The Lake Moo plain (44∘37′29′′ N, 9∘32′25′′ E), about 0.15 km2 wide, is located near the boundary between the Emilia–Romagna and Liguria regions (Piacenza province, Italy; Fig. 1a) in the upper valley of the Nure stream at an altitude of 1130 m above sea level (a.s.l.). A high tree cover density characterizes the present-day catchment area (i.e., the total woodland cover is 65.55 %; Corticelli et al., 2011), with a high richness of plant species and an exceptional concentration of protected mountain taxa of particular phytogeographical interest. The vegetation landscape shows the widespread occurrence of Fagus sylvatica, locally interrupted by grazing areas and blueberry moorlands with the presence of rare Juniperus nana and Sorbus chamaemespilus. Reforestation of Pinus nigra is also documented (Fig. 1b).

Present-day climate conditions primarily reflect the interaction of the prevalent atmospheric flow and the mountain range surrounding the Lake Moo plain. The proximity to the Ligurian Sea (about 40 km SW) makes the location particularly favorable to orographic precipitation enhancement in the case of moisture-laden southerly flow. The most precipitation is usually observed in spring and autumn, with a marked peak in early autumn. In this period we observe a particular synergy of midlatitude synoptic disturbances, becoming more frequent towards the cold seasons, and strong convective systems still developing over the warm Mediterranean Sea at the end of summer and in autumn (Grazzini et al., 2020); this is a situation very favorable for the genesis of extreme precipitation events in the area.

The study area mainly consists of extensively fractured serpentinites that represent the accreted fragments of the Ligure–Piemonte oceanic basin separating the European and Adria plates during the Middle–Upper Jurassic (Marroni et al., 2010). The ultramafic rocks are bordered by polygenic breccias made of blocks of limestones or marly limestones embedded within a fine-grained matrix (Mt. Ragola Complex, late Santonian–early Campanian; Elter et al., 1997).

The geomorphological landscape includes flat areas and steep slopes located at different altitudes. On the former, marshy environments commonly occur recording the last filling phases of small lacustrine basins, some of which still exist as Lake Moo and Lake Bino (Figs. 1–2). However, the origin of Lake Moo is still a matter of debate as some authors point to a glacial origin (Elter et al., 1997; Marchetti and Fraccia, 1988; Carton and Panizza, 1988), while others have interpreted the basin as the expression of Holocene deep-seated gravitational slope deformations (Geological, Seismic and Soil Service of the Emilia–Romagna Region, 2012) mainly controlled by changes in the incision rates of the Nure stream (Elter et al., 1997). The complexity of past and present morphological processes is likely enhanced by the superposition of lithological units with strong mechanical contrast, such as ophiolites and the underlying a predominantly clayey unit (Elter et al., 1997).

Taking into account the criteria indicated by Gilli et al. (2013) and Schillereff et al. (2014), Lake Moo shares several advantageous characteristics for the reconstruction of Holocene flood activity:

steep slopes (average inclination of 24∘) composed of deposits highly susceptible to erosion (i.e., polygenic and monogenic breccias with a pelitic matrix – the Mt. Ragola Complex; Elter et al., 1997; Fig. 2);

The field campaign led to the acquisition of two sedimentary cores (S1 and S2), 14 and 6 m long, respectively, and one trench (T1) to investigate the sedimentary succession capped by the high-density flood deposits formed during the recent HIP event (13–14 September 2015). The location of the cores and the trench benefited from a detailed geomorphological map (Fig. 2) and a high-resolution reflection seismic survey (Fig. SUP1 in the Supplement), both originally produced for this research. The latter provided useful information about the lake-basin-floor morphology and the thickness of the infilling succession.

The two cores, S1 and S2, were obtained using a continuous drilling system that guaranteed an undisturbed stratigraphy and a high recovery percentage (about 90 %). A trench (6 m long, 3 m wide, 2 m deep) was excavated between the two coring sites, allowing for a detailed analysis of the most recent deposits.

We focused our attention on the longest core (S1) that shows the complete record of the Lake Moo infilling succession, as it reaches the ophiolite bedrock. Therefore, all the laboratory analyses (grain size, radiocarbon and palynological analysis) were undertaken on core S1, while the other core (S2) and the trench were used to support the stratigraphic reconstruction. Elevation and geographic coordinates of reference core S1 were acquired using a Garmin eTrex 10 GPS receiver: 1120.2 m a.s.l. and 44∘37′25′′ N–9∘32′43′′ E, respectively.

