Trunk river and tributary interactions recorded in the Pleistocene–Holocene stratigraphy of the Po Plain (northern Italy)

Tributary rivers can contribute significantly to alluvial‐plain construction by supplying large volumes of clastic material. Their relation to the main axial river strongly influences sediment deposition and preservation. The Po Plain is fed by the Po River and a dense network of transverse tributaries draining the nearby Alpine and Apennine chains. Stratigraphic, sedimentological, petrographic and geochemical analyses on 38 cores permitted detailed differentiation of Po and Apennine sedimentary units. Po River deposits are vertically stacked channel‐belt sand bodies with high contents of quartz–feldspar and metamorphic rock fragments, combined with high chromium levels. These sand bodies, 20 to 30 km wide, are replaced southward by finer‐grained deposits that represent the distal Apennine tributary‐rivers system. Apennine sands, confined in narrow ribbons, show lower quartz–feldspar contents, abundant sedimentary lithics and lower chromium levels. In the last 870 kyr, the boundary between the Po and the Apennine sediment delivery systems shifted along a north–south axis in response to distinct controlling factors. A 20 km northward shift of the Po channel belt, possibly related to a tectonic event, is recorded across a regional unconformity dating to the Marine Isotope Stage 12/11 transition. High sediment supply rates during glacial‐lowstand periods widened the Po channel belt southward towards the Apennine domain for a few kilometres. The Last Glacial Maximum channel‐belt sand body, 30 km wide and 40 m thick, progressively narrowed northward after the glacial culmination. During the Holocene, channel patterns became avulsive and distributive. Narrow channel belts (<3 km) formed along the Po River branches, and abundant swamp and poorly drained‐floodplain muds were preserved in interfluvial areas. Activation and deactivation of the Po branches resulted in sharp narrowing and widening of the area available for Apennine‐rivers sedimentation. This work provides insights into tributary‐trunk river relations which control grain‐size distribution and compositional characters of subsurface deposits.


INTRODUCTION
Continental successions are valuable archives of information on past environmental changes that occurred in response to distinct forcing factors. Vertical changes in fluvial-channel stacking patterns have been observed in ancient and Quaternary alluvial successions worldwide (Cleveland et al., 2007;Labourdette & Jones, 2007;Kasse et al., 2010;Blum et al., 2013;Mancini et al., 2013;Newell et al., 2015;Campo et al., 2016). The relative influence of distinct controlling factors on these variations has been discussed in several studies (Schumm, 1993;Martinsen et al., 1999;Blum & T€ ornqvist, 2000). Glacio-eustasy has been regarded as the main factor controlling alluvial sedimentation after the Early-Middle Pleistocene transition, ca 900 kyr BP (Busschers et al., 2005;Gibling et al., 2005;Blum & Aslan, 2006). In these studies, stratal architecture was compared with a well-known climatic and eustatic history reconstructed through independent global and local proxies (e.g. Dansgaard et al., 1993;Waelbroeck et al., 2002). Late Quaternary investigations are also advantaged by the detailed knowledge of the drainage system, which permits reliable reconstructions of the source-to-sink patterns of sediment dispersal. The mutual relations between the trunk river and its tributaries have been explored in previous studies, which addressed: (i) the differences in composition between tributaries and trunkriver sediments (Sinha et al., 2009;Garzanti et al., 2011;Tentori et al., 2016Tentori et al., , 2018; (ii) the geomorphic response of tributaries to axial river incision (Leeder & Stewart, 1996;Posamentier, 1998Posamentier, , 2001Posamentier & Allen, 1999;Green, 2009); (ii) wetland formation due to tributary fan progradation in trunk river valleys (McCarthy et al., 2011); (iii) erosion of the toes of tributary fans by trunk-river lateral migration (Bishop, 1995;Larson et al., 2015); (iv) tributary-trunk river interactions due to hydrological changes over a few decades (Meade & Moody, 2010;Xu, 2016). Kvale & Archer (2007) and Vis et al. (2008) described the sedimentological characteristics and internal architecture of tributary incised valleys. The preservation potential of tributary networks has been discussed in several papers (Fielding et al., 2012, and references therein).
Through the analysis of the stratigraphic relations between Po and Apennine river deposits, this work aims at evaluating: (i) the extent to which tributary rivers may contribute to sediment accumulation in alluvial plains; and (ii) how sedimentological and compositional characters of the preserved sediments are influenced by the interactions between the trunk river and its tributaries through time.
The Pleistocene-Holocene stratigraphy of the Po Basin was reconstructed in a 4000 km 2 wide sector of the Po Plain, ca 70 km away from the modern coastline, with the aims of examining patterns of sediment dispersal and depicting major shifts in sediment provenance between the Po-Alpine and the Apennine river systems. For this purpose, 38 cores were analysed and correlated along six stratigraphic cross-sections, together with 300 stratigraphic data (published core descriptions, piezocone penetration tests and well logs). A 450 m deep cross-section explored the stratigraphy of the Po Supersynthem, an unconformity-bounded unit dated to the last 870 kyr. Five cross-sections focused on the Late Pleistocene-Holocene depositional architecture (last 50 kyr). This time window, encompassing the last glacial period and the present interglacial, offers the opportunity to study the response of the trunk river and of its tributaries to high-magnitude climate and eustatic changes.

