Eric _Moore_MS_BGSU_Metadata_Summer_2016


Macroinvertebrate detritivores directly and indirectly affect ecosystem processes and nutrient cycling, but the effects of detritivore interactions on ecosystem processes remain poorly understood. Furthermore, mounting evidence emphasizes the importance of understanding the effects and implications of environmental disturbance on interactions between soil organisms, including invasive species and climate change.
In this study, we asked: Is there evidence of competitive or complementary interactions between a native detritivore and two non-native detritivores? Additionally, can detritivore species-specific and interspecific interaction effects on soil ecosystem processes and microbial activity, be moderated by temperature?
To answer these questions, we performed a mesocosm experiment which included three detritivore species, a millipede native to North America (N. americanus) and two introduced earthworm species (L. rubellus and L. terrestris). We fully replicated a Simplex mixture design using these species under two temperature treatments, ambient and warmed 3.3°C.
We expected to observe species-specific and complementary effects of the study organisms due to differences in functional traits. Furthermore, we anticipated that temperature would alter species interactions, and warming would exert a disproportionately greater negative effect on surface dwelling millipedes. Overall, we expected invasive earthworm effects to overwhelm the effects of millipede presence.
An interaction between L. rubellus and N. americanus predicted an increase in litter mass lost and microbial biomass C, indicating a potential complementarity effect. N. americanus reduced both NAG and BG enzyme activity. L. terrestris reduced NAG and BG activity at warmed temperature, but increased activity at ambient temperature. An interaction effect between L. rubellus and L. terrestris also predicted a reduction in NAG activity. Earthworm biomass was significantly reduced over the duration of the experiment regardless of temperature treatment. L. terrestris significantly increased NH4+ leaching, and detritivores did not significantly affect carbon (CO2) efflux.
There was no evidence to indicate that interspecific interactions between these detritivores are moderated by temperature. However, these results indicate that N. americanus may exert some biotic resistance to invasion pressures by L. rubellus and L. terrestris. Future experiments may consider manipulations of food resources, additional trophic levels, and physical soil characteristics to parse out underlying mechanisms.

Core Areas: 
Short name: 

These data were collected with the purpose of better understanding interspecific interactions between native and non-native forest floor detritivores. Furthermore, these data were used to better understand the role temperature plays in moderating detritivore interactions and the effects of detritivores on ecosystem processes and microbial communities


Mesocosms were established on a covered porch at the University of Michigan Biological
Station (UMBS) in Pellston, MI, USA (45-35.5°N, 84-43°W). UMBS soils (used in this
experiment) are acidic (pH 4.8), and classified as mixed, frigid Entic Haplorthods of the Rubicon
series (92.9% sand, 6.5% silt, 0.6% clay). Local dominant tree species include bigtooth aspen
(Populus grandidentata), red maple (Acer rubrum), northern red oak (Quercus rubra), paper
birch (Betula papyrifera), Eastern white pine (Pinus strobus), and American beech (Fagus
grandifolia). Average (1979-2010) summer (June-August) temperature and precipitation at
UMBS is 17.8 °C and 219 mm respectively (Nave et al., 2011). Experimental Design A total of 40 mesocosms were constructed and subjected to one of two temperature
treatments, ambient or elevated 3.3°C above ambient (which corresponds to IPCC projections for
northern Michigan (IPCC, 2013)), resulting in 20 mesocosms per temperature treatment. Within
each temperature treatment, detritivore assemblages comprised of two invasive earthworm
species (L. rubellus and L. terrestris) and a native millipede (N. americanus) were introduced to
mesocosms following a Simplex design (Cornell, 2011) (Figure 1).
The Simplex design, which has been successfully implemented in other studies on
diversity-functioning relationships among soil detritivores (Andriuzzi et al., 2015; O’Hea et al.,
2010; Piotrowska et al., 2013; Sheehan et al., 2006; Sheehan et al., 2007; Sheehan et al., 2008),
allowed us to manipulate detritivore biomass on a continuous spectrum. Biomass was used as a
proxy for species richness and evenness, and explicitly manipulated on the assumption that, in
soil systems, biomass is intrinsically linked to an organism’s functional impact (Bradford et al.,
2002; Bílá et al., 2014; Turnbull et al., 2014; van Geffen et al., 2011). By controlling biomass as
a measure of evenness, the Simplex design further allowed us to probe into the strength of
species interactions (Petchey & Gaston, 2002; Kirwan et al., 2007).
The purpose of the Simplex design is to assess mixture performances of species
combinations including: single species monocultures, ‘centroids’ where each species contributes
equally to total detritivore biomass, and intermediary combinations of biomass (Figure 1). Single
species monocultures and centroids were duplicated to account for losses due to mortality, and
three detritivore-free mesocosms served as controls under each temperature treatment.
Replicating the other species combination treatments was not necessary because they encompass
a continuous range of species co-occurrences as opposed to factorial combinations, which allows
the data to be analyzed using multiple regression. The Simplex design is advantageous to
diversity-functioning studies as it makes possible the ability to differentiate (a) species identity
effects from the effects of interspecific interactions as well as (b) interaction effects from
biomass effects on ecosystem processes (Kirwan et al., 2007; Kirwan et al., 2009).
Each mesocosm containing detritivores received approximately 6 grams of detritivore
biomass (which is comparable to the higher end of total detritivore biomass densities observed at
this site, (Crumsey et al., 2014)). Total biomass was relatively consistent across mesocosms with
small discrepancies between ideal values dictated by the Simplex design, and actual values (±
1.8% on average, Range: 0.808 g). Discrepancies in biomass by species for each combination
were also relatively small (± 4.5% on average). The number of individuals in each bucket was
also kept as even as possible (± 1 N. americanus individual, ± 1 L. terrestris individual, ± 1 to 2
L. rubellus individuals) across treatments to reduce variation associated with the number of intra-
and interspecific interactions.