Facies characterization of the Lake Moo sedimentary succession was mainly performed by integrating the observable macroscopic physical characteristics (i.e., grain size, sedimentary structures, Munsell chart color and types of bounding surfaces) with the grain size data available for the reference core S1. The interpretation in terms of depositional environments also benefited from the precise position of the core(s) with respect to the spill point of Lake Moo and the application of the facies tract concept. The latter strongly supported the interpretation of the coarse-grained intervals encountered throughout the studied succession. Specifically, a facies tract is defined as the association of genetic facies that can be observed within the flood deposits along with the downslope motion transformation (Lowe, 1982; Mutti, 1992; Mutti et al., 1996). Thus, the application of the facies tract concept, if framed into the stratigraphy of the core(s), may allow for the reconstruction of how flood deposits changed through time at a fixed location (Figs. 3, 4).

The abundance of wood remains and peaty layers within the sedimentary succession under examination supported the development of a robust chronological framework. A total of 12 samples were selected from the reference core S1 and radiocarbon dated at the CEDAD Laboratory of the University of Salento (Italy). The data are available in Table TS1 (in the Supplement), and the conventional 14C ages were converted into calendar years using the OxCal software version 3.10 (Reimer et al., 2013).

An age–depth model was also constructed using linear interpolation. Coarse-grained intervals were not excluded from the processing because of the difficulty to evaluate the exact thickness of the deposits interpretable as instantaneous events. Indeed, the upper limit of coarse-grained intervals commonly appears transitional toward finer sediments, and, in a few cases, some intervals for a total thickness of 107 cm were not recovered.

In order to refine facies characterization and highlight past vegetation changes at the study site, pollen analyses were carried out on 14 samples collected from the core S1. We focused the analyses on two key stratigraphic intervals formed during periods of well-known different paleoclimate (and paleotemperature) conditions: (i) Interval 1 (ca. 10.5–9 m of core depth), covering the period between ca. 9600 and 7300 cal yr BP, thus centered on the Holocene Thermal Maximum – HTM (Renssen et al., 2012), and (ii) Interval 2 (ca. 5.5–4.5 m of core depth), recording the final stages of HTM and the following cooling period between ca. 5500 and 3800 cal yr BP. In these two stratigraphic intervals, samples were collected from fine-grained layers following a mean sampling resolution of ca. 30 cm.

A standard methodology already tested for pollen substrates was applied with some minor modifications (Lowe et al., 1996). The method includes a series of laboratory treatments: about 8–10 g of sample was treated in 10 % Na pyrophosphate to deflocculate the sediment matrix; a Lycopodium tablet was added to calculate the pollen concentration (expressed as pollen grains per gram), the sediment residue was subsequently washed through a 7 µm sieve. The sample was resuspended in HCl 10 % to remove calcareous material and subjected to Erdtman acetolysis; a heavy liquid separation method was then introduced using Na metatungstate hydrate with a specific gravity of 2.0 and centrifugation at 2000 rpm for 20 min. Following this procedure, the retained fractions were treated with 40 % hydrogen fluoride for 24 h, and then the sediment residue was washed in distilled water and after in ethanol with glycerol; the final residue was desiccated and mounted on slides using glycerol jelly and finally sealed with paraffin.

Identification of the samples was performed at ×400 magnification, and only difficult pollen, such as Triticum, Avena or Hordeum types, was observed at ×1000 magnification.

Determination of pollen grains was based on the Palinoteca of the “Centro Agricoltura Ambiente – CAA G. Nicoli” laboratory (Italy), atlases and a vast amount of specific morpho-palynological literature. Names of the families, genus and species of plants conform to the classifications of Italian Flora proposed by Pignatti (2017–2019) and European Flora (Tutin et al., 1964–1993). The pollen terminology was based on Berglund and Ralska-Jasiewiczowa (1986), Faegri and Iversen (1989), and Moore et al. (1991) with slight modifications that tend to simplify plant nomenclature. The term “taxa” is used in a broad sense to indicate both the systematic categories and the pollen morphological types (Beug, 2004). For each sample, at least 500 pollen grains were counted, and the identified taxa have been expressed as percentages of the total pollen sum that includes only terrestrial pollen, no fern spores and no aquatic plants.

On the basis of vegetational and ecological characteristics, the following main pollen groups were identified: conifers (Pinus–Abies alba), deciduous trees (this group includes quercetum taxa – Quercus, Carpinus betulus, Corylus avellana, Fraxinus, Ostrya carpinifolia, Tilia and Ulmus + other deciduous trees), meadow (this group mainly comprises Fabaceae and Asteroideae, Caryophyllaceae, Cichorioideae and Poaceae), anthropogenic indicators (e.g., Cerealia, Chenopodium, Convolvulus arvensis, Plantago, Urtica) and “alia”, which includes all taxa excluded from previous groups. The group of hygro + aquatic plants was also distinguished. It includes hygrophilous herbs (e.g., Cyperaceae), helophytes and hydrophytes (i.e., Lemna, Junchus, Nymphaea, Phragmites, Potamogeton and Sparganium), which are considered a good proxy for humid conditions typical of wetlands.

