Bosco Pisano is a patch of woodland located on the slopes of the Iblei Mountains, nowadays a Nature 2000 site (ZSC ITA090022 – Bosco Pisano). The landscape unit is a mosaic where open woodlands and pasturelands alternate. The vegetation is dominated by the presence of Olea europaea var. sylvestris-dominated formations (EUNIS habitat type G2.41), Quercus suber (EU Habitat Directive 9330) and Quercus pubescens (EUNIS G1.732) woods, pseudo-steppe with grasses and annuals of the Thero-Brachypodietea (EU Habitat Directive 6220), Sarcopoterium spinosum phryganas (EU Habitat Directive 5420), and Western Mediterranean and thermophilus scree (EU Habitat Directive 8130). Bosco Pisano lies on volcanic rock outcrops (Carbone et al., 1986) with prevailing very shallow andic brown soils and lithosols, where rock outcrop is abundant (Fierotti et al., 1988). The area chosen as case study, 780 m² large (Figure 2), includes an unusual almost monospecific wild olive cluster (Figure 3). It is a unique example of wild olive woodlands that might have been much more widespread in the past and currently represented by few fragmented patches occurring occasionally in the entire region.

As part of the study, an assessment was made of the tree morphology and grafting events since the olive trees in Bosco Pisano, even though wild olive, have the typical multistemmed shape of coppiced cultivated individuals (as for instance in Cozzo del Lampo, cf. 2.2). The olive wood was likely used for fuelwood, fodder-branches, and timber. Coppicing practices (the cut of tree branches and trunks at or near the ground level) are already mentioned in written sources from the 17^th century and there may also be charcoal kilns in the vicinity (Garfì and Di Pasquale, 1988). Historical reconstructions of the local land use (cf. Di Pasquale et al., 2004) dates from the 17^th century onwards. Until the last century, the local landscape spatial patterns probably resembled present-day grazing areas, such as the Dehesas or Montados (in Spain and Portugal respectively). During the 1940s and 1950s, this territory became virtually abandoned, except for cork harvesting and grazing. The area is today used for overwintering of animals, as part of a transhumance practice (cf. Garfì and Di Pasquale, 1988).

Di Pasquale et al. (2004) suggest that the wild olive dominated stands are probably the most ancient and stable units, when compared to other more recent forest communities, as the macchia formations (estimated as c. 60-year-old), resulting from the decline of coppicing, and the cork oak stands, most likely originated in its current settings at the end of the 19^th century, now largely senescent and unable to rejuvenate due to the lack of natural regeneration owed to overgrazing.

Garfì and Di Pasquale (1988) further suggest that this woodland is a relict of wider oak forests, existing until the classic age in Sicily (i.e., Diodorus Siculus, Bibliotheca, 4–84). This wood made of wild olive trees represents a now forgotten form of past land use of the olive, and for this reason is an important case study.

Cozzo del Lampo is a hill of approximately 1.28 km², located in a rural area of inner Sicily, and characterized by the abundance of large size (c. 12–15 m in circumference) domesticated century-old olive trees (Figure 4). These agroecosystems are Mediterranean formations of Olea europaea var. europaea (EUNIS G2.91), with presence of Mediterranean subnitrophilous grass communities (graminoid formations that may cover post-cultural or pasture lands, EUNIS E1.61), sub-Mediterranean deciduous thickets (EUNIS F3.2) and diss formations (EUNIS E1.433). Along the slopes of the hill, these agroecosystems are intermixed with Mediterranean tall-grasses (EUNIS E1.4), evergreen sclerophyllous scrubs (EUNIS E1.2A) and riparian thickets close to the water streams (EU Habitats Directive 92D0). The hill, as a mosaic of Olea europaea var. europaea groves in terms of land use, is then surrounded by cereal and fallow fields. Since the 1950s post-war land reorganization in Sicily, these olive groves have been managed as small family holdings maintained for self-consumption and/or recreation.

Cozzo del Lampo is an elevated area with slopes and/or protruding rocks of the Terravecchia Formation (i.e., interbedded quartzite pebbles, sandstone and shale sequences, cf. Grasso and Pedley, 1988). The size and configuration of the olive trees suggest that they are century-old, though no exact dating is available. The olive trees are found in places where it was extremely hard or impossible to plough the land manually.

