USDAUnitedStates

Departmentof An Interpretation ofAgriculture

Forest Landscape Structure fromService

North,Central Historic and Present LandForestExperiment

Station Cover Data in the EasternGeneralTechnical

Report NC.192 Upper Peninsula of_ Michigan

Janet Silbernagel, Jiquan Chen, Margaret R. Gale, Kurt S. Pregitzer, and John Probst

Outwash

Bedrock-controlled

Lowland Sand Lake Plain

Morainal Origin

o

o .

North Central Forest Experiment StationForest Service--U.S. Department of Agriculture

1992 Folwell AvenueSt. Paul, Minnesota 55108

Manuscript approved for publication July 31, 19971997

An Interpretation of Landscape Structurefrom Historic and Present Land Cover Data

in the Eastern Upper Peninsula of Michigan

Janet Silbernagel, Jiquan Chen, Margaret R. Gale,Kurt S. Pregitzer, and John Probst

Policy makers and planners dealing with broad of ecological units and strengthen the frame-scale issues must account for landscape and work on which both finer resolution landscaperegional scale factors in their design. In light decisions may be based. Already there haveof contemporary broad-scale environmental been rigorous efforts linked with physicalissues, landscape pattern analysis has become geography to describe land surface patternsa necessity for both resource managers and (GoUey 1995). But, many of these studies,scientists. Questions such as ... "how will regardless of the investigator's perspective,managing a 40-ha tract affect adjacent areas have been of limited extent (Pastor andor regions?..." have promoted new approaches Broschart 1990, Simpson et al. 1994), lowto society's dilemmas like air and water qual- resoluUon (Dunn et aL 199 I, Hulshoff 1995,

: ity, species and habitat losses, and increased O'Neill et al. 1988), or of narrow temporal scalecarbon dioxide levels (King 1993). Maps of (Luque et aL 1994, Muller and Middletonregional- and landscape-scale ecological units 1994, Simpson et a/. 1994, Turner andare available to, and now frequently used by, Ruscher 1988). What is needed are works thatpolicy decision-makers and land managers to can increase spatial and temporal extent whileapproach these sorts of issues. Regional and maintaining high resolution so that resultslandscape ecological units based on biophysi- can be applied to a larger area and provide acal factors provide an ecologically based longer frame of reference, yet sun be meaning-framework for land use decisions. However, ful for site level assessments.

for many of these ecological units, measurabledescriptions of pattern, composition, and other There has been significant interest in land-CharacterisUcs are lacking. Measures that scapes surrounding the Great Lakes, wherequantify and describe landscape structure land use has had a particularly large influence(pattern and composition) improve knowledge on the lake ecosystems (e.g., Upper Great

Lakes Biodiversity Committee; Lake SuperiorBinational; SOLEC, in press). Fluctuations inhuman population in the Lake States since

Janet SIIb rnagel is with the Department of European settlement have played a role in' Horticulture and _dscape Architecture, development of current forest patterns and

Washington State University, Pullman, WA have been the focus of several research pro-99164. grams. For instance, Frelich (I 995) docu-

mented the amount of "old forest" in historic

Jlquan Chen, Margaret Gale, and Kurt and present forests of the Lake States, includ-Pregitzer are with the School of Forestry and ing amounts present in various land cover

Wood Products, Michigan Technological types. He defined "old forest" as both oldUniversity, Houghton, MI 49931. growth (forests dominated by long-lived spe-

cies beyond 120 years) and old-seral forestsJohn Probst is with the Forestry Sciences (dominated by short-lived species that areLaboratory, USDA Forest Service, North relaUvely old at age 80). The results showedCentral Forest Experiment Station, that about 5.2 to 8.3 percent of the _e

• Rhinelander, WI 5450 I-9128 States forest is now old forest, compared with

an estimated 68 percent before European II. BACKGROUNDsettlement. Frelich also found that in thelargest single block of remaining unlogged A. Study Areaforest in the Lake States, Minnesota's Bound-ary WaterS Canoe Area Wilderness, the spatial 1. Geographic location: The study area is

•pattern, age structure, and species composi- the Eastern Upper Peninsula of Michigan,tion have been significantly altered since the comprising six counties: Alger, Chippewa,1800's. Still, Frelich's study was limited to Delta, Luce, Mackinac, and Schoolcraft; andassessment of "old forest," was coarse in 1,795,524 ha _. The area is bordered by threeresolution (errors of +/- 50 percent for forest Great _es--Huron, Michigan, and Supe-areas 400 ha in size or less), and did not rior--and is characterized by lake-moderatedquantify spatial pattern to augment additional climates and land types including hardwoodlandscape analyses, forest on glacial moraines; pine barrens on

outwash and beach ridges; coastal marshes;The distinction of this study from important interior bogs, fens, and swamps; and llme-works like Frelich's and others' lies in its stone outcrops (Peterson 1986). The area liesgeographic size, its sequence of time, and its completely within the Northern Great Lakesprovisionalassessment units. We chose Ecological Section (map I), (Barley eta/. 1994,landtype association groups as the basis for McNab and Avers 1994).comparison to control for broad physiographicpatterns, while strengthening the knowledge 2. Physiographic setting: The Northernbase of these units and their value as planning Great Lakes Section is a level to gently rollingelements (McNab 1996, Sims et al. 1996). lowland of fiat outwash or lacustrine plainMany previous landscape assessments have with dune fields near the Great _es. Par-been based on administrative or arbitrary tially buried end moraines protrude from the