The 14 samples prepared for pollen analysis were also investigated for the identification of microcharcoals. Microanthracological analysis has been used to track past changes in fire history, likely connected to anthropogenic activities (Vescovi et al., 2010). A point-count estimation of microscopic charcoal abundance was carried out, and charcoal fragments encountered during pollen counting were recorded in four size classes based on the long axis length: 10–50, 50–125, 125–250, >250 µm (Whitlock and Millspaugh, 1996; Clark, 1982; Patterson et al., 1987; Whitlock and Larsen, 2001; Fisinger et al., 2008). The former two classes were interpreted to be windblown transported, hence giving information about regional fire events, whereas the latter two were considered the result of local vegetation burning (Vittori Antisari et al., 2015).

Pollen and microcharcoal results are presented in Sect. 5 to support facies stratigraphy and to discuss controlling factors on flood activation.

Central to our study is the availability of a reliable (paleo)temperature dataset for the chronological period recorded by the Lake Moo sedimentary succession. In this respect, the Holocene paleoclimate reconstruction conducted by Samartin et al. (2017) on the nearby Lake Verdarolo represents a very important reference. The Lake Verdarolo site is located at 1390 m a.s.l., 270 m higher and 54 km SE from Lake Moo (Fig. 1), in a very similar climatic context. The authors reconstructed the mean July air temperature using a chironomid-based inference model developed through a combination of data extracted from more than 200 lakes in Norway and the Swiss Alps (Heiri et al., 2011). This vegetation-independent paleotemperature reconstruction is the first for the northern Apennines, and it agrees with that of Gemini Lake (1350 m a.s.l. elevation and 48 km SE from Lake Moo) as well as several other records coming from central Italy (Samartin et al., 2017).

Modern temperature and precipitation time series (1961–2018) at Lake Verdarolo and Lake Moo were derived from the gridded high-resolution dataset of Emilia–Romagna (Eraclito4), described in Antolini et al. (2016). Trend estimation and a Mann–Kendall significance trend test were computed with the pyMannKendall package (Hussain et al., 2019).

The age–depth model, based on 8 out of 12 14C dates available for the S1 core (Table TS1 in the Supplement), indicates that the sedimentary succession above the bedrock encompasses almost all of the Holocene period (about the last 10 kyr; Fig. 6). Three ages (at 9.60, 3.10 and 2.40 m of core depth) were excluded as they show evidence of contamination (i.e., ages younger than 1950 CE), while one sample collected at 2.35 m was not datable due to a very low organic matter content. The selected calibrated ages are stratigraphically coherent and mainly derived from wood fragments and peaty deposits. In order to reduce the potential bias due to long-lived plants and reworking processes, small twigs were selected (Oswald et al., 2005) from fine-grained deposits and identified when possible (e.g., Pinus and Abies alba). Due to the small dimension of the basin and the ophiolite prevalence in the drainage area (Fig. 2), we consider the hard-water effect on peats negligible.

The resulting age–depth model suggests the occurrence of four main stratigraphic intervals in terms of sedimentation rates. The lowermost interval (about 9.5–7.3 cal kyr BP) denotes a mean accumulation rate of ca. 0.50 mm yr-1 that increases up to ca. 2.38 mm yr-1 within the overlying section dated around 7.3–5.5 cal kyr BP. A change in the accumulation rate occurs at around 5 m with a drop to ca. 0.3 mm yr-1 between the 5.5 and 4.0 cal kyr BP interval. Between ca. 4.5 and 3 m, the deposit shows an uncertain chronology due to the lack of reliable radiocarbon dates and the occurrence of an unconformity surface at around 4.5 m, at the contact between Units 2 and 3 (Figs. 3, 6). During recent times (the last 146–14 cal yr BP), the accumulation rate has increased, reaching ca. 12 mm yr-1.

