This research stemmed from a project begun in 1978 by Dr. Robert Moore, Department of Biology, Montana State Univeristy (MSU), to investigate arthropods associated with elk carcasses in the Lamar Valley of Yellowstone National Park. A large die-off of elk during the winter of 1977/78 resulted in an abundance of carcasses. The 1978 work consisted of pit-fall trap sampling of arthropods immediately adjacent to six elk carcasses and the corresponding sampling of arthropods 40 m distant from each of the carcasses.

The 1993 research attempted to replicate the 1978 project with certain methodological modifications. The purely descriptive focus of the 1978 project was expanded to become a hypothesis test.

Study Site:

Yellowstone National Park

Barbee and Varley (1985) describe Yellowstone National Park, including the surrounding wilderness (a total area of about 73 million km²), as likely being the largest relatively undisturbed natural ecosystem in the temperate zone. They state that YNP is an ecological island containing some of North America's highest concentrations of once continental-ranging wildlife. For these reasons, YNP is an ideal location to study a native-Nearctic, large-carcass, carrion community.

The park is located at the conjunction of Montana, Wyoming, and Idaho, USA. It was established in 1872, accepted as a Biosphere Reserve in 1976 and as a World Heritage Site in 1978 (Bishop 1993). The park is a forested, volcanic plateau and lies in the southern portion of the northern Rocky Mountains (44°08'-45°07'N., 109°10'-111°10'W.) containing elevations from 1586 to 3408 m (loc. cit.; Dirks & Martner 1982). The park covers an area of ca. 9000 km² (Romme & Knight 1982). Precipitation, mostly in the form of snow, ranges from 258 mm/yr at the north west entrance of the park to an estimated 2000 mm/yr in the southwest of the park (loc. cit.). Temperatures range from a mean -12 °C in January to 13 °C in July, measured at Yellowstone Lake (loc. cit.). However, temperatures below 0 °C have been measured in every month of the year within the park, even at the northwest entrance (elev.= 1586 m) (Despain 1990).

The area within the park is 80% forested and of the forested area, ca. 60% is composed of subalpine fir (Abies lasiocarpa) or lodgepole pine (Pinus contorta), (Despain 1990). Plant communities also include semi-arid sagebrush and grassland steppe (<1800 m elev.) and alpine tundra above the elevation of 2900 m, (Whitlock 1993). An estimated 1,100 species of vascular plants grow in the park (Despain 1990).

The most relevant biotic feature to this study is the park's ungulate biodiversity. Seven species are found within YNP: wapiti (Cervus elaphrus), mule deer (Odocoileus hemionus), bison (Bison bison), moose (Alces alces), bighorn sheep (Ovis canadensis), pronghorn (Antilocapra americana), and white-tailed deer (Odocoileus virginianus), (Craighead 1978).

However, the current density of bison within YNP may not reflect the historical (pre-settlement) density. Evidence suggests that pre-european settlement bison were rarely present in the Agropyron, intermountain Province along the eastern border of which YNP lies (Mack & Thompson 1982). Evidence for this hypothesis includes historical and fossil records, the morphology, ecology and physiology of the dominant vegetation type and is found in an analysis of the dung beetle communities of the regions (Howden 1966; loc. cit.).

Historical aspects of the geology of YNP relevant to this study include the volcanic activity which has resulted in massive eruptions every 600,000 years that must have periodically reset the area's biological diversity to near zero. The glacial ice caps that have periodically buried the area during the Pleistocene also must have greatly reduced the area's biodiversity (Despain 1990; Whitlock 1993). The current assemblage of communities is relatively young, having originated only 14,000 years ago, with the last deglaciation (loc. cit.). These factors, plus the relatively dry, high altitude and northern latitude influences on the park, combined with the effects of frequent burns (e.g the 1988 fires) support the prediction that YNP would have a lesser species diversity compared to similarly-sized and protected areas in lower latitudes and/or elevations that have experienced less frequent and/or severe local extinctions (Klopfer 1959; Klopfer & MacArthur 1960; Janzen 1967, Sanders 1968, 1969; Wilson 1974; Rohde 1978a,b).

Study Site: Lamar Valley

The research was conducted within the northern-most portion of the Lamar River Valley, east of Tower Junction, west of the Lamar River Canyon (45°55°N. 110°18-24°W.) (Fig.1). This region is unique within the park as a result of it receiving some of the park's lowest levels of precipitation, its general lack of forest-cover, its lower elevations (Despain 1990), and its association with winter herds of elk and bison.

This Northern Range of YNP has been the focus of much biological research. Since 1988, there have been at least three meetings of scientists conducting research and/or monitoring on Yellowstone's Northern Range. The reports of these meetings, summarizing work in progress and spanning the years 1988 to 1990 include over 40 different studies (Singer 1988, 1989; Anon. 1990). Not one of these investigations incorporated invertebrate taxa, a point addressed as a shortcoming within the 1988 report (Singer 1988). To my knowledge, this study is the first to focus on the terrestrial invertebrate fauna within Yellowstone's Northern Range.

Site Histories

For each site, I provide descriptions taken verbatim from the 1978 field notes and include location data for the 1993 sites (see also Figures 1 & 2). I provide habitat type, habitat type distribution within Yellowstone National Park, common plant species and soil condition for each site, as detailed by Despain (1990). Also included are data on the degree of burn that resulted from the 1988 fires within a 50 m square surrounding the Universal Transverse Mercator coordinates of each site. These burn data were provided by YNP/National Biological Survey Research Biologist D. G. Despain (in lit.).

It should be noted that the attempts to place the 1993 traps in the same locations as the 1978 traps were simply best-guesses based on the 1978 field note descriptions of the sites. Because of the somewhat imprecise 1978 descriptions (see Site Histories below) and the changes in site appearance that have occurred during the 15 year interim, I can only be confident that the 1993 traps were placed within ca. 50 m² of the 1978 locations. Tables 1 and 2 provide a synopsis of important site information.