• boundaries (e.g., 7.5' USGS quadrangle maps), lowlands roughly paranel to the lakes. A largerather than an ecological frameworlc Without portion of the area is covered by Pleistocenesuch a framework, it is not possible to account stratified drift, mostly outwash sand. Lacus-for variation in the physical environment, trine deposits occur between morainal ridges

and are widespread in the Eastern UpperI. OBJECTIVES Peninsula. Pleistocene and Holocene sand

dunes occur near the Great Lakes (Martin

The objective of this report was to: 1957). Silurian limestones and dolomites arelocally exposed along _es Huron and Michi-

Compile quantitative InformatWn, or landscape gan, while Cambrian and Precambrian sand-metrics, to supplement existing qua!itative stone outcrops occur along Lake Superior.descr_tlons of land type associations in the The dominant geomorphic processes for thisstudy area, Compare the distribution of land- area are fluvial erosion, transport, and deposi-scape measures among landtype association tion; lake-shore erosion and deposition; andgroups in hlstorlc and present landscapes, minor dune construction (Albert 1995, McNab

and Avers 1994).This study occurred in the context of a

larger research project whereby historic and 3. Land ownership: Much of the Easterncontemporary landscape variability was de- Upper Peninsula is public land including thescribed and quantified, landscape change was Hiawatha National Forest, Lake Superior Stateassessed over a 150-year time period, cultural Forest, Seney National Wildlife Refuge, Pic-settlement patterns were assessed over a tured Rocks National Lakeshore, and3,000-year time period, and the landscape Tahquamenon State Parl_ Other large landchange was discussed in terms of the combi- holdings belong to The Nature Conservancynation of biophysical and cultural factors, and three major industrial landowners. To-This report addresses the first portion of the gether these holdings make up about 12research: description and quantification of percent of Michigan's land area in one of thehistoric and contemporary landscape patterns, most undeveloped portions of the Eastern

United States.

-

• _ One hectare equals 2.471 acres.

2

B. Ecological Classification yet still provide enough information to identify. differences in ecological potential or trends

Because of the infinite complexity inherent in and to improve coordinated managementany ecosystem, land areas are often catego- among multiple land owners or regulatoryrized into discrete units so that biophysical agencies (Cleland pers. comm., ECOMAPcomponents can be measured and understood 1993).(Allen and Start 1982, I_jn and Udo de Haes1994, Urban eta/. 1987). Unified functional Through an interagency effort, a first approxi-ecosystems are divided and categorized into marion of landtype associations (LTA's) weresimilar and dissimilar pieces at various scales, mapped across the Eastern Upper Peninsulainthe interests of description and understand- to help coordinate natural resource steward-ing (Rowe 1996, 1991). A hierarchical frame- ship (map I). The LTA's were identified inwork of ecological units was developed for the accordance with concepts and methods de-United States, from which .units for this study scribed in The National Hierarchical Frame-were based (ECOMAP 1993). The Framework work of Ecologlcal Units (ECOMAP 1993, Banwas modeled after the multi-factor, multi-scale and Padley 1993, Uhlig and Jordan 1996).approach to understanding ecosystem struc-ture and function pioneered in Baden- The delineation process is iterative, using both

I Wurttemberg, Germany (Albert 1995, Barnes top-down (regionalization) and bottom-up1984, Spurr and Barnes 1980). At the broad- (classification) information. At this stage, theest scale of the framework, global ecosystems delineations have been based primarily onwere identified, within which a series of finer intuitive and visual methods. However, bot-scale ecoregions are nested (Barley et a/. 1994, tom-up ecological classification and researchBarley 1996). Next, Albert et aL (1986, 1995) such as in this report contribute to, refine,mapped regional landscape ecosystems of and validate the landscape boundary Ioca-Michigan, Minnesota, and Wisconsin to the tions. Similarly, macroscale research (e.g.,subsection level, based on climatic and broad regional climatic data) is underway that willphysiographic factors. The finest level units of provide quantitative top-down contributions.their work provided the framework for subdivi- Preliminary delineation criteria includedsion into local landscape ecosystem units terrain (major slope and elevation changes orcharacterized by topography, microclimate, escarpments, variability of slope) fromsoft, and vegetation (Albert et al. 1986). 1:250,000 USGS topography maps; geology

(bedrock type, surficial geology, major glacialHI. METHODS features, paleoshorelines) from Anderton