Structural setting and subsurface stratigraphy
The Po Plain represents the foredeep of the Southern Alps and of the Northern Apennines, filled with Pliocene-Quaternary sediments (Fig. 1A). The Alpine and Apennine thrust belts developed in response to the convergence between the African and European plates, starting in the Cretaceous (Carminati & Doglioni, 2012) and still progressing at the rate of 3 to 8 mm/yr (Devoti et al., 2011).
The Northern Apennines developed since the Neogene on the hanging wall of a west-directed subduction zone (Livani et al., 2018, and references therein). The most external portion of the Northern Apennine accretionary wedge is buried beneath the Po Plain (Fig. 1B) and is composed of three arcuate systems of thrust-related folds (Fantoni & Franciosi, 2010;Turrini et al., 2014). From west to east: the Monferrato, Emilia and Ferrara-Romagna folds. The Ferrara-Romagna folds have been tectonically active since the Early Pliocene (Toscani et al., 2009;Boccaletti et al., 2011;DISS Working Group, 2018). Particularly, some faults of the Ferrara-Romagna Arc (i.e. Mirandola thrust system) were activated during the 2012 Emilia earthquake (Bonini et al., 2014) with fault plane solutions indicating dominantly compressional mechanisms (Pondrelli et al., 2012;Scognamiglio et al., 2012).
The Upper Po Synthem (UPS in Fig. 1C) is partitioned into five smaller-scale depositional successions (transgressive-regressive cycles of Amorosi et al., 2004;Fig. 1C). Beneath the modern coastal plain each succession consists of shallow-marine and coastal sediment bodies overlain by alluvial deposits (Fig. 1C). Close to the Apennine margin and beneath the Po River, the landward equivalents of these transgressiveregressive successions are composed of muddominated overbank strata that transition upward to regionally extensive fluvial channel bodies. Pollen series from 150 m deep cores (Amorosi et al., 2004(Amorosi et al., , 2008 and electron spin resonance dates (Ferranti et al., 2006) document that shallow-marine coastal bodies were deposited during late transgression and highstand of interglacial periods, whereas laterally extensive channel-belt sand bodies accumulated during forced regression and lowstand of glacial periods .

Surface geology of the drainage system
The Po River originates in the Western Alps and flows easterly for 652 km (Fig. 1A). It receives water and sediments from 141 tributaries, which drain a cumulative area of about 75 000 km 2 . The western part of the Po drainage basin, including large sectors of the Western and Central Alps and of the western Apennines, is characterized by extensive outcrops of crystalline-metamorphic and ophiolite complexes (Fig. 1A). In contrast, Mesozoic carbonate rocks are extensively exposed in the Southern and Eastern Alps (South-Alpine units in Fig. 1A).
The southern part of the study area is fed by three major Apennine rivers: Secchia, Panaro and Reno (Fig. 1A). These rivers have remarkably smaller catchment areas (<2300 km 2 ), mainly composed of Cretaceous tectonically deformed clays with carbonate blocks (Ligurian Units in Fig. 1A; Remitti et al., 2011) and of Palaeocene to Neogene coastal-marine sandstones and marls (Epiligurian Units in Fig. 1A; Ricci Lucchi, 1987). Pliocene shallow-marine clays and subordinate Pleistocene coastal sands crop out at the Apennine margin (Post-Evaporitic Units in Fig. 1A, Ricci Lucchi et al., 1982). The Secchia and Panaro rivers are tributaries of the Po River, whereas the Reno River, which flows nowadays into the Adriatic Sea, has been a tributary of the Po River in the past centuries, as highlighted by historical maps (available online at http://geoportale.regione. emilia-romagna.it).

The stratigraphic dataset
This study relies upon sedimentological, petrographic and geochemical analyses of 38 cores (locations shown in Fig. 2). Core length, generally <50 m, permitted the detailed investigation of Late Pleistocene and Holocene units (last 50 kyr). Only cores MIR and MED, 127 m and 111 m long, respectively, penetrated the entire Po Supersynthem (last 870 kyr). Cores were retrieved with a rotary-wash drilling method, which allowed high percentages (>90%) of sample recovery for fine-grained material. This technique, however, does not permit the preservation of sedimentary structures in sands. Cores were described in terms of lithology, grain-size, colour, consistency and accessory material (macrofossils, vegetal remains, peat and carbonate concretions). In situ tests include pocket-penetrometer measurements and estimates of the CaCO 3 content through reaction to HCl. Additional facies analysis was carried out in a 5 m long and 3 m deep trench (San Carlo trench, Fig. 2), where sediment deposited by a palaeo-Reno River, active during the 18 th Century, have been exposed .
The stratigraphic dataset also includes 40 core descriptions, 145 well logs (logged lithological sections and downhole wireline logs) and 122 piezocone penetration tests (CPTUs). Core descriptions refer to boreholes drilled for the Geological Mapping of Italy to scale 1 : 50.000 project (depth <180 m) and provide high-quality information on lithology, colour, consistency, reaction to 10% HCl, pocket-penetrometer measurements, and radiocarbon dates. Basic lithological descriptions (i.e. sand versus mud) deriving from the analysis of well cuttings were used to reconstruct the geometry of fluvialchannel bodies at depths between 50 m and 500 m. Lithological information (cuttings), resistivity logs and palaeontological data from six exploration wells (depth >500 m) were used as anchor points to track the base of the Po Supersynthem (see Stratigraphy of the Po Supersynthem section). Piezocone tests were interpreted based on Amorosi et al. (2017a) after calibration with nearby cores (see Facies association section).

