New mammalian local faunas from the first ca. 80 ka of the Paleocene in northeastern Montana and a revised model of biotic recovery from the Cretaceous–Paleogene mass extinction

ABSTRACT The earliest phases of mammalian recovery following the Cretaceous–Paleogene (K–Pg) mass extinction are incompletely known but crucial to understanding the development of modern terrestrial ecosystems. Here we report new mammalian faunal data from three vertebrate microfossil assemblages in the Hell Creek region of northeastern Montana, the deposition of which we constrain to within the first 28–80 ka of the Paleocene using new stratigraphic observations within a high-resolution chronostratigraphic framework. We quantified the taxonomic diversity among these three assemblages and five other assemblages from both the Hell Creek region and Denver Basin, together spanning the first ca. 300 ka post-K–Pg mass extinction. Our results allowed us to sub-divide the established ‘disaster’ and ‘recovery’ phases of recovery into the following sub-phases: (i) early disaster, characterized by the presence of ‘dead clades walking,’ high relative abundance of bloom taxa, and the appearance of post-mass-extinction immigrants, (ii) late disaster, characterized by a reduction in the number of ‘dead clades walking,’ continued high relative abundance of bloom taxa, and a more diverse assemblage of immigrants, (iii) early recovery, characterized by decreased relative abundance of bloom taxa, and continued immigration, and (iv) late recovery, characterized by the onset of in situ diversification. We note important differences in the pattern and timing of mammalian faunal succession between the Hell Creek and Denver Basin, suggesting that post-K–Pg mammalian recovery was spatially heterogeneous. Our results provided a new model for post-K–Pg mammalian biotic recovery that can now be tested with other earliest Paleocene assemblages across western North America.


INTRODUCTION
The Cretaceous-Paleogene (K-Pg) mass extinction 66 million years ago (Ma) is a pivotal event in evolutionary history, resulting in the formation of the mammalian-dominated terrestrial ecosystems present today (Simpson, 1937).The Paleocene biotic recovery of mammals transitioned into an evolutionary radiation, particularly of placentals (but see dos Reis et al., 2014 for discussion of molecular evidence suggesting a latest Cretaceous radiation for placental mammals), associated with increases in body mass (Alroy, 1999;Smith et al., 2010;Wilson, 2013), taxonomic richness (Lillegraven, 1972;Maas & Krause, 1994;Wilson, 2014;Wilson et al., 2021), and ecological disparity (Chen et al., 2019;Grossnickle & Newham, 2016;Grossnickle et al., 2019;Halliday & Goswami, 2016;Wilson, 2013).This important interval in mammalian history has been studied using a variety of taxonomic and, to a lesser extent, ecological proxies (e.g., Archibald, 1983;DeBey & Wilson, 2014;DeBey & Wilson, 2017;Lillegraven & Eberle, 1999;Lyson et al., 2019;Maas & Krause, 1994;Smith et al., 2018;Wilson, 2013Wilson, , 2014)).However, those proxies have often been applied over coarse spatiotemporal scales (i.e., regional-to-continental and hundreds of thousands to millions of years), such that local, ecologically relevant patterns of interest are obscured because they are aggregated across biogeographic provinces, environmental gradients, and long temporal intervals.A key step toward obtaining higher-resolution patterns of the K-Pg biotic recovery and radiation of mammals is the detailed description of multiple fossil assemblages and their geological context (sedimentology, stratigraphy, and geochronology) within a single, local framework.
The localities in this study-Morales 1, Herpijunk Promontory, and Carrie Padgett 6-are briefly mentioned by Archibald (1982), but paleontological and geological fieldwork conducted in the last 12 years has led to larger fossil sample sizes from these localities, refined understanding of their stratigraphic relationships, and better integration of their fossil assemblages into the high-resolution geochronological framework for the study areas.Indeed, we constrain the age of these localities, which are all located within a ∼2 km 2 area, to the first ca.80 ka after the KPB within C29r; this offers the highest resolution view yet known of the initial phase of post-K-Pg mammalian recovery in the Western Interior of North America.To better understand the temporal pattern of post-K-Pg mammalian recovery, we quantify the taxonomic diversity of these fossil assemblages and compare them with those from other early Puercan (Pu1) local faunas from the Hell Creek region (Archibald, 1982;Lofgren, 1995;Clemens, 2002;Wilson, 2014;Sprain et al., 2018), which have been integrated into the same geochronological framework.We also make comparisons to fossil assemblages from the farther afield Denver Basin (Eberle, 2003;Dahlberg et al., 2016;Middleton & Dewar, 2016) to explore the heterogeneity of biotic recovery patterns across the Western Interior.

Biotic Recovery
Biotic recovery is a complex evolutionary and ecological process that lacks a single, clear definition.We broadly describe biotic recovery as the return to pre-mass-extinction or similar levels of diversity (Erwin, 1998;Kauffman & Harries, 1996;Sahney & Benton, 2008).This process of biotic recovery is often characterized as a three-phase model derived from marine settings (Erwin, 1998).Phase 1, often referred to as the 'lag,' 'survival,' or 'disaster phase,' is the immediate aftermath characterized by low evenness and low taxonomic richness, along with high relative abundance of ecological generalist bloom taxa and low relative abundance of immigrant refugia species, taxa that were likely present nearby but did not appear in the fossil record until after the disaster (Erwin, 1993(Erwin, , 1998;;Kauffman & Harries, 1996;Harries & Kauffman, 1990;Schindler, 1990;Schubert & Bottjer, 1992;Smith et al., 2018).Phase 2, the 'recovery phase,' is typically characterized by the appearance of new taxa that are phylogenetically closely related to previously documented taxa (presumably byproducts of in situ diversification of refugia taxa or local survivors), a continued increase in taxonomic richness and relative abundance of non-bloom taxa, a decreased relative abundance of bloom taxa, and the reappearance of Lazarus taxa (Kauffman and Harries, 1996).Lastly, Phase 3, the 'recovered phase,' marks the return to, or exceeding of, pre-mass-extinction levels of richness and evenness (Erwin, 1998;Kauffman and Harries, 1996).We acknowledge that the previous descriptions of the three-phase model were derived from marine settings, and as such, the patterns observed may differ from those observed in the terrestrial realm.Therefore, we refine this model of biotic recovery to better fit the patterns observed in our terrestrial communities.We refer to the refugia species as immigrant taxa, due to the lack of resolution on their origins and the possible reoccupation of their original habitat (see the Faunal Recovery Groups section for information on taxonomic groupings; see Kauffman and Harries, 1996, for a more in-depth assessment of refugia species).Moreover, we subdivide the phases based on our diversity analysis and stratigraphic interpretations of the three described localities, along with other localities from the Hell Creek region and the Denver Basin (chosen because its chronostratigraphic framework is comparable to that of the Hell Creek).

Field Methods
The University of California Museum of Paleontology (UCMP) and the University of Washington Burke Museum of Natural History and Culture (UWBM) field crews collected the fossils included in this study over the last 40 years.Bulk sediment samples were collected from the fossiliferous horizons of the three vertebrate microfossil localities: Morales 1 (UCMP locality V77128 = UWBM locality C1671), Herpijunk Promontory (UCMP V77129 = UWBM C1153), and Carrie Padgett 6 (UCMP V77124 = UWBM C1360).The UWBM field crews collected over 2000kg of bulk sediment across all three localities; mass estimates of the bulk sediment collected by the UCMP crews are not available but are likely comparable.The bulk sediment samples were underwater screen-washed both in the field and then in the laboratory, using nested boxes with #18-and #30-size meshes (0.98 mm and 0.54 mm openings, respectively) and following procedures in Cifelli et al. (1996).Volunteers and students picked fossils from the resulting sediment concentrate with the aid of dissecting microscopes (10× magnification) at the UCMP and UWBM facilities (Cifelli et al., 1996).Here, we describe 398 mammalian specimens from these collections; specifically, those isolated premolars and molars and a few dentary and maxillary fragments that are identifiable to the genus level or lower.Other mammalian specimens (189), mainly premolars and tooth fragments, were found as well.Specimen descriptions and images can be found in Supplementary File 1 (Figs.S1-S12, Tables S1-S18, Supplementary File 1).The geological descriptions for the localities are based mainly on our own field observations and measurements but also on the published literature (e.g., Archibald, 1982;Smit & Van der Kaars, 1984).We measured stratigraphic sections for the localities using a Jacob's staff and Abney level.When beds could not be physically traced on foot, we calculated stratigraphic distances from marker beds (e.g., coal layers) using high-resolution (<1 m accuracy) Trimble R2 GNSS receivers (Wide Area Augmentation System enabled).