Field surveys were carried out in collaboration with farmers and pruners (cf. Ferrara and Ingemark, 2023; Ferrara et al., 2024b). Trunk and spout morphology was assessed, together with crown shape and size, past grafting and pruning events if visible on the tree. Some trees have likely been grafted from the roots of previous wild olive individuals (sensuLa Mantia, 2005), since no signs of grafting is evident on the trunks. Several consecutive coppicing (tree cuts at or near the ground level) and pollarding (tree cuts higher at the trunk level) events are visible, resulting in the current multi-stemmed and multi-branches habit of these century-old olives. Both coppicing and pollarding as pruning techniques have the aim to foster self-renewal in the tree (Rackham, 2018). According to the size and shape criteria specified by Schicchi and Raimondo (2011), the olive trees in Cozzo del Lampo could potentially be as old as 800 or 1,000 years (Figure 5). Finally, their spatial arrangement is distinctive compared to more recent olive orchards, resembling a mosaic of diverse configurations unique to this area (Ferrara et al., 2024a).

The Morello valley, where the hill is located, has been inhabited since prehistory (Giannitrapani et al., 2014). Archaeological surveys in one olive orchard were carried out by Tegerdal Hune (2022), resulting in the location of an undated site. Giannitrapani et al. (2014) suggest an abandonment phase in the entire valley from late Antiquity to the 14^th century. This was followed by slow repopulation and, from the 18^th century, the consolidation of an extensive feudal estate, characterized by intensive cereal cultivation (Verga, 1993; Carocci, 2010). A redistribution of land to peasants took place after the end of World War II. Based on recent collection of oral histories and geonarratives (cf. Ferrara et al., 2024b), until c. 1950 fruit trees (such as Pyrus pyraster L. (Burgsd.), Punica granatum L., Prunus spinosa L. subsp. spinosa, to name a few) were cultivated in between these olive trees, resembling the intercropping system of the “Mediterranean garden”. Outmigration from the area remains a problem today (Fondazione Migrantes, 2022), though small-scale farming activities persist, mainly involving cereal and livestock production (Ferrara and Lindberg, 2023).

The third study area presented in this paper is located in a rural district called Malìa, in the Madonie Mountains, known as a biodiversity hotspot due to remnants of semi-natural vegetation (Baiamonte et al., 2015). The Olea europaea var. europaea formations (EUNIS habitat type G2.91) in the Malìa study area are found in permanent mesotrophic pastures and aftermath-grazed meadows (EUNIS E2.1) dominated by Mediterranean subnitrophilous grass communities (EUNIS E1.61), Calicotome infesta (EUNIS F5.515), Ampelodesmos mauritanica dominant formations (EUNIS F5.53), Medio-European rich-soil thickets (EUNIS F3.11) and Spartium junceum scrubs (EUNIS S53). The study area surveyed has an extension of approximately 550 m² (Figure 6), and it is embedded in a tangled web of Quercus pubescens (EUNIS G1.732) Quercus suber (EU Habitat Directive 9330) woods and Italic poplar galleries (EUNIS G1.314). Here too, during fieldwork, we assessed the morphology and grafting events visible on the century-old olive trees. Evidence of past grafting, done on semi-mature wild and feral trunks at more than 1.5 meters from the ground (Figure 7), suggests an adaptation of protect cultivated olives to grazing animals. Based on their trunk size and overall physiognomy, these olive trees are estimated to be century-old.

Numidian Flysch, a rare Oligocene to mid-Miocene deep-marine sandstone and mudstone formation (Hubert Thomas, 2011; Pinter et al., 2016), emerges until the surface in certain places, making the topography very steep and soil layers shallow. In this agroecosystems, the spatial distribution of century-old olive trees appears random, growing directly from and among the bedrock, even in the steepest parts of the hill slope. In addition, numerous olive trees are located in the most remote and inaccessible places (Figure 8).

Local reconstructions of Holocene vegetation and fire dynamics show the expansion of open grasslands from the 5^th millennium BCE onwards, on top of previously closed woodlands (Tinner et al., 2016). This grassland expansion, alongside other evidence, has been linked with human activities (Belvedere and Forgia, 2010; Forgia et al., 2013, 2021). Unfortunately, documentation for the more recent period remains scarce.

From interviews with local residents, the area had a similar socio-economic history as Cozzo del Lampo. It was part of a large estate, which after the end of World War II was parceled into small plots, bought by the farmers already working the land. Malìa represents a past multifunctional space, for cultivation and grazing, while also providing fuelwood and timber (cf. Ferrara et al., 2023), which suggests this case study to be representative of an old agroforestry model (similar to those in Antiquity), characterized by the high presence of grazing activity and sparsely planted olive trees.