Sediment provenance analysis
Petrographic and geochemical analyses of samples collected at depth <50 m allowed the recognition of the Po versus Apennine provenance of Late Pleistocene and Holocene sediments in the uppermost two cycles of unit UPS (Fig. 1C). Sediment provenance interpretations also relied upon petrographic data from Lugli et al. (2004;TAV cores) and Fontana et al. (2015Fontana et al. ( , 2019cores Fig. 2. Study area with location of the analyzed cores, the sampling sites for petrographic and geochemical analyses, and the trace of the stratigraphic cross-sections of Figs 6 to 9. The elevation (metres below modern sea level) of the 450 kyr unconformity and the projection of buried tectonic structures is also shown (modified after Martelli et al., 2017). SC1-2 and MB1-2), and geochemical data from Amorosi et al. (2016;cores CNC, ROV and ZMB) and Bruno et al. (2018;cores EM19, MN1 and Novi). Information on the sediment provenance of older units derive from Amorosi & Sammartino, 2018, cores MIR and MED) and from the Sheets 181, 202 and 203 of the Geological Map of Italy to scale 1 : 50.000.
Petrographic analyses were carried out on 19 sand samples collected from four cores (CV1, CV2, CV3 and Modena). For each sample, the bulk sand was used for qualitative petrographic observations, whereas point counting was carried out on the fine sand fraction (0.125-0.250 mm) separated by dry sieving. Modal analysis was performed according to the Gazzi-Dickinson method designed to minimize the dependence of the analysis on the grain size (Zuffa, 1985). At least 300 grains were point counted for each section. Components not related to the original sand composition, such as authigenic carbonate nodules and other particles originated from soil erosion, and organic matter and penecontemporaneous shell fragments were excluded from final calculations. The modal analyses were compared with detrital modes from modern rivers: Po (four samples, this work), Reno (four samples, Fontana et al., 2015), Secchia, Panaro and its minor tributaries (42 samples, Lugli et al., 2004Lugli et al., , 2007. Geochemical analyses were performed on 60 sand-to-clay samples collected from 10 cores (CNC, EM6, EM15, MF, MED, MIR, ZMB, RN, ROV and SFP, Fig. 2). Sixty-eight additional near-surface (<1.5 m) samples portrayed on the Pedogeochemical Map of Regione Emilia-Romagna (Amorosi et al., 2014) were used as reference for sediment-provenance characterization of subsurface deposits. All samples were oven-dried at 50°C, powdered and homogenized in an agate mortar and analyzed by Xray fluorescence (XRF) spectrometry for major element oxides, loss on ignition, and trace metals using a Philips PW 1480 spectrometer (Koninklijke Philips N.V., Amsterdam, The Netherlands). The matrix correction methods reported in Leoni et al. (1986) were followed. Certified reference material, including samples BR, BCR-1, W1, TB, NIM-P, DR-N, KH and AGV-1 (Govindarajiu, 1989) was also analyzed. The estimated precision and accuracy for traceelement determinations was 5%. For elements with concentrations <10 ppm, the accuracy was 10%.

Stratigraphic correlation and absolute chronology
The stratigraphy of the Po Supersynthem (last 870 kyr) was investigated down to 500 m depth, along a south-north oriented, 75-km long crosssection between the Apennine foothills and the town of Mantova (dashed line from core 219S2 to MN1 in Fig. 2). Nineteen core logs, 62 waterwell logs and six exploration well-logs were correlated, with a mean distance between the core locations of 1.1 km. The Late Pleistocene-Holocene depositional architecture (last 50 kyr) was investigated along four south-north oriented cross-sections (AA' to DD', Fig. 2) and a westeast oriented transect (EE' in Fig. 2). Sections AA' to EE' are 50 to 80 m thick and 40 to 60 km long. The mean distance between the core locations is ca 1.0 km.

Facies associations
Five facies associations were identified in the analysed cores and in the S. Carlo trench (location shown in Fig. 2). These are characterized as follows.

Fluvial channel (FC)
Description. This facies association is composed of 3 to 20 m thick gravel and sand bodies with erosional or sharp base and fining-upward (FU) grain-size trend (from coarse to fine sand). Gravel bodies, observed only in the southern cores (Modena and TAV, Fig. 2), are composed of sub-rounded, heterolithic cobbles and pebbles in a sandy or silty matrix. In northern cores, grey (2.5YR 7/1) coarse sands form >8 m thick sediment bodies, with occasional pebble layers at the base (Fig. 3B). In cores MIR and MED, at depths >90 m, sands are grey-peach (5YR 8/2), with heterolithic pebbles and cobbles (mainly dolostone and porphyry with diameter <5 cm). In core sands sedimentary structures are not preserved. High-angle planar and trough crossstratification has been observed in yellowish sands (2.5Y 8/2) of the San Carlo trench (Fig. 3A). Body fossils are absent and wood fragments are seldom encountered. In CPTUs adjacent to the analysed cores, tip-resistance values (q c ) are in the range of 3 to 20 MPa, with an overall decreasing upward trend. Pore pressure (u) values are negative.
Interpretation. Based on the lithology and grain-size, this facies association is interpreted as deposited in a high-energy environment. Sharp or erosional bases, basal pebble-layers, FU grain-size trends, traction structures and scattered wood fragments are common features of fluvial-channel deposits (Allen, 1963). Gravels observed in southern cores, close to the Apennine foothills, are interpreted as braided-river deposits (Miall, 1985(Miall, , 1996Bridge, 1993). The lack of preserved sedimentary structures in cores does not permit a detailed facies attribution. High q c values are typical of coarse-grained material (Schmertmann, 1969). Negative u values reflect rapid dissipation of the excess pore pressure generated by the penetration of the cone (Campanella et al., 1982). The presence of dolostones and porphyries in northern cores at depth >90 m suggests provenance from the Southern Alps .