Taxonomic Metrics
Relative abundance and other taxonomic metrics were calculated to allow for quantitative comparisons across localities.
Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-2) Relative abundance was calculated as the number of identifiable specimens (NISP) of a taxon divided by the total number of specimens (identifiable to genus or lower) in the assemblage.We used NISP to represent individuals because the vertebrate microfossil assemblages described here were transported at least some distance by fluvial action prior to deposition and association of elements from the same individual is unlikely (Badgley, 1986;Rogers & Brady, 2010).Taxonomic richness was assessed using three metrics: raw richness, rarefied richness, and shareholder quorum subsampling (SQS) (Alroy, 2010).For each metric, richness was calculated at both the species and genus level to account for difficulties in assigning species-level identifications to certain isolated specimens.We compared the richness values of the three assemblages described in this paper with those from the latest Cretaceous Flat Creek local fauna (combined assemblages from Flat Creek 3 and 5; Wilson, 2014) along with three other earliest Paleocene assemblages from the Hell Creek region (Z-Line and Luck O Hutch [Smith et al., 2018] and Worm Coulee 1 [Wilson, 2014]) and two from the Denver Basin (Gars Galore [Dahlberg et al., 2016] and Littleton [Middleton & Dewar, 2004]), all of similar age.
To correct for differences in sampling intensity of assemblages, we calculated rarefied richness with 95% confidence intervals for each assemblage using the function rarefy() from the R package vegan, which uses the subsampling methodology of Hurlbert (1971).We ran rarefaction with sampling level N = (smallest sample size-1).To account for differences in evenness structure of assemblages, we also subsampled with SQS following Holland (2015) based on Alroy (2010).To further account for differences in evenness, we ran SQS excluding the most common taxon (Alroy, 2010) and chose our quorum level (q) based on the assemblage with the lowest value (Luck O Hutch in all cases; Tables S19, S20, Supplementary Data 1).
We calculated two other taxonomic diversity indices at genus level that incorporate both richness and relative abundance of taxa in each assemblage: Simpson's evenness (1-D; Simpson, 1949) and Pielou's evenness (J'; Pielou, 1966).These indices were chosen due to their relative ease of interpretation and emphasis on different aspects of relative abundance; Simpson's evenness emphasizes abundant taxa, whereas Pielou's evenness emphasizes rare taxa (Magurran, 2004; Tables S19, S20, Supplementary Data 1).We also assessed faunal dissimilarity between assemblages at the genus level using the Bray-Curtis dissimilarity metric (BC) (Lance & Williams, 1967).BC performs well with small sample sizes but tends to emphasize abundant taxa (Krebs, 1989).We followed the double standardization method outlined in Donohue et al. (2013) and Smith et al. (2018) to balance the influence of abundant and rare taxa (Faith et al., 1987;Krebs, 1989).We conducted non-metric multidimensional scaling at the genus level to visualize the associations between assemblages and genera.

Faunal Recovery Groups
We categorized each taxon from the post-KPB mammalian local faunas into one of four faunal recovery groups: 'dead clades walking,' bloom taxa, immigrant taxa, or byproducts of in situ diversification (Table S21, Supplementary Data 1).The term 'dead clades walking' (DCW; Jablonski, 2002) refers to taxa whose survival across an extinction event was short-lived.We define DCW as taxa present in the Lancian that locally survived the K-Pg mass extinction but attained only low relative abundance in the Pu1 and did not persist beyond that subinterval (e.g., Cimexomys spp.).Bloom taxa were also local survivors, typically ecological generalists, but found in high relative abundance during the disaster stage of biotic recovery (Kauffman & Harries, 1996;Erwin, 1998).We follow Wilson (2014) in categorizing Mesodma spp., Thylacodon montanensis, and Procerberus formicarum as bloom taxa.Immigrant taxa are defined as having made their first local appearance post-KPB and lacking a probable ancestor among Lancian local faunas in the same region (Clemens, 2002(Clemens, , 2010;;Weil & Clemens, 1998).The archaic ungulates Baioconodon, Protungulatum, Oxyprimus, and Mimatuta and the eucosmodontid multituberculate Stygimys, all of which first appeared immediately after the KPB, are considered immigrants (but see Archibald et al., 2011 andKelly, 2014 for possible pre-KPB immigration events of certain archaic ungulates).The byproducts of in situ diversification are taxa (e.g., Procerberus cf.P. grandis) that are probable descendants of immigrant taxa or local survivors (Clemens, 2010;Eberle, 2003;Weil & Clemens, 1998).This group includes members of clades that appear after the first appearance of the clade (e.g., eucosmodontid multituberculates, arctocyonid and periptychid archaic ungulates).Although these taxa could possibly represent later immigration events rather than in situ diversification, they differ from the taxa that we designate as 'immigrants' because they have congeners present in older local faunas from the region and in certain cases a demonstrated phylogenetic connection (McComas & Eberle, 2016;Weil, 1999).Although categorization of taxa into faunal recovery groups is imperfect, it provides a useful means to characterize the phases of post-KPB biotic recovery.

GEOLOGICAL BACKGROUND OF THE HELL CREEK REGION
The Hell Creek region of Garfield and McCone counties in northeastern Montana has exposures of the Hell Creek Formation (mostly uppermost Cretaceous) and Tullock Member of the Fort Union Formation (lower Paleocene).In this region, the Hell Creek Formation is ∼90 m thick and is characterized by drab-colored (e.g., gray, purple) mudstones and sandstones, typically with gradational contacts between meter-scale strata (Archibald, 1982;Fastovsky, 1987;Hartman et al., 2014).The Tullock Member is ∼80 m thick and is characterized by brightcolored (e.g., yellow, tan) mudstones and sandstones with sharp, clearly defined contacts between decimeter-scale strata and by common and laterally continuous low-grade coal or lignite layers (coals, hereafter) (Archibald et al., 1982;Fastovsky, 1987;Fastovsky & Bercovici, 2016).The contact between the Hell Creek and Fort Union formations is defined as the base of the stratigraphically lowest, laterally continuous coal layer, which is termed the 'Z coal' (Brown, 1952;Calvert, 1912;Collier & Knechtel, 1939;Moore et al., 2014; however, note the presence of the Null Coal [or Tonstein Lignite] in the Hell Creek Formation, ∼25-30 m below the formational contact in some sections in eastern Garfield and western McCone counties [Lofgren, 1995;Rigby & Rigby, 1990;Sprain et al., 2015]).At many sites in central Garfield County, including our study areas, the Z coal immediately overlies the 'impact claystone' associated with the Chicxulub bolide impact, characterized by some combination of an iridium anomaly, shocked quartz, and glassy spherules (Alvarez et al., 1980;Clemens & Hartman, 2014;Moore et al., 2014).As such, in our study areas the Z coal is referred to as the Iridium Z (IrZ) coal, and the Hell Creek-Fort Union contact is considered roughly coincident Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-3) with the KPB (see Moore et al., 2014 for a review).In eastern Garfield County and western McCone County, the KPB impact layer has not been identified, and the uppermost ∼1 m of the Hell Creek Formation is interpreted as early Paleocene in age (see Lofgren [1995] and Smith et al. [2018] for discussion; see also Tobin et al. [2021] for possible boundary identification).

Morales 1
This locality occurs on a ∼3 m-tall, isolated hillock in the Hauso Flats area (Figs. 1, 2).Vertebrate microfossils and gastropod and bivalve fossil shell fragments are found as surface float on the hillock slope and the surrounding flat at ground level.The basal 175 cm of the hillock is a dark gray, massive, silty claystone with irregular, iron staining (orange-red) and carbonized plant debris (∼0.1-1.0 cm maximum dimension) (Fig. 3, Unit 1).Overlying this basal unit is a 30 cm-thick, greenish-gray, silty claystone characterized by iron staining, slickensides, sparse 5-20 cm deep root traces with minimal branching, and sparse carbonized plant debris (∼0.1-0.5 cm in maximum dimension)-all of these features are characteristic of the hydromorphic paleosols found in other nearby sections of the Hell Creek Formation (Fig. 3, Unit 2; Fastovsky and McSweeney, 1987).Vertebrate microfossils are hosted within a generally subangular, very-fine-grained sandstone with rare fine to mediumsized grains; the sandstone forms an irregular contact with the underlying green-gray, silty claystone unit of the Hell Creek Formation.The vertebrate microfossil-bearing sandstone also incorporates lenses of bivalve/gastropod shell-hash, carbonized plant debris, and abundant, pebble-sized mud rip-up clasts.This layer varies in thickness laterally from 1 cm to 10 cm (∼5 cm on average) and thins to the southwest, consistent with the dip direction of the overlying strata (Figs.2A 3, Unit 3).Above the fossilbearing unit is a 20 cm-thick, thinly laminated, silty claystone unit with carbonized plant debris (Fig. 3, Unit 4).Overlying Unit 4 is a 30 cm-thick layer alternating every ∼5 cm between a yellow, very-fine sandstone, similar to the fossil-bearing horizon though it lacks fossils, and a laminated, silty claystone (Fig. 3, Unit 5).These alternating coarse-to-fine lithologic units dip at an angle of ∼5°to the southwest, consistent with lateral accretion surfaces that characterize point bar deposits (Fig. 2A; Thomas et al., 1987;Rogers, 1998).The uppermost unit is a 45 cm-thick, black-dark gray, thinly laminated claystone that also dips at an angle of ∼5°to the southwest.This unit is characterized by abundant carbonized plant debris and becomes more fissile upsection.
Approximately 150 m north-northwest of Morales 1 across a flat is Iridium Hill (Figs. 1,2B), where Alvarez et al. (1983) and colleagues identified an iridium anomaly in a thin, reddish claystone (i.e., 'impact claystone') immediately underlying the Z coal (IrZ coal).The dark gray and greenish-gray, silty claystone units that underlie the IrZ coal and impact claystone at Iridium Hill can be traced laterally to the Morales 1 hillock (Fig. 2, Unit 2).The IrZ coal is not present in the hillock at Morales 1, although we would expect its presence based on a horizontal projection of the layer from Iridium Hill.We interpret the fossiliferous unit of Morales 1 and the overlying strata to represent channel-point-bar deposition during channel migration in the earliest Paleocene.In this model, the channel scoured through the IrZ coal and uppermost Hell Creek Formation at that location.This interpretation is consistent with previous stratigraphic interpretations of the strata near the KPB in the Hell Hollow and Hauso Flats areas (Archibald, 1982;Smit & van der Kaars, 1984), and to date no definitively Cretaceous-aspect fossils have been recovered from Morales 1.

