The extinction of non-avian dinosaurs at the KPB is the most profound example of megafauna being abruptly removed from fluvial environments. In the Western Interior of North America (Western Interior, hereafter), the disappearance of dinosaur megafauna coincides with a shift in fluvial sedimentary facies that is mappable in outcrop and thus formalized in stratigraphic nomenclature denoting contacts between uppermost Cretaceous (e.g., Lance and Hell Creek) and Paleocene (e.g., Fort Union) formations3,5,12. Direct evidence of the Chicxulub asteroid impact frequently coincides with those formational contacts (Fig. 1), especially in the Williston Basin, which hosts the most complete continental KPB sections in the world12. There, a thin, red claystone interval at the base of the lowest coal or lignite (simply 'coals', hereafter) of the Fort Union Formation (i) frequently yields elevated iridium anomalies, shocked quartz, and altered microtektite spherules12 and (ii) is a few cm below a volcanic ash bed coincident in age with Chicxulub tektites from Haiti13,14. Similarly abrupt KPB facies shifts -- sharp changes in lithology, fluvial architecture, and floodplain deposition that are stark in outcrop appearance -- have been documented throughout the Western Interior, and they may be a global phenomenon, but the present incompleteness of continental KPB records elsewhere prevents that unequivocal assessment here (see Supplementary Discussion). The current consensus is that, despite their synchronicity, KPB facies shifts are unrelated to the EKME and instead reflect chance overprinting of a global catastrophic event on what was otherwise a time-transgressive paleoenvironmental trend15,16,17,18,19. Here, we report on five newly discovered KPB sections from the Western Interior that further support the coincidence between continental facies shifts and the EKME, prompting us to reevaluate pre- and post-KPB sedimentology15,16,17,18,19,20,21 and the mechanisms underlying those geological changes.

It is remarkable that, across ~30° latitude (~37°-65° N), ~15° longitude (~104°-117° W) (Fig. 1), and diverse paleoenvironmental settings, continental KPBs in the Western Interior can be reliably discovered by targeting obvious facies shifts. To test the prevalence of that phenomenon, we sampled for iridium anomalies at formational contacts that were plausible KPB candidates based on paleontological evidence (see 'Methods'), yet captured distinct paleolandscape settings. These included: Lance-Fort Union formational contacts in the Bighorn Basin, Wyoming (Supplementary Fig. 1; Supplementary Fig. 2), a region bounded by Laramide mountain belts, and the Hell Creek-Fort Union formational contact at Mill Iron, near Ekalaka, Montana (Supplementary Fig. 3), a region far from elevated topography (Fig. 1). We discovered elevated iridium anomalies, which we attribute to the Chicxulub asteroid impact and thus the KPB, at all five of the sampled formational contacts (Supplementary Table 2). These represent the first KPBs identified in the Bighorn Basin and the southernmost KPB in the Williston Basin (Fig. 1). Together, they reinforce the pattern that continental KPBs invariably coincide with a facies shift across disparate paleoenvironments (see Supplementary Discussion).

These observations prompt us to challenge the prevailing notion that KPB facies shifts are independent of the EKME. Instead, we propose a new hypothesis: dinosaur megafauna were ecosystem engineers during the Late Cretaceous, and their sudden extinction caused an abrupt shift in continental facies at the KPB that persisted into the early Paleocene (the 'Dinosaurs as Ecosystem Engineers' hypothesis). We demonstrate further support for this hypothesis below, focusing especially on the Williston Basin since it hosts most of the known continental KPB sections, has superb three-dimensional outcrop exposures that most other KPB sections lack, and has featured prominently in debates about the EKME.

Continental facies shifts at the KPB most obviously reflect two aspects of fluvial systems: (i) river-meander-belt size and stability and (ii) the amount of clastic sediment (clay, silt, sand) deposited on the distal floodplain. Uppermost Cretaceous meander belts are characteristically narrow and unstable, delivering a steady supply of clastics to the distal floodplain; lower Paleogene meander belts are characteristically wide and stable, starving the distal floodplains of clastics and permitting the accumulation of concentrated organic debris. Those differences between pre- and post-KPB fluvial systems are epitomized by the Hell Creek/Lance and Fort Union formations, respectively, of the Williston (Hell Creek) and Powder River (Lance) basins (Fig. 2). The Lance/Hell Creek and Fort Union formations are all meandering-river deposits, evidenced by the preponderance of fine-grained clastics, erosive and asymmetrical channel bases, abandoned-channel deposits (oxbow lakes), and lateral-accretion beds (Fig. 2; Supplementary Fig. 4). The Hell Creek and Lance formations are characterized by (i) predominantly small (~1-3-m thick, ~5-10-m wide), single-storied fluvial-sandstone beds (small, short-lived river channels or crevasse splays), with large (~5-10-m thick, ~0.1-0.5-km wide), multistoried fluvial sandstone beds being rare, interbedded with (ii) massive, green-gray mudstone beds that exhibit root traces, slickensides, gleying, and gray-yellow mottling (immature, water-logged soils; i.e., hydromorphic paleosols). The lower Fort Union Formation is characterized by (i) large (~8-30-m thick, 0.1-1-km wide), multistoried fluvial sandstone beds (large, amalgamating, proximal river channels) and 'variegated beds' (associated point bars; see below) and (iii) coals (swamp-dominated, organic-rich distal floodplains).