Crevasse and lev ee (CL)
Description. This facies association includes <3 m thick sediment bodies, laterally or vertically associated with FC deposits, composed of: (i) fine to silty sand with FU trend; (ii) coarsening-upward (CU) silty to fine sand; and (iii) sand-mud alternations at centimetre to  Table 1. decimetre scales. Colour is grey (7.5YR 8/1) or beige (10YR 8/2). Body fossils are absent. FU sand bodies have small lateral extent (<20 m), concave-up base and flat upper surfaces. Crossstratification was commonly observed in outcrop and wood fragments were encountered locally. CU sand bodies, a few tens of metres wide, have nearly tabular shape with faint convex-up upper boundary. Sub-horizontal laminae due to thickening-upward coarser strata are observed. In sand-mud intercalations, mud strata are structureless and bioturbated, with sparse carbonate concretions and root traces. Sandy horizons have sharp or erosional lower boundary and are laminated or cross-bedded (CL in Fig 3A). Sand-mud couplets gently slope towards adjacent floodplain muds. Thickness and sand-mud ratio decrease downslope. In CPTU tests, q c is <10 MPa, whereas u < u 0 (static equilibrium pore pressure).
Interpretation. Sedimentological features and the close association with the fluvial-channel bodies allow interpretation of these deposits as channel-related facies. Particularly, sand bodies with erosive base and internal FU trend are interpreted as crevasse channels (Singh, 1972), whereas sheet-like sand bodies with CU trend are referable to crevasse splays. Sand-mud alternations grading into adjacent floodplain muds could represent channel lev ees (Alexander & Prior, 1971), formed by proximal overbank deposition. Bioturbation and root traces testify to incipient pedogenesis between successive overflow events. Sparse carbonate concretions likely reflect local groundwater-table fluctuations.
Well-drained floodplain (WDF). Description. This facies association is composed of thick (>10 m), subtly varying successions of hardened clayey silt and silty clay. Palaeosols were recognized at discrete stratigraphic intervals, marked by the presence of diagnostic horizons (A, Bk and Bw, Fig. 3C and Table 1). A horizons, up to 1 m thick, are dark brown (7.5YR 6/2) and show no or faint reaction to HCl. Bk horizons are typified by abundant carbonate concretions in the form of coalescent nodules and coatings. Bw horizons exhibit faint mottles of Fe and Mn oxihydroxides. The observed palaeosols show A-Bk-Bw, A-Bk or A-Bw profiles. Pocket penetrometer values are 2 to 3 kg/cm 2 , with higher values (>3 kg/cm 2 ) in Bk horizons. Fossils are absent. In CPTUs, q c and f s (lateral friction) are in the range of 1 to 3 MPa and 50 to 150 kPa, respectively. Pore pressure is generally greater than u 0 .
Interpretation. The dominance of fine-grained material suggests deposition in a low-energy interfluvial environment (Collinson, 1978). Palaeosols with A-Bk or A-Bk-Bw profiles have been interpreted as Inceptisols (Soil Survey Staff, 1999;Buol et al., 2011); palaeosols with A-Bw profiles as Entisols (Soil Survey Staff, 1999;Kraus, 1999). The mobilization of carbonates and their transfer into the soil profile suggest low groundwater table during exposure (well-drained floodplain). Unlike deeply weathered reddish soils (McCarthy & Plint, 1998;Abels et al., 2013;Kraus et al., 2015), these palaeosols show evidence of only incipient pedogenesis (Amorosi et al., 2014. In Inceptisols, the characteristics of secondary calcite (evolutionary stages 2 and 3 of Gile et al., 1981;Machette, 1985) are consistent with exposure periods of a few thousand years (Bruno et al., 2020a). Entisols likely result from exposure periods of a few hundred years (Kraus, 1999). Low q c and u > u 0 are typical of finegrained material (Robertson et al., 1986). Relatively high f s is referable to pedogenic processes .
Poorly drained floodplain (PDF) Description. This facies association is composed of soft, light-grey (2.5YR 8/1) clayey silt and silty clay, ca 3 m thick, with faint horizontal lamination (Fig. 3B). Isolated carbonate concretions are seldom encountered. Fossils and oxides-hydroxides were not observed. Pocket penetrometer values range between 1 and 2 kg/cm 2 . In CPTUs, q c and f s are in the range of 0.8 to 1.2 MPa and 20 to 50 kPa, respectively. Pore pressure is greater than u 0 . This facies association is commonly sandwiched between welldrained floodplain and swamp deposits.
Interpretation. This facies association has been interpreted as deposited in a poorly drained floodplain, with water table fluctuating close to the topographic surface. This interpretation is consistent with its stratigraphic position between well-drained floodplain and swamp deposits. A relatively high level of the groundwater table likely prevented eluviation-illuviation and oxidation. The formation of sparse and isolated carbonate concretions is likely due to water-table fluctuations.
Interpretation. The dominance of fine-grained material suggests deposition in a distal low-energy environment. Light grey colour of muds reflects poor degradation of organic material. The preservation of undecomposed plant material and the lack of oxides and secondary carbonates, together with low q c and f s values, suggest deposition in an environment with persistently high water table (for example, swamp; Diessel, 1992; Richardson & Vepraskas, 2001;Stolt & Rabenhorst, 2011).

Po versus Apennine sediment provenance
The Po Plain is a multi-sourced system, where Alpine and Apennine lithostratigraphic units display distinctly different compositional signatures. Provenance of Late Pleistocene and Holocene sediments has been assessed by comparing modal compositions of detrital grains and geochemistry of the cored samples with those of modern sediments from the Secchia, Panaro, Reno and Po rivers (Figs 4 and 5).

Sand petrography
Modern sands from the Apennine rivers (Secchia, Reno, Panaro and minor streams, Fig. 4A) have an overall litharenitic composition. The lithic association includes sedimentary siliciclastic grains (siltstones and shales) and carbonate lithics (largely micritic limestones and calcite spars). Shale grains are consolidated, wellrounded, with iso-orientation of the clay minerals, revealing their detrital origin. Modern Po River sands in the study area are distinguishable from Apennine sands by their significantly higher content of quartz-feldspar (Fig. 4A), coarsegrained metamorphic rock fragments, micas and heavy minerals. Sedimentary lithics are subordinate and consist of carbonate, with minor siltstone and shale grains. Carbonate lithics are more abundant downstream of the confluence with the Panaro River. Serpentinite and volcanic grains are minor components. The composition of Holocene sands is variable. Sands from southern cores (Modena, TAV, CV1, CV2 and CV3, location in Fig. 2) are similar to those of the modern Secchia and Panaro rivers. By contrast, sands from the northern cores (MB1 and MB2) show a quartz-feldsparrich composition similar to the modern Po River. Sands of the 18th Century Reno River (S. Carlo trench and cores SC1-2) have markedly lower contents of lithics, in particular shale and carbonate grains, compared to the modern Reno compositional field (Fig. 4A).
Late Pleistocene samples from cores Modena, TAV, CV1, CV2 and CV3 are within the compositional fields of modern Apennine rivers, but show an overall lower amount of sedimentary lithics and higher quartz and feldspar contents (Fig. 4B). This enrichment is particularly significant for cores Modena, SC1 and SC2, which partly overlap the compositional field of the modern Po River. Samples from the northernmost cores (MB1 and MB2) yielded the highest quartz and feldspar contents. Sand grains appear relatively unaffected by alteration and corrosion.
Petrographic data reflect substantial differences in extent and lithology of the drainage catchments between the Po and Apennine rivers. The major trunk river (Po) carries detritus well-mixed from wider sectors of the Alpine and Apennine orogens over a longer distance (Fig. 1A). As a result, the Po sand is mainly composed of quartzrich and feldspar-rich grains and metamorphic rock fragments. On the contrary, Apennine rivers have relatively smaller catchments and supply detritus mainly from sedimentary strata. In particular, the Ligurian units represent the primary source of shales and carbonates to the river network. The overall higher quartz and feldspar contents locally observed in Apennine Pleistocene samples seem to be primarily controlled by weathering conditions during the last glacial stage. Denudation, erosion and accelerated transport were probably responsible for the disintegration and microfracturing of lithic sedimentary grains, thus promoting an indirect increase of quartz and feldspar in sands (Lugli et al., 2007).