Phytoliths are used in this paper to reconstruct past vegetation, land use and broader environmental dynamics (Pearsall and Trimble, 1984; Strömberg et al., 2007; Aleman et al., 2014; Feng et al., 2017). Phytoliths have also been used to study erosion patterns and paleoclimate (see review in Qader et al., 2023). In old soils, phytoliths are preserved due to silica resistance to decay (Pearsall, 2015). Although phytoliths have an enormous potential in studying and assessing agricultural soils, such field of research is underdeveloped (though see Meunier et al., 1999; Fernández Honaine et al., 2006; Liang et al., 2015; Hussain et al., 2023; Pokrovsky et al., 2024; Benvenuto et al., 2025 for positive examples). Plants and soils act as a Si ‘filter’ (cf. Vandevenne et al., 2012, 243), transforming the dissolved silica (DSi) originated from mineral weathering into biogenic silica (BSi). Such biogenic silica is then returned to the soil when plants decay. Along this cycle, vegetation and soils changes caused by specific land use activities, such as the removal of biomass during harvest, affect the Si quantity and distribution (Barão et al., 2020). Through phytoliths analysis, we can thus trace how vegetation composition and assemblages have been shaped by land use practices. By approaching the historical olive agroecosystems present in the study areas as containing legacies of their past-present agricultural and silvicultural use, we developed a methodological approach that would allow investigating this biocultural heritage. By ideally following the roots of century-old olive trees underground, we explored the memory of the soil through the analysis of phytolith assemblages (Piperno, 2006). By comparing the relative frequencies through time of all diagnostic phytoliths (Strömberg, 2004), first of all we intended to observe if distinctive assemblages along the stratigraphic profiles could lead to distinguish patterns of change in vegetation and habitats structure (Strömberg et al., 2013). Secondly, we considered if the correlation between different vegetation communities represented by these assemblages could offer new interpretative perspectives about past land use practices (included successional trajectories along gradients of human disturbance, cf. Witteveen et al., 2023), climatic and broader environmental trends over the long term in these agroecosystems.

Therefore, testpits were opened in each of the case study areas, for the extraction of phytoliths from soil layers following a protocol developed by Mazuy et al. (2024) specifically for these types of agro-ecological contexts, as detailed in Section 3.2. Where available, ¹⁴C dates were modeled to provide a relative age-depth.

The identification of phytolith morphotypes (Figure 9 for examples) to categorize into assemblages has followed local phytoliths ecology (Section 3.3) and phytolith assemblages ratios have been interpreted according to four different indices (Section 3.4), in order to provide indication of specific plant communities structures, composition and variation patterns along each profile and when comparing the three different sites.

Soil sampling for phytoliths extraction and analysis was adapted to the specific geological features in each study area and the spatial composition of their century-old olive trees.

In Bosco Pisano, two profiles of 20 cm depth were opened (Figure 10a, Table 1), resulting in two samples. Sampling here aimed to estimate tree cover density through phytoliths signature in Olea wood formations (cf. 2.1 for habitats characterization), so to employ it as a local tree cover threshold when interpreting the results from the other two case studies.

In the two study areas Cozzo del Lampo and Malìa, five testpits 1.5 m deep were opened in each area, and soil sampled every 10 cm, resulting in a total of 75 samples per study area (Table 1). In Cozzo del Lampo, the testpits followed the slope transect with an interval of 40 meters (Figure 10b). In Malìa, a transect was not possible due to the shallow soil depth in recurrent places. The testpits were placed near olive trees in an area with relative soil thickness and were local residents said there had been cereal and/or legumes cultivation and pasture (Figure 10c, Table 1).

Stratigraphic analyses were carried out for the profiles excavated in Cozzo del Lampo and Malìa (Supplementary Material S1). Three charcoal samples from Cozzo del Lampo and two samples from Malìa were submitted for radiocarbon dating at the Ångström Laboratory, Uppsala University. Dates were calibrated by the laboratory using IOSACal v0.4.1 and the IntCal20 calibration curve (Reimer et al., 2020). A Bayesian age-depth model was produced in R (R Core Team, 2021) using the package Bacon (Blaauw et al., 2022) and the uncalibrated dates for Cozzo del Lampo samples are published already in Ferrara and Wästfelt (2025).