Herpijunk Promontory
This vertebrate microfossil locality (Herpijunk, hereafter) occurs on a ledge, ∼5 m above the base of a thick exposure of upper Hell Creek and lower Tullock sediments approximately 700 m east-southeast of Morales 1 (Figs. 1, 2).Vertebrate microfossils weather out onto the ledge, with some fossils eroding down the hillside onto the sandy flat below (Fig. 4A).
The fossil-bearing unit at Herpijunk is a 15 cm-thick claypebble conglomerate that forms the base of a 1.5 m-thick sequence of alternating ∼10 cm-thick sandstone and ∼0.5-1 cmthick mudstone beds that dip ∼5°to the west, consistent with lateral accretion bedding (Figs. 4,5,.The fossilbearing horizon forms a sharp, irregular (likely erosional) contact with an underlying 20 cm-thick, massive, light greengray, silty-claystone that is rich with carbonized plant debris and root traces (Fig. 5, Unit 8).Above the basal fossil-bearing pebble conglomerate, the coarse-grained units fine upwards vertically from medium-grained sands to very fine sands and up-dip from medium-grained sands to sandy siltstones.The entire lateral accretion deposit is overlain by a 1.2 m section of thinly laminated, black (with rust staining) silty claystone (Fig. 5, Unit 12).For a full stratigraphic log of the Herpijunk locality see Fig. 5.
The Herpijunk locality is one point along the base of a laterally extensive channel deposit recognized by a lateral accretional surface in the outcrop across its extent (Fig. 4).Tracing the base of these deposits ∼65 m to the northeast, this unit eventually crosscuts an exposure of the IrZ coal (Fig. 4B)-which hosts a confirmed impact claystone and has yielded the highest iridium anomaly in the region (∼11.2ppb; Smit & van der Kaars, 1984)-and the lateral extension of the MCZ coal (see 'Geochronological framework for Hell Hollow and Hauso Flats areas' section below), before grading upwards into horizontal beds ∼2 m above the IrZ coal.On the basis of this sedimentological and stratigraphic evidence, we interpret the Herpijunk fossiliferous horizon as a lag or single-event deposit near the base of a point bar formed by an earliest Paleocene channel as it scoured through the IrZ and MCZ coals and into underlying Hell   Ickert et al. (2015).The Iridium Hill square marks the location of a KPB section with a so-called 'impact claystone' layer, which is characterized by an iridium anomaly that is associated with the bolide impact (Alvarez et al., 1980).The Nirvana square marks an additional location of the MCZ lateral extension identified by Ickert et al. (2015)  The inclined, lateral-accretion beds that characterize both Morales 1 and Herpijunk and crosscut the IrZ coal are present at lowermost Tullock exposures across the Hauso Flats area (along a nearly E-W transect) and can be traced to other vertebrate microfossil-bearing exposures geographically intermediate between those two localities.Given that (i) the depositional scenario for Herpijunk is nearly identical to that for Morales 1, (ii) both localities were deposited in the lowest portions of the Tullock Member, (iii) they are less than 1 km from each other, and (iv) we are able to trace the same point bar facies from Morales 1 to Herpijunk, we hypothesize that both vertebrate microfossil localities were formed in meanders of the same channel complex, which we term the Hauso Flats channel.

Carrie Padgett 6
This locality is at the base of a ∼8 m-high and ∼7 m-diameter grass-covered knob of fine-to-medium-grained, moderately well-indurated sandstone (Fig. 6); vertebrate microfossils are found as surface float near the base of this exposure and on the surrounding flat.The stratigraphically lowest unit exposed at Carrie Padgett 6 (Unit 1) is a 60 cm-thick, dark gray, silty claystone with popcorn weathering (Fig. 7, Unit 1).That same dark gray claystone can be physically traced to exposures ∼100 m northwest of Carrie Padgett 6, where it is overlain by the Z coal and underlain by green-gray, silty claystone with light orange mottling (∼5-10 cm in diameter) and abundant ∼20 cmdeep root traces (Fig. 6).The Z coal (see Tobin et al., 2021 for further discussion) has been traced laterally across the Hell Hollow area and hosts a bentonite that has been radioisotopically and geochemically correlated with the bentonite from the IrZ coal at Iridium Hill in the adjacent Hauso Flats area (Renne et al., 2013;Ickert et al., 2015;Sprain et al., 2015Sprain et al., , 2018)).The fossil-bearing unit is a 15 cm-thick, very-fine-to fine-grained sandstone that hosts abundant pebble-sized, mud rip-up clasts, sulfur nodules (likely secondary modification of organic precursors), and 0.5-2.0cm-diameter coal fragments (Fig. 7, Unit 2).This fossiliferous unit forms an abrupt, likely erosional contact with the dark gray claystone below.Overlying the fossiliferous unit is a white, fine-to-medium grained sandstone with 0.25-0.50m-thick trough cross beds.We interpret this unit, which is 8 m-thick and extends to the top of the section at this outcrop (Fig. 7, Unit 6), as a channel deposit.Because this channel deposit overlies Hell Creek muds and the IrZ coal is missing in this section, we interpret the Carrie Padgett 6 locality as a channel-lag or single-event deposit at the base of an early Paleocene channel that scoured through the IrZ coal and into underlying Hell Creek strata.This interpretation is consistent with previous studies (Archibald, 1982;Weaver et

Geochronological Framework for Hell Hollow and Hauso Flats Areas
The coal beds distributed throughout the Tullock Member typically contain multiple, sanidine-rich bentonites.Many of these bentonites have been sampled in the Hell Hollow and Hauso Flats areas and have yielded high-precision 40 Ar/ 39 Ar ages (Renne et al., 2013;Sprain et al., 2015Sprain et al., , 2018;;Swisher et al., 1993).The distinct Pb isotopic composition of the coal bentonites sometimes allows for even more precise correlations of potentially discontinuous coal layers than is possible from radioisotopic ages alone (Ickert et al., 2015).There has also been a significant amount of high-resolution paleomagnetic work in the Hell Hollow and Hauso Flats areas (LeCain et al., 2014;Sprain et al., 2018;Swisher et al., 1993).The impact claystone, characterized by an iridium anomaly and associated with the bolide impact, has been identified in the Hauso Flats area and potentially in the Hell Hollow area (see Tobin et al., 2021 for further discussion; Alvarez et al., 1980;Clemens & Hartman, 2014;Moore et al., 2014).We use the results of those studies and the observed stratigraphic relationships (Weaver et al., 2022a) of the relevant beds to constrain the depositional ages of the fossil localities in this study.
Three dated marker beds, or their lateral equivalents, are preserved in our study areas: the Iridium Z (IrZ), McGuire Creek Z (MCZ), and Hauso Flats Z (HFZ) coals.Those marker beds allow us to arrange the deposits hosting the mammalian fossil assemblages of interest (Morales 1, Herpijunk, and Carrie Padgett) in relative temporal order and to place them within the first ca.28-80 ka after the KPB (Fig. 8).A bentonite within the IrZ coal, stratigraphically a few centimeters above the impact claystone, has yielded a pooled mean 40 Ar/ 39 Ar age of 66.052 ± 0.008/0.043Ma, which is treated as the age of the KPB (Renne et al., 2013;Sprain et al., 2018;Swisher et al., 1993).The MCZ coal, which was first identified ∼75 km to the east of our study areas in the McGuire Creek area of McCone County (Lofgren, 1995), hosts a bentonite that has yielded a slightly younger (28 ± 18 ka) pooled mean age of 66.024 ± 0.014/0.044Ma (Sprain et al., 2018).The MCZ bentonite from McCone County has been correlated to exposures in Garfield County based on a distinct Pb isotope signature that was detected by Ickert and colleagues (2015) in a carbonaceous-shale-hosted bentonite exposed in and around the Hauso Flats area; we call this the 'MCZ lateral extension' (see Fig. 1, square symbols).The highest relevant marker bed is the HFZ coal.This ∼1.5-2 m-thick coal is exposed in both the Hauso Flats and Hell Hollow areas (among other places) and can be visually traced between those two study areas (Figs. 2,4,6).Bentonites hosted within the HFZ coal have been correlated, both geochemically and radioisotopically, across areas where it is exposed and have yielded an age of 65.973 ± 0.020/0.047Ma (Fendley et al., 2019;Ickert et al., 2015;Sprain et al., 2015Sprain et al., , 2018)).