Mesocosm Construction and Maintenance Mesocosms were constructed using 5-gallon plastic buckets (height 14.5”, diameter at top 11.9”, diameter at bottom 10.33”). A circular hole 6 cm in diameter was cut into the center of the bottom of each bucket and covered with 2 mm mesh to allow for adequate drainage and to
prevent organisms from escaping. Funnels were then attached underneath each hole and fixed
with removable plastic bottles that were used to collect leachate (Figure 2). The buckets were
placed underneath a porch overhang and suspended 3 feet off the ground between wooden planks
(Figure 3).
Soil was collected during the second week of May 2016. Litter and surface detritus were
raked away before soil was removed in layers by shovel. Collected soil was separated into B, E,
and mixed O/A horizons and placed into separate plastic bins by layer for 24 hours. Each layer
was then sieved through ¼ inch wire, which removed large rocks and macroinvertebrates, but
maintained much of the naturally occurring soil aggregates in each layer. Each layer was
homogenized separately using the cone and quarter method (Gerlach et al., 2002). This process
involves: piling soil onto a plastic tarp forming a cone shape, raking quartered sections of the
pile towards four opposing directions, and shoveling the distributed soil around to other quarters
to evenly disperse the soil before reforming the original cone. The cone and quarter method was
applied to each layer 3 times, and each layer was visually inspected for residual
macroinvertebrates before being added to the mesocosms. Four liters of B horizon were added
first to each bucket, followed by four liters of E horizon and finally seven liters of O/A horizon.
These proportions deviate slightly from field proportions. More organic O/A horizon was
included to support detritivore and microbial populations in the mesocosms. The mesocosms
were then allowed stabilize for four weeks prior to start of the experiment. During this time, 200
mL of distilled water was added to each mesocosms weekly to prevent drying and death of
microbial communities.

Circles (25.4 cm diameter) were cut out of the accompanying bucket lids so that only the
lip remained; this allowed for ease of removal and replacement when measuring respiration.
Mesh screening (1 mm) was fitted to cover the frame to allow for opening air exchange, as well
as to prevent organisms from entering and leaving the mesocosms.
Mixed diversity litter composed of six species (see Study System for species descriptions)
from a forest stand burned in 1954 was collected, dried, and stored in trash bags in a climate
controlled laboratory prior to the start of the experiment. Litter bags, composed of 10 g dried,
leaf litter representative of dominant local hardwood species and 4 g of quaking aspen wood
chips, were constructed using fish netting (hole width approximately 12.7 mm). This material
was used as it facilitated large detritivore movement through the litter bags.