(1993), Mart/n (1957), and Farrand (1982);A, Data Assembly soil (communities of softs, drainage, texture,

depth to bedrock, source and mode of deposi-1, Landtype association: Landscape eco- tion) from county-level soft association maps;systems, defined as "a cluster of interacting localized climate (lake effect wind-blown sandecosystems repeated in similar form through- and wave action, major cold air drainages,out," (Crow 1991; Forman and Godron 1986; etc.) from Santer (1977); large lakes or openJordan et oL, in prep.) provide a spatial foot- water systems; watersheds (drainage pattern,print that can account for mesoscale physical water level, and fluctuation variability); andenvironmental factors. They are a useful size vegetation (LANDSAT 1984 image, pre-Euro-for many analyses because they are commonly pean settlement vegetation from General Landbased on easily identifiable physiographic Office surveyors notes, aerial photography);factors, they are wen-suited to scales of mod- existing subsection maps; plot data; andem inventory technology (remote sensing, GIS, personal experience (Jordan et oL, in prep.).etc.), and they can be easily recognized fromautomobile or airplane (Smith and Carpenter For this study, LTA's were intuitively grouped1996). Boundaries separating landscapes are into the following eight categories based ondistinct, as often seen in satellite images, due dominant physiographic simflariUes: bedrock-to geomorphology and human activity (Forman controlled, lowland sand lake plain, morainal1995). Within a relatively short time, land- origin, outwash, clay lake plain, ground mo-scape ecosystems can be delineated and raine, transitional/complex, and inland lakedescribed across large areas (multiple States), (Gary et aL 1974). The groupings are compa-

rable to Michigan's Land Types mapped by the4

Michigan Department of Natural Resources Pre-European settlement landscape data based(Santer 1977). We used only four of the eight on the GLO survey notes were used as thegroupings: bedrock-controlled, lowland sand historical dataset of land cover for this studylake plain, morainal origin, and outwash for (Ewert, pers. comm.; Albert, pers. comm.; MNFIanalysis because we believed they were the 1994). Surveyors of Michigan in the earlymost internally consistent and because they 1800's took detailed notes on the location,occurred repeatedly across the study area species, and diameter of each tree used to mark(map 2). section lines and comers. Their notes also

detailed landscape features such as lakes,2. Present land cover: Land cover data wetlands, fields, natural disturbances, trails,derived from Landsat Thematic Mapper (TM) and settlements. For our study, advantages ofsatellite imagery were used to measure present these records were: (I) they are available in

•landscape structure. The TM imagery was digital format for the entire Upper Peninsula ofclassified into land cover categories for the Michigan, (2) original field work was conductedentire Upper Peninsula with 60-m resolution, according to a predetermined plan, and (3) theMaclean Consultants, Ltd. of Houghton, surveys consUtute a systematic aligned sampleMichigan, obtained Landsat Thematic Mapper of the historic forest species and thus are(TM) imagery for the Upper Peninsula of useable for quantitative analysis (Bourdo 1956,Michigan from Eosat, Inc., in contract with the Price 1994).Michigan Department of NatUral Resources, toclassify TM images into land cover categories Land cover maps were compiled by the Michi-(Maclean 1994). Images were selected to gan Natural Features Inventory on 1:24,000provide: (1) complete coverage of the Upper USGS topographic maps. Section line informa-Peninsula; (2) minimal cloud cover; (3) cover- tion recorded by land surveyors in the earlyage no earlier than 1990; and (4) data from the 1800's was plotted onto the USGS maps. Covercorrect time of the growing season for best type boundaries between each section line were

._ species differentiation (june I through July interpolated using elevation lines, surface•31). Classification accuracy was presented in geology maps, and other early vegetation maps;the form of contingency tables that showed the resulting resolution was definable to 259both errors of omission and errors of commis- ha, although many smaller polygons weresion (Maclean 1994). The result was a com- delineated. Cover classes were based on ex-plete land cover classification for the entire panded Michigan Resource Information SystemUpper Peninsula with an average 90.2-percent (MIRIS) land cover codes. Small cover types notcorrect (accuracy) within a 95-percent confi- transected by the surveyors were probably notdence interval, recorded. Bias in tree species selected by

surveyors has been documented (Bourdo 1956)•3, Historic land cover: General Land Office and was assumed in map production (MNFI•(GLO) survey records have been commonly 1994). Nevertheless, previous investigationsused in historic vegetation and landscape have concluded that the GLO records could beassessment studies. Early uses of these used to reconstruct historic vegetation (Bourdo

records involved reconstruction of pre-Euro- 1956, Curtis 1959). In fact, in a study by Pricepean settlement forests in the form of maps (1994) of composition, structure, and distur-and compositional descriptions (Curtis 1959, bance regimes of historic forests in westemVeatch 1959). The suitability of survey records Chippewa County, Michigan (within the Easternfor.quantitative analysis of historic vegetation Upper Peninsula), tests found only moderatewas first documented by Bourdo (I 956). bias toward the 15- and 20-cm diameter classesSubsequent studies by Lorimer (1977, 1980), in the mixed conifer/deciduous lowland forest