The sudden, geographically widespread appearance of coals is the most conspicuous feature of lower Paleogene facies in the Western Interior and defines the base of the Fort Union Formation. Even in the absence of coals, however, the inorganic sedimentology of the Fort Union Formation is readily distinguished from the underlying Hell Creek or Lance formations, largely due to the 'variegated beds' -- brightly colored, thinly interbedded claystone, siltstone, and fine-grained sandstone beds that are often under- and overlain by coals and were thus likely deposited laterally and coevally to coal deposits elsewhere. These 'variegated beds' have historically been interpreted as pond or lake deposits, and Fort Union deposition has thus been attributed to various mechanisms to increase 'wetness' in the early Paleocene, resulting in mostly ponded-marshy accumulation of clastic and organic sediments. We collectively term these proposed mechanisms, which include increased precipitation and/or a rise in water table, the 'Wet Paleocene' hypothesis.

New observations from our work in the Hell Creek region of northeastern Montana, in conjunction with previous research, cast doubt on the 'Wet Paleocene' hypothesis: (i) Hell Creek Formation paleosols denote the prior existence of a very high water table, and (ii) most 'variegated beds' of the Fort Union Formation are not pond deposits but rather lateral-accretion deposits associated with large, broadly meandering rivers (Fig. 2; Supplementary Fig. 4). There is substantial support for the first point -- most overbank deposits in the Hell Creek Formation represent ponds or hydromorphic paleosols. The second point, to our knowledge, has not been fully appreciated. In the dozens of outcrops studied across the eastern Montana portion of the Williston Basin, we found that the 'variegated beds' consist of alternating fine- and coarse-grained sediments that dip at ~5-25°, coarsen down-dip, frequently grade into proximal channel-sandstone deposits, and exhibit paleoflow indicators (e.g., ripple marks) that are perpendicular to dip direction (Fig. 2). Further, although paleoflow directions remain roughly east-northeast across the KPB in our study area (Hell Creek mean = 109°, n = 166; Fort Union mean = 63.5°, n = 310), the standard deviations from those mean paleoflow directions are substantially higher in the Fort Union Formation (SD = 69.3°) compared to the Hell Creek Formation (SD = 52.9°), suggestive of a more pronounced meandering system in the early Paleocene. Thus, instead of representing pond or lake deposits, the 'variegated beds' represent point-bar deposits (and potentially counterpoint bar deposits, see Supplementary Discussion) associated with the lateral migration of fluvial-channel meanders. This interpretation is consistent with the available paleoclimatic data, which indicates neither an abrupt nor persistent change in precipitation intensity across the KPB (see Supplementary Discussion).

Transgression of the Cannonball Sea (Fig. 1) is most commonly invoked to explain the 'Wet Paleocene' interpretation of KPB facies change in the Williston Basin (the 'Marine Transgression' hypothesis), largely based on instances of minor (cm-m scale) stratigraphic offset between purported KPBs and the Hell Creek-Fort Union formational contact (i.e., the facies change is time-transgressive, not synchronous with the KPB). However, most of those KPBs are identified using non-impact-related tools or proxies like palynology or carbon-isotope stratigraphy, which are sensitive to reworking and lithology and thus insufficiently precise to place the KPB at ≤ 2 m scales (this has been noted previously). Palynology is still an important tool for constraining the approximate stratigraphic position of the KPB at the outcrop scale, and pollen is often the best tool available in certain locations, despite its propensity for remobilization by fluvial action (see Supplementary Discussion for a fuller discussion about the merits and shortcomings of using pollen or carbon isotopes to place the KPB).

Although some impact-identified KPBs are located a few cm below, within, or atop a coal (Fig. 1; Supplementary Table 2), those putative examples of time-transgressive KPB facies change assume that coals are the sole sedimentological indicator of that change. Yet, a preponderance of coals in lower Paleocene strata is merely one symptom of the systematic shift in the nature of fluvial systems at the KPB (Fig. 2). Indeed, coal-bearing units do occur in the Late Cretaceous, including infrequently in the Hell Creek Formation; that some KPBs would be preserved in Cretaceous coal swamps that persisted into the early Paleocene should therefore be expected (Supplementary Fig. 5; see Supplementary Discussion).