Depositional Age of the Vertebrate Microfossil Localities in the Hell Hollow and Hauso Flats Areas
We interpret all three of the focal localities in this study as hosted within channel deposits that scoured through, and are thus geologically younger than, the IrZ coal (and by extension the KPB).At the Herpijunk locality section, the Hauso Flats channel also scoured through a stratigraphically higher, ∼10 cm-thick carbonaceous shale that can be laterally traced to the MCZ lateral extension at the Nirvana section ∼830 m to the east-southeast (Ickert et al., 2015;Sprain et al., 2018;Figs. 3, 8).Given that the top of the Hauso Flats channel lies only ∼2 m above the IrZ and ∼10-20 cm above the MCZ lateral extension, we interpret the deposition of the Herpijunk fossiliferous horizon to have been shortly after the deposition of the MCZ coal.Thus, we bracket the age of the deposition of the Herpijunk FIGURE 4. A, a south-facing view of the outcrop containing the Herpijunk Promontory vertebrate microfossil locality (white star); the ∼6.4 m-thick sandstone unit that is ∼5.5 m above the locality is interpreted as the Carrie Padgett channel complex; the Herpijunk microfossil locality is within the Hauso Flats channel deposit that crosscuts both the MCZ and IrZ coals and sits below a sandstone bed that is lithologically nearly identical to the sandstone characteristic of the Carrie Padgett channel deposit (Weaver et al., 2022a).B, close-up, east-facing view of the section highlighted with a box in A, showing the Hauso Flats channel deposit that hosts the Herpijunk vertebrate microfossil locality scouring the MCZ coal; although not shown, the Hauso Flats channel also crosscuts the IrZ coal to the right.See pickaxe in the foreground for scale.Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-8) fossiliferous horizon between the ages of the MCZ and HFZ coals but recognize that it is likely towards the older end of that age bracket, as suggested by the stratigraphic proximity of the top of the Hauso Flats channel and the MCZ coal (Fig. 8; 66.024 ± 0.014/0.044Ma and 65.973 ± 0.020/0.047,respectively; Sprain et al., 2018).The Morales 1 locality section does not preserve the MCZ lateral extension, presumably because there are not enough overlying strata preserved on the hillock or because the Hauso Flats channel scoured through the MCZ lateral extension locally.Nevertheless, because we interpret this vertebrate microfossil concentration, like Herpijunk, as having been deposited in the Hauso Flats channel complex, we also bracket the age of the deposition of the Morales 1 fossiliferous horizon between the ages of the MCZ and HFZ coals and likely towards the older end of that age bracket (Fig. 8; 66.024 ± 0.014/0.044Ma and 65.973 ± 0.020/0.047,respectively).The Carrie Padgett 6 locality is at the base of the much thicker Carrie Padgett channel complex (Archibald, 1982;Weaver et al., 2022a), the top of which is ∼8 m above the IrZ coal (and thus stratigraphically above the top of the Hauso Flats channel deposits).An MCZ-equivalent coal or carbonaceous shale has not yet been identified in the Hell Hollow area.However, at the Herpijunk locality section in the Hauso Flats area, there is a ∼6.4 m-thick sandstone bed that is lithologically nearly identical to the Carrie Padgett sandstone (as described in Weaver et al., 2022a), the base of which is ∼4 m above the MCZ-equivalent carbonaceous shale and the Hauso Flats channel (Fig. 5).This extension of the Carrie Padgett channel in the Hauso Flats area does not scour the MCZ lateral extension, which implies that either (1) the Carrie Padgett channel deposit in Hauso Flats is younger than that exposed in Hell Hollow or (2) the Carrie Padgett channel did not scour as deeply in Hauso Flats as it did in Hell Hollow.In either case, we interpret the Carrie Padgett channel as having been formed after deposition of the Hauso Flats channel, MCZ coal, and IrZ coal given its stratigraphic position above those three units.Previous work has also demonstrated that the Carrie Padgett channel crosscut, and was thus deposited after, the Hell Hollow channel (Weaver et al., 2022a), which is the sedimentary unit that hosts the wellsampled Worm Coulee 1 vertebrate microfossil locality  (Archibald, 1982;Wilson, 2014).Nonetheless, in the Hell Hollow area, the Hell Hollow channel deposit, host to the Worm Coulee 1 local fauna, spans the same stratigraphic interval as the Carrie Padgett channel, and therefore it too likely formed after the deposition of the MCZ coal and Hauso Flats channel (Weaver et al., 2022a; but see Archibald, 1982 for a different interpretation).We interpret the depositional age of the Carrie Padgett fossiliferous horizon as younger than the Hauso Flats channel but still bracketed between the ages of the MCZ and HFZ coals (Fig. 8; 66.024 ± 0.014/0.044Ma and 65.973 ± 0.020/0.047,respectively; Sprain et al., 2018).
Taken together, all the localities in our study and some of those previously studied from these areas (Archibald, 1982;Wilson, 2014) were deposited between 28 ka and 80 ka after the KPB.We can arrange them into a temporal succession from oldest to youngest: (1) Herpijunk and Morales 1, (2) Worm Coulee 1, and (3) Carrie Padgett (Weaver et al., 2022a;Fig. 8).Note that the fossil concentrations associated with channel deposits may have been sourced from any of the strata (and possibly multiple strata) within the stratigraphic interval of the channel deposit (Rogers & Brady, 2010;Weaver et al., 2022a).Thus, the temporal constraints of the fossil assemblages may differ from the temporal constraints of the channel deposits discussed here.Because the paleochannels associated with our fossil concentrations scoured into underlying Hell Creek strata, although not deeply (0.5-1 m), latest Cretaceous fossils might have been incorporated into our assemblages; however, thus far we have not documented any definitively Cretaceous-aspect fossils in them to indicate reworking of vertebrate remains from below the KPB.

Other Pu1 Western Interior Localities
From adjacent western McCone County come two additional points of faunal comparison: the Z-Line local fauna (from UCMP locs.V84193 and V84194) and the Luck O Hutch (LOH, hereafter) local fauna (from UCMP loc.V88036).The taxonomic composition and diversity of both local faunas have previously been described in detail (Lofgren, 1995;Smith et al., 2018).Their localities are near each other in the McGuire Creek area and have been integrated into their local chronostratigraphic framework (Sprain et al., 2015(Sprain et al., , 2018;;Smith et al., 2018).The KPB impact claystone has not been identified in McCone County, but a carbonaceous shale that underlies the Z-Line localities is interpreted as the KPB on the basis of vertebrate biochronology and carbon isotope stratigraphy (Arens et al., 2014; but see Tobin et al., 2021 for an alternative view on identification of the KPB using carbon isotopes).The Z-Line localities are thus bracketed between the KPB and the MCZ coal (Fig. 8; Smith et al., 2018) and are the oldest localities considered in this study.The LOH locality is stratigraphically above the MCZ coal and below the C29r/C29n boundary (65.724 ± 0.013/0.044Ma; Sprain et al., 2018).Thus, it is younger than the Z-Line localities, but it is not yet possible to definitively determine the age of LOH relative to the Worm Coulee 1 and Carrie Padgett localities-their corresponding bracketed temporal intervals overlap substantially (Fig. 8; Smith et al., 2018).Nevertheless, because the channel deposits, which host the localities, only slightly overlap in stratigraphic position (Fig. 8), we suggest that the LOH local fauna is probably younger than the Hell Hollow and Carrie Padgett local faunas.
More geographically distant points of faunal comparison come from the Denver Basin's Gars Galore and Littleton local faunas (DMNH loc.2560 and UCM locs.77267 and 77283, respectively).
Gars Galore yields a Pu1 local fauna that is from 9 m above the locally identified KPB (Dahlberg et al., 2016).The estimated age of the Gars Galore locality is ca.65.893 Ma or ∼128 ka after the KPB, based on the median sediment accumulation rate derived from the new age model in Clyde et al. (2016); this age is within the age range of LOH in the Hell Creek region.The Littleton local fauna has been interpreted as 100-300 ka after the KPB (Lyson et al., 2019), which is younger than all the Hell Creek region local faunas discussed in this study, except possibly LOH (Smith et al., 2018).

Raw and Subsampled Richness
To compare taxonomic richness among the local faunas of interest, we calculated raw richness, rarefied richness, and shareholders quorum subsampling (SQS) at both the species (Tables S18, S19, Supplementary Data 1) and genus level (Fig. 9).Out of the newly described local faunas Herpijunk has the highest values (13 spp., 9 gen.; Fig. 9), similar to those from other Pu1 local faunas from both the Hell Creek region (Worm Coulee 1 and Luck O Hutch) and Denver Basin (Gars Galore; Fig. 9).Z-Line and Carrie Padgett have the lowest raw species-and generic-richness values (10 spp., 8 gen.and 10 spp., 7 gen., respectively; Fig. 9).The Littleton local fauna, which is likely the youngest of the Pu1 local faunas included here, has substantially higher richness values for all metrics at both the species and genus level (raw richness: 20 spp., 15 gen.; Fig. 9).The rarefied and SQS values of all the Pu1 local faunas, except the Littleton local fauna, are much lower compared with the values of the Lancian Flat Creek local fauna (28 spp., 21 gen.).

Evenness Metrics
We calculated Simpson's evenness (Fig. 10) and Pielou's evenness (Tables S19, S20, Supplementary Data 1) at the genus level to investigate changes in community structure of the local faunas, specifically the relative abundance structure.Higher values for these indices indicate a local fauna with a more even distribution of relative abundances across its constituent taxa; whereas lower values indicate a local fauna with an uneven distribution of relative abundances (i.e., predominated by a few taxa).The Lancian Flat Creek local fauna has high evenness (0.85 for Simpson's evenness), whereas the Pu1 local faunas' values have mostly lower evenness and fall into three groups: (1) the five local faunas deposited in the earliest Paleocene (<130 ka post-KPB, ca.65.893 Ma) except Worm Coulee l have low evenness (0.30-0.40 for Simpson's evenness; Fig. 10), significantly less than that for Flat Creek; (2) the LOH local fauna (ca.28-328 ka post-KPB) has moderate evenness (0.50 for Simpson's evenness; Fig. 10), although not significantly greater than the values in Group 1; and (3) the Worm Coulee 1 and the Littleton local faunas have higher evenness (0.75 and 0.89 for Simpson's evenness, respectively; Fig. 10), significantly greater than those for all the other Pu1 local faunas and, in the case of the Littleton local fauna (ca.100-300 ka post-KPB), statistically indistinguishable from that for the Lancian Flat Creek local fauna.Thus, the Pu1 local faunas had low evenness through the first 80-100 ka post-KPB except for the anomalously high evenness for Worm Coulee 1, and there FIGURE 8.A schematic of the stratigraphic relationships among Tullock Member channel deposits from the Hell Creek region discussed in this paper.The thickness of each channel deposit and each coal layer is to scale, but the breadth of each channel deposit is not.The ages of the coals are from Sprain et al. (2018) and are presented as the weighted mean age with uncertainties, excluding systematic sources.The entire represented section is in magnetochron C29r (Smith et al., 2018;Sprain et al., 2018).The diagonal hatching of the Hauso Flats channel represents the lateral accretion beds.The overlapping of the Carrie Padgett channel on the Hell Hollow channel represents the temporal sequence of the deposition of Hell Hollow channel sediments first, followed by an erosive regime, and then the deposition of the Carrie Padgett channel sands (Weaver et al., 2022a).The KPB carbonaceous shale is a ∼10 cm layer that coincides with a negative carbon isotope excursion that was interpreted by Arens et al. (2014) as marking the KPB.The stratigraphy bracketing Jack's channel differs from that depicted for the other sections because the upper bracket is not represented (see Smith et al. [2018] for a more detailed explanation).
Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-12) was an increase in evenness between 100 and 300 ka post-KPB, with higher values for both Luck O Hutch and the Littleton local fauna (non-significant and significant, respectively).Values for Pielou's evenness largely follow the same pattern (see Supplementary Information).