I Canham (1978), Canham and Loucks (1984), type. Therefore, Price concluded that GLONoss (1985), and Whitney (1986) described the records from western Chippewa County wereuse of GLO notes and other historic records in largely free of bias toward any specific diametercharacterizing historic forests and critiqued classes or tree species.the various approaches. In a more recentapplication of historic data in the Lake States, Each layer was compiled in digital format andBaker (1991) simulated effects of settlement georeferenced to Universal Transverse Mercatorand fire suppression by using a GIS-based projection on a Unix platform ARC/INFO Geo-spatial model and historical data on fire size graphic Information System. A legend defining

• and frequency, the cover types was included with both the

5

lhistoric and present coverages (tables 1 and Metrics can be categorized according to what2). Because the two legends were not identi- they measure. Some metrics measure pattern,cal, they had to be combined so that cover others measure composition. We used thetypes in one coverage matched those in the FRAGSTATS Spatial Pattern Analysis Programother, or in some cases, a cover type was (MacGarigal and Marks 1993) to quantify andrenamed (table 3). Of the total 24 cover types describe patterns within and among LTAin the combined legend, I0 were selected for groups in the historic and present land covers.analysis based on the following criteria: the In this report, we focus on the following land-

type was present in both time periods in at scape metrics: number of patches (N), totalleast two of the LTA groups; the types repre- area (TA), mean patch size (MPS), patch sizesented a broad range of land cover classes coefficient of variation (PSC_, largest patch

indicative of the region and pertinent to the index (LPI), patch density (PD), and Shannon'sobjectives-of this study (e.g., mixed conifer and diversity index (SHDI), the combination ofcedar were selected, but white spruce was which we believed best represented aspects ofnot); based on our familiarity with the two landscape structure. Largest patch indexdatasets, the type seemed to be relatively indicates the contribution of the largest patchconsistent between and within the two to total area. Patch size coefficient of variationdatasets (e.g., herbaceous shrub and provides a measure of relative variation. Patchopenlands were inconsistent, particularly in density is the number of patches per 100 ha,the historic set); or the type was of particular and Shannon's diversity index indicates theimportance to land use change questions (e.g., amount of different class types in a landscape.agriculture). Equations for each of the indices used are

provided in Appendix A.B. Data Analysis

Class statistics and indices calculated for thisIndices measuring several components of report included class area (CA), N, MPS,t

landscape pattern and composition together landscape similarity index (LSIM), LPI, PD,are a means to quantify structure and change PSCV, and mean shape index (MSI). Land-(Hess 1994, Mflne 199 I, O'Neill et al. 1988). scape similarity index measures the proportionIndices and basic metrics (measures of pattern of a class area to the total landscape area.and composition) can be computed for three Mean shape index is a ratio of patch area toscales: patch, class, and landscape. Patch- patch perimeter. Selected landscape and classscale metrics are primarily used as building calculations were tabulated and graphed byblocks to calculate class and landscape LTA group.metrics, and relate to an individual patch orstand of a relatively homogenous (consistent) There is a limitation to conclusions that maycover type. Class statistics relate to individual be drawn from comparing historic to presentland cover classes, or cover types, based on data due to methodological and resolution

• dominant overstory species within a landscape differences between the two. The objectives ofand include: number of patches, percent of a this paper do not include comparisons overclass in a landscape, mean patch size of the time, and inappropriate inferences should beclass, and average perimeter/area ratios avoided.(shape indices). Class metrics are known fortheir application in estimating levels of frag- IV. RESULTS AND DISCUSSION! ,

mentation or in determining the degree towhich the matrix (most connected) class is A. Landscape Compositionfragmented by another class type. Landscape

J statistics relate to the landscape or LTA group 1. Historic Landscape Composition: No#

as a whole and include number of patches, areas were mapped as "urban" in MNFI'stotal area, mean patch size, and patch size transcription of the GLO maps as we have forvariability; diversity of classes, dominance of the present cover; rather, a separate map layerone class over others, and shape, or fractal was created by MNFI showing cultural settle-indices. Landscape-scale metrics are used to ments, primarily Indian villages. Non-veg-examine landscape diversity and structure of etated areas did not occur in the outwashan entire landscape, or to compare landscapes group, and agriculture did not occur in any ofof different time periods or location the four groups. Herb/openland occurred only

• (MacGarlgal and Marks 1993).

7

in outwash, and shrub land did not occur in in the present land cover were quite dense," any of the groups studied. Oak cover type was and dominated by cedar. In the lowland

absent in an LTA groups. Neither dry nor wet group, however, cedar stands were less dense,hardwood-conifer mixes were found in the or included other species.historic datasets. Red pine was not found inthe bedrock group, and white spruce was B. Landscape Measuresabsent from the historic lowland landscape(map 3a, table 4). 1. Historic Landscape Metrics: Of the four

.