Table 1. Sites listed by pit-fall trap type present during the 1978 and 1993 field seasons, in the Lamar Valley of Yellowstone National Park. X = 4 pitfall traps, blank = no traps placed.

site no.	1978 carcass	1978 control	1993 carcass	1993 control	1993 check
1	X	X			X
2	X	X	X	X
3	X***	X***			X
4	X	X	*	*	*
5	X	X			X
6	X	**		**
7			X	X

*: Site not present in 1993, see site 4. below.
**: Control pit-fall traps of site 6 were not replicated because the 1978 controls were those of site 1, due to the proximity of these two sites, see site 6 below.
***: 1978 site 3 had only 3 carcass and 3 control pit-fall traps.

Table 2. Site cover, characteristics, elevation and burn degree from the 1988 fires. Burn data is the percent unburned within a 50 m² area surrounding the UTM coordinates of the sites.

site no.	cover	characteristics	elev.	burn
1	open	S river slope, sagebrush	1867m	96% unburned
2	open	S slope, sagebrush	1940m	100% unburned
3	forest	Doug.-fir, mesic, nr. river	1830m	72% unburned
4	forest	N slope, Douglas-fir	1824m	no data*
5	mixed	sagebrush/ Douglas-fir	1851m	80% unburned
6	open	S river slope, sand/gravel	1867m	no data*
7	open	nr. lake, mesic, grassland	1864m	100% unburned

* : Sites not replicated in 1993, burn data irrelevant.

Figure 1 shows the study region and the seven sites within the northern Yellowstone-Lamar River Valleys. Figure 2 shows the same area indicating the successional recovery stage of habitats from the 1988 fires.

1978 field notes: "30 March. Elk is on a southeast facing slope abouit 120 meters above Yellowstone River and about 200 m. north of bridge over Y. River just east of Tower Junction. Much bare ground, some grass bunches, sagebrush. Warm, open location. Dry site."

1993 location data: USA: WYOMING: Yellowstone National Park. Lamar Valley, Bridge over Yellowstone River, east from Tower Junction. UTM: 49.74.200 N., 5.46.960 E., lat/long.: 44°55.33 N., 110°24.32 W. Down-slope direction 100° ESE, elevation: 1867 m, ca. 120 m (along slope) above river, Four pitfall traps within 2 m of two trees (Juniperus sp.).

Habitat type (Despain 1990): Artemisia tridentata, Festuca idahoensis. Distribution of this habitat type in YNP: A moist shrub community common in the Yellowstone-Lamar River Valleys, covering most of the terrain above 2080 m; occurring where sagebrush is present in the park. This type occupies 8% of YNP's non-forested area and is found in elevations up to 2900 m.

Common plant species (loc. cit.): Artemisia tridentata, Festuca idahoensis, Koeleria cristata, occasional Agropyron spicatum, Chrysothamnus nauseosus, Artemisia frigida and Geum triflorum.

Soil condition (loc. cit.): Of Andesitic origins.

1988 Burn data (Despain in lit.): 4% Canopy burn, 96% unburned.

Notes: In 1978 this site had carcass and control traps but in 1993 because no carcass was present, only check traps were placed.

1978 field notes:"30 March. 1 mile east along highway from Lamar River Bridge. Elk is on a south-facing slope in sagebrush-grassland habitat. It is in a dry, sun-exposed site."

1993 location data: USA: WYOMING: Yellowstone National Park. Lamar Valley, north mouth of the Lamar River Canyon, on the north side of the road. UTM: 49.74.400 N., 5.54.200 E., lat/long.: 44°55.40 N.,110°18.85 W. Downslope direction: due south, elevation: 1940 m, walk ca. 300 m, 23° N from the gate at the Slough Creek road turn-off.

Habitat type (Despain 1990): Artemisia tridentata (-up to 2 m in height), Festuca idahoensis, phase: Geranium viscosissimum. Distribution of this habitat type within YNP: This type occurs within the middle to upper elevations (up to 2900 m) of the Yellowstone-Lamar River Valleys in the northern portion of the park, occupying 20% of non-forested area within YNP.

Common plant species (loc.cit.): Moister than the Artemisia tridentata/ Festuca idahoensis [see site 1] habitat type; generally having a greater standing crop of grasses and forbs; also includes Danthonia intermedia, Bromus carinatus, Agropyron caninum, Carex raynoldsii, Geranium viscosissimum, Helianthella uniflora, Potentilla gracilis, Eriogonum umbellatum.

Soil condition (loc.cit.): Deep, moist, fertile soils of Andesitic origin.

1988 Burn data (Despain in lit.): 100% unburned.

Notes: In 1993 four pitfall traps were placed around the remains of an elk, ca. one to two weeks dead. The carcass consisted only of bones and tissue residues. Four control traps were placed 40 m (312° WNW) from the elk carcass.

1978 field notes:"6 April. Cow 1.5 years old near Yellowstone River (8 m from bank) on east side of the river near the high-water mark. It is at the base of a north slope with sagebrush and scattered Douglas Fir & Lodgepole Pine. The cow is wedged against the base of a Lodgepole Pine and near a down juniper. Cool site".

1993 location data: USA: WYOMING: Yellowstone National Park, Lamar Valley, Bridge over the Yellowstone River, east of Tower Junction. East side of the river, UTM: 49.74.200 N., 5.47.120 E., lat/long.: 44°55.30 N., 110°24.25 W. Elevation: 1830 m, (ca. 1m above river), base of north slope, level ground, tree cover.

Habitat type (Despain 1990): Pseudotsuga menzisii, Symphoricarpos albus. Distribution of this type in YNP: within the Yellowstone and Lamar River Valleys at lower to middle elevations (1830 m-2140 m) in warmer areas; along low slopes and benches; deep, moist soils with northern to eastern aspects, representing 3% of the area within the park.

Common plant species (loc.cit.): Populus tremuloides, Pinus contorta, Pseudotsuga menziesii, Symphoricarpos albus, Spiraea betulifolia, Mahonia repens, Prunus virginiana, Amelanchier alnifolia, Calamagrostis rubescens, Achillea millefolium, Smilacina racemosa.