Litter bags were randomly assigned to buckets, and mesocosms were randomly assigned
temperature and detritivore treatments. Heat bulbs (Phillips® 250W 120V) were suspended
above the surface of mesocosms to raise air temperature 3.3 °C (± 0.25°C) above ambient at the
surface of the soil in the elevated temperature treatments (Figure 4). A single heat lamp was used
to warm 4 mesocosms, so mesocosms were clustered into groups of 4 for a total of 10 groups of
4, with 5 groups under each temperature treatment (Figure 3). These groups were treated as
blocks in the statistical analyses.
I-Buttons (DS1925, Maxim Integrated, San Jose, CA) were used to monitor temperature
changes over the duration of the experiment. I-Buttons were placed in each mesocosm (40 total)
at the soil surface at the edge of the bucket furthest from the heat lamps and bucket cluster

Mesocosm moisture content was monitored using gravimetric methods. The mass of each
bucket with soil at field moisture capacity was measured at the start of the experiment and was
adjusted weekly using DO water as needed. Volumetric water content of each mesocosm was
only recorded at the conclusion of the experiment in order to avoid altering detritivore burrowing
and efflux measurements. Detritivore Assemblages Detritivores were hand collected from forests surrounding UMBS during late May 2016.
L. rubellus and L. terrestris were retrieved using an electroshock extraction method (Crumsey et
al., 2014). Due to difficulties finding mature L. terrestris individuals in late Spring, individuals
purchased from bait shops were used as supplements. To avoid biases towards using purchased
versus collected L. terrestris individuals in mesocosms, all L. terrestris individuals were pooled
in single terrarium and selected randomly when assigned to mesocosms. All detritivores were
housed in the climate controlled Lakeside Laboratory (21°C) at UMBS and separated into
terrariums by species for two weeks prior to the start of the study. Each terrarium contained soils
and litter representative of UMBS forests.

Before being introduced to mesocosms, earthworms were identified to species, and only
individuals that appeared healthy were used. Due to difficulties in collecting adult L. terrestris
with low enough biomass requirements for some Simplex proportions dictated by the design,
juveniles were used as necessary. To avoid the possibility of confusing L. rubellus juveniles with
L. terrestris, only juveniles that were observed leaving burrows in habitat types primarily
dominated by L. terrestris during electroshocking were used in the study on the assumption that
these burrows were early indications of anecic burrowing behavior. At the conclusion of the
study, the matured juveniles were identified and confirmed as L. terrestris in all cases.

The earthworms and millipedes were placed in containers with moist paper towels and
allowed to clear their guts for 24 hours prior to being introduced to the mesocosms. The mass of
each organism was recorded before being placed on the mesocosm surface soil and allowed to

Temperature treatments began one week after detritivore treatments to allow for adequate
acclimation. Casting, molting, and soil surface mortality were observed and recorded throughout
the duration of the experiment.

Carbon Efflux was measured weekly using a LI-6400 (LICOR, Lincoln, NE).