LTA groups studied, the lowland group was2. Present landscape composition: The largest in total area (509,511 ha), with 5,409present coverage was substantially more patches, and the outwash group was smallestdetailed due to the higher resolution from the (198,831 ha), with 1,640 patches (table 5). InThematic Mapper (TM)classification (map 3b. the historic landscape, LPI was highest in theHowever, a few cover types were absent in the moraine group (10.280), with a mean patchLTA groups considered (table 4). There were size of 121 ha and lowest in the lowland groupno oak, balsam fir, or hemlock types in any of (2.757, MPS = 94 ha), although the other threethe present datasets. In the original present groups (bedrock, lowland, and outwash) werelegend, most conifer types were coded into two relatively close (table 5, fig. I). Because LPI inclassifications: <70-percent but >40-percent the lowland group was lower than in the othercanopy closure contributed by that species, groups, we can determine that the lowlandand >70 percent canopy closure contributed MPS was less affected by a single large patchby that species. In the LTA groups assessed, (table 5). The four groups had relatively closecedar occurred only in the >70-percent canopy values for PD and SHDI. PD was highest inclosure class, except for the lowland group, the lowland group (1.062) and lowest in thewhere it occurred only in the <70-percent bedrock group (0.63 I). Outwash had theclass. In other words, in the bedrock, mo- highest diversity historically (SHDI = 2.069),mine, and outwash groups, all cedar patches and lowland had the lowest (0.520).

12

I0

8o_

m_

x 6"o

4

°

0 I ,' IBedrock Lowland Moraine Outwash

LTA group

_ Largest patch index __ Patch density _ Shannon's diversity index

Figure 1.---Values for three landscape level _Ldlces: largest patch index (LPI),patch• density (PD), and Shannon's Diversity Index (SHDI) calculated by LTA group on

the historic land cover.8

" 2. Present Landscape Metrics" Similar outwash had more mixed pine than otherresults were obtained for the present land- groups, with no one class being as dominant.scape, in which the lowland group had the The lowland group was historically moregreatest number of patches (63,994), and the dominated by mixed conifer and wetlands (fig.remaining three groups all had approximately 3a).25,000 patches. LPI was highest in the mo-raine group (5.63 I), but due to the higher In the present moraine landscape, northernresolution, MPS was only I 1 ha (fig. 2). Low- hardwoods were most prevalent (LSIM). Inland had the lowest LPI again (I. 170, MPS = 8 present bedrock, northern hardwoods andha). MPS for all four groups was between 8 mixed conifer were still important, but agricul-and 11 ha. Present PD was highest in the ture was as well, rather than cedar. Thelowland group (12.560) and lowest in the present outwash landscape was quite wellmoraine group (9.056), although all four distributed among several classes includinggroups reflected high PD due to higher resolu- northern hardwoods, mixed pine, mixedtion data. Diversity for all groups in the conifer, white pine, wetlands, and agriculture.present landscape was low (0.578 - 0.615) The lowland group, presently, was weighted to(table 5, fig. 2). mixed conifer and northern hardwoods, prima-

rily, followed by wetlands and mixed pine (fig.C. Class Measures 3b).

Northern hardwoods, especially, and mixed Classes with high LPI historically in the too-conifer to a lesser extent were the most preva- raine group were northern hardwoods, mixedlent (LSIM) of the 10 classes analyzed in the conifer, mixed pine, and cedar. The wetlandshistoric moraine landscape. Historic bedrock class had the highest LPI in historic lowland.had a relatively high amount of mixed conifer, In the bedrock and outwash groups, the

• northern hardwoods, and cedar, while historic classes were less driven by single large

Figure 2._Values for three landscape level indices: largest patch index (LPI),patch denslhj (PD), and Shannon's Diversity Index (SHDI) calculated by LTAgroup on the present land cover.

I0

0 :_,.........................................................................................................................................................................................................................................................................................................................................................

° 20-i. ................................................................................................................................................................................................................................................................................................................. i.i .................I1

lo i !i

,.,- -4- _....R..... t--- _;.......... i .... _....Non-veg Agricul . N.hdwds Aspen Low hdwd Wetlands W.pine Mix pine Cedar Mix conifer

Cover class

_ Bedrock _ Lowland _ Moraine Outwash

Figure 3a.--Landscape s/m//ar/ty index (LSIM) calculated by cover type for each LTA group• using the historic landscape dataset (MNFI 1994). LSIM measures the proportion of one

cover hjpe in area per the total landscape (LTA group) area.