Further, magnetostratigraphy indicates that Cannonball Sea transgression began no earlier than chron C29n and occurred most substantially during chron C28r, thus postdating the KPB (in the middle of chron C29r) by ca. 300-900 ka (Supplementary Fig. 5). Marine transgression should also result in high-accommodation features in the continental record, such as hydromorphic paleosols, low channel:floodplain ratios, and single-story channels; a pattern that more closely matches the Hell Creek Formation (which does interfinger with marine deposits to the east) than the Fort Union Formation (Fig. 2; Supplementary Fig. 5). Lastly, the 'Marine Transgression' hypothesis fails to explain why facies changes analogous to those in the Williston Basin coincide precisely with the KPB in geographically distant areas like the Alberta and Powder River basins, rather than tracking the advance of Cannonball Sea (Supplementary Fig. 5), or why facies changes occur at the KPB in areas like the Bighorn Basin that were separated from the rest of the Western Interior by Laramide arches (Fig. 1). Thus, the preponderance of evidence casts doubt on the 'Marine Transgression' hypothesis; the KPB and lower Fort Union more likely coincided with a marine regression or lowstand, both locally and globally.

The Laramide Orogeny has also been proposed to explain KPB facies shifts, but that process was spatially diachronous and insufficiently abrupt to generate the observed stratigraphic patterns across such a large geographic range (Fig. 1) -- Laramide deformation advanced from the southwest to the northeast of the Western Interior over the course of the Late Cretaceous through early Paleogene, and there is no evidence of a continent-wide pulse of basement-cored block uplifts coincident with the KPB. Landscape denudation via Chicxulub-associated wildfires, in contrast, could be both abrupt and widespread, but such a mechanism cannot explain the persistence of those facies changes throughout the lower Fort Union Formation for a temporal duration exceeding 1 million years. Thus, no previous hypotheses adequately explain the observed facies shifts across the KPB (see Supplementary Discussion).

We propose that the sedimentological and stratigraphic features that define Upper Cretaceous versus lower Paleogene continental strata are the result of ecosystem engineering by dinosaur megafauna in the Cretaceous and their disappearance in the Paleogene (Fig. 3). The 'Dinosaurs as Ecosystem Engineers' hypothesis better explains the synchroneity, style, and persistence of continental facies change and is linked to a causal mechanism that is well-documented in Quaternary terrestrial ecosystems. Mammalian megafauna profoundly affect landscapes and vegetation, but dinosaurian megafauna were many times larger, and the spatial extent of ecosystem engineering scales positively with body size. Even smaller herbivorous dinosaurs likely reached body masses of ~1000 kg, and some (e.g., Triceratops horridus) may have approached 15,000 kg, so their impact on terrestrial vegetation structure and fluvial dynamics was likely profound.

Not only were individuals of these species massive, but taphonomic studies indicate most ceratopsians and hadrosaurs were herd animals. These herds, along with rarer large-bodied predators (e.g., Tyrannosaurus rex, ~8000 kg), likely maintained open or patchily forested habitats during the Late Cretaceous via trampling and uprooting. Consequently, fluvial systems would flood and avulse frequently, resulting in: (i) short-lived meander belts and (ii) a frequent supply of clastic sediment to the distal floodplain, consistent with the (i) small, single-storied fluvial channels, thin crevasse-splay sandstone beds, and (ii) hydromorphic paleosols that characterize Upper Cretaceous strata. When those megafauna went extinct, so too did their influence on vegetation cover, allowing dense forests to take root, resulting in stable and long-lived river-meander belts that would have: (i) allowed prolonged reworking of channel meanders and (ii) starved the distal floodplain of clastics thus promoting organic accumulation, consistent with the (i) large, multistoried sandstone channels with abundant adjacent lateral-accretion bedding and (ii) laterally continuous coals that characterize lower Paleocene strata (Fig. 3).

Whether or not non-avian dinosaurs were in decline prior to the EKME remains debated. Nevertheless, in the Western Interior, large-bodied dinosaurs, especially ceratopsians, persisted up to the KPB and their disappearance during the EKME was geologically abrupt. Indeed, in nearly all of the KPB localities discussed here, dinosaurian megafauna are well-known from below the boundary and disappear at the KPB. Further, proposals of dinosaur decline specifically refer to taxonomic richness, not relative abundance -- the former is germane to evolutionary interpretations, but the latter is most relevant to ecological impacts, and there is no evidence that large-bodied dinosaurs were numerically depauperate prior to the EKME. Lastly, although mammals underwent adaptive radiations in the Late Cretaceous and increased in body size quickly in the earliest Paleocene, mammals remained relatively small for much of the early Paleogene; as such, mammals did not likely substantially impact the structure of standing vegetation until the Eocene or Oligocene.

