the most sensitive and easily observable indicator of terrestrial ecosystem response to current climate change (i.e. advancement of spring events; Peñuelas and Filella 2001; Lavorel and Garnier 2002; Parmesan and Yohe 2003; Root et al. 2003; Parmesan 2007; Ding et al. 2013; Guo et al. 2015; Piao et al. 2019).

Shifts in the phenology of interacting species may induce shifts in their synchrony, with cascading consequences for communities, ecosystem dynamics, and ecosystem services (Yang and Rudolf 2010; Morellato et al. 2016; Kharouba et al. 2018). For example, phenological mismatch between the plant species Corydalis ambigua Cham. and Schlecht (Papaveraceae) and its pollinator led to a reduction of seed production owing to low pollination service (Kudo and Ida 2013). Other ecosystem services, such as carbon storage can be affected by phenological modifications. For instance, Keenan et al. (2014) showed that carbon uptake by forests might improve due to phenological shifts caused by climate change.

Variations in plant phenology can result from differences in soil structure and chemical composition including nutrient concentrations (Rossiter 1978; Wielgolaski 2001; Petraglia et al. 2014). Such modifications of soil characteristics can arise from agricultural activities. Agriculture is a major component of global changes (Vitousek 1994) and has important impacts on soils and ecosystems that outlast the duration of agricultural activity (Austrheim and Olsson 1999; McLauchlan 2006), notably by altering abiotic (e.g. nutrient content, pH, soil structure, water content) and biotic (e.g. species composition and structure, seed bank) conditions (Hobbs et al. 1988; Römermann et al. 2005). Phosphorus (P), one of the main components of most fertilizers, is a major determinant of plant growth and productivity (Aerts and Chapin 1999; Güsewell 2004). Phosphorus concentration can influence the phenology of annuals, which commonly display delayed flowering and maturity in response to low P (Rossiter 1978; Ma et al. 2002; Nord and Lynch 2008). Soil clay and silt contents, often shifted by agricultural activities, can also affect plant phenology by altering water retention (Wielgolaski 2001; Al Majou et al. 2008). Wielgolaski (2001) found that high clay content may favour high soil moisture, resulting in earlier bud break in Betula pubescens Ehrh. In contrast, Mediterranean annual species may shorten their life cycle due to water stress, with low soil moisture resulting in earlier flowering transition (Aronson et al. 1992). However, whether former agricultural activities may shift phenological patterns of re-established plant communities remains unknown.

This study focuses on the impact of former agricultural activities on the phenology of xeric grassland plant communities located in southeastern France, in the Plain of La Crau. La Crau is a protected socio-ecosystem hosting endemic species (e.g. the hedgehog grasshopper Prionotropis rhodanica Uvarov; Foucart and Lecocq 1998) and unique species-rich plant assemblage (Devaux et al. 1983). The intact steppe vegetation is dominated by perennial species Brachypodium retusum (Pers.) P.Beauv. and Thymus vulgaris L. which represent approximately 50% of the biomass (Buisson and Dutoit 2006), and hosts a large diversity of annuals (Buisson et al. 2006). The durability of the ecosystem is closely linked with itinerant sheep grazing, which has been the traditional land-use for several centuries (Buisson and Dutoit 2006). We aim to evaluate the phenological patterns of plant communities that grew on different soil type by comparing (1) intact native grasslands, hereafter named steppe, and (2) abandoned fields, hereafter named formerly-cultivated plots. Römermann et al. (2005) reported important changes in the composition of plant communities but also in the physical and chemical soil characteristics of the formerly-cultivated plots; thereby a shift in phenological patterns is expected. At the community level, reproductive strategy (i.e. annuals and perennials) may influence the phenological pattern of the community (Burgheimer et al. 2006; Meng et al. 2016). Annual plants reproduce once and die (i.e. all resources are used for reproduction with none saved for the following season), while perennial plants reproduce repeatedly and cycle through vegetative and reproductive phases. Annual plant phenology may respond faster to environmental variations than that of perennials (i.e. annuals tend to flower earlier in response to climate change; Fitter and Fitter 2002; König et al. 2018), due to their faster generation times and evolution rates (Laroche and Bousqet 1999). The two groups are therefore analyzed separately. Also, because (1) phenological response to environmental variations can be species-specific (Bisigato et al. 2013) and (2) differences in phenology between formerly-cultivated and steppe communities can result from variations in soil conditions but also variations in species composition, we considered phenological patterns of Bromus rubens L. and Lobularia maritima (L.) Desv., respectively the annual and perennial species recorded in both plot types with the highest occurrence.