60 [

: ! ]I

i i504...........................................................................................................................................................................................................................................................................................................................I

40J t, I 1

• - _ -. !• ." _,_ t t

> 30-.J !i" ".....................................................................................................................................................................................................................................|

20-!...................................................................................._ i

Non-veg Agricul N'.I_dwds Aspen Lowhdwd Wetlands W.pine MIXpine Cedar MixconiferCoverclass

Figure 3b.--Landscape similarity index (LSIM) calculated by cover type for each LTA group• using the present landscape dataset (Maclean 1994). LSIM measures the proport/on of

one cover hjpe in area per the total landscape (LTA group) area.11

patches. In historic bedrock and outwash, is weighted per unit area. In historic bedrock,classes with high LPI were northern hard- mixed conifer, cedar, and northern hardwoodswoods, mixed conifer, cedar, and mixed pine had the most variable pattern. Historic(table 6). In present bedrock and outwash, outwash was similar with mixed conifer,classes with high LPI included northern northern hardwoods, and mixed pine havinghardwoods, agriculture, mixed pine and white the highest PSCV. But, in historic moraine,pine (table 7). Of all four groups in the northern hardwoods and wetlands had thepresent coverage, wetlands had the highest LPI most variable size pattern, while wetlands andin the lowland group, indicating larger interior mixed pine had high PSCV in historic lowland,¢wetland areas within the lowland group. (table 6). Presently, the moraine and outwashOverall, the moraine group had the largest groups had variable patterns of mixed andsingle patches among an cover types, but white pine and northern hardwoods. Howevelespecially in northern hardwoods, which in present bedrock, northern hardwoods andsuggests a greater amount of core areas/ agriculture were most variable, and in presenlinterior habitats in the moraine landscapes, lowlands, northern hardwoods had high PSC_In the bedrock and moraine landscapes, MPS (table 7).was very high in the northern hardwoodsclass, but both groups also had high LPI and Mean shape index (MSI) was fairly even in th_PSCV, which implies there might be a few very historic landscape among landscape groupslarge patches and many sman patches in these and cover classes, except that the non-veg-groups, etated had much higher MSI values in all bu

the outwash group (fig. 5a). Present shapeThe greatest number of patches for any one indices were also quite even among groupsclass type was 6,076 patches of mixed pine in and classes, with non-vegetated class gener-

• the present lowland landscape (tables 6 and ally slightly higher. In the present landscap7). This was believed to be a reflection of the high MSI for non-vegetated areas was attriblowland sand ridge-swale complexes found uted to the inclusion of roadways within th_within this LTA group, in which a mosaic spectral class, which have high perimeter tcWetland/bog landscape is interspersed by a area ratios, and therefore, high MSI. Asperpattern of linear sand ridges vegetated with was generally the lowest in MSI of the 10Jack pine and other seral species. Patch classes, and all landscape groups reflecteddensity indicates how fine the pattern of nearly the same MSI value in aspen (fig. 5blandscape is, or how much it is fragmentedinto many small patches. PD is measured as D. Assumptions About the Data andthe number of patches per 100 ha. In all LTA Inherent Bias or Limitationsgroups, except lowlands, historically andpresently, mixed conifer had the highest PD. This project used three different geographlIn the lowland group, mixed pine, wetlands, datasets, each developed by different met]and northern hardwoods had high PD (tables As a result, each land cover layer had uni,6 and 7). Also, in the historic bedrock group, limitations. The LTA coverage, discussednon-vegetated areas had very high PD, even previously, was derived from intuitive malthough the mixed conifer class was highest procedures in which the boundaries repr_(fig. 4a). PD was much lower in historic data sented intangible ecological transitions.due to the resolution differences and should result, neither map error nor accuracy c¢not be compared to the present data. In the be estimated for this layer. Nevertheless,present landscape, all LTA groups had high a meaningful framework for landscape a.,mixed conifer PD, although in the lowland ments and planning, and one that can b,group, wetlands and mixed pine were also strengthened by studies of this type. Thquite high. All four groups reflected slmilar units, like regional ecosystems, should tmoderate PD values in the agriculture, north- considered hypotheses for testing, devel,ern hardwoods, aspen, and mixed pine classes from ecological theory and knowledge of(fig. 4b). landscape. By using the units for studi,

such as this, we can identify units thatWe refined our measure of landscape pattern functionally different versus those thatby considering how regular or variable the respond to landscape disturbances in spattern is. PSCV measures patch size start- ways (Albert 1995).dard deviation divided by mean patch size and12

0.5

0.45

0.4

0.35

® 0.3

, >_0.25

0.2

0.15

0.1

O.05 _ I - - _..__- -o : : : g'-'- : _'- i :__- : '_ I

Non-veg Agrtcul N.hdwds Aspen Lowhdwd Wetlands W.plne MIXpine Cedar MixconiferCoverclass

_Bedrock _ Lowland _Moraine _Outwash

Figure 4a.--Patch density, the number of patches per 100 ha, calculated by cover type as afunction of LTA group using the historic landscape dataset (MNH 1994).

" 1.5

_=¢o> 1

' . . r_i

• ,

0.5

o l INon-veg Agricul N.hdwds Aspen Lowhdwd Wetlands W.pine Mixpine Cedar Mixconifer

, Coverclass

_ Bedrock _ Lowland _ Moraine _ Outwash

Figure 4b.---Patch density, the number of patches per 100 ha, calculated by cover type as afunction of LTA group using the present landscape dataset (Maclean 1994).

13

4[

3.5 -_.........................................................................................................................