Phytolith extraction was done at the Paleobiology Laboratory (Department of Earth Sciences, Uppsala University), following the protocol developed in Mazuy et al. (2024). This protocol allows extracting quantities of biogenic silica suitable for phytolith morphotypes analysis, from soils and contexts with low phytoliths concentrations. The samples have been first deflocculated through magnetic stirring. Thereafter, the fine fraction (200 μm) is treated with hydrochloric acid 33% (to remove carbonates) and then boiled in a hot bath (80°C) with potassium hydroxide 10% for 15 minutes (to remove organic matter). Phytoliths are then extracted by heavy liquid flotation (using sodium polytungstate, SPT at 2.35 density) and in a hot bath (80°C) with hydrogen peroxide 30% for 1 hour to remove further organic matter.

Being collected on the island of Sicily, where the volcano Etna has been active for thousands of years, our soil samples contain a high amount of cryptotephra, which is silica glass (GSi). The density of cryptotephra is similar to that of phytoliths and biogenic silica more in general (BSi) (≤ 2.3 g cm-3), consequently silica glass is also separated together with phytoliths during the heavy liquid extraction process. In our specific case, the final BSi residue represents thus the sum of phytoliths and cryptotephra. Consequently, whatever quantitative assessment of the silica fraction extracted from our samples would be misleading in terms of specifying the exact amount of phytoliths content. A further separation between tephra particles and phytoliths would have been necessary to obtain such specific measures, which however could have caused partial loss of the already phytolith-low content of the studied samples (cf. Mazuy et al., 2024). Since cryptotephra is clearly distinguishable from phytoliths under microscope and being mainly interested in analyzing relative abundance of phytolith morphotypes and their assemblages variability to gain qualitative information about changes in vegetation composition, land use and plant behavior, the quantitative determination of the biogenic silica content in our samples was out of the scope of the work presented in this paper.

Though different parts of a plant produce diverse phytolith morphotypes, several phytoliths can be commonly identified to subfamily level (Table 2). Phytoliths attributed to dicotyledonous (plants having two embryonic leaves in their seeds, such as trees and shrubs) from temperate areas, as our case studies, are globular (or spheroid) (Bozarth, 1992; Alexandre et al., 1997; Albert et al., 1999; Runge, 1999; Delhon et al., 2003) and tracheary (Bozarth, 1992; Alexandre et al., 1997; Madella et al., 1998; Albert et al., 1999; Runge, 1999; Delhon et al., 2003; Jarl and Bruch, 2023). They can also be jigsaw (Carnelli et al., 2004; Kawano et al., 2006; Gu et al., 2008; An and Lu, 2015; Potì et al., 2019; Hayashi et al., 2021) and blocky (Tsartsidou et al., 2015; An, 2016; Boixadera et al., 2016; Ntinou and Tsartsidou, 2017; Burguet-Coca et al., 2020; Kraushaar et al., 2021; Tencariu et al., 2022). Even relatively modest percentages of these morphotypes can be interpreted as substantial, as dicots have a low phytolith production (Carnelli et al., 2004; Tsartsidou et al., 2007).

Grasses (monocotyledon plants, i.e. having only one embryonic leaf in the seeds) of the Poaceae family produce elongate, papillate, acute bulbosus and bulliform flabellate forms (Twiss et al., 1969; Fredlund and Tieszen, 1994; Piperno, 1988; Ball et al., 2001; Neumann et al., 2019). Elongate phytoliths present morphological variations in vegetative parts (the elongate sinuate type is formed in leaves, and simple elongate in stems). Elongate dendritic and papillate forms are produced by the flowering parts of Poaceae (glumes, lemma and palea, also named inflorescence). Dendritic long cells are commonly associated with the chaff of domesticated Pooideae, such as wheat, barley and oat (Rosen, 1992; Ball et al., 1999; Portillo et al., 2006; Albert et al., 2008). However, dendritic forms are also found in wild grasses (Novello and Barboni, 2015), thus their attribution must be carefully assessed.

Acute bulbosus is a morphotype produced in the interior part of the hair cells, above all in grasses (Alexandre et al., 1997; Barboni et al., 2007). Being produced in sedges and dicots as well, this morphotype is not considered here diagnostic. Nonetheless, it has been included in the counted morphotypes due to the limited amount of phytoliths in certain layers correlated with the stratigraphy (cf. Supplementary Material S1).

Bulliform flabellate and blocky morphotypes are produced in the bulliform cells of leaves (Mader et al., 2020), which usually silicify in a later stage of plant life (Moulia, 1994). Bulliform flabellate phytoliths allow leaves to bend to avoid excessive water loss, and their formation is directly influenced by environmental conditions such as high evapotranspiration (Bremond et al., 2005b; Novello et al., 2012).