A positive link emerges correlating Holocene (July) paleotemperatures, precipitation intensity reconstructions and flood history (Fig. 7) at the study site, as the stratigraphic units showing the highest frequency of coarse-grained flood deposits (i.e., Units 2, 4 and 5) invariably fall into periods of higher temperatures (i.e., HTM and the post-LIA time). This strongly supports the hypothesis that greater warmth favors the occurrence of extreme precipitation events, probably more frequently at the end of the summer–autumn as already reported by other authors (e.g., Marcott et al., 2013; Giguet-Covex et al., 2012). Although the process attribution and the geographical uniformity of the HTM wet phase are still a matter of debate, the abundance of coarse-grained flood deposits within the Lake Moo record around 9.5–5.5 kyr BP suggests a precipitation increase also identified in other sites of the central and southern Mediterranean (Magny et al., 2012b).

However, local factors such as changes in vegetation cover (i.e., tree cover percentage) due to human disturbance and fires could have significantly contributed to enhancing slope erosion and then influenced the Lake Moo sedimentary record, partly weakening our reconstruction of the temperature–HIP–flood activity relationship at the Holocene scale. In order to overcome this issue, we explored and compared the pollen and microcharcoal content of two key intervals (I1 and I2 in Figs. 7, 10) that are characterized by a different flood record (high versus low flood activity) formed during consecutive periods of distinct climate conditions: the HTM and the following cooling. The difference in mean temperature between I1 and I2, obtained by averaging the corresponding chronological intervals in the Verdarolo paleotemperature curve, is +1.3 ∘C, while the maximum difference reaches +3.1 ∘C (Fig. 8).

For the post-LIA interval, in particular since 1950 CE, no significant contribution to debris mobilization can be attributed to vegetation landscape changes along the slopes as Lake Moo is a protected area characterized by dense forests (Corticelli et al., 2011; Sect. 2.1). Consistently, the most recent pollen sample (P14 yielding an age of 146–14 cal yr BP; Fig. 7) shows a tree cover of about 80 %.

The main considerations arising from the in-depth analysis of I1 and I2 are the following.

For I1 (ca. 9.5–7 kyr BP; 10.5–9 m of core depth), the woody component (i.e., tree cover percent) is invariably higher than the herbaceous one, and it shows a value of 80 % on average. The conifers, mainly represented by pines and silver fir, are dominant with respect to the deciduous trees (e.g., Corylus avellana and Quercus), reaching a maximum of 95 % within the P4 sample dated between 9.3 and 9.1 cal kyr BP. Interestingly, Abies alba shows the highest percentages (ca. 15 %–55 %) within the chronological interval that corresponds to the warmest period of the HTM (Figs. 7, 10; Samartin et al., 2017). Moreover, A. alba reaches the peak of ca. 55 % during the HTM wettest conditions (around 7.5–7.3 cal kyr BP; I1 humid period in Fig. 7), as reconstructed by several authors (Regatieri et al., 2014; Zanchetta et al., 2011; Zhornyak et al., 2011) from the Apuan Alps speleothems. This trend is consistent with the fact that A. alba is a warm-temperature tree with a preference for high moisture availability (Tinner et al., 2013), as clearly documented within the nearby Lake Greppo record (Vescovi et al., 2010, confirming higher rainfall accumulation already highlighted by lake-level dynamics from the Jura Mountains and central–northern Italy; Magny et al., 2004, 2009, 2012b). Although no specific inferences can be made on past water levels of Lake Moo, it is interesting to note that the low amount of hygrophytes and aquatics (Fig. 7) is fully consistent with highstand conditions. Anthropogenic indicator taxa (Fig. 7), as well as the anthropogenic index (AI; Fig. 10), show very low values throughout the interval, documenting a negligible human impact on the area in accordance with the extremely low content of microcharcoals (0.01–0.13 mm2 g-1; Fig. 10). Besides, the coarser microcharcoals (>125 µm) are not recorded, suggesting no significant local fires. A slightly higher amount of anthropogenic indicator taxa occurs within sample P6, dated between 8070 and 7880 ca yr BP; this is interpreted as an effect of categorization difficulties about spontaneous taxa such as Artemisia vulgaris type, Chenopodium, Plantago and Urtica.

Palynological data, combined with the stratigraphic interpretation of the S1 core, support our hypothesis that warmer and wetter conditions typical of the HTM led to a high amount of HIP and then high flood activity in the basin, independently of local factors. The presence of an extensive vegetation cover along the slopes implies the need for even more intense rainfall to trigger a debris mobilization that leads to the deposition of thick, coarse-grained layers containing pebbles with remarkable diameters (LM6, LM7 and LM8 in Fig. 10).

The paleoenvironmental–paleovegetation features characterizing the Interval 2 persists up-core, in correspondence to Unit 3 that records apparent flood inactivity (Figs. 3, 7). The uppermost sample P14 likely tracks a strong increment of A. alba.