Sediment geochemistry
The distinctive difference in drainage-basin bedrock composition between the Po River catchment and Apennine (Secchia, Panaro and Reno) river watersheds (Fig. 1) is strongly reflected by trace-element geochemistry. Weathering of peridotite, gabbro, basalt and ophiolitic rocks, cropping out extensively in the Western Alps and at the north-western tip of the Apennines (Fig. 1), delivers large volumes of Cr-rich and Ni-rich detritus to the Po River system (Amorosi et al., 2014). A typical element ratio that was tested successfully in modern Po Plain sediments for the discrimination of Po River versus Apennine rivers provenance composition is Cr/V, irrespective of sandy versus muddy lithologies (Amorosi & Sammartino, 2007;Amorosi et al., 2014). An equally effective normalization factor of geochemical data that has been used to emphasize the role of parent-rock composition is the Cr/ Al 2 O 3 ratio (Greggio et al., 2018). The Cr/V ratio, adopted in this paper, has been commonly used as a key index for sediment provenance from mafic and ultramafic rocks (Garver et al., 1996;Lu zar-Oberiter et al., 2009). In the Po Plain, this ratio exhibits distinctly higher values along the Po River, whereas very low levels are invariably recorded in near-surface samples of its Apennine tributaries (Secchia, Panaro and Reno rivers in Fig. 5).
In the study area, Late Pleistocene and Holocene samples from the southernmost cores (EM15 and RN in Fig. 2) exhibit distinctively low Cr/V levels that plot invariably into the compositional fields of Apennine river sediments, across all grain sizes (Fig. 5). In contrast, samples from cores ZMB, CNC, ROV, MED, MIR and MF, despite being located up to 25 km south of the modern Po River (Fig. 2), show a characteristic Po River affinity, with consistently higher Cr/V values (Fig. 5). Core samples from EM6 and SFP plot considerably off of this general trend and display mixed (Po and Apennines) geochemical signatures (Fig. 5).
The spatial distribution of individual core samples and their relation to stratigraphic architecture is shown and discussed in the next sections.