Soil condition (loc.cit.): Glacial till derived from volcanic and sedimentary rocks including basalt, andesite, limestone and sandstone.

1988 Burn data (Despain, in lit.): 12% canopy burn, 12% mixed burn, 4% undifferentiated burn, 72% unburned.

Notes: This was the only forested site used in 1993. This site lacked a carcass in 1993, thus only check traps were placed. In 1978 only three pit-fall traps were used for the carcass and the control because of a tree against which the carcass lay.

1978 field notes: "6 April. Calf, just below old road on north facing slope about 200 m east of Yellowstone River bridge (nr. Tower Jct) on north side of highway. Moist, small stand of Douglas fir, roses & deep herbaceous vegetation".

Notes: This site was not replicated in 1993 because attempts to locate it failed. It seemed that the site had eroded into the river. However, the habitat type description from 1978 indicates it probably shared the habitat type of site 3.

Figure 1. Study site, Lamar Valley. The seven trapping sites are numbered.

 

Figure 2. Study site, Lamar Valley, the extent of the 1988 burn. Successional cover types are shown. (3)=postfire, young Pseudostuga menziesii, (5)=recently burned, Pinus contorta expected to colonize, (7)=nonforested, (8)=300+ year-old Pinus contorta, Picea engelmannii, Abies lasiocarpa, and whitebark pine. (16)=Pseudostuga menziesii/non-forested, (17)=Pseudostuga menziesii, (23)=aspen, (24)=evenly-aged Pseudostuga menziesii, (29)=dense, small diameter, Pinus contorta, (35)=150 to 300 year old Pinus contorta, closed-canopy, (40)=water.

1978 field notes: "6 April. Bull--resting underneath a single Douglas fir on a slightly sloping bench (north slope) about 80 m east of the Yellowstone River, just north of the highway bridge near Tower Jct. Relatively moist site."

1993 location data: USA: WYOMING: Yellowstone National Park, Lamar Valley, Bridge over the Yellowstone River, east of Tower Junction. 80 m east from the east side of the bridge, ca. 20 m north of road, UTM: 49.74.120 N., 5.50.200 E., lat/long.: 44°55.29 N., 110°24.07 W. Almost level bench, (old road), elevation: 1851 m.

Habitat type (Despain 1990): Same habitat type as site 1. However, this site differed from site 1 by its proximity to habitat type of site 3, its highly disturbed soil (old road bed) its greater number of trees, and its lack of sun-exposed slope. Site 5 appeared more moist than site 1.

1988 Burn data (Despain in lit.): 8% canopy burn, 12% mixed burn, 80% unburned.

Notes: This site lacked a carcass in 1993 so only check traps were placed.

1978 field notes: "2 June. Mature cow, died night of 5/30/78. 50 m north of elk #1 [site 1] in old hot spring (geyserite) eroded gully".

Notes: In 1978 the control traps of site 1 were used as controls for site 6 because of the proximity of the two carcasses. No traps were placed for site 6 in 1993 because this site lacked a carcass in 1993 and because the 1978 control traps for site 6 were those of site 1. This area had very little vegetation and was primarily loose sand/gravel. Site 6 was poorly represented in the data (Appendix B).

1993 location data: USA: WYOMING: Yellowstone National Park, Lamar Valley, east of Junction Butte, ca. 90 m from south-east corner of Trumpeter lake, (largest of two lakes). Elevation: 1864 m (6110'), UTM: 49.73.440 N., 5.50.200 E., lat/long.: 44°54.84 N., 110°21.96 W.

Habitat type (Despain 1990): The habitat type of site 7 was the same as that of site 2 except site 7 was more moist due to the nearby lake, a nearby stream and its location within a valley bottom. Also, fewer sage and more grasses were present.

1988 Burn data (Despain in lit.): 100% unburned.

Notes: This site had a bison carcass in 1993 that was ten days old when traps were placed, only bones, hair and tissue residues were present. Control traps were placed (340°), 40 m from the carcass in identical slope and vegetation aspects. This site was not used in 1978.

Carrion Source Taxa

The two species of vertebrates used as carrion sources in this study, elk (Cervus elaphus: Cervidae) and bison (Bison bison: Bovidae), are both mammals of the order Artiodactyla (Jones 1985). Elk once ranged across much of the northern hemisphere but experienced a severe range reduction during the late 1800s due to human activities (loc. cit.). Elk are gregarious and migratory. Individual weights range from 228 to 375 kg. (loc. cit.). A vast literature exists documenting aspects of this species' biology (e.g. Singer 1988; Samuel et al. 1991 and references therein).

Bison are gregarious animals that once ranged from Canada to Mexico although the greatest concentrations of these animals were found in the Bouteloua gracilis (mid-continental prairies) Province, east of the Rocky Mountains of North America (Daubenmire 1978; Mack & Thompson 1982). It has been estimated that prior to the arrival of Europeans there were ca. 70 million in the New World (Jones 1985). However, by 1889 fewer than 1000 individuals remained (loc. cit.). Females can reach masses of up to 500 kg and males up to 910 kg (loc. cit.). These two species' long histories in the Nearctic and their once large and dense populations suggest they are good candidates for an investigation of a native Nearctic, large-carcass, carrion community.

Study Taxa

The choice to limit this investigation to the order Coleoptera was made for practical and ecological reasons. To adequately investigate any invertebrate fauna, due to the difficulty of identifying most taxa, some degree of taxonomic specialization is required. Specialists in given taxonomic groups are frequently needed to identify species. However, the inaccessibility or lack of specialists for some groups combined with the difficulties and resources involved in obtaining the assistance of these specialists often handicaps efforts. For these reasons, I chose to restrict the investigation to an insect order that is fairly well-represented by specimens and within the taxonomic literature resources at Montana State University. In addition, the experience of Dr. Ivie, Dr. Gustafson, and LaDonna Ivie with beetle communities in Glacier National Park for the past 5 years provided a strong taxonomic, theoretical and methodological base for this project. The choice of Coleoptera also resulted from the ecological importance of this taxon. No animal order contains more described species than the order Coleoptera. Erwin (1982) estimated the beetles represent at least 40% of all Arthropod species. Wilson (1987) sites a figure of 290,000 described species for the Coleoptera and 1.03 million for all animals, thus, ca. one of every three described animal species is a beetle. Of all the described species of life on earth (1,413,000 spp.), one in every five is a beetle (loc. cit.).