Bottles of collected leachate were analyzed at the end of the mesocosm stabilization period, at four weeks, and at mesocosm harvest, for a total of three separate collection dates. The day before each collection date, each mesocosm received 500 mL of DO water and bottles were
collected 24 hours later. Each sample was acidified with 6N TMG HCl and filtered prior to analysis (Fisherbrand™ Glass Fiber Circles, Whatman GF/F). Ammonium (NH4+) concentration was analyzed using the automated phenate method (Bran + Leubbe® AA3). Nitrate (NO3-)
concentration was determined using IC (ThermoFisher™/Dionex™ Integrion HPIC™); column set AS-11 HC (4 micron), isocratic separation using 30 mM KOH as eluent. Total leachate NH4+ and NO3- are reported as µg/L and mg/L respectively. Mesocosm Harvest The experiment ran for a total of 8 weeks and mesocosms were harvested randomly over
a three-day period.
On the first day of harvest, litter bags were removed from the surface of the soil and
visually inspected to detect if any of the study organisms were inside the bags. The litter bags
were then sealed inside plastic bags to ensure no litter was lost prior to drying. Two soils cores at
a depth of 15 cm were taken from the soil surface; one core was extracted from directly beneath
the litter bag, and another core was taken randomly around the circumference of the surface soil.
When detritivore burrows were present at the soil surface, the second core was targeted on those
areas. The reasoning behind this comes from evidence that suggests the chemical and physical
properties of detritivore burrows and casts are species-specific and affect microbial community
functioning as well as soil ecosystem dynamics (Jégou et al., 2001). Millipede molts were not
collected, but recorded, due to similar evidence which suggests that arthropod molts affect
microbial community functioning (Cabib, 1987). Soil samples were placed in large plastic bags,
homogenized for 5 minutes, and stored in a 4°C refrigerator.
On the second harvest day, half of the mesocosms were destructively sampled to remove
surviving organisms. The remaining half of mesocosms were harvested on the third day. The
remaining organisms were then placed in containers with moist paper towels for 24 hours to
clear their guts after being extracted from the mesocosms. After 24 hours, the final biomass of
each organism was recorded to determine change in biomass. Enzyme Assays Immediately following collection, soils were placed in sealed plastic bags and stored at
4°C for six days. Soils were then placed into large coolers with ice and transported to the
University of Toledo for enzyme analysis. Hydrolytic (β-1,4-glucosidase (BG) and N
acetylglucosaminidase (NAG)) and oxidative (phenol oxidase (POX) and peroxidase (PER))
microbial enzyme activity were measured using fluorimetric methods six days after soil
collection following the methods outlined in (Saiya-Cork et al., 2002). The oxidative enzyme
levels were found to be below detection limits, and are therefore not reported here.
Enzyme activity is typically used as a proxy for microbial nutrient demand (Schimel &
Weintraub, 2003). The main function of BG is to hydrolyze cellobiose (a dissacharide derivative
of cellulose) to release glucose for microbial C-acquisition (Sinsabaugh, 2005). NAG exhibits an
analogous function, but acts on chitin, hydrolyzing chitobiose (a nitrogenous glucosamine dimer)
to monomers for microbial N-acquisition (Sinsabaugh et al., 2008). Enzyme activity is reported
as nmol per hour per gram dry soil.
Fluorometrically labelled enzyme substrates were prepared one day prior to assays using
4-methylumbelliferone (MUF) to make 4-MUF-β-D-glucopyranoside and 4-MUF-N-acetyl-β-D
glucosaminide for BG and NAG respectively. The average soil pH of 10 randomly sampled
mesocosms (~ 5.3) was used as the reference point for preparing the modified universal buffer
(MUB) pH. Approximately 1 g (± 0.1 g) of each sample was added to 125 mL plastic bottles and
slurried with ~125 mL of MUB. Each sample then was poured into a wide mouth plastic
container and was continuously stirred. While stirring, 200 μL aliquots of each sample was
pipetted into 96-well microplates for a total of 16 replicates per sample per assay. Plates were
allowed to incubate for 5 hours at 20°C prior to reading. Following the incubation period,
fluorescence was measured for each well using a BioTek Synergy HT microplate reader (BioTek
Instruments Inc., Winooski VT, USA). Microbial Biomass C and N Extractable microbial C (DOC) and N (DON) were determined using the chloroform
fumigation-extraction technique outlined in Brookes et al. (1985) and modified by Scott-Denton
et al. (2006). Approximately 5 g (± 0.1 g) of each soil sample was added to separate centrifuge
tubes and extracted with 25 mL 0.5 M K2SO4. Two sets of centrifuge tubes per sample as well as
blanks were prepared to account for fumigated and non-fumigated samples. Prepared centrifuge
tubes were placed horizontally on a shaker table and allowed to shake for 1 hour. One set of
centrifuge tubes was placed under a fume hood with 2 mL ethanol-free chloroform where it was
allowed to fumigate for 24 hours. Additionally, both sets of extracts were filtered using
Whatman #1 filter paper and diluted at a 1:10 ratio prior to TOC analysis. The diluted extracts
were analyzed using a Shimadzu TOC-Vcpn total organic carbon analyzer with a total N module
(Shimadzu Scientific Instruments Inc., Columbia, MD, USA). Microbial biomass carbon and
nitrogen were determined by subtracting DOC and TN in the unfumigated samples from the
DOC and TN in the fumigated samples. Organic C and N (DOC and DON) are reported as µg
per gram dry soil.

Following removal from the mesocosms, litter bags were placed in a drying oven at 70°C
for 24 hours. Soil that had adhered to the bags was carefully removed before the mass of the
litter was determined.

Data sources: 

Automated Phenate (Bran + Leubbe® AA3)
IC (ThermoFisher™/Dionex™ Integrion HPIC™); column set AS-11 HC (4 micron), isocratic separation using 30 mM KOH as eluent
BioTek Synergy HT microplate reader (BioTek Instruments Inc., Winooski VT, USA)
Shimadzu TOC-Vcpn total organic carbon analyzer with a total N module (Shimadzu Scientific Instruments Inc., Columbia, MD, USA)
I-Buttons (DS1925, Maxim Integrated, San Jose, CA)
Heat bulbs (Phillips® 250W 120V)
LI-6400 (LICOR, Lincoln, NE)

Northern Michigan Rodent Distribution


Distribution of rodents in northern Michigan documented by Dr. Phil Myers, et al.