Stratigraphy of the Po Supersynthem
Middle Pleistocene stratigraphy Above the folded and faulted Plio-Pleistocene substratum, the Po Supersynthem has variable thickness, from ca 100 m close to the Apennine margin and above the culmination of the Mirandola Anticline, to ca 400 m in the depocentres. The base of the Po Supersynthem, observed in cores 219S2 and MIR and in deep exploration wells, is characterized by the upward transition from coastal to continental deposits (Fig. 6).
In the northern sector, laterally extensive fluvial-channel sand bodies alternate with mudprone strata (Fig. 6). In core MN1, the upward transition from mud to sand is marked by a decline in pollen from warmth-loving species (i.e. broad-leaved trees), with a parallel increase in species (Pinus, mountain trees and herbs) indicating cooler climatic conditions (Amorosi et al., 2008). Two mud horizons, which record the maximum expansion of warmth-loving forests, were assigned to MIS 7 and MIS 5e (Fig. 6) based on stratigraphic correlations with coeval coastal sediments . Compositional data from cores MED, MIR and CNC show that channel-belt sand bodies numbered as '2' to '4' in Fig. 6, ca 30 km wide, were deposited by the Po River flowing in an axial position. The oldest sand body of unit UPS (number '1' in Fig. 6) is preserved only within a syncline. The youngest sand body results from the amalgamation of channel complexes 4a and 4b (Fig. 6) mainly accumulated during the last glacial episode. Towards the north, Po River sands are laterally juxtaposed with sand bodies deposited by Alpine rivers (see cores EM19 and MN1; Fig. 6). Towards the south, 35 to 40 km from the Apennine foothills, Po sands pass to floodplain muds delivered by the Apennine rivers. In this mud-prone succession, the physical tracking of palaeosols was possible only within the upper 50 m, where high-quality core data are available (see Late Pleistocene-Holocene stratigraphy section). Fluvial-channel lenses deposited by Apennine rivers (see provenance data from cores Modena, TAV181, CV1 and CV3, Fig. 6) are generally thinner and narrower than the Po-sourced sediment bodies. A thin, organic-matter-rich horizon marks the base of MIS 5e deposits (Fig. 6). At depth >50 m, basal surfaces of subsynthems can be tracked only tentatively due to poor quality of well-log descriptions (see dashed lines in Fig. 6). A recurrent alternation of coarse and fine-grained sediments is also observed close to the Apennine margin, where vertically stacked gravel bodies are the dominant feature.
Alluvial deposits of the Lower Po Synthem (LPS in Fig. 6) are folded and pinch out towards the culmination of the Mirandola anticline, where they are less than 10 m thick. Compositional data from cores MED and MIR show a strong affinity with modern Alpine rivers, such as Oglio and Mincio . The coeval Po fluvial-channel sediments likely correspond to two sand bodies encountered in well Carpi1, ca 25 km north of the Apennine foothills, at ca 360 m and 250 m depth ('A' and 'B' in Fig. 6, respectively), which Late Pleistocene-Holocene stratigraphy The Late Pleistocene-Holocene stratigraphy is explored here to unravel the mutual relations between the Po River and its southern tributaries across the last glacial cycle and the Present Interglacial.
In the northern part of the study area, a 40 m thick channel-belt sand body was encountered at depths >10 m (Figs 7, 8 and 9) and tracked continuously along strike for at least 30 km. Its southern boundary is located ca 20 km south of the modern Po River. Petrographic and geochemical data from 89 samples indicate sediment supply from the Po River (stratigraphic position of samples in Figs 7, 8 and 9). The local occurrence of thin mud lenses permits its subdivision in three smaller-scale aggradationally stacked sand bodies ('4a','4b' and '4c' in Figs 7,8 and 9). Based on sparse radiocarbon data and on the stratigraphic relations with overlying and underlying organic-matter-rich horizons, sand body 4b accumulated during MIS 3 and MIS 2 (between ca 52 and 14 kyr BP), whereas 4c was dated to the MIS 2 -MIS 1 transition (ca 14-9 kyr BP). Sand body 4a is tentatively assigned to MIS 4 based on the correlation scheme of Fig. 6. Thicker and more laterally persistent mud strata separate 4a from 4b (Fig. 9B). Older sand bodies attributed to MIS 6 are locally encountered at the bottom section in Fig. 9B.
Holocene (MIS 1) sand bodies are vertically amalgamated onto older sands and show lateral transition to muddy facies associations (Figs 7, 8 and 9). These sand bodies were deposited by the Po River (cores ZMB, SP, BUR and MB1-2) and by Apennine tributaries (SFP and SC1-2). Holocene mud-prone deposits in core EM6 (Fig. 9B) show alternating contributions from Po and Apennine sediments sources.
South of the Po channel belt, muds are dominant and fluvial-channel bodies are lens-shaped (Figs 7, 8, and 9). Fifty-three compositional data denote an Apennine sediment provenance. Laterally extensive gravel and sand sheets were encountered only at the southern edge of the study area. The Late Pleistocene interval, almost entirely composed of WDF muds, is marked by laterally extensive (>20 km) Inceptisols, locally replaced by coeval fluvial-channel deposits or eroded by younger channels. Two palaeosols, dated to 29-25 and 14.7-10.6 kyr BP (bold red and grey lines respectively in Figs 7, 8, and 9), mark the base of MIS 2 and MIS 1 deposits (Figs 7, 8 and 9). The 'red' palaeosol caps a series of closely spaced palaeosols that are radiocarbon dated to MIS 3. The 'grey' palaeosol marks the Pleistocene-Holocene boundary. The elevation of the Late Pleistocene palaeosols is highest close to the Apennine foothills and above the culmination of the Mirandola anticline. The MIS 3 palaeosol-bounded units are thin (<2 m) and nearly tabular. The unit bounded by the 'red' and 'grey' palaeosols is 4 to 12 m thick, with minimum values above the Mirandola Anticline (Figs 7, 8 and 9).
Holocene palaeosols are typically discontinuous, more immature (Entisols), and bound lensshaped mud-dominated units with convex-up upper boundaries and maximum thickness of 12 to 15 m. PDF, SW and CL facies are abundant. FC facies are isolated and of small lateral extent.

Middle Pleistocene evolution of Po and Apennine river systems
Vertically stacked channel-belt sand bodies are the dominant features of Middle and Late Pleistocene stratigraphy in the central Po Plain. Radiocarbon dates from the uppermost sand body ('4b' in Fig. 6) indicate its accumulation during the last glacial stage. Deposition of older channel-belt sand bodies during glacial stages is demonstrated by the increase in pollen from taxa typical of cold climates at the upward transition from mud-prone strata to laterally extensive sand bodies (Vittori & Ventura, 1995;Amorosi et al., 2008). The presence of four channel belt sand bodies ('1' to '4', Fig. 6) above a surface dated to ca 450 kyr BP (Martelli et al., 2017) suggests their deposition during the last four glacial stages (i.e. MIS 10, 8, 6 and 4-2). This age attribution is supported by the correlation with fluvial-channel bodies exposed in the Enza section, just 30 km west of the study area (location in Fig. 10), which accumulated at glacial culminations (Gunderson et al., 2014). The age of channel belts A and B is constrained between the ages of the basal and upper surface of the Lower Po Synthem (870 kyr and 450 kyr, Muttoni et al., 2003;Gunderson et al., 2014;Martelli et al., 2017) regionally mapped on a seismic basis. Sand body 'B', overlain by the 450 kyr surface, was tentatively assigned to MIS 12, based on stratigraphic correlation with coeval channel bodies exposed in the Enza Section (Gunderson et al. 2014). A palaeo-Po River course flowing closer to the Apennine chains has also been postulated by Garzanti et al. (2011), who documented a southward progradation of Alpine fans during MIS 12.
The location of boundaries between Po and Apennine deposits, identifiable within glacial stratigraphic intervals as a mud-sand contact, fluctuated in a north-south direction during the last 870 kyr. Lateral shifts <10 km are recorded from MIS 8 to the Present (see sand bodies '2', '3', '4a' and '4b', Fig. 6), whereas a northward shift of more than 20 km is recorded across the 450 kyr unconformity (from B to 1, Fig. 6). Alpine sediments from the LPS in cores MED and MIR  indicate that the Po River flowed south of the Mirandola anticline (ca 30 km south of the modern Po River) before 450 kyr BP (Fig. 10A). This reconstruction is corroborated by compositional data reported in previous studies, which documented the presence of Po sands close to the Apennine foothills (cores 181S3 and 203S9 in Fig. 10; Ceriani & Giulio, 2008;Albertini et al., 2009). Sand body '1', tentatively attributed to MIS 10, is confined north of the Mirandola thrusts (Figs 6 and 10B). Interestingly, Apennine sediments were not encountered in cores MED and MIR. Thus, this area was bypassed by the Apennine rivers during MIS 10 (Fig. 10). Alternatively, the Apennine sediments were not preserved at the anticlinal culmination. The preservation of oldest UPS deposits only within syncline structures and onlapping geometries onto the UPS lower boundary (Fig. 6) testify to a possible accelerated growth of the Mirandola Anticline around 450 kyr BP, at the LPS-UPS transition. In contrast, trunk-river sedimentation spread above the anticline culmination since at least MIS 8 (see sand bodies 2 to 4b, Figs 6 and 11), suggesting a possible decelerated anticline growth.