In addition and perhaps related to the large number of species, the beetles appear to display the greatest trophic diversity of any animal order (M.A. Ivie pers. comm.). Unlike most animal orders whose species tend to specialize and be restricted to a certain trophic level (e.g. Lepidoptera within the 1° consumer level), beetles are found in virtually every trophic level (loc. cit.). Also, because beetles are holometabolous, for many species of beetle, two ecological niches are filled. Therefore, the "niche richness" of a beetle community equals up to twice the species richness. For these reasons, it is very likely that understanding the beetles of an ecosystem provides more information about that ecosystem than understanding any other single animal order. It has consequently been proposed that beetles are one of the best choices for potential terrestrial ecological indicators as a result of their trophic and species diversity (loc. cit.).

1978 Methods

Field

Between 05 May and 27 August, 1978, MSU researcher Robert E. Moore assisted by Clifford Youmans, Sharon Rose, Sue Fullerton, William Baker, and others sampled arthropods associated with elk carcasses. They used pit-fall traps to sample adjacent to and 40 m distant from six elk carcasses in the Lamar Valley. Their pit-fall traps consisted of 12 oz. Solo(R) plastic cups, with smaller Solo(R) coffee cups, minus their bottoms, suspended within and from the larger cups' rims. The sampling area of these traps measured 09 cm in diameter. Soapy water was placed in the cups as a killing agent. Generally, (see Table 1 and Site Histories) four traps were placed, aligned with compass directions, within 1 m of a carcass and a similar configuration of four control traps were placed 40 m distant from each carcass. Judging from photos taken of the traps and carcasses in 1978, the pit-fall traps' rims were not always flush with the ground, but were ca. 0-2 cm above the substrate. The 1978 trap periods were variable in duration (Table 3).

Table 3. Pit-fall trap periods for 1978, Lamar Valley, Yellowstone National Park.

trap period	no. days in trap period	trap open	collected
1	6	04 MAY	10 MAY
2	5	10 MAY	15 MAY
3	7	15 MAY	22 MAY
4	3	22 MAY	25 MAY
5	4	25 MAY	29 MAY
6	4	29 MAY	02 JUN
7	3	02 JUN	05 JUN
8	4	05 JUN	09 JUN
9	3	09 JUN	12 JUN
10	3	19 JUN	22 JUN*
11	3	26 JUN	29 JUN*
12	3	03 JUL	06 JUL*
13	3	10 JUL	13 JUL*
14	3	17 JUL	20 JUL*
15	3	27 JUL	30 JUL*
16	4	03 AUG	07 AUG*
17	3	10 AUG	13 AUG*
18	3	17 AUG	20 AUG*
19	3	24 AUG	27 AUG*
mean + SD :	3.784 + 1.087
total days:	70

* : Note sample periods are not consecutive, dates during which traps were inactive separate sample dates.

The samples were collected by pouring them from the pit-fall trap cups into glass vials or bottles, depending on the quantity of sample, in the field. Destroyed pit-falls or cups plugged with hair from the carcasses were frequently mentioned in the field notes. No consistent coding of the condition of each trap was used, so variation in trap catch quality could not be correlated with impaired trap function. The samples were stored under refrigeration prior to their sorting and transfer into 70% ethanol. Appendix B displays the number of intact samples recovered from the 1978 field season.

Lab

All the field samples were sorted during 1978 in the lab to remove non-arthropod materials. All arthropods of a given sample and a corresponding sample label were stored in 70% ethanol within rubber-stoppered, glass vials. These vials were stored in the Montana State Entomology Collection (MTEC) and ethanol levels were maintained during the 15 years between 1978 and 1993.

1993 Methods

Field

Between 13 May and 14 September, 1993, Michael Ivie, LaDonna Ivie, Derek Sikes, and Matthew Becker used pit-fall traps to sample adjacent to an elk and a bison carcass, 40 m distant and within 3 non-carcass sites that were used during 1978. The non-carcass sites were used to augment the information provided by the control traps at carcass sites. A low ungulate mortality coupled with extremely high vertebrate scavenger pressures (mainly coyotes, Canis latrans) limited the available carcasses in 1993. Table 4 compares the 1978 field methodology to the 1993.

Table 4. Comparison of experimental designs.

item	1978	1993
Traps	12 oz. Solo (R) beer cups w/ Solo (R) coffee cup minus bottom as funnel. Diameter of opening = 9 cm bottle.	2-liter pop bottles w/top inverted as funnel w/ 12 oz. Solo (R) beer cup as catchment inside. Diameter of opening = 11 cm
Trapping liquid	soapy water	propylene glycol (PG) (10% formalin)
Rain roofs	absent	absent
Collection schedule	19, 3-7 day (x=3.8 + 1.1 day) non- continuous intervals.	17, 7-9 day (x=7.3 + 0.6 day) continuous intervals.
Collection of sample	soapy water w/sample poured into vials/ jars & refrigerated	PG w/sample poured through filter cloths,cloths & refrigerated in Whirl-Pacs(R).
No. & Trap type	6 carcasses X 4 traps 5 controls X 4 traps =48 traps	2 carcasses X 4 traps 2 controls X 4 traps 3 checks X 4 traps =28 traps

The 1993 pit-fall traps were fashioned from 2-liter carbonated beverage bottles. The upper portions were cut cleanly off at the point where the bottle's straight sides curve inwards. These top portions were inverted to create funnels that fit snugly within the lower portions. The sampled surface area measured 11 cm in diameter. Twelve-ounce Solo(R) collecting cups were placed within each trap to catch all materials entering via the funnel opening. The collecting cups were used because they allowed removal of captured insects without disturbing the traps. The traps were sunk into the ground so that their rims were flush with the substrate. The cups within each trap were filled with ca. 150 ml of propylene glycol (10% formalin). The propylene glycol acted as a killing agent and preserved the captured insects. Formalin was added to deter vertebrates from drinking the liquid. Propylene glycol is an FDA-approved food additive and, if consumed, harmless to the protected vertebrate species in the area.