We use museum and other collection records to document large and extraordinarily rapid changes in the ranges and relative abundance of 9 species of mammals in the northern Great Lakes region (white-footed mice, woodland deer mice, southern red-backed voles, woodland jumping mice, eastern chipmunks, least chipmunks, southern flying squirrels, northern flying squirrels, common opossums). These species reach either the southern or the northern limit of their distributions in this region. Changes consistently reflect increases in species of primarily southern distribution (white-footed mice, eastern chipmunks, southern flying squirrels, common opossums) and declines by northern species (woodland deer mice, southern red-backed voles, woodland jumping mice, least chipmunks, northern flying squirrels). White-footed mice and southern flying squirrels have extended their ranges over 225 km since 1980, and at particularly well-studied sites in Michigan’s Upper Peninsula, small mammal assemblages have shifted from numerical domination by northern species to domination by southern species. Repeated re-sampling at some sites suggests that southern species are replacing northern ones rather than simply being added to the fauna. Observed changes are consistent with predictions from climatic warming but not with predictions based on recovery from logging or changes in human populations. Because of the abundance of these focal species (the 8 rodent species make up 96.5% of capture records of all forest-dwelling rodents in the region and 70% of capture records of all forest-dwelling small mammals) and the dominating ecological roles they play, these changes substantially affect the composition and structure of forest communities. They also provide an unusually clear example of change that is likely to be the result of climatic warming in communities that are experienced by large numbers of people.