Late Pleistocene and Holocene evolution of the Po and Apennine river systems
The great thickness of the Po alluvial succession (up to 420 m in last 870 kyr, Fig. 6), the aggradational stacking of channel-belt sand bodies, and the relatively poor development of palaeosols reflect high subsidence rates (Posamentier & Vail, 1988;Wright & Marriott, 1993;Blum et al., 2013) that exceed 2 mm/yr in syncline and coastal plain areas (Carminati & Martinelli, 2002;Bruno et al., 2017b). In these depocentres, a thick (up to 70 m) sediment succession accumulated during the last glacial episode ( Fig. 6; see also Campo et al., 2020). Within this overall high-accommodation trend, vertical changes in: (i) fluvial-channel stacking patterns; (ii) lateral extent, maturity and continuity of palaeosols; and (iii) relative proportion of poorly drained floodplain and swamp facies, reflect variations in accommodation linked to the climaticeustatic history of the area. The Late Pleistocene succession records a progressive widening of the Po channel belt towards the glacial culmination, from MIS 4 (sand body '4a', Figs 6 and 12) to MIS 3-MIS 2 ('4b', Figs 6 and 12). Sand body 4b accumulated in a high-sediment supply setting due to its connection with the Alpine fluvio-glacial system (Fontana et al., 2014) and can be correlated from west to east for more than 120 km (Campo et al., 2016;Amorosi et al., 2017a;Bruno et al., 2017a), Laterally continuous channel-belt sand bodies from ancient (Labourdette & Jones, 2007) and late Quaternary successions ('amalgamated valley fills' of Blum et al., 2013) have been commonly reported in continental sequence stratigraphic models, as part of the 'low accommodation systems tract' (Wright & Marriott, 1993;Currie, 1997;Catuneanu et al., 2009). In the Late Pleistocene succession of the Po Plain, low accommodation is also demonstrated by low sedimentation rates ) and the lateral association of sand body 4b with a set of closely spaced Inceptisols with great lateral continuity, which developed in a well-drained environment. Similar series of vertically stacked, aggradational palaeosols bracketing the sequence boundary (McCarthy & Plint, 2013) have been described as 'soil zones' (Morton & Suter, 1996), 'pedocomplexes' (Hanneman & Wideman, 2006 or 'fluvial aggradational cycles' (Atchley et al., 2013). The widening of the Po channel-belt resulted in the cannibalization of mud deposits, which were transferred 400 km downstream of the study area to the lowstand prograding complex (Pellegrini et al., 2018), and in the reduction of the potential area for Apennine-rivers sedimentation. Preserved Apennine sediments are thin palaeosol-bounded overbank strata, whose deposition was controlled by millennialscale climate oscillations (Bruno et al., 2020a). From the analysis of 80 radiocarbon dated soil profiles, Bruno et al. (2020a) documented that palaeosol burial occurred during cold periods, when a more open herbaceous vegetation favoured erosion in the Apennine drainage basins and sediment transfer to the alluvial plains.
The post-glacial succession, between 14 and 10 kyr BP, records the progressive expansion of the Apennine sedimentation area due to the narrowing of the Po channel ('4c' in Fig. 12). That narrowing was the result of sediment trapping in proglacial lakes after retreat of the South-Alpine glacier (Fontana et al., 2014). The onset of the Holocene is marked by the upward decrease in fluvial channel width and connectivity (Figs 7, 8 and 9), with a parallel increase in frequency and volume of crevasse, poorly drained floodplain, and swamp deposits, associated with discontinuous Entisols. These sedimentological changes were accompanied by a sharp increase in sedimentation rates . These elements testify to the transition of the Po River channel from laterally migrating to avulsive/distributive, analogous to the changes observed in several alluvial systems worldwide (Lewis et al., 2001;Briant et al., 2005;Gibling, 2006;Kasse et al., 2010;Blum et al., 2013;Newell et al., 2015). This transition is an indirect consequence of marine transgression (Aslan & Autin, 1999;Phillips, 2011;Stouthamer et al., 2011), which dramatically reduced the area available for sediment storage and the river longitudinal gradients. Decreases in stream power and the capability to erode banks and make space for sediment storage led to river aggradation, crevassing and avulsions (Blum et al., 2013). In particular, channels became deeper, narrower and avulsive as they entered their backwater reach (Blum et al., 2013). Around 7.0 kyr BP, at maximum marine ingression, the area formerly occupied by the Po channel belt turned into the innermost edge of a wave-dominated estuary Fig. 11B). The abundance of swamp and poorly drained facies in the interfluvial areas is consistent with this reconstruction. Mid-Late Holocene delta progradation was characterized by frequent nodal avulsions and the consequent activation and deactivation of distinct Po River branches . Narrow (<3 km) channel belts developed in these short time intervals (Fig. 13). Avulsions occurred also in the Apennine sector, as testified to by compensationally stacked alluvial lobes. The abundance of poorly drained and swamp deposits in this area is likely related to rapid and localized vertical aggradation, which induced high groundwater table and waterlogging in adjacent interfluvial areas. The temporary deactivation of the southernmost Po river branches permitted the intrusion of Apennine sediments into the Po River domain (Fig. 13). The vertical alternation of sediments with contrasting geochemical compositions observed in core EM6 (Fig. 9B) reflects alternating contribution of the Po and the Apennine rivers at the boundary between their respective domains.