Faunal Dissimilarity Metrics
The Bray-Curtis (BC) dissimilarity values indicate how dissimilar two local faunas are in taxonomic composition and relative abundance structure.Higher values indicate a greater degree of dissimilarity between local faunas.The Littleton local fauna is the most dissimilar among our selected Pu1 local faunas (BC: 0.67-0.81at the genus level; Table 1).This pattern is not driven by inter-basin distance-the Littleton and Gars Galore local faunas are highly dissimilar from each other as well (0.77 at the genus level).Worm Coulee 1 is also highly dissimilar to the other assemblages (BC: 0.44-0.74;see Discussion for more detailed explanation).Comparisons among all other Pu1 local faunas yield low to moderate dissimilarity, indicating similar taxonomic composition and relative abundance structure (BC: 0.17-0.46;Table 1).
These results are graphically reflected in the non-metric multidimensional scaling (NMDS) analysis (Fig. 11).Coordinate 1 shows the high dissimilarity between the Littleton local fauna (Fig. 11, far right) and all other Pu1 local faunas (center).Coordinate 2 does not segregate among the other Pu1 local faunas; however, some genera that are found exclusively at Worm Coulee (Prodiacodon and Acheronodon) and Luck O Hutch (Prodiacodon and Periptychidae indet.)plot in the upper quadrant.Moreover, genera found exclusively at Morales 1 (Puercolestes) and Gars Galore (Maiorana) plot towards the lower margin.