Specific grass silica short cell phytoliths (hereafter GSSCPs, Neumann et al., 2019) can be diagnostic of the following Poaceae subfamilies (Table 2). C3 grasses in cool moist/temperate climate (Pooideae) are associated with rondels (Twiss et al., 1969; Fredlund and Tieszen, 1994; Piperno and Pearsall, 1998; Barboni and Bremond, 2009). Rondels are produced also in the Arundinoideae (cf. Tencariu et al., 2022). The trapezoid (Barboni and Bremond, 2009) and Crenate morphotypes (Twiss et al., 1969; Fredlund and Tieszen, 1994; Barboni et al., 2007) are mainly produced in Festucoid grasses (Tsartsidou et al., 2007).

Chloridoideae are C4 grasses found in dry and hot environments. These are dominated by saddle forms (Piperno, 2006; Madella et al., 2016) and rondels found together with saddles (cf. Bamford et al., 2006; Barboni and Bremond, 2009). Warm and wet environments tend to support C4 Panicoideae where species have many morphotypes as bilobate (Twiss et al., 1969; Fredlund and Tieszen, 1994; Barboni and Bremond, 2009), polylobate (Twiss et al., 1969; Fredlund and Tieszen, 1994; Neumann et al., 2019) and cross morphotypes. Bilobate phytoliths can also occur in a few Festucoid grasses and some Chloridoid grasses (as pointed out by Metcalfe, 1960; Danu et al., 2020).

Phytoliths were identified and counted using light microscopy at ×400 magnification. Their morphotypes were categorized into plant taxonomic groups according to the International Code for Phytolith Nomenclature (ICPN) (Neumann et al., 2019) and the PhytCore online database (Albert et al., 2016). In addition, the UMR 7264 CEPAM CNRS-Université Côte d’Azur (Nice, France) was consulted for identifications. Only those recognized morphotypes listed in the International Code for Phytolith Nomenclature (ICPN) (Neumann et al., 2019) were counted, with a minimum of 200 phytoliths per sample. Such an amount is attested to give statistically interpretable assemblages (cf. Strömberg, 2009; Zurro Hernández, 2018), above all in soil samples coming from specific contexts (e.g., farming systems, arboriculture) where the concentration of phytoliths can be lower if compared with their higher concentration in soils of very specific environments, such as grasslands and archaeological sites (Cabanes et al., 2011).

The relative abundance of each phytolith morphotypes per soil layer was calculated based on the total sum. Along each excavated profile, phytolith assemblages were analyzed using four indices, to assess if the diachronic variability in their ratios could be informative of changes in vegetation structure and compositions, as well as land use practices and plant behavior to climatic disturbances.

The long/short cell index measures the ratio of long cells vs short cells. Short cells phytoliths are impregnated with silica as soon as they form (Sangster, 1970), while silicification of long cells become more intense with age (Rencheng et al., 2017). Thus, the ratio of long cells to short cells in a phytolith assemblage may provide information about the composition of grasses in terms of mature versus young grasses (Delhon et al., 2024). The formula used to calculate the long/short cell index here is Long/short cell index = long cells (elongate entire + sinuate + dentate + dendritic)/short cells (GSSCPs).

The Fs-index (Bremond et al., 2005b, 2008) can provide information on local microhabitats and/or climatic conditions, since an abundance of bulliform flabellate phytoliths is correlated with an increase in evapotranspiration and/or prolonged water stress (cf. Bremond et al., 2005b; Mader et al., 2020).

The formula used to calculate the Fs-index here is Fs-index = bulliform/GSSCPs (Bremond et al., 2005b, 2008).

The Dicot/Poaceae index (D/P index) measures the ratio of dicotyledons morphotypes (D) versus Poaceae morphotypes (P). This index allows to estimate the tree cover density (Alexandre et al., 1997), since based on the assumption that open habitats have a higher proportion of grass phytoliths (Strömberg et al., 2018). The lower the D/P value, the more open the habitat. In Mediterranean environments, 0.1 is the lower limit in the D/P index to distinguish between dicotyledon- and grass-dominated assemblages (Delhon et al., 2003).