Controlling factors of trunk-tributary river shifts and implications for fine versus coarsegrained sediment storage
In large alluvial systems, trunk rivers are commonly regarded as the main agents of sediment deposition, whereas the contribution of tributaries is considered highly subordinate (Aslan & Autin, 1999;Busschers et al., 2007). Recent works highlight the role of tributary rivers in the evolution of alluvial and coastal-plain landscapes (Aalto et al., 2003;Fielding et al., 2012;Simms & Rodriguez, 2015;Tentori et al., 2021) and in the storage of fine-grained material (Kvale & Archer, 2007;Vis et al., 2008). In the Po valley, which in the study area extends for 90 km along strike, transverse tributaries have considerable space for the deposition of sediment supplied by two active orogens (i.e Southern Alps and Northern Apennines) parallel to the trunk river. A similar configuration characterizes the Ganges system, where tributaries draining the Himalayan belt (Singh et al., 2008) and the Indian Craton (Sinha et al., 2009) supply considerable amounts of sediment to the mainstream. The alluvial record of the Po Basin hosts evidence of episodic lateral shift of the Po River over the last 870 kyr, with consequent enlargement or shrinking of the South-Alpine and Apennine river domains. Of particular interest is the boundary between the Po and the Apennine sedimentary areas because it is marked by a sharp change in grain-size (Figs 6 to 9) and sediment composition (Figs 4 and 5). This boundary was displaced in a south-north direction by three main processes that acted at different timescales and responded to distinctly different controlling factors.
1 A northward shift of the Po channel belt by more than 20 km at the MIS 12-MIS 11 transition (Figs 6 and 10), implied a substantial enlargement of the Apennine domain, with the accumulation of considerable amounts of finegrained material from then onward. A structural control is proposed for this reorganization of the fluvial network, recorded across a regional unconformity characterized by onlap contacts in the study area (Fig. 6) and in other parts of the basin (Martelli et al., 2017). The LPS-UPS boundary, dated to the MIS 12-MIS 11 transition, also corresponds to a marine transgression at termination V (Lisiecki & Raymo, 2005) which may also have contributed to a substantial reorganization of the river network. Further chronological, palaeogeographical and structural data are required to discriminate the relative role of eustasy and tectonics. 2 Changes in the width of the Po channel belt driven by glacial-interglacial climate oscillations resulted in a periodic variation of the area available for Apennine rivers sedimentation. High sediment supply from the Alpine fluvio-glacial and Apennine river systems and low accommodation during falling stage-lowstand periods resulted in widening of the channel-belt and its displacement towards the Apennine domain. These climatically induced variations resulted in lateral shifts of the boundary between the Po and the Apennine domains of less than 10 km (Fig. 10). Fig. 13. Sediment provenance of Late Holocene exposed sediments showing lateral juxtaposition of the Po and Apennine deposits (modified after Amorosi et al., 2014). Location in Fig. 2. 3 During interglacial periods, nodal avulsions of the trunk river resulted in sharp widening and narrowing of the area available for Apennine river sedimentation. These variations at the 10 2 -10 3 yr timescale are highlighted within single cores by the alternation of deposits with contrasting composition (core EM6, Fig. 9B) and by the presence of Apennine ribbons within the Po sedimentary domains (Fig. 13). This study argues for a dominant autogenic control on Po River avulsions during the Holocene , although short-lived climate oscillations may have enhanced river instability in some periods (Cremonini et al., 2013).

CONCLUSIONS
Interaction of tributaries with a trunk river can influence grain-size distribution and composition of sediment stored in alluvial plains. In the Po Plain, the main river deposited aggradationally stacked sand bodies, 20 to 30 km wide and up to 40 m thick, separated by overbank mud strata. Po sands show high contents in quartzfeldspar and metamorphic rock fragments, combined with high chromium levels. To the south, Apennine deposits are coarse-grained only close to their valley outlets, where they form laterally continuous gravel and sand bodies. Downstream, muds are dominant, and sands are confined in narrow ribbons. Sands delivered by Apennine rivers are readily distinguishable from the Po sands due to their lower quartz-feldspar content, abundant sedimentary lithics and lower chromium levels.
From 870 to 450 kyr BP, when the Po River flowed close to the Apennine chain, Apennine river deposition was restricted to a relatively narrow area. Soon after 450 kyr BP, the Po channel belt shifted northward by more than 20 km, probably in response to a prominent tectonic event. Since then, Apennine rivers have deposited large volumes of fine-grained material over a considerably larger area than before.
This area periodically narrowed and enlarged in response to glacial-interglacial oscillations. In particular, Po channel-belt widening during glacial periods reduced the area for Apennine rivers sedimentation. A 25 to 30 km wide channelbelt sand body was deposited by the Po River between ca 52 and 14 kyr BP. Coeval Apennine muds to the south are punctuated by vertically stacked Inceptisols that attest to short-lived phases of no sedimentation driven by millennial-scale climate oscillations.
The post-glacial succession records the progressive narrowing of the Po channel belt and the transition to an avulsive-distributive Po River pattern in the study area. Narrow (<3 km) channel belts formed along the Po River branches and abundant swamp and poorly drained-floodplain muds were preserved in interfluvial areas. The temporary deactivation of the southernmost Po river branches resulted in the sharp widening of the area available for Apennine river sedimentation and permitted the intrusion of Apennine sediments in the former domain of the Po River. Alternating sedimentation from the Po and the Apennine rivers is recorded at the boundary between the two domains, where sediments with contrasting petrographic and geochemical signatures are vertically superposed or laterally juxtaposed.
Combined stratigraphic and sediment provenance analyses are particularly effective in the investigation of Quaternary alluvial successions, where the reconstructions of the patterns of sediment dispersal typically rely upon a robust chronology, a detailed knowledge of the drainage system and reliable sediment-provenance markers.