Species included
We chose this assemblage of 8 species of small forest rodents for 4 reasons. First, each species reaches a distributional limit within or close to the northern Great Lakes region. Second, each is commonly captured by the techniques most widely used by collectors. Records of other species are available but were acquired through the use of trapping or hunting techniques that have not been employed consistently across the 150 years of collecting in this region (e.g., firearms and large traps are seldom used in recent collections, and mist nets for the capture of bats did not become available until the last half of the 20th century). Third, we focused on woodland species because trapping since 1980 has concentrated heavily on forest habitats, and consequently their record is stronger than that of mammal assemblages in other habitats. Fourth, these species are relatively common and frequently captured, often in the same trap-lines. We did not consider a few species that are extremely rare in the region (e.g., woodland voles, Microtus pinetorum) or that seldom enter woodlands (southern bog lemmings, Synaptomys cooperi; meadow voles, Microtus pennsylvanicus; grassland jumping mice, Zapus hudsonius).
Additionally, we report widespread changes in the distribution of common opossums. Opossums are a southern species whose range has extended gradually northwards since the early 20th century (Gardner & Sundquist, 2003).
Data sources
Records from 1978-2008 came primarily from extensive live-trap sampling by field crews from the University of Michigan, Michigan State University, and Miami University. The purpose of these surveys was to document the current distribution and relative abundance of species of small mammals, and all captures were recorded. In almost all cases localities are believed to be accurate to within less than 500 m (Appendix). Questionable species identifications were confirmed using molecular techniques (Appendix). When identifications could not be confirmed, animals not readily identified using field characters were eliminated from the analysis (64 out of 10,273 Peromyscus and 11 out of 293 Glaucomys were deleted).
Recent opossum records were based on field observations and especially, records of road-killed animals made from 2006-present. Coordinates of road-killed animals were recorded using a GPS unit.
Most records prior to 1978 came from the specimens and field notes housed in the University of Michigan Museum of Zoology and the Michigan State University Museum. Additional specimen records were obtained from the MaNIS network http://www.manisnet.org (Appendix). Error in estimating locality coordinates varied widely (Appendix). We examined the estimated error associated with the coordinates of each specimen with the intent of eliminating records whose error overlapped either previously reported range limits or boundaries of the geographic regions on which comparisons of community composition are based (Appendix). A few records were not mapped because their estimated errors were extremely large, but in every case specimens were unambiguously assignable to one of the geographic regions of the study. For some critical records with uncertain localities, we were able to reduce estimated error considerably by referring to field notes and/or published descriptions of collecting expeditions.
We examined and verified the identifications of all museum specimens that suggested significant changes in distribution.
A few records were also provided by individual collectors or taken from published papers. In most cases they involve unexpected findings, usually occurrences outside of the normal range of a species (e.g., Ozoga & Verme, 1966; Haveman, 1976; Stormer & Sloane, 1976; Wells-Gosling, 1982). These records provide documentation of range expansion and are included below in maps and calculation of range change, but as no information was usually provided on what other species were trapped, these records were excluded from analyses of faunal composition.
Regions included
Published range maps of species of mammals in Michigan suggest a transition between a fauna associated with the oak hickory woodlands and savannahs typical of the southern part of the state, and a northern fauna associated with northern hardwood and coniferous forests (Hall, 1981; Baker, 1983). At the time these maps were compiled, northern and southern faunas met in the middle of the Lower Peninsula, in a region (“tension zone”) that is characterized by differences in soils and a transition from a more southern to a more boreal flora (Fig. 1; Medley & Harman, 1987). Our focus is on changes concentrated to the north of this zone, and consequently we restricted our attention to records north of 44oN latitude (Fig. 1).
A number of islands are found in Lakes Michigan, Superior, and Huron. Many are inhabited by small mammals, and extensive collection records are available for some. These islands have little or no opportunity to receive immigrants from the mainland, and the composition of their fauna likely reflects the species present when the islands were isolated by rising water as the lakes first formed, nearly 10,000 years ago. Records from islands separated from the mainland by at least 10 km (Beaver, High, Hog, Timm’s, Squaw, Whiskey, Trout, Gull, Garden, N and S Manitou, N and S Fox, Bois Blanc, Isle Royale) were not considered in this analysis. Further, we eliminated 4 sites in the northern Lower Peninsula, because since 1978 they were visited repeatedly, often several times a year, to obtain specimens or to follow the populations of particular species. Including them would have strongly biased the analyses in the direction of conditions at those sites, and for inferences concerning regional community composition, would represent a form of pseudoreplication (Hurlbert, 1984). These sites (and the area each encompasses) are as follows (Fig. 1):
1) 45.168 – 45.1775oN, 84.375 - 84.401oW (2.1 km2)
2) 45.088 – 45.1147 oN, 84.402 – 84.425 oW (5.34 km2)
3) 45.271 – 45.296 oN, 84.416 - 84.443 oW (5.88 km2)
4) 3 line transects, 300-500 m in length, at the University of Michigan Biological Station: 45.546°N, 84.667°W; 45.5567°N, 84.7015°W’; 45.4894°N, 84.6849°W.
Huron Mountains
Repeated collections made at a few sites in the Huron Mountains are especially informative. The Huron Mountains are a series of low granitic hills (maximum elevation 600 m) near the Lake Superior shoreline in the central Upper Peninsula of Michigan (Fig. 1). Approximately 7,300 ha are owned by a private association, the Huron Mountain Club, whose members support research on their property through the Huron Mountain Wildlife Foundation. This area includes a 2600 ha Nature Research Area of primary (never logged) forest. The Huron Mountain Wildlife Foundation has funded 3 surveys of the mammals of the region. The first, a comprehensive survey of vertebrates by Richard Manville, was carried out from autumn 1939 through summer 1942 (Manville, 1947, 1949). To sample small mammal populations, Manville set up 8 quadrats chosen to represent the habitats of the region. Each quadrat comprised an 11 x 11 trapping grid (30 ft between traps). Manville used live traps and trapped for 5 consecutive days 4 times over the course of the study. He deposited extensive series of voucher specimens in the collections of the University of Michigan Museum of Zoology, and we have confirmed his identifications of Peromyscus. In 1972-1973, John Laundre also conducted small mammal censuses in the Huron Mountains, trapping at or near the same locations as Manville and using similar techniques (Laundre, 1975). Unfortunately, his report does not list numbers of individuals of most species captured, and we are therefore unable to include his records in the analyses of relative abundance reported here. Nor have we been able to locate voucher specimens. His account, however, is useful in documenting the presence/absence of species in 1972-3 compared to other time periods. In 2004-2005, the survey was repeated by Allison Poor (Poor, 2005). Poor used live-trapping techniques similar to those of Manville and Laundre and located most of her quadrats at or very close to the same sites. Poor, however, trapped for 3 days/sampling period, taking 2 samples in 2004 and 1 in 2005. Like Manville, she recorded all captures, and she deposited vouchers (mainly tissue samples) in the University of Michigan Museum of Zoology.
Time periods
Preliminary examination of maps and capture records suggested that for small mammals, change in distributional patterns accelerated during the late 20th century. While these preliminary results also suggested some differences among species in the timing of change, to simplify comparisons of SFR assemblages we arbitrarily chose to compare collections made from 1883 (when the first records were obtained) through 1980 with those made from 1981 to the present.
Data analysis
A total of 14,076 records of the 8 focal species of SFRs from north of 44oN latitude were used in the analyses reported below. Of these, 4,808 came from museum catalogues and records taken from the literature, and 9,268 from our sampling. These records include 4,099 captures from 564 localities recorded during the period 1883-1980, and 9,977 captures from 591 localities from 1981-2007. The focal small forest rodents make up 96.5% of all captures of forest-dwelling rodents (including tree squirrels and rare species) and 70% of all captures of forest-dwelling small mammals (including the above species plus shrews and moles). For opossums, we included 94 capture records from MaNIS, 163 records from a survey of road-killed animals carried out in 1968 (Brocke, 1970), and 281 records from a similar survey done in 2006-8.