Newly Described Local Faunas
We assign all three mammalian local faunas described here-Morales 1, Herpijunk, and Carrie Padgett-to the Pu1 interval zone of the Puercan NALMA based on the presence of characteristic taxa (Table 2, e.g., Stygimys kuszmauli, Procerberus formicarum, Thylacodon montanensis, Oxyprimus erikseni, Protungulatum donnae; Lofgren et al., 2004).Previous work also supports this designation (Archibald, 1982;Archibald et al., 1987;Wilson, 2014).The Pu1 interval zone constrains these local faunas to the first ca.300 ka of the Paleocene (Lofgren et al., 2004).Using the chronostratigraphic framework for the Hell Creek region, we further constrain the deposition of these mammalian fossil assemblages between the first ca.28 ka and 80 ka of the Paleocene; this offers the highest resolution view yet known of the initial phase of post-K-Pg mammalian recovery (Fig. 12).These local faunas resemble each other and previously described post-KPB 'disaster' local faunas including Z-Line and Worm Coulee 1 (Smith et al., 2018;Wilson, 2014).The taxonomic composition is fairly consistent among them, as shown by the low taxonomic dissimilarity values and clustering in the NMDS plot (Table 1, Fig. 11).Each of the three local faunas has low evenness, particularly at the genus level, which is typical of 'disaster' local faunas (Harries & Kauffman, 1990;Erwin, 1993;Wilson, 2014;Smith et al., 2018).This pattern is driven by the high relative abundance of the multituberculate genus Mesodma (> 70% of the individuals in each assemblage; Fig. 11, Table 2).Mesodma has been interpreted as a local survivor of the K-Pg mass extinction that proliferated in the aftermath with increases in relative abundance from ∼20-40% pre-K-Pg to ∼70-80% post-K-Pg (bloom taxon); its numerical abundance was perhaps due to some combination of its small body size, generalized diet, and reproductive strategy (Clemens, 2002;Weaver et al., 2022b;Wilson et al., 2012;Wilson, 2013Wilson, , 2014)).Two other taxa previously identified as post-KPB mammalian bloom taxa, the metatherian Thylacodon montanensis and the eutherian cimolestid Procerberus formicarum (Wilson,FIGURE 9. Genus-level richness estimates for select Hell Creek region (filled circles) and Denver Basin (open circles) mammalian assemblages.The vertical sequence of localities reflects the depositional age of each locality, which may differ from the age of the associated fossil assemblage.Morales 1 and Herpijunk are bracketed to reflect that they are from the same channel (the Hauso Flats channel) and in turn likely the same age.The Denver Basin locality ages are based on published ages (Dahlberg et al., 2016;Lyson et al., 2019), which we have re-calibrated to the age used for the KPB in the Hell Creek region.The ages of coals are in millions of years ago and reflect the weighted mean ages without uncertainties (see Sprain et al., 2018).A, raw richness values; B, rarefied richness values are shown with 95% confidence intervals (calculated using standard error); C, shareholder quorum subsampling (SQS) richness values calculated using Alroy's ( 2010 (Dahlberg et al., 2016;Lyson et al., 2019), which we have re-calibrated to the age used for the KPB in the Hell Creek region.The ages of coals are in millions of years ago and reflect the weighted mean ages without uncertainties (see Sprain et al., 2018).Note the relatively low and stable levels of evenness during the Pu1 with the exception of Worm Coulee 1 and Littleton local faunas.Abbreviation: Cret, Cretaceous.2013, 2014), occur in the three local faunas described here but in low relative abundances (<5% and 10%, respectively) inconsistent with their bloom taxon designation.Immigrant taxa, including the archaic ungulate genera Baioconodon, Protungulatum, Oxyprimus, and Mimatuta and the eucosmodontid multituberculate Stygimys kuszmauli, also occur in low relative abundances (each < 5% of individuals) like in other Pu1 'disaster' faunas (Lofgren, 1995;Smith et al., 2018;Wilson, 2014).The high relative abundance of the bloom taxon Mesodma, the high degree of similarity among the Morales 1, Herpijunk, and Carrie Padgett local faunas, and their low evenness suggest that they sample what has broadly been termed the disaster interval of the post-KPB biotic recovery (Fig. 13).
From Post-KPB Disaster to Recovery Phase in the Hell Creek Region The Worm Coulee 1 local fauna has long stood as the standardbearer for the post-KPB disaster phase of the recovery in the Hell Creek region (Archibald, 1982;Archibald et al., 1987;Clemens, 2002;Wilson, 2014).By comparison with the pre-KPB Lancian local faunas (e.g., Flat Creek 5), the disasterphase mammalian fauna has thus been characterized as having (i) low species richness (Fig. 9; Wilson, 2014), (ii) substantial taxonomic turnover (e.g., disappearance of ≥ 75% of Lancian taxa; Wilson, 2014), and (iii) highly uneven community structure (Fig. 10; Wilson, 2014), driven by a few local bloom taxa (Mesodma thompsoni, Procerberus formicarum, and Thylacodon montanensis) with high relative abundances and a substantial number of immigrant taxa (seven archaic ungulate species and Stygimys kuszmauli) with low relative abundances (Clemens, 2002;Clemens, 2010;Wilson, 2014).Now, with additional sampling and description of mammalian assemblages from the lowermost Paleocene, greater precision on the age of these assemblages, and quantitative analysis of the diversity of these and other assemblages from the first 300 ka post-KPB (= Pu1 interval zone), we can refine our model of the disaster and recovery phases of mammalian recovery in the Hell Creek region.We focus on the following Pu1 local faunas from the region: Morales 1, Herpijunk, and Carrie Padgett (this study) as well as Z-Line and Luck O Hutch (Lofgren, 1995;Smith et al., 2018) and Worm Coulee 1 (Archibald, 1982;Wilson, 2014).
Early Disaster Sub-Phase-The Z-Line local fauna from the McGuire Creek area is from the oldest deposits among the Hell Creek region Pu1 local faunas (Fig. 8).It is distinct from the Worm Coulee 1 local fauna as well as the other Pu1 local faunas.The Z-Line local fauna has some of the lowest richness values, although not significantly less, for all three metrics at the genus level (raw = 6 genera, rarefied = 4.8 genera, and SQS = 2.6 genera; Fig. 9).Among the immigrant taxa that characterize  2018) classified as 'dead clades walking' (Jablonski, 2002).Nonetheless, in our NMDS plot the local fauna does not segregate from other Pu1 local faunas, driven primarily by the abundance of Mesodma (Fig. 11).We hypothesize that these faunal differences are evolutionarily significant, reflecting an earlier stage of post-KPB recovery than do the other Pu1 local faunas.Several of the following hypotheses could explain some or all these faunal differences, but we contend that they are not as well supported by the available data: (i) sampling of the Z-Line local fauna is still relatively incomplete; (ii) there is a taphonomic bias against largerbodied archaic ungulates at Z-Line; and/or (iii) as the easternmost of the Hell Creek localities, Z-Line captures a distinct depositional environment.Indeed, the combined sample size of the Z-Line assemblages (N = 116) is considerably less than those of Morales 1 (N = 248) and Worm Coulee 1 (N = 920); nevertheless, we largely discount the sampling hypothesis (i) because the rarefaction curve for Z-Line (Smith et al., 2018:fig. 5) is nearly asymptotic compared with those for other local faunas with similar sample sizes, indicating that more sampling probably would not substantially alter the faunal composition.The size-bias hypothesis (ii) does not seem likely considering that Mimatuta sp., which is similar in size to the other archaic ungulates, and other larger vertebrate fossils (e.g., champsosaur vertebrae) have been recovered from the Z-Line localities.Addressing hypothesis (iii), the Z-Line localities are hosted in a fine-to-medium-grained channel deposit, which is lithologically similar to the depositional environments of all the other localities under consideration here, and the size and taxonomic makeup of their vertebrate microfossil assemblages are consistent with the other Pu1 localities reported here, suggesting they had similar taphonomic origins (Rogers et al., 2017;Rogers & Brady, 2010).Thus, although it is impossible to entirely reject that the distinctness of the Z-line local fauna simply reflects paleoenvironmental heterogeneity, those paleoenvironmental differences must have been subtle enough so as not to leave an appreciable impression on the sedimentology of the site.Therefore, our working hypothesis is that the Z-Line local fauna represents the earliest documented stage of post-KPB recovery, which we refer to as the 'early disaster sub-phase' and bracket between 66.052 ± 0.008/0.043and 66.024 ± 0.014/0.044Ma or the first 28 ka following the KPB (Sprain et al., 2018;Fig. 13).We characterize the early disaster sub-phase by the presence of multiple 'dead clades walking,' a high relative abundance of bloom taxa (specifically Mesodma), and very low richness and relative abundance of immigrant taxa (Fig. 13).However, another mammalian assemblage in the Hell Creek region, Constenius (UCMP loc.V96268), may sample the same temporal interval as Z-Line.TABLE 2. Mammalian faunal lists for the Pu1 local faunas in the Hell Creek region used in this study.The number of specimens of each taxon and the corresponding percent relative abundance (in parentheses) are presented for each local fauna.Note: The V-number below the locality name is the UCMP vertebrate locality number.Abbreviations: B, taxon designated as a bloom taxon; D, taxon designated as a 'dead clade walking'; I, taxon designated as an immigrant; ID, taxon designated as the product of in-situ diversification; * specimens whose identification has not been confirmed or has been questioned by previous work.
Emphasizing that this assemblage has not been fully examined and described, preliminary inspection by one of us (GPWM) suggests it has better representation of immigrant taxa than does Z-Line.Comprehensive study of the Constenius assemblage and its local stratigraphy as well as additional sampling at Z-Line will be critical to sharpening our view of the early disaster subphase and the possibility of spatial heterogeneity within the Hell Creek region.FIGURE 12.The number of genera in each faunal recovery group for each of the included local faunas.The vertical sequence of localities reflects the depositional age of each locality, which may differ from the age of the associated fossil assemblage.Morales 1 and Herpijunk are bracketed to reflect that they are from the same channel (the Hauso Flats channel) and in turn likely the same age.The Denver Basin locality ages are based on published ages (Dahlberg et al., 2016;Lyson et al., 2019), which we have re-calibrated to the age used for the KPB in the Hell Creek region.The ages of coals are in millions of years ago and reflect the weighted mean ages without uncertainties (see Sprain et al., 2018).We classified each taxon into one of four faunal recovery groups: 'dead clade walking' (DCW), bloom taxon, immigrant taxon, or a byproduct of in situ diversification (Table 2; see Methods and Supplemental Information for a more detailed explanation).Abbreviation: Cret., Cretaceous.
Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-17) Late Disaster Sub-Phase-The Morales 1, Herpijunk, Carrie Padgett, and Worm Coulee 1 local faunas are all from deposits between the MCZ and HFZ coals and are all interpreted to be younger than the Z-Line local fauna (Fig. 8; 66.024 ± 0.014/ 0.044 Ma and 65.973 ± 0.020/0.047,respectively).The genus Mesodma is the most common taxon in all four local faunas.Moreover, only one 'dead clade walking' is present (Cimexomys spp.), and at least three of the characteristic Pu1 immigrant taxa (Protungulatum donnae, Baioconodon spp., Mimatuta spp., Oxyprimus erikseni, and Stygimys kuszmauli) are present in low relative abundance in all the local faunas.The four local faunas also plot near each other in our NMDS plot, a pattern driven by similarities in taxonomic composition and relative abundances (Fig. 11).We interpret the faunal similarities among these four local faunas to mean that they represent the same phase of post-KPB recovery.We refer to it as the 'late disaster sub-phase' as it is further along the process of biotic recovery than the early disaster sub-phase represented by the Z-Line local fauna (Fig. 13).The late disaster sub-phase is characterized by the continued  12 for generic richness values for each local fauna).We classified each taxon into one of four faunal recovery groups: 'dead clade walking' (DCW), bloom taxon, immigrant taxon, or a byproduct of in situ diversification (Table 2; see Methods and Supplemental Information for a more detailed explanation).For each sub-phase, the relative abundance of individuals of each faunal recovery group is represented by the width of its corresponding-colored bar (green = dead clades walking, yellow = bloom taxa, blue = immigrant taxa, purple = in-situ diversification).Note that the relative abundance of bloom taxa decreased while the relative abundance of immigrants increased from the early disaster through late recovery.Taxonomic richness also increased from early recovery into late recovery with the onset of in situ diversification.Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-18) high relative abundance of bloom taxa, reduction in the number of 'dead clades walking,' and the first appearance of a more taxonomically rich assemblage of immigrant taxa.
Although we assign the Worm Coulee 1 local fauna to the late disaster sub-phase, it differs from other late-disaster local faunas in several ways.The bloom taxa Thylacodon montanensis and Procerberus formicarum (Fig. 11) have higher relative abundance in Worm Coulee 1 (18% and 32%, respectively; Wilson, 2014) than in Morales 1, Herpijunk, and Carrie Padgett (<10% for both taxa).Also, the Worm Coulee 1 local fauna records the presence of taxa that are otherwise only known from younger deposits, including an undescribed periptychid (Periptychidae indet.)and the multituberculates Catopsalis alexanderi and Acheronodon garbani (but see Lofgren 1995 for other possible early immigration events of Catopsalis alexanderi).Several hypotheses could explain some or all of these faunal differences: (i) Worm Coulee 1 has a taphonomic bias towards smaller-bodied mammalian taxa; (ii) the Hell Hollow channel, which hosts the Worm Coulee 1 assemblage, incorporated fossils from across multiple heterogeneous paleoenvironments; (iii) Worm Coulee 1 potentially sampled a younger vertebrate microfossil source than Morales 1, Herpijunk, and Carrie Padgett 6; and (iv) the differences are an artifact of differences in sampling intensity.We reject the size-sorting hypothesis (i) based on the presence of a broad size range of fossils from mammalian and non-mammalian taxa.The high relative abundance of bloom taxa suggests that the Worm Coulee 1 local fauna is of similar age to the other late disaster local faunas (Fig. 13).However, the early appearances of younger aspect taxa (Periptychidae indet., Catopsalis alexanderi, and Acheronodon garbani) and differences in the most abundant bloom taxa suggest a potentially younger age that prevents us from fully rejecting hypothesis (iii).We are also unable to reject hypotheses (ii) and (iv).Worm Coulee 1 is within a more laterally expansive channel complex than Morales 1, Herpijunk, and Carrie Padgett, which are all found within relatively narrow meander belts (Weaver et al., 2022a).If the meander belt that makes up the Hell Hollow channel deposit reworked underlying sediments from a broader geographic area, then it is possible that the difference in faunal composition is the result of Worm Coulee 1 hosting fossil concentrations accumulated from a wider array of different paleoenvironments (Hypothesis (ii); Rogers & Brady, 2010;Weaver et al., 2022a).This could potentially explain the differences in abundance of bloom taxa (Thylacodon montanensis and Procerberus formicarum) and the early appearances of younger aspect taxa that potentially resided in more fringe environments not sampled by other channels.Moreover, Worm Coulee 1 has the largest sample size (N = 920) of all the Hell Creek region Pu1 localities, potentially more closely approximating the true relative abundance structure of Pu1 local faunas and the first appearances of younger-aspect taxa (Hypothesis (iv)).Despite these differences, we still attribute Worm Coulee 1 to the late-disaster subphase due to the high relative abundance of bloom taxa, the presence of 'dead clades walking,' and overall faunal similarity with other late-disaster local faunas (Fig. 11).Continued collecting and taphonomic study of Morales 1, Herpijunk, and Carrie Padgett and other spatially adjacent localities will allow more robust testing of these hypotheses.
Early Recovery Sub-Phase?-TheLuck O Hutch (LOH) local fauna is bracketed between 66.024 ± 0.014/0.044Ma and 65.724 ± 0.013/0.044Ma (ca.28-328 ka post-KPB; Sprain et al., 2018).These temporal brackets overlap those of the Carrie Padgett and Worm Coulee 1 local faunas, but we interpret LOH as likely younger based on the depth of the channel deposit (see 'Other Pu1 Western Interior localities' and Fig. 8).We must also acknowledge that the low sample size of LOH (55 specimens) reduces our ability to fully characterize the early recovery subphase.Nevertheless, like the other Pu1 local faunas, LOH has a high relative abundance of the bloom taxon Mesodma (Smith et al., 2018) and, in turn, clusters with the late-disaster subphase local faunas in the NMDS plot (Fig. 11).Of the other recognized bloom taxa, Thylacodon montanensis is absent and Procerberus formicarum is very rare (2% relative abundance) in the LOH local fauna.Additionally, the evenness of the LOH local fauna is greater than that of the other Pu1 local faunas in the region, except for Worm Coulee 1, although this difference is only statistically significant relative to Morales 1 (p = 0.006; Fig. 10).Younger-aspect taxa, such as Periptychidae indet.(not Mimatuta spp.) and ?Prodiacodon crustulum, occur in the LOH local fauna as well.Accordingly, Smith et al. (2018) proposed that LOH represents a later stage of post-KPB recovery than the other Pu1 local faunas, the 'early recovery sub-phase,' which is characterized by lower relative abundance of bloom taxa and higher evenness (Kauffman & Harries, 1996;Erwin, 1998;Smith et al., 2018).The Harley's Point locality (UCMP V77087) from Garfield County temporally overlaps the LOH locality, bracketed between 65.946 Ma and 65.912 Ma (ca.105-139 ka post-KPB; Wilson Mantilla et al., 2021).Although the Harley's Point local fauna has not been fully described, the appearance of two purgatoriid plesiadapiform species suggests that it might also represent the early recovery sub-phase (Wilson Mantilla et al., 2021).Likewise, the Coke's Clemmys locality (UCMP V88046) from McCone County also temporally overlaps the LOH locality, bracketed between 65.802 ± 0.116/ 0.125 and 65.724 ± 0.013/0.044Ma (ca.250-328 ka post-KPB; Sprain et al., 2018;Smith et al., 2018).Similar to LOH, Coke's Clemmys sample size is small (58 specimens) but the appearance of the purgatoriid plesiadapiform species Purgatorius cf.P. coracis suggests it also represents the early recovery subphase (Smith et al., 2018).Formal descriptions of the Harley's Point local fauna and additional sampling at LOH and Coke's Clemmys will allow us to better understand and provide a formal designation for the early recovery sub-phase.Until then, we refer to the LOH local fauna as tentatively representing the early recovery sub-phase due to the presence of the younger aspect taxa Periptychidae indet.and ?Prodiacodon crustulum, along with the fact that it was likely deposited after the late disaster sub-phases localities (Smith et al., 2018).
Late Recovery Sub-Phase-In the Hell Creek region, there is a 231-ka temporal gap between the well-documented local faunas of the disaster phase (e.g., Worm Coulee 1) that are less than 80 ka post-KPB and the previously studied local faunas of the recovered phase (e.g., Garbani local fauna; Weaver et al., 2022a) that are 512-934 ka post-KPB.This gap is partly filled by the tentative early-recovery local faunas such as LOH, Coke's Clemmys, and Harley's Point, although these local faunas would benefit from additional sampling or in the case of Harley's Point formal study.We suspect that within the rest of that gap lies what we call the 'late recovery sub-phase,' which represents the faunal transition from the early recovery subphase to the recovered phase (more detailed explanation in forthcoming section).We infer that this sub-phase would be characterized by the onset of in situ diversification, absent in the LOH local fauna, along with continued immigration.Through ongoing targeted fossil prospecting, geochronological sampling, and description of existing fossil collections, we aim to fill this gap in the Hell Creek region.As discussed below, we assign the Littleton local fauna from the Denver Basin to this late recovery sub-phase, although we acknowledge that compositional differences with older Hell Creek region local faunas could reflect differences in biogeography as well as recovery sub-phase (see Denver Basin Late Recovery Sub-Phase below).Moreover, it is possible that the localities described herein simply reflect 'disaster' and 'recovery' phases and that the taxonomic differences among the local faunas are driven by sampling of spatial heterogeneity.Nonetheless, these differences occur more or less in stratigraphic order and fit logically with the expectations of ecological recovery theory.We expect that this is not the case and that the local faunas represent different subphases of biotic recovery.