Our study suggests that past cultivation practices can substantially affect the composition of Mediterranean xeric grassland communities with differences visible after 30 years of abandonment, but alter phenological patterns only slightly. Species composition is clearly distinct between steppe and formerly-cultivated communities (Fig. 2, Supplementary file 2), which is consistent with studies carried out in the same area (Römermann et al. 2005; Helm et al. 2019). Steppe communities are dominated by the perennial steppe grass B. retusum, while this species is absent from formerly-cultivated plots communities, where meso- to oligotrophic grassland species, such as A. geniculata, B. rubens, and Rostraria cristata (L.) Tzvelev are overabundant (Fig. 2, Supplementary file 2; see Helm et al. 2019 for community composition analysis). The failure of B. retusum to re-establish in formerly-cultivated plots is attributed to (1) low seed production (Vidaller et al. 2019a; Buisson et al. 2021), (2) dispersal limitation due to a mainly clonal reproduction in this ecosystem (Dureau and Bonnefon 1998), (3) competition with formerly-cultivated plot species (Coiffait-Gombault et al. 2012; Buisson et al. 2015), (4) the absence of safe sites for rhizomes because stones were removed for cultivation (Caturla et al. 2000; Buisson et al. 2015), and/or (5) grazing pressure in the seedling establishment stage (Vidaller et al. 2019a). Altogether, our results on species composition reflect expected dynamics after soil disturbance in this system.

Despite differences in species composition, our results highlight a convergence in phenological patterns of annuals, perennials, B. rubens, and L. maritima between plot types (Figs 3–6). Some differences can however be noticed. The reproductive phenology of annuals and B. rubens appears slightly in advance in the steppe. The peak of annuals carrying flower buds occurs two weeks earlier (Fig. 3C) and the largest increase of fruiting annuals occurs one week earlier (week 4 to 5 vs week 5 to 6; Fig. 3E) in steppe than in formerly-cultivated plots. Our results also suggest that flower bud formation of the annual grass B. rubens started earlier (i.e. 92% vs 44% of individuals with flower buds at week 1; Fig. 5C, Supplementary file 1) and that the fruiting peak occurred one week earlier in steppe plots than in formerly-cultivated plots (Fig. 5D). Bromus rubens phenology however displays a high variability within plot types (Fig. 5), implying that one isolated difference should be interpreted with caution. Repeating the experiment with a larger sample size might be necessary to confirm these results. The phenological advance of annual species in the steppe is unexpected, since the higher clay content, which is likely to preserve soil humidity and lower the P recorded in the steppe soil (Römmermann et al. 2005; Helm et al. 2019), should have delayed reproductive phenology of annuals compared to formerly-cultivated plots (Aronson et al. 1992; Wielgolaski et al. 2001; Nord and Lynch 2008). Indeed, low soil phosphorus decreases root-length but phosphorus accumulation is compensated by a phenological delay (Nord and Lynch 2008). One explanation might be the stone cover, higher in steppe plots than in formerly-cultivated plots from which they were removed for cultivation (Buisson and Dutoit 2004): stones can increase soil temperature and store heat during the day which is released at night (Devaux et al. 1983), thus leading to some phenological advance. At the community level, the occasional differences in the phenology of annuals may also be caused by differences in species composition between plot types (i.e. species-specific differences).