Additional sources specimen records:

Crider JE (1979) A Wildlife Inventory of the Sturgeon River Wilderness Study Area. M.S. thesis, Michigan Technological University, Houghton MI.

Teresa Friedrich -- UMMZ field notes

Allen Kurta -- personal communication to pm

Sean Maher-- UMMZ field notes

Manville RH (1947) The vertebrate fauna of the Huron Mountains, Marquette County, Michigan. Ph.D. thesis, University of Michigan, Ann Arbor, Michigan, 263 pp.

Rosa Moscarella -- personal communication to pm

pmrecords -- UMMZ field notes

Allison Poor -- UMMZ field notes; Poor AP (2005) Changes in the Small Mammal Fauna of the Huron Mountain Club, Marquette County, Michigan: an Effect of Global Warming? MS thesis, University of Michigan, Ann Arbor, Michigan, 62 pp.

Skillen R (2005) Changes in the Distribution of Michigan’s Flying Squirrels. MS thesis, Michigan State University, 105 pp.

Data sources: 
Northern Michigan Rodent Distribution
Quality Assurance: 

Data sources for museum records—Specimen information was obtained through the MaNIS network (http://www.manisnet.org) from the following museums: California Academy of Sciences; Cornell University (CU); Field Museum of Natural History (FMNH); Florida Museum of Natural History (FLMNH); Harvard University Museum of Comparative Zoology (MCZ); Los Angeles County Museum of Natural Science (LACM); Louisiana State University (LSUMZ); Michigan State University Museum (MSU); Museum of Natural Science, Royal Ontario Museum (ROM); San Diego Natural History Museum (SDNHM); Santa Barbara Museum of Natural History; Texas A&M University, Texas Cooperative Wildlife Collection (TCWC); Texas Tech University Museum (TTU); United States National Museum of Natural History (USNM); Universidad Nacional Autonoma de Mexico, Instituto de Biologia (IBUNAM); University of Alaska Museum (UAM); Museum of Vertebrate Zoology, University of California (MVZ); Museum of Natural History, University of Kansas (KU); University of Michigan Museum of Zoology (UMMZ); University of Minnesota Bell Museum of Natural History (MMNH); University of Puget Sound Slater Museum of Natural History; University of Utah Museum of Natural History; Burke Museum, University of Washington (UWBM); Museum of Southwestern Biology, University of New Mexico (MSB).
Error—Records compiled from Museums and our sampling programs are subject to multiple sources of error. Assessing and correcting error, insofar as possible, is a difficult and time-consuming process, but it is essential; records such as these cannot simply be downloaded and incorporated into research (Williams et al., 2002; Chapman, 2005a). Further, when error is suspected but cannot be checked, it is important to consider what effect it might have on an analysis. Error that is unbiased with respect to the questions being considered is less of a problem than error that systematically skews an analysis one way or another. Unbiased error may make patterns harder to detect or obscure them entirely, but it is unlikely to create patterns where none exist.
Our goals in this paper are to identify distributional shifts and changes in local assemblages of small mammals and to discuss possible causes of change. With respect to those goals and the particular collections on which we relied, we addressed the following areas of uncertainty with regard to each record:
1. problems with species identifications
2. problems associated with the precision of location
3. collector bias– why properly identified and georeferenced collections might still provide a misleading picture of community composition.
1. Problems with species identifications
Most small mammal species in the northern Great Lakes region are fairly easy to identify in the field or as museum specimens, but among the forest rodents on which we focus, field identification of 2 pairs (Peromyscus, woodland deer mice and white-footed mice; Glaucomys, southern flying squirrels and northern flying squirrels) is sometimes difficult. For museum records, we examined and verified the identifications of all specimens that suggested significant changes in distribution. For field records, since 1980 almost all field identifications were made by the authors or by assistants trained by us. Identifications of most questionable Peromyscus were confirmed by electrophoretic examination of salivary amylase alleles (Aquadro & Patton, 1980) or restriction fragment length polymorphism (Poor, 2005).
2. Problems associated with the precision of location
Coordinates for localities associated with our field work (post 1980 records) were either georeferenced directly using GPS units, or located on maps (usually to quarter-quarter section) and later georeferenced using Topozone (http://www.topozone.com) and/or Google Earth (http://earth.google.com).
For most other specimens, localities, including (when available) latitude, longitude, and estimated coordinate error, were downloaded from http://www.manisnet.org/. For localities that were missing coordinates and/or estimates of coordinate uncertainty, we used Topozone, Google Earth, and an assortment of local maps to supply coordinates. Missing coordinate uncertainties were calculated using the MaNIS Georeferencing calculator (http://www.manisnet.org/gci2.html; see also Chapman & Wieczorek, 2006). In some instances, we were able to refine coordinates and/or reduce uncertainty significantly by using field notes, papers published by collectors, and in the case of recent collections, first-hand knowledge of the sites where collections were made.