3 ............ R ........................................................................................................................................................... _

t ti]' i• l i

>: 2!........;! ..................................................................................................................................................................................i

1._! i,ii: !i i1!......i! .............................................................i ...........I .............................i•-I!!i_i_:_ :i ! i

Non-veg Agricul N.hdwds Aspen Low hdwd Wetlands W.pine Mix pine Cedar Mix conifer

Cover class

. _Bedrock _ Lowland _Moraine _Outwash

Figure 5a.---Meen shape index, a perimeter to area ratio, cal_J!_ted by cover type as afunction of LTA group using the historic landscape dataset (MNFI 1994).

4

3.5 'I" .............................................................................................................................................................................................................

!

!o3 2.5 ....t.............................................................................................................................................................................................................................

I2- i....................................................................................................................................................................................................................

. N a,__'NN. 1.5 i ................................ ....... i

.,, 1

Non-veg Agricul N hdwds Aspen Low hdwd Wetlands W.pme Mix pine Cedar Mix conifer

Cover class

Figure 5b.---Mean shape index, a perimeter to area ratio, calculated by cover type as afunction of LTA group using the present landscape dataset (MNFI 1994).

14

Sampling completeness of the land cover hardwoods and mixed conifer were the most. datasets was another concern. The historic prevalent cover types of the 10 studied, his-

coverage was based on the equivalent of a toricaUy and currently. Northern hardwoodssystematic aligned sampling strategy. The were especially prevalent in the morainesurveyors walked and recorded trees along the group, while the mixed conifer type was moretownship and section survey lines (1-mile prevalent in the bedrock group. Wetlands andintervals). The transcribed digital data from mixed pine, in addition to northern hard-MNFI, however, contained little attribute woods, were also prominent in the lowland

• information, or legend about the cover types group. In the outwash group, these types werederived from the original line tree data, except also present but mixed and white pine werethe compositional information derived from more prevalent. Largest single patches (LPI)this study, and its resolution was definable were found to be redundant to LSIM, and

' only to 1 square mile. These data limitations therefore were not assessed to the same extentare not uncommon to landscape research, as other indices. However, in comparisons of

•however. Frelich's (1995) study of "old forest" mean patch sizes (MPS) of classes, LPI can bechanges in the Lake States over the past 150 used to gage how much MPS is determined by

_years relied on historic vegetation maps by a single large patch. LPI also may suggestVeatch (1928), who inductively mapped his- areas with more interior habitat. The highesttoric Vegetation from GLO notes and his LPI values were found in northern hardwoodsknowledge of soil patterns at a broader scale within the moraine group. Patch density wasthan the maps by MNFI. useful in comparing the scale of pattern within

landscapes. The bedrock group had the lowestThe present coverage was derived from differ- patch density, or coarsest (least patchy) pat-ent methodology: classification of sateUite tern historically, yet in general, mixed conifer,imagery, which comprehensively recorded land which was a dominant class in the bedrockcover in the total study area based on the group, tended to have higher patch densityspectral signature of vegetation types within than other cover types. Generally, cover typesthe image area, rather than a sample. Resolu- in the moraine group had the most variabletionwas as high as 60 m to an accuracy of (least regular) pattern, but no one cover class90.2 percent. However, some types, such as was consistently more variable than others.non-vegetated, had an accuracy as low as28.6-percent correct. Furthermore, the Shape indices, based on perimeter to area

. present coverage, like the historic, lacked a ratios, were relatively consistent across land-compositional/structural legend for the cover scape groups and cover classes, with theclasses. Neither historic or present datasets exception of the non-vegetated class. Non-described, or were intended to describe, verti: vegetated areas had higher shape indices incal structure. Nor did they provide detailed both historic and present coverages. In thespecies composition information, beyond the present coverage, this finding was attributedcompositional assessment included in this to inclusion of roadways, which have highstudy, as would maps developed from plot perimeter/area values in the non-vegetatedsamples and quantitative ecological classiflca- cover class. However, it is not clear whattion techniques, factors would have contributed to high shape

indices in historic non-vegetated areas., V. CONCLUSIONS

i ' Diversity, based on Shannon's Diversity Index,This study has quantified landscape patterns was highest in the outwash group and lowest•for four physiographically based landtype in the lowland group of the historic land-association groups across the Eastern Upper scapes. In present landscapes, diversity wasPeninsula of Michigan from historic (1840's) very similar among the four groups; it wasand current (1990-1992) datasets. Prevalence only slightly higher in the lowland group andor dominance of cover classes was based on lower in the moraine group than in others.class area (CA) and landscape similarity index(LSIM), or class area weighted by total land- Conclusions about a specific cover type mayscape area. In the four groups studied-- be drawn by referring to tables 6 and 7. Thesebedrock-controlled, lowland sand lake plain, tables should be a reference for land managers

•morainal origin, and outwash--northern and planners making decisions based on

15

knowledge of actual landscape structure. The Albert, Dennis A.; Denton, Shirley R.; Barnes,broad landscape data in this report may be Burton V. 1986. Regional landscape eco-used in conjunction with s/milar landscape systems of Michigan. Ann Arbor, MI: Uni-studies in the Lake States or in other areas versity of Michigan, School of Natural Re-with similar biophysical landscape units to sources. 32 p.estimate patterns beyond northern Michigan.