Phenological patterns of perennials appear overall more similar between plot types than annuals (Fig. 4). A high percentage of perennials (Fig. 4A) and of the perennial herb L. maritima (Fig. 6A) remained in vegetative state during the 8 weeks of the study. The peak of perennials carrying flowering buds occurring on week 3 is not followed by an increase in flowering individuals the subsequent weeks, which may be explained by flower bud consumption by sheep (Fig. 4B). Grazing may also blur the phenological patterns of L. maritima (i.e. consumption of flower buds and flowers resulting in very low proportions of flowering/fruiting individuals), especially in steppe plots where no flowering peak occurred (Fig. 6B, D). The highest grazing pressure on annuals and L. maritima detected on the steppe plots may be due to the high percent cover of B. retusum, which has a low palatability, and thus more grazing on all other species (Meuret et al. 2013).

The distinct phenological patterns between annuals and perennials may be explained by differences in reproductive/survival strategies. In the Mediterranean Basin, summer is the most stressful season for plants with mild to severe drought and high temperatures (Lionello et al. 2006). Most species reach their flowering peak in spring, when temperatures are already warm, but some species also flower in autumn or throughout the year (Picó and Renata 2001), like L. maritima (Bourrelly et al. 1983). Annuals flowering in spring must complete their reproductive cycle before periods of high water deficits, and have a short seed maturation period resulting in early dispersal dates (Segrestin et al. 2018). In this sense, we found that most annual species are in a reproductive stage at mid-June (< 4% of annuals in vegetative state at week 8; Fig. 3A). Many perennial species, in contrast, remain in vegetative state over the 8-week study (i.e. min. 57% in steppe plots and 43% in formerly-cultivated plots; Fig. 4A). The high proportion of perennials remaining in vegetative state in spring may be due to perennial species (1) that do not flower every year because of unfavourable environmental conditions, (2) being in a juvenile phase and unable to flower yet, and/or (3) flowering later in the season. For instance, B. retusum, the dominant perennial in steppe plots, rarely blooms the first year (Vidaller et al. 2018) and may flower later in the season when subjected to grazing (Vidaller et al. 2019b). The perennial herb L. maritima, representing respectively 18% and 29% of perennials in steppe and formerly-cultivated plots, generally flowers for ten months (from September to late June) with the peak of the flowering period in autumn (Bourrelly et al. 1983; Bosch et al. 1997; Picó and Renata 2000). Accordingly, we found that L. maritima has a high percentage of individuals in vegetative state during our 8-week spring study (> 50% except at week 4; Fig. 6A).

Overall, phenological patterns of annuals and perennials as well as the annuals/perennials ratio did not differ much between plot types, implying a convergence of annuals/perennials ratio and phenology in communities re-established after disturbance. The convergence of communities at the phenological and reproductive strategy levels suggests that deterministic assembly rules drive community assembly at these levels (i.e. environmental conditions determine the types of available niches and therefore the species that can fill them; Fukami et al. 2005; Helsen et al. 2012). The convergence in high percentage of annuals abundance (close to 70% in both plot types) most likely results from stressful environmental conditions, such as harsh Mediterranean climate (i.e. dry and windy climate with high inter-annual variability tend to favour annuals; Madon and Médail 1997) and recurrent grazing (i.e. grazing favours annuals; Díaz et al. 2007). In addition, in xeric environments, phenological patterns are intimately related to temperature (Rathcke and Lacey 1985), water availability (Abd El-Ghani 1997) but also grazing (Tadey 2020), which could explain the convergence in phenological patterns in our study system.

Our results highlight that past cultivation practices can affect the composition of Mediterranean xeric grassland communities with differences visible after 30 years of abandonment, but alters only slightly phenological patterns. We suggest that the harsh Mediterranean climate and sheep grazing most likely acted as strong habitat filters constraining community assembly at the phenological level. We do not exclude that other biotic interactions (e.g. plant-pollinators interaction, inter-specific plant competition for pollinators) may have influenced the phenological composition of the communities established after the disturbance. Quantifying their importance would require further investigation.

The data that support the findings of this study are available from the corresponding author, M.C.M. Hess, upon reasonable request.