Testing the Generality of the Mammalian Recovery Model
Although the Hell Creek region has become a focal point for studies of the K-Pg mass extinction and biotic recovery, we recognize that it is a single location and may not accurately reflect patterns more broadly at the continental or global scale.However, there are few places in the world with the fossil record and geochronological framework to permit comparisons of high-resolution patterns of post-KPB mammalian recovery (see reviews of global records in Clemens, 2001Clemens, , 2010)).Thus, we investigate the generality of our model of biotic recovery in the Western Interior using the Denver Basin as a single point of comparison but hope that in the coming years we can expand those comparisons.The Denver Basin is located ∼920 km south of the Hell Creek region and hosts multiple well-described Pu1 local faunas within a comparable geochronological framework (e.g., Clyde et al., 2016;Fuentes et al., 2019;Lyson et al., 2019).We focus on the Gars Galore local fauna (Dahlberg et al., 2016) and the Littleton local fauna (Eberle, 2003;Middleton, 1982;Middleton & Dewar, 2016) because both have been interpreted as representing Pu1 faunas but differing in their relative ages.
Denver Basin Late Disaster Sub-Phase-The Gars Galore mammalian assemblage (DMNH loc.2560) consists of 108 specimens, of which 69 are identifiable to genus level or lower (Dahlberg et al., 2016).The age of the Gars Galore locality is ca.128 ka (ca.65.893 Ma) after the KPB, based on its stratigraphic height above the locally identified KPB and application of a median sediment accumulation rate (Clyde et al., 2016).This age falls within the tentative early recovery sub-phase of our new model; however, the Gars Galore local fauna differs substantially from the Luck O Hutch local fauna.The Gars Galore local fauna has lower species richness (9 spp.), lower evenness, and a very high relative abundance of the bloom taxon Mesodma (81%), all of which are more in line with early-disaster and late-disaster sub-phase local faunas from the Hell Creek region (Figs.9, 10).The Gars Galore local fauna is also distinct in that it lacks cimolestids (e.g., Procerberus formicarum or Puercolestes simpsoni), which are common in Pu1 local faunas from the Hell Creek region (Archibald, 1982;Wilson, 2014) and in other, younger Pu1 local faunas from the Denver Basin (Eberle, 2003;Middleton & Dewar, 2004).It is unlikely that their absence in the assemblage reflects a taphonomic size bias, given that taxa both smaller (e.g., Mesodma) and larger (e.g., Baioconodon) than the common Pu1 cimolestids are present.Moreover, specimens from Gars Galore were collected using screen-washing equipment and techniques similar to those used in the Hell Creek region (Dahlberg et al., 2016), and the depositional environment is a fluvial system similar to the localities in the Hell Creek region (Dahlberg et al., 2016).The Gars Galore local fauna also lacks the appearance of multiple younger-aspect taxa found at the tentative early recovery local faunas (e.g., Conacodon).Overall, the Gars Galore local fauna, despite its younger age, is more similar in composition to the Hell Creek region disaster phase local faunas than to the tentatively early recovery local fauna of Luck O Hutch (Figs. 9, 10).These differences with the earlyrecovery sub-phase local fauna from the Hell Creek region might reflect spatial heterogeneity of the biotic recovery process, uncertainty in the age estimate of Gars Galore, the incomplete sampling of the Gars Galore local fauna, or some combination of those factors.Until those factors are more fully evaluated, we provisionally assign the Gars Galore local fauna to the late disaster sub-phase.
Denver Basin Late Recovery Sub-Phase-The Littleton local fauna consists of assemblages from two localities, the Alexander locality (UCM loc.77267) and South Table Mountain (UCM loc.77283) (see Middleton & Dewar, 2004 for a discussion of whether these two localities should be combined into a single local fauna).These localities fall within magnetochron C29r, like other Pu1 localities in the Hell Creek region and Denver Basin (Eberle, 2003;Middleton & Dewar, 2004;Lyson et al., 2019).The age of the local fauna is otherwise poorly constrained but has been interpreted to be 100-300 ka after the KPB (Lyson et al., 2019), likely younger than Gars Galore and almost all the selected Hell Creek region local faunas, except perhaps Luck O Hutch (though we interpret the LOH as older than Littleton).We would thus expect the Littleton local fauna to represent either the early recovery or late recovery sub-phase of biotic recovery.Indeed, it preserves the bloom taxon Mesodma but in much lower relative abundance than any of the other Pu1 local faunas (<3%; Eberle, 2003;Middleton & Dewar, 2004).It has higher evenness and taxonomic richness (20 spp.) than the older Pu1 local faunas in the Denver Basin and Hell Creek region and is comparable in those metrics to the latest Cretaceous (Lancian) Flat Creek local fauna (Figs. 9, 10).The Littleton local fauna also has a greater species richness of multituberculates and archaic ungulates than the other Pu1 local faunas, and many of those taxa are more derived representatives of the mammalian families that characterize older Pu1 local faunas.For example, Conacodon of the Littleton local fauna is a more derived periptychid than either Mimatuta or Maiorana (Archibald et al., 1983), of which the latter two are present among most of the older Pu1 local faunas examined here; this implies either in situ diversification or an immigration event (Archibald, 1982(Archibald, , 1983)).We interpret this higher taxonomic richness and evenness, higher relative abundance of immigrant taxa, and byproducts of in situ diversification of the Littleton local fauna as indicators of a more advanced sub-phase of biotic recovery -the late recovery sub-phase (Fig. 13).We also acknowledge that, as is the case for Gars Galore, some differences with Hell Creek region local faunas might also reflect spatial heterogeneity in the biotic recovery process.Other studies have shown that by the early Paleocene there were patterns of north-south provincialism (e.g., Hovatter & Wilson, 2019;Smith et al., 2018;Weil, 1999) as well as diversity gradients related to topography (Johnson et al., 2003).For example, floras in the western portion of the Denver Basin closer to the foothills of the Rocky Mountains tend to be more taxonomically rich than their more eastern, coastally adjacent counterparts (Johnson et al., 2003).Thus, the greater taxonomic richness and evenness of the Littleton local fauna relative to more northern and eastern local faunas such as Luck O Hutch could relate to latitudinal and/or paleoenvironmental differences as well.The high richness of archaic ungulates in the Pu1 local fauna of the Great Divide Basin in Wyoming might also reflect a closer proximity to upland environments (Atteberry & Eberle, 2021;McComas & Eberle, 2016).That said, we underscore that these faunas have wide age brackets that might give a false sense of contemporaneity.Along with the need for additional sampling of these local faunas, higher precision geochronology is required to further resolve our understanding of the heterogeneity in these patterns.

CONCLUSION
Taken together, our new model (Fig. 13) proposes a more granular look at the disaster and recovery phases of post-KPB biotic recovery in mammalian communities.Analyses of the Morales 1, Herpijunk, and Carrie Padgett local faunas along with other recently described local faunas within a high-resolution chronostratigraphic framework allowed us to identify key Claytor et al.-Post-K-Pg mammalian biotic recovery (e2222777-20) sub-phases of biotic recovery: early disaster, late disaster, early recovery, and late recovery.We suggest that immigrants began colonizing vacated ecospace immediately after the KPB (early disaster) but did not fully establish themselves across the region until after the disappearance of most genera classified as 'dead clades walking' (late disaster; Fig. 12).The disaster phase in total spans at least the first 80 ka following the KPB and possibly up to 128 ka in the Denver Basin.One possible explanation for the length of the disaster phase includes environmental instability related to Deccan Trap volcanism, supported by evidence of a pulse in volcanic activity that roughly coincides with the interval (Schoene et al., 2019(Schoene et al., , 2021; but see Sprain et al., 2019 for a different interpretation).The length of instability is also supported by the paleobotanical record, which shows a return to pre-KPB levels of taxonomic richness in the region after 80 ka, suggesting a return of stable environmental conditions (Wilson Deibel, 2022).Moreover, the difference in the length of the disaster phase in the Hell Creek region and Denver Basin may be a result of spatial variability in the return to pre-KPB levels of floral richness (Lyson et al., 2019).
Our results suggest that the transition between the disaster phase and recovery phase occurred between 100 ka and 300 ka following the KPB and was mostly driven by increases in both generic richness and relative abundance of immigrant taxa, displacing bloom taxa and 'dead clades walking.'In situ diversification of taxa likely lagged, only ramping up after the first 300 ka following the KPB.Ultimately, our study underscores the importance of evaluating biotic recovery at finer temporal and geographic scales such that local, ecologically relevant patterns of interest are discernible.This was highlighted by the comparisons with the Denver Basin local faunas, which suggest biotic recovery was not spatially uniform.Comparisons with other regions within the Western Interior, including the Hanna and Great Divide Basins of Wyoming, could further test the generality of the model and explore how differences in regional landscapes, climate, or other environmental variables (e.g., vegetation) might have influenced the timing of post-KPB mammalian recovery and give us a more complete view of what led to the mammalian-dominated terrestrial ecosystems that characterize the modern world.
Abbreviations DMNH, Denver Museum of Nature & Science, Denver, CO, U.S.A.; UCM, University of Colorado Museum, Boulder, CO, U.S.A.; UCMP, the University of California Museum of Paleontology, Berkeley, CA, U.S.A.; UW, the University of Washington, Seattle, WA, U.S.A.; UWBM, the University of Washington Burke Museum of Natural History and Culture, Seattle, WA, U.S.A.