The traps were arranged in the same configuration as those of 1978. Four traps were clustered within 2 m of each other, forming a square with one trap for each of the cardinal compass directions. Also, as in 1978, four control traps were similarly configured 40 m away from carcass traps. Three trap categories were used in 1993, carcass traps, control traps and check traps. Check traps differed from control traps by lacking corresponding carcass traps.

The samples were collected an average of every 7.3 days (Table 5). Specimens and debris were removed from the propylene glycol by the use of filter cloths. A 15 cm² square piece of fabric-netting was placed over a hardware cloth frame and the preservative and specimen mixture was poured through the cloth thereby separating the materials from the preservative. The preservative was returned to the traps for reuse unless rain diluted, whereupon it was replaced. Each filter cloth and its sample was folded once and placed with a typed label into a Whirl-Pac(R) plastic bag. The use of these bags allowed the removal of most air from the bags, thus maintaining the sample completely in residual preservative.

Each trap was inspected and reset if erosion had occurred, to ensure the rim was flush with the substrate. A trap condition code was entered into the field notebook and onto the sample labels to record whether the trap showed evidence of having its efficiency impaired during the sample period. The following code was used: (1) trap in perfect condition, (2) trap physically perfect but flooded, (3) trap physically impaired (e.g entrance blocked by debris), (4) trap and sample destroyed (e.g. removed by a coyote). These codes allowed later evaluation of the quality of the data (Appendix F).

Table 5. Pit-fall trap periods during 1993, Lamar Valley, Yellowstone National Park.

trap period	no. days in trap period	open	collected
1	9	13 MAY	22 MAY
2	8	22 MAY	30 MAY
3	8	30 MAY	07 JUN
4	8	07 JUN	15 JUN
5	7	15 JUN	22 JUN
6	7	22 JUN	29 JUN
7	7	29 JUN	06 JUL
8	7	06 JUL	13 JUL
9	7	13 JUL	20 JUL
10	7	20 JUL	27 JUL
11	7	27 JUL	03 AUG
12	7	03 AUG	10 AUG
13	7	10 AUG	17 AUG
14	7	17 AUG	24 AUG
15	7	24 AUG	31 AUG
16	7	31 AUG	07 SEP
17	7	07 SEP	14 SEP
mean + SD =	7.294 + 0.571
total days =	124

Lab, Sample Sorting

Beetles were removed from the 1978 samples that had been stored in the Montana Entomology Collection and the 1993 field collections, and were stored with labels in 70% ethanol in a Nalgene(R) scintillation vial (volume=25 ml).

The 1993 samples were processed by M. Becker and myself. The filter cloths were rinsed under a gentle water stream in a metal testing sieve (mesh size: 250 µm, Tyler equivalent: 60 mesh) to remove dirt and specimens from the cloths. The specimens were then sorted in water within a sorting tray under a dissecting microscope. In addition to the beetles, Orthoptera and non-formicid Hymenoptera were saved.

Lab, Identifications and Specimen Counts

Some introduction is required regarding the practice of parataxonomy as defined by D. Janzen (1991). Generally, samples are first sorted to a high taxonomic level, such as order, then identified to morpho-species by a parataxonomist and finally identified to species by an appropriate taxonomist/systematist. The use of parataxonomy in this manner is becoming more common as the need grows for fast and inexpensive but thorough investigations of highly diverse faunas. The use of some degree of parataxonomy is almost required for large faunas, such as those of most insect communities. In practice, each time a new morpho-species was encountered, it was mounted and used as a voucher in a reference collection. One 8.5 x 3.5 cm unit tray (generally ca. 20 specimens) would be filled with mounted voucher specimens to allow a taxonomist/systematist to evaluate the variation included in the morpho species concept. More details are provided on the reference collection under Specimen Storage and Retrieval.

In generating the identifications and counts, I emptied each sample into a sorting tray, associated all the species with voucher morpho species in the reference collection and recorded the number of individuals of each. Most species were so processed under a dissecting scope while in ethanol, although some required mounting. For each sample, I recorded the corresponding data (see Data Handling below) by hand along with a code indicating the location of the specimens (mounted collection or ethanol collection).

Names of species were obtained by comparison with previously identified materials from the YNP reference collection, MSU's Glacier National Park collection or from the MTEC. In many cases, the taxonomic literature was consulted. Some specimens that could not be identified by simple visual comparison of morphology were mounted, labeled and put aside, with appropriate notes made that indicated the specimens existed but were pending identifications. These specimens were eventually determined by an appropriate taxonomist and the corresponding identities were recorded. To the extent possible, appropriate systematists checked my morpho species concepts, and corrections were made as necessary. In some cases, I reexamined samples to confirm previous determinations to assure accuracy in difficult cases.

Data Handling

The data were coded numerically for later analyses. Each family and species of beetle was coded by an identification number [e.g. 004-003 =family #004 (Carabidae); species #003 (Carabus taedatus Fabricius) (see Appendix A)]. Data were recorded by adding the number of specimens per species per sample to the species/family code (e.g. 004-003-015 = 15 specimens of Carabus taedatus). The data identifying each trap sample were recorded as a heading under which the species/number of individuals would be listed, one species per line.

The family-species list was continuously updated and refined whenever a new species was added to the data. The new species would be given the next available identification number within its family. Every number used, even if later synonymized, was retained on the family-species list to prevent future mistakes resulting from ID number errors.