Importantly, the immediate consequences of the Chicxulub asteroid impact on standing vegetation cover were relatively short-lived and only evident at the cm scale in the rock record. Although plant taxonomic recovery from the EKME likely took thousands to millions of years, lower taxonomic richness does not preclude forest density or canopy structure. Indeed, the warm, monsoonal climates of the earliest Paleocene of the Western Interior would have allowed the establishment of dense forest vegetation in a matter of decades to centuries post-KPB. The 'Dinosaurs as Ecosystem Engineers' hypothesis can therefore explain the pace, style, and persistence of the continental KPB facies shift and is consistent with the vertebrate and plant fossil records (see Supplementary Discussion).

The facies shifts described above reflect landscape-scale changes in fluvial geomorphology that are likely synchronous with the KPB; yet, KPB-coincident facies changes are not identical in mountain-proximal versus mountain-distal portions of the Western Interior (Fig. 4). Those in the Williston Basin are the best studied, and they broadly match the KPB facies shifts observed in other mountain-distal basins, such as the Powder River, Alberta, and Brackett basins (Fig. 4; but see Supplementary Discussion). The KPB in the Willow Creek Formation of the Alberta Foothills was deposited off the flank of the Sevier thrust belt (Fig. 1) and thus reflects a unique facies change reflective of locally well-drained, semiarid conditions. KPBs in the Raton Formation of the Raton Basin (Fig. 1) mark the contact between the underlying 'lower coal zone' and overlying 'barren series' and superjacent 'upper coal zone'; the former is characterized by single-storied channels and hydromorphic paleosols, whereas the 'barren series' is dominated by multistoried, laterally aggraded, channel-sandstone beds. KPB facies shifts in the Bighorn Basin closely match those in the Raton Basin (Fig. 4; see Supplementary Discussion), differing mainly in an absence of coals in the underlying Lance Formation. KPBs from the Denver Basin may exhibit a facies change like those in the Raton and Bighorn basins; however, the only impact-identified Denver Basin KPBs come from a drill core and a poorly exposed outcrop in the central and eastern basin, respectively, precluding precise assessment of the KPB facies change currently (but see Supplementary Discussion). We contend that those similarities between the Raton, Bighorn, and potentially Denver basins reflect their proximity to emerging Laramide mountain belts (Figs. 1 and 4).

That continental KPB facies shifts in mountain-proximal settings are not identical to those in mountain-distal settings is expected given that (i) the effects of vegetation and ecosystem engineering on fluvial geomorphology, and (ii) the drivers of river avulsion, both of which are central to the 'Dinosaurs as Ecosystem Engineers' hypothesis, differ markedly in mountain-distal versus mountain-proximal settings (Fig. 3). Those differences could explain the abrupt appearance of laterally expansive and vertically thick proximal channel deposits overlying the KPB in places like the Raton and Bighorn basins (see Supplementary Discussion).

The 'Dinosaurs as Ecosystem Engineers' hypothesis explains both the geographically widespread coincidence between the KPB and continental facies shifts and provides a mechanism for those changes. Nonetheless, more empirical data are needed to robustly evaluate its legitimacy; in particular, more quantitative assessments of fluvial geomorphology and plant-canopy structure across the KPB throughout the Western Interior. Our hypothesis is falsifiable and testable in numerous ways, such as: (i) Continental, impact-identified KPBs should always be closely associated with a sedimentary facies change; (ii) Forest canopy structure, which is commonly inferred via light-sensitive proxies, should be relatively open in the latest Cretaceous and relatively closed in the earliest Paleocene; and (iii) Fluvial deposits overlying the KPB should reflect more broadly meandering and temporally long-lived rivers relative to those below the KPB, at least in mountain-distal settings.

If true, the 'Dinosaurs as Ecosystem Engineers' hypothesis has important implications for the evolution of terrestrial ecosystems. For example, early flowering plants were weedy opportunists that thrived in frequently disturbed environments, and avulsion-prone fluvial systems of the Cretaceous maintained by dinosaur megafauna may have promoted their early ecological proliferation. The closing of forest canopies predicted by the 'Dinosaurs as Ecosystem Engineers' hypothesis is also consistent with observed increases in angiosperm-seed size after the KPB, attributed to the emergence of dense forests following the extinction of dinosaur megafauna. Lastly, most Mesozoic mammals were likely terrestrial or semifossorial, matching expectations for an open rather than densely forested habitat. Widespread emergence of dense forest cover in the Paleocene may have therefore prompted the post-KPB rise of arboreal and frugivorous mammals, such as primates.

In sum, this work highlights the need for continued research on the Cretaceous-Paleogene transition in continental settings. It is increasingly apparent that abiotic drivers like climate change and tectonics have shaped the distribution and evolution of life; yet, as demonstrated here, life itself can also sculpt our planet. By bearing these relationships in mind, we may find that Earth's climate, landscapes, and biota have been tightly entwined throughout geologic history.