FIGURE 1 .
FIGURE 1. A, map of the Hell Creek region in Garfield and McCone counties, Montana, U.S.A. (modified from Weaver et al., 2022a); both Z-Line (UCMP V84193 and V84194) and Luck O Hutch (UCMP V88036) are located at McGuire Creek.B, satellite imagery of the study areas (Google Earth) showing the fossil localities Morales 1 (UCMP V77128), Herpijunk (UCMP V77129), Carrie Padgett 6 (UCMP V77124), and Worm Coulee 1 (UCMP V74111) (circles) and relevant geochronology localities (squares).The Ickert MCZ square marks the location of the MCZ lateral extension identified byIckert et al. (2015).The Iridium Hill square marks the location of a KPB section with a so-called 'impact claystone' layer, which is characterized by an iridium anomaly that is associated with the bolide impact(Alvarez et al., 1980).The Nirvana square marks an additional location of the MCZ lateral extension identified byIckert et al. (2015) and a location of a KPB section.The road running northwest between Nirvana and Carrie Padgett 6 separates the Hauso Flats and Hell Hollow study areas.Abbreviation: MCZ, McGuire Creek Z.
FIGURE 1. A, map of the Hell Creek region in Garfield and McCone counties, Montana, U.S.A. (modified from Weaver et al., 2022a); both Z-Line (UCMP V84193 and V84194) and Luck O Hutch (UCMP V88036) are located at McGuire Creek.B, satellite imagery of the study areas (Google Earth) showing the fossil localities Morales 1 (UCMP V77128), Herpijunk (UCMP V77129), Carrie Padgett 6 (UCMP V77124), and Worm Coulee 1 (UCMP V74111) (circles) and relevant geochronology localities (squares).The Ickert MCZ square marks the location of the MCZ lateral extension identified byIckert et al. (2015).The Iridium Hill square marks the location of a KPB section with a so-called 'impact claystone' layer, which is characterized by an iridium anomaly that is associated with the bolide impact(Alvarez et al., 1980).The Nirvana square marks an additional location of the MCZ lateral extension identified byIckert et al. (2015) and a location of a KPB section.The road running northwest between Nirvana and Carrie Padgett 6 separates the Hauso Flats and Hell Hollow study areas.Abbreviation: MCZ, McGuire Creek Z.
al., 2022a) that showed that the Carrie Padgett channel deposit crosscuts both the older, early Paleocene-age Hell Hollow channel deposit, which hosts the Worm Coulee 1 vertebrate microfossil locality (UCMP V74111), and the IrZ coal.The Carrie Padgett channel deposit hosts several other vertebrate microfossil localities, of which only the Carrie Padgett 1 locality (UCMP V79100) has thus far yielded mammalian fossils.Eight mammalian fossils have been identified and cataloged from the site (UCMP 294522, 294524, and 294525: Mesodma sp.; UCMP 294523 and 294527: Mesodma thompsoni; UCMP 294526 Mesodma formosa; UCMP 150007: Procerberus formicarum; and UCMP 150027: Mimatuta minuial), although we have not had the opportunity to confirm the taxonomic identifications of the latter two specimens.Because the Carrie Padgett 1 locality is in close geographic proximity to Carrie Padgett 6 locality (∼135 m northeast), is from the base of the same channel deposit, and is sedimentologically identical to the fossiliferous

FIGURE 2 .
FIGURE 2. A, a south-facing view of the hillock containing the Morales 1 (UCMP V77128; white star) vertebrate microfossil locality and surrounding stratigraphy (person for scale); the white arrows highlight the ∼5°southwest dip of the unit containing the Morales 1 vertebrate microfossils, which we interpret as the basal Tullock Member locally.B, a wide east-southeast view of the Hauso Flats area and stratigraphy relevant to our interpretations of the vertebrate microfossil locality and the Herpijunk vertebrate microfossil locality (UCMP V77129); Morales 1 and Iridium Hill are roughly 150 meters apart.Haziness in the photographs is due to smoke from nearby wildfires during the 2021 summer field season.

FIGURE 5 .
FIGURE 5. Stratigraphic section of the vertebrate microfossil locality Herpijunk (UCMP loc.V77129 = UWBM loc.C1153).The carrot (^) indicates Unit 10.The asterisk (*) indicates Unit 15.The top of the stratigraphic section does not indicate the top of the local stratigraphy but rather the end of stratigraphic logging.

FIGURE 6 .
FIGURE 6. A, an east-southeast-facing view of the outcrop containing the Carrie Padgett 6 vertebrate microfossil locality (white star), the Carrie Padgett channel, and the overlying HFZ coal.B, an east-facing view of the outcrop containing the Carrie Padgett 6 vertebrate microfossil locality (white star), the Carrie Padgett channel, and the overlying HFZ coal.

FIGURE 7 .
FIGURE 7. Stratigraphic section of the vertebrate microfossil locality Carrie Padgett 6 (UCMP loc.V77124 = UWBM loc.C1360).The asterisk (*) indicates Unit 5. Note the vertical scale break between 1 m and 8 m.The top of the stratigraphic section indicates the top of the local stratigraphy.
FIGURE 9. Genus-level richness estimates for select Hell Creek region (filled circles) and Denver Basin (open circles) mammalian assemblages.The vertical sequence of localities reflects the depositional age of each locality, which may differ from the age of the associated fossil assemblage.Morales 1 and Herpijunk are bracketed to reflect that they are from the same channel (the Hauso Flats channel) and in turn likely the same age.The Denver Basin locality ages are based on published ages(Dahlberg et al., 2016;Lyson et al., 2019), which we have re-calibrated to the age used for the KPB in the Hell Creek region.The ages of coals are in millions of years ago and reflect the weighted mean ages without uncertainties (seeSprain et al., 2018).A, raw richness values; B, rarefied richness values are shown with 95% confidence intervals (calculated using standard error); C, shareholder quorum subsampling (SQS) richness values calculated using Alroy's (2010) methodology.Note the relatively stable levels of richness across the Pu1 local faunas with the exception of the Littleton local fauna.Abbreviation: Cret, Cretaceous.

FIGURE 11 .
FIGURE 11.Plot of Coordinates 1 and 2 of the non-metric multidimensional scaling (NMDS) analysis of select Pu1 mammalian local faunas using the Bray-Curtis genus-level dissimilarity matrix.Circles indicate local faunas from the Hell Creek region (filled circles) and Denver Basin (open circles).Triangles indicate select genera present within the local faunas.Note that the Littleton local fauna plots farthest away from all localities along Coordinate 1, whereas the Coordinate 2 axis primarily separates singleton taxa on the margins from more common taxa in a central cluster.

FIGURE 13 .
FIGURE 13.Proposed model of post-KPB mammalian recovery patterns in the Western Interior of North America based on local faunas of the Hell Creek region and Denver Basin.Local faunas represented by numbers: 1, Flat Creek; 2, Z-Line; 3, Morales 1; 4, Herpijunk; 5, Carrie Padgett; 6, Worm Coulee 1; 7, Gars Galore; 8, Luck O Hutch; 9, Littleton.Black numbers represent local faunas from the Hell Creek region, gray numbers represent local faunas from the Denver Basin.Each local fauna was placed into a biotic recovery sub-phase based on its depositional age and similarity in community structure (see Discussion for more detailed explanations).The age range of each sub-phase is in thousands of years (ka) post-KPB and reflects the proposed age of the constituent local fauna.The asterisk for Gars Galore indicates there is a difference between the vertical placement on the timeline (100 ka) and its estimated age (128 ka).The generic richness for each sub-phase (open diamonds) is based on the corresponding value of the local fauna or the mean value for the five local faunas in the late disaster sub-phase (see Fig.12for generic richness values for each local fauna).We classified each taxon into one of four faunal recovery groups: 'dead clade walking' (DCW), bloom taxon, immigrant taxon, or a byproduct of in situ diversification (Table2; see Methods and Supplemental Information for a more detailed explanation).For each sub-phase, the relative abundance of individuals of each faunal recovery group is represented by the width of its corresponding-colored bar (green = dead clades walking, yellow = bloom taxa, blue = immigrant taxa, purple = in-situ diversification).Note that the relative abundance of bloom taxa decreased while the relative abundance of immigrants increased from the early disaster through late recovery.Taxonomic richness also increased from early recovery into late recovery with the onset of in situ diversification.

TABLE 1 .
Taxonomic dissimilarity values of select Pu1 mammalian assemblages.The values in the top panels were calculated using Bray Curtis metric.Values in the left panels were calculated at the genus level, and those in the right panel at the species level.Abbreviations: BC, Bray Curtis; CP, Carrie Padgett 1 and 6 combined; GG, Gars Galore; HJ, Herpijunk; LF, Littleton Fauna; LH, Luck O Hutch; M1, Morales 1; WC, Worm Coulee 1; ZL, Z-Line Q and Z-Line-E combined.