The data were entered into a non-document Wordstar(R) file (ASCII format), one sample per line. Numerical codes were used to incorporate trap sample information, such as date, trap condition, site, etc., for each sample entry. Appendix F contains all the data within the computer file, with an explanation of the codes used.

When two or more species identification numbers were found to have been mistakenly employed for a single species (synonyms), all but one of the names associated with those numbers on the family-species list were replaced with a "-same as XXX-XXX". Prior to analysis, the data file was corrected to eliminate such multiple ID numbers.

When two or more species were confused as one, upon detection of the problem, one of the species would be given the next available number and all previously examined specimens would be found, inspected and reidentified. The previously recorded data would then be corrected.

Specimen Storage and Retrieval

Specimens were mounted and labeled on pins in standard museum form to establish a voucher-series, constituting a full 8.5x3.5 cm unit tray (ca. 20 specimens). Once a voucher-series of a given species had been mounted, subsequent specimens of that species were identified, counted and returned to 70% ethanol within their sample vial. This system required that no data exist that were not represented by specimens in either the mounted or unmounted collection. If specimens were lost, destroyed or otherwise unretrievable, the data associated were excluded from analyses. This was necessary because without the specimens to vouch for the data it was impossible to verify the data (typographic or other errors could not be found if there was no absolute correspondence between the voucher specimens and the data).

All specimens within the two collections (unmounted & mounted) were organized to provide easy retrieval. The mounted specimens were organized taxonomically. The unmounted specimens were kept within their trap sample vials which were organized by trap (order always N,S,W,E) within a site and by site (order always: 1978: site 1,2,3,4,5,6; 1993: site 1, 3, 5, 7(bison), 7(control), 2(elk), 2(control)) within a trapping period (date). This organization allowed easy retrieval of ethanol-stored specimens when taxonomic questions arose. When they did, the computer file was searched automatically to locate all samples that contained the species of interest. Voucher specimens of all species were deposited in the MTEC.

Data Verification

To minimize errors and establish a series of checks and balances, specific error-checking procedures were employed. There were three tiers in which information was stored and different error reduction techniques were required for each: (1) the voucher specimen collection, (2) the primary hand-written data and (3) the computer data file.

The errors that originated in the first tier usually stemmed from two sources: (1) the mistaken identification of conspecific specimens as different species and (2) the mistaken identification of non-conspecifics as belonging to the same species. The first type of error was often associated with species that displayed great intraspecific variation. The second type of error was usually associated with sibling-species that were morphologically similar enough to appear a single species (e.g Sphaeridium scarabaeoides and S. lunatum). Such mistakes could not be found without the help of experts or the use of the taxonomic literature.

To reduce the number of these errors, I frequently employed available taxonomic information. For example, the most common silphid species in the Lamar Valley, Thanatophilus lapponicus (Herbst), could be confused with a rare, alpine species, Thanatophilus coloradensis (Wickham). By being aware that the alpine species might occur in my samples I was able to screen every specimen of the common species to ensure proper identification.

Only by employing traditional taxonomy could identifications approach an error-free level. For groups in taxonomic chaos, such as the majority of the Aleocharine Staphylinidae, the literature is of little use, causing species identifications to be prohibitively difficult, even with specialist assistance.

The errors that originated in the second tier, the primary, hand-written data file, were of two kinds: (1) misidentifications and miscounts. The misidentifications were caught by a computerized cross reference of the identification numbers in the data with the numbers on the family-species list. Any numbers in the data that did not exist in the family-species list were flagged for verification. This method did not catch misidentified species ID numbers that corresponded to valid species numbers but caught all misidentifications that did not correspond to established species. Errors of this nature should have been infrequent enough and randomly distributed throughout the data to be considered "noise" that would not alter the outcome of analyses. Miscounts were not systematically checked for (to do so would require nothing less than recounting all the specimens), but instead a simple visual inspection of the abundances of each species would catch most order-of-magnitude miscounts (e.g. 10 specimens of a rare Coccinelid when only a single one was captured).

Another method that caught some errors involved the use of computer programs written by Dr. Gustafson to find data entry errors. These programs checked, for example, that each sample record had no species entered more than once. Once found, errors like this could be corrected by referring to the hand-written data and specimens.

Each sample entered into the computer file included a count of the number of species in the sample. The computer program summed the number of species entered in each sample to verify that the actual sum corresponded to the recorded sum. Any counts that did not match the number of species listed were flagged for verification.

The errors that originated in the third tier were of only one kind -- typographic errors, and these were easily found. Because the computer file was simply a copy of the primary hand written data file, a simple, visual cross-reference of each entry between the two data files caught the vast majority of typographic errors.

After all verification and error-corrections were completed, a final correspondence check was performed. This was done to ensure that the specimens of randomly chosen individuals in the computer file could be found and that they had the proper ID number and the same name as listed in the family species list.

Hypotheses Tested and Analyses

For the hypotheses tested, the null and alternative hypotheses state:

H0: The presence of a large vertebrate carcass does not affect the local occurrence, richness and abundance of Coleopteran taxa [as measured by pit-fall sampling], in comparison to an otherwise similar area lacking a large vertebrate carcass.

H1: The local occurrence, richness and abundance of Coleopteran taxa [as measured by pit-fall sampling], are affected by the presence of a large vertebrate carcass in comparison to an otherwise similar area lacking a large vertebrate carcass.

These hypotheses were tested by the comparison of species richness (ANOVA), species occurrence (similarity index), total beetle abundance (ANOVA), total species abundance (Chi²), species discovery curves, species diversity indices, temporal beetle abundance plots, species rank-abundance plots between carcass and control data sets and principal components analysis. To investigate these hypotheses, 1978 and 1993 data were both combined and analyzed separately. The 1993 check sites 1,3 and 5 were excluded from most analyses because they unbalanced the data (20, 1993 control/check trap units versus 8, 1993 carcass trap units, see also Table 1).