Even though the D/P index has a standard formula (D/P) (cf. Alexandre et al., 1997), the variables indicating Poaceae and Dicots have been interpreted by scholars in slightly different ways (for a detailed review cf. Ferrara, 2024). Building on the standard D/P index (called here D_a/P index, considering as diagnostic of dicots only spheroid and tracheary morphotypes), Ferrara (2024) has developed also a local D/P index (D_b/P index), which includes also the blocky (blocky polyhedral and cubic) and jigsaw morphotypes as diagnostic of dicots from local samples.

The formulas used to calculate the D/P index are the following:

The inflorescence/culm-leaves index measures the ratio of Elongate dendritic phytoliths (from inflorescence bracts) versus Elongate entire and Elongate sinuate (from culms and leaves) (Piperno, 1988; Tsartsidou et al., 2007; Delhon et al., 2020). The ratio of the index varies according to the presence of the whole plant or residues from straws and spikelets, and this information can be indicative of local agricultural practices (e.g., harvest) and/or cereal processing (Delhon et al., 2020).

The formula used is Inflorescence/culm-leaves index = elongate dendritic + dentate/elongate entire + sinuate.

Results are presented in diagrams drawn using the software TILIA 3.03 (Grimm, 1993) (Section 4).

Two testpits (20 cm deep) were opened in two locations with a different density of tree cover (Figure 10a): Profile 1 is characterized by a dense tree cover; Profile 2 is located in a more open space but still inside the olive wood. Phytoliths are abundant in both the samples from Bosco Pisano, which is somewhat surprising considering its vegetation (woodland), but less surprising if we take into consideration the low historical anthropogenic impact on the soil of this area.

Pooideae morphotypes dominated with the high percentage of short cells phytoliths (on average 34% vs long cells 46%). The most common morphotype are rondel and trapezoid. These phytoliths were formed mainly in the leaves and the stems of grasses. Phytoliths from grass inflorescence (dendritic) are not present in the phytolith records. Phytoliths from woody and herbaceous dicotyledons are present in the two samples representing, on average, 9% (including also the blocky morphotype, cf. Section 3.3 for D_b/P index specifications), nonetheless, phytoliths from grasses are still dominant. Morphotypes related to other monocotyledons are low, with C4 grass phytoliths at only 1.25%.

An analysis of phytolith morphotypes assemblages was done using the D_a/P and D_b/P indices and the results are indicative. Overall, both indices result above the standard threshold of 0.1 for woodland estimation, except for D_a/P of Profile 2 (less dense tree cover), which is 0.08. Furthermore, Profile 1 (with dense tree cover) has higher values of both indices than Profile 2 (D_a/P = 0.13 vs 0.08; D_b/P = 0.34 vs 0.17). Results from Bosco Pisano, despite lack of soil depth, allowed for a calibration of the D/P index in relation to the other case study areas.

Phytolith morphotypes are abundant in most of the samples throughout the five profiles from Cozzo del Lampo. The only layers where it was not possible to extract phytoliths were samples taken in eroded bedrock (Profile 2, 56–150 cm; Profile 4; Profile 5, 92–150 cm) (cf. Strömberg et al., 2018). We begin here giving an overview of the results from all profiles analyzed and then focus on specific profiles.

In Malìa, the only ¹⁴C age obtained from collected charcoal is very recent, consequently it was not possible to date any of the profiles. Even with the absence of a tentative chronology, the phytolith record in Malìa still provides diachronic information.

Phytoliths are abundant in most samples, apart from those layers with eroded bedrock (Profile 1, 60–150 cm; Profile 2; Profile 3, 80 cm and 100–150 cm; Profile 4, 50 cm and 80–150 cm; Profile 5, 150 cm). Detailed stratigraphic descriptions of each single profile are provided in Supplementary Material S1.

Similar to Cozzo del Lampo, all the phytolith records in Malìa are characterized by a morphological predominance of grasses from the Pooideae subfamily (C3), given by the high percentage of short cells phytoliths (31.4%) vs long cells (46.76%). The most common morphotypes recognized are rondel and trapezoid, formed mainly in the leaves and the stems. Phytoliths from grass inflorescence (dendritic) are present in low percentages, values which are however higher than in Cozzo del Lampo (3% vs 0.6%). Morphotypes related to other monocotyledons are scarce (0.45%). Phytoliths from woody and herbaceous dicotyledons are omnipresent in small amounts (5.83%). Wood and bark types (spheroids) are more abundant (5.71%) than from leaves (tracheary, 0.11%).

The assemblage analysis of the phytolith record from Malìa indicates specific environmental and land use history dynamics, which can be correlated with alternate periods of land abandonment. Profiles 4 and 5, described in details below, represent the most indicative cases of these dynamics.