For SFR species, collection records are available from 977 identifiable localities in Michigan north of 44oN latitude (records whose coordinate uncertainty overlapped the 44th parallel were eliminated). Of these, 936 (95.8%) had uncertainties < 20 km. The maximum uncertainty for any locality was for 2 records that could be restricted only to the Upper Peninsula. Because analyses of community composition involved combining records of specimens captured over large geographic areas (Lower Peninsula, Upper Peninsula, Huron Mountains), placement of any locality at any point within even the largest area of uncertainty did not change its geographic area. Thus, we were able to include specimens from all 977 localities.
For documentation of range change, we examined all localities with estimated uncertainty > 20km to determine if the area of uncertainty overlapped the edge of the known distribution when the collection was made. This was never the case; most records suggesting range extensions were from our own (post 1980) surveys, and the estimated errors associated with their localities are small.
3. Collector bias–properly identified and georeferenced collections might still provide a misleading picture of community composition
Collecting is usually done for a particular purpose. Collectors sometimes have the goal of determining species composition and abundance in a community, and they record everything they capture. Collectors are often, however, looking for particular species or sampling particular habitats. They may or may not take exemplars of other species or record them in their notes. If specimens are taken, their numbers are likely to be biased in favor of the species under study. Collectors are also likely to keep or report specimens that surprise them because they represent rare or unexpected finds, while perhaps under-representing or ignoring common species. They sometimes take only “vouchers,” specimens placed in collections to document the presence and identification of a particular species at a locality. Collection of vouchers is seldom done in the context of community composition. Further, collections may be biased geographically. Areas that are easily accessible or particularly attractive may be over-represented, while remote or, at the opposite extreme, heavily urban areas are often less frequently collected.
It is possible to address some or all of these problems, but the methods and effectiveness of doing so depend on the sources of data and the goals of a study. Here, we are fortunate in several respects. Almost all collections after 1980 were made by us or by our students, and they include a full list of animals captured. Further, the majority of earlier records were accumulated from the first half of the 20th century, when the goal of many collectors was explicitly to document the total fauna of the areas collected (e.g., Wenzel, 1911; Dice & Sherman, 1922; Hatt, 1923; Dice, 1925; Green, 1925; Blair, 1941; Appendix Fig. 1).
By restricting our study to a subset of species (small forest rodents, SFRs) that are often found together and are likely to be captured using trapping methods widely employed by collectors, we minimize both the tendency of collectors to favor certain habitats and biases introduced by evolving collection methods.
Because collecting effort is seldom documented for early collections, differences among collections might simply represent the intensity of the collecting effort. We therefore compared SFR assemblages in 2 ways. First, we examined the abundance of each species relative to other SFR species in the collection. Relative abundance analyses, however, rely on collectors reporting all individuals of each species. This assumption is met by post-1980 collections but perhaps not by some earlier ones, Consequently, we further compared the results of relative abundance analyses to “occurrence analyses” that require only that collectors report the presence of each species captured (see Methods and Discussion, above). The general agreement between relative abundance and occurrence analyses suggests that their collections give us a reasonable and consistent picture of SFR assemblages at the time the collections were made.
Bias might also result from changes in the geographic pattern of collecting. If early collecting had concentrated in one area and late collecting in another, differences in the SFR assemblages found would not be surprising. Both the Upper and Lower Peninsulas, however, were widely collected during both time periods (Appendix Fig. 2) and consequently, geographic bias favoring one region or another is unlikely to be significant. Further, over 230 collectors contributed to the records comprising our database. The bias of an individual collector for a particular place or species is unlikely to have a large effect.
Finally, for mammals, the susceptibility to capture varies widely among species. Estimates of the relative abundance of a species obtained from trapping records may be strongly affected by its propensity to enter traps as well as by its actual representation in the community. Here, however, we focus on change over time. We cannot be certain that, for example, the abundance of white-footed mice in collections (38% of the SFRs captured prior to 1980) means that they made up precisely 38% of the actual SFR community. The fact, however, that after 1981 their relative abundance increased to 78% demonstrates that their representation in the SFR community has increased dramatically (Table 1).