The Chi² analyses tested the null hypothesis that a given species would be equally likely to occur (or would occur in equal abundance) in a carcass versus a control trap, adjusted for the number of trap units per type. The ANOVAs compared the means between categories, such as the mean number of beetles in all pooled carcass traps versus the mean number in all pooled control traps. In these instances, the ANOVAs were equivalent to standard parametric t-tests.

The Chi², ANOVA, and PCA analyses were programed and run by Dr. Gustafson. The assistance of Dr. Gustafson was necessary because of the non-matrix format in which the data were stored. The data were not stored in a matrix format because the data were sparse, that is, a matrix format would have required a cell for every species, trap and date combination, most of which would be filled with zeros. These zeros were removed to condense the data, thereby making storage and handling more efficient. A single matrix for this data set would require 494,395 cells (445 species=rows x 1111 trap units=columns). Dr. Gustafson had previously written FORTRAN programs specifically designed to analyze data stored in this non-matrix format.

Identification of
Carcass-Associated Species

I used similar criteria to identify 1993 carcass-associated species with one major difference. Species had to be present more frequently in carcass traps at both carcass sites, but only one site needed to be significantly associated (whereas, because 1978 had six versus 1993's two carcass sites, 1978 species had to display significant associations at two or more sites). Two exceptions were made for species that are documented saprophages, Sphaeridium scarabaeoides LeConte (Smetana 1978) and Heterosilpha ramosa (Say) (Anderson 1985) which were included in the 1993 list in spite of their presence at only one site.

Alpha Diversity

The use of species diversity indices in ecology has received much criticism (Green 1979; Magurran 1988). To produce such indices, species richness and relative species abundance are combined into a single value. Because one cannot obtain the information that was used to create a diversity index from the index itself, the index therefore represents a reduction in information. Critics argue that species rank-abundance plots provide all the information that diversity indices simultaneously represent and hide (Green 1979; Magurran 1988).

However, despite the criticisms and the turmoil surrounding the use of such indices I see at least one advantage to their use. Although the comparison of species rank-abundance plots will provide more information than the comparison of species diversity indices, indices are much simpler means to store information for quick comparisons, particularly between different studies.

A problem that may outweigh this small advantage stems from the great number of incompatible indices that have waxed and waned in popularity. The erratic use of these indices has produced the opposite of the intended effect. One cannot compare species diversity between studies that employed different indices. Green (1979) reviews myriad arguments against the use of such indices. However, he also provides the few scenarios in which the use of an index would be appropriate. The use of an index seemed justified for this study because I was interested in comparing two communities in a multitude of ways and I wanted one of the ways to provide a quick means for future comparison of results.

Magurran (1988) provides a good review of the pros and cons associated with numerous indices. From Magurran's analyses and those of others (Southwood 1978; Taylor 1978) a single diversity index, the log series alpha, seemed the best choice as a universal diversity statistic. Unlike twelve other indices, the alpha index has all the following attributes: it has good discriminant ability between data sets, low sensitivity to sample size, is more responsive to species richness than dominance, it is simple to calculate, and is widely used (Magurran 1988).

I provide the following information on the index employed for this study. The log series alpha index was obtained using the equation

alpha=(N(1-x))/x where x is obtained from

S/N=[(1-x)/x][-ln(1-x)] with S being the total

number of species and N being the total number of individuals (Magurran 1988).

Estimate of Species Richness

The first-order jackknife species richness estimate developed by Heltshe and Forrester (1983) was used for this investigation. It is obtained from the following equation:

Where, JESR = Jackknife estimate of species richness, y = total number of species observed in all trap units, n = number of trap units and k = number of species that occur in only one trap unit, regardless of their abundance. Note that with large n, the term n-1/n approaches unity, so the equation simplifies to JESR = y+k.

This estimator was chosen for a number of reasons including it being a non-parametric technique thereby requiring no distributional assumptions. Baltanás (1992), in an analysis of the performance of three estimators of species richness, including the first-order jackknife, concluded that this estimator was the least biased and more frequently achieved estimates within + 10% of the true value than the others. Palmer (1990, 1991) also compared the performances of estimators. Of the eight estimators that Palmer tested, the first-order and second-order jackknife estimators were tied for best overall. The first-order was the most precise (i.e. produced estimates closest to true values) and the second-order jackknife was the most unbiased (e.g. the percentage of overestimates was closest to the ideal of 50%).

Palmer (loc. cit.) states that more testing of estimators is needed to verify their performance in the real world. Many questions regarding their value remain unanswered (Slocomb & Dickson 1978; Edwards 1993). However, Baltanás (1992) concluded that although no estimator has been shown to be perfect in all situations tested, they provide better estimates of true species richness than do unaltered direct counts of observed species. This is because the number of species observed in a study can be considered only a lower boundary on the actual species richness (Heltshe & Forrester 1983). It is theoretically possible for estimated species richness to produce different ratios than observed, (e.g. Observed S1> Observed S2, whereas Estimated S1< Estimated S2). This argues for the use of estimated values in this investigation because, although unlikely, the estimated species richness of the fauna with the lower observed richness could be greater than the estimated species richness of the fauna with the higher observed richness.

Beta Diversity

The Morisita-Horn similarity coefficient was used to test the hypothesis that the carcass and control data sets of each year were random samples from identical faunas versus the alternative that the faunas are not identical (Wolda 1981). Magurran (1988) reviews the relative merits of this index in comparison to other popular indices. The Morisita-Horn index was chosen because (1) it is based on abundance data and thus performs superiorly to all qualitative (binary) indices, such as the Jaccard or the Sorenson (Smith 1986; Magurran 1988), and (2) of all the quantitative indices evaluated by Wolda (1981) and Smith (1986), versions of the Morisita-Horn index performed the best, partially because this index has been shown to be the least dependant of all indices available on sample size and diversity (Morisita 1959; Wolda 1981).

The Morisita-Horn index to compare data set A with data set B is calculated from the equation:

where aN= total number of individuals in data set a and ani = number of the individuals in the ith species in data set A.

This index is interpreted as the probability that two randomly drawn individuals from both faunas sampled will belong to the same species, relative to the probability of randomly drawing two individuals of the same species from either of the two faunas alone (Horn 1966).

One of the assumptions of the experimental design was that the proximity of the four traps placed as controls and around each carcass would increase the probability that the set of four traps would sample a single fauna. Each trap within a set of four acted to increase the redundancy (i.e. each trap sampled the same fauna) of the system thereby reducing the severity of lost data if a trap were destroyed. To test the assumption that each set of four traps sampled only one fauna I used the Morisita-Horn similarity index to compare individual trap samples pooled over the season.

Principal Components
Analysis

This analysis finds uncorrelated, linear combinations of the variables (species) which explain the greatest variance within the data. The first three or four of these combinations (termed principal components) together, often explain the majority of the total variation and frequently can be interpreted as biologically important variables. Because the principal components are uncorrelated, their interpretation is often quite straightforward.

The PCA used in this study was based on a correlation matrix rather than a covariance matrix. This equalized the variances of the species (correlations range from -1 to 1, whereas covariances range from 0 to infinity) thereby preventing species with unusually great variances from distorting the picture of community composition (Pielou 1984; Gustafson pers. comm.). However, species with fewer than 30 individuals in the data set were excluded from the analysis. This exclusion reduced the number of species from 445 to 73. Such exclusion was justified because there were 72 traps (observations) in the total data set and including all 445 species (variables) would have created two large problems: (1) the computational effort alone would have been prohibitive and (2) because the variances were standardized the enormous number of singletons would have contributed an excessive degree of noise (spurious correlations) and highly reduced the interpretability of the results. Optimally, PCA is performed with more observations (samples) than variables (species), an unlikely situation in beetle community ecology research (Gustafson pers. comm.).

1993 Experimental
Design Constraints

Aspects of the experimental design were suboptimal. Although numerous differences existed in methods between 1978 and 1993, the test of the primary hypothesis did not require that these years be comparable. The two year's data sets were made to be similar not for comparative purposes, but to reduce the costs in time and money involved in the research (had the number of sites, the number of traps and the geographic location been entirely different in 1993 than in 1978, the catch would have included more singleton species and had a greater species richness which would have translated to a more sparse data set and a greater cost in time and money, respectively). The comparison of interest involved the carcass and control data sets. These were comparable within years. However, some improvements on the design are worth mentioning.

Any design to document an effect of an impact into a habitat has certain, standard requirements (Green 1979; Stewart-Oaten et al. 1986). For example, replicate samples should be taken within each combination of time and location. Although the trapping design for this study involved the use of four pit-fall traps (replicates within a site) it is questionable whether the different sites qualified as true replicates. That is, each of the seven sites shared many physiognomic traits but did differ in possibly significant aspects. The ideal design would have included two to three close replicate sites of each habitat including slope and other aspects to control for site effects in the data.

However, the control trap design did employ within-site replication, assuming the control region was indeed of identical composition as the carcass region. Yet another constraint resulted from the clustering of control traps. Any deviation in the background habitat between the control cluster and the carcass cluster might have confounded results. For a hypothetical example, a host-plant near the carcass traps and common to the habitat but absent from the small region of the control traps, could have resulted in a phytophagous beetle having been captured significantly more frequently in the carcass traps than the controls. A naive interpretation of the corresponding statistical results would lead one to conclude that this phytophage was associated with carcasses. This problem would have been reduced if the control traps had been separated widely, (e.g. each 40 m from the carcass and 40 m from each other). To account for this limitation, when necessary, results were produced that included only species present at multiple sites (thus excluding single-site spurious associations). This correction worked because it was improbable that a single spurious association would occur at multiple sites. This correction unfortunately excluded species possibly displaying non-spurious associations that were present at only one site--a further problem that would have been minimized by having had multiple site replicates of each site type.

Additional confounding factors include the validity of the control traps as true "controls". It could not be determined whether the presence of a carcass 40 m distant affected the control trap data. A superior design would have used transects of traps extending outward from the carcass (similar to Bornemissza's design (1957)), thereby quantifying the effect that distance from the carcass has on the biota.

The previous two problems in design might have been surmountable had traps been active prior to the introduction of the carcass into the habitats. This is the recommended procedure in environmental impact experimental design (Stewart-Oaten et al. 1986). By having traps active before a carcass is introduced, the pre and post-carcass data can be compared to obtain a more valid documentation of the changes in community structure due to the carcass. In such a scenario the pre-carcass data controls for location effects. However, to control for temporal changes the distant control traps are also required. The controls in a distant local isolate the effect of time but they are subject to variation from location. Between the two, both time and location effects would be accounted for. However, this study was limited in this regard because carcasses could not be manipulated, they were simply found where vertebrate scavengers left them.

An important aspect to ecological sampling is understanding the biases associated with one's sampling device(s) (Green 1979). It has been argued that pit-fall sampling may not measure population size, but rather, may measure only the site-specific activity of individuals and should not be used to quantitatively sample or compare communities (e.g. Greenslade 1964; Southwood 1966). Also, the same species may be differentially susceptible to trapping due to factors such as the environment of the trap. These limitations must be considered when interpreting results from pit-fall trapping studies. An increase in activity cannot be distinguished from an increase in species abundance, occurrence or richness. For this reason, any interpretation of this studies' results must consider the alternative hypothesis (that carcasses may not affect actual richness, abundance and occurrence of taxa, but rather simply affect the activity of species).

Also, although it may be obvious, it should be noted that the data collected are not a random sampling of all populations of beetles in the study region, but rather a selective sample composed of those species that are more likely to be captured by pit-fall traps. Some species of beetles are unlikely to be caught by pit-fall traps, but may play important local community roles. The sole reliance on one sampling method prevented a complete picture of the species diversity. However, because all the data collected were equally biased, this bias in no way invalidates the analyses used to test the hypothesis.