Allen, T.F.H.; Start, Thomas B. 1982. HIerar-ACKNOWLEDGMENTS ©hy: perspectives for ecological complex-

ity. Chicago, IL: University of Chicago Press.Partial support and funding were provided by 30 p.the Michigan Department of Natural Re-sources, the National Council for Air and Anderton, John A. 1993. PaleoshorelineStream Improvement, The Nature Conser- geoarchaeology in the Northern GreatVancy, Michigan Technological University, the Lakes: Hiawatha National Forest. HeritageHiawatha National Forest, and Washington Program Monograph No. 1.State University. Recognition for developmentof spatial land cover datasets goes to Dr. A. Barley, Robert G. 1996. Multi-scale ecosys-Maclean and Maclean Consultants, Ltd., and tern analysis. Environmental Monitoring &to Dr. D. Albert of Michigan Natural Features Assessment. 39:21-24.Inventory. Dr. E. Padley and J. K. Jordan,with the Eastern Region of the Forest Service, Barley, I_G.; Avers, P.E.; King, T.; McNab,and I_ L. Watson, initially developed the W.H., eds. 1994. Ecoregions and subre-landtype associaUon mapping on Forest gions of the United States. Washington,Service lands in this region, and collaborated DC: U.S. Department of Agriculture, Foreston expansion of the map units across the Service, ECOMAP Team. 1 mapstudy area. We also appreciate the vision of (1:7,500,000).D. Howlett, Ecosystem Team Leader, to extendthe landtype association mapping beyond the Baker, William L. 1991. Effects of settlementboundaries of the Hiawatha National Forest. A and fire suppression on landscape strue-number of other people on the Hiawatha ture. Ecology. 73: 1879-1887.National Forest, especially W. F. Spinner, D.Evans, J. Schultz, D. Gerdes, K. Seleen, and Ball, Janet; Padley, Eunice A. 1993. DraftI_ Brittain greatly advanced the landtype delineation of landtype associations inassociation delineation program. AppreciaUon the Eastern Upper Peninsula of Michigan.also goes to J. Johnson and j. Durbin, of On file, Hiawatha National Forest.Mead Paper, Inc., in Escanaba, Mich/gan, andto Q. Zhang, C. Herlevich, Q. Zhao, and M. Barnes, Burton V. 1984. Forest ecosystemHyslop of Michigan Technological University, classification and mapping in Baden-for GIS preparaUon, and Dr. N. Crookston, of Wurttemberg, West Germany. In:the Intermountain Forest Experiment Station Bockheim, J.G., ed. Proceedings of thein Moscow, for computer support. Special symposium on forest land classificaUon:

•thanks to Dr. T. Crow of the Forestry Sciences experience, problems, perspectives; 1984Laboratory, North Central Forest Experiment March; Madison, WI. NCR-102. 276 p.Station, in Rhinelander, Wisconsin for his'support and many suggestions on this project. Bourdo, Eric A., Jr. 1956. A review of the

General Land Office survey and of its useLITERATURE CITED in quantitative studies of former forests.

Ecology. 37(4): 754-768.Albert, Dennis ,_ 1995. Regional landscape

ecosystems of Michigan, Minnesota, and Canham, Charles D. 1978. CatastrophicWisconsin: a working map and classiflca- windthrow in the hemlock-hardwoodtion. Gen. Tech. Rep. NC-178. St. Paul, MN: forest of Wisconsin. Madison, WI: Unlver-U.S. Department of Agriculture, Forest sity of Wisconsin. 94 p. M.S. thesis.Service, North Central Forest ExperimentStation. 250 p.

16

Canham_ Charles D.; Loucks, Orie L. 1984. GoUey, Frank B. 1995. Reaching a landmark.Catastrophic windthrow in the pre-settle- Landscape Ecology. 10(1): 3-4.merit forests of Wisconsin. Ecology. 65(3)"803-809. Hess, George. 1994. Pattern and error in

landscape ecology: a commentary. Land-Crow, Thomas I_ 1991. Landscape ecology: scape Ecology. 9(I)" 3-5.

the big picture approach to resourcemanagement. In: Decker, Daniel J., et a/., Hulshoff, Maureen I_ 1995. Landscape indi-eds. Challenges in the conservation of bio- ces describing a Dutch landscape. Land-lOgical resources: a practitioner's guide, scape Ecology. 10(2): 101-111.Boulder, CO: Westview Press: 55-65.

Jordan, James K.; Padley, Eunice; Cleland,Curtis, JohnT. 1959. Vegetation of Wiscon- David T.; Gates, Joseph ,_; Hoppe, David J.;

sin. Madison, WI: University of Wisconsin Kempf, Leonard S.; Leuelling, Barbara;Press. 657 p. ' Seleen, Kirstin L.; Shadis, David A.;

Sflbernagel, Janet. In prep. Landtype asso-Dunn, Christopher P.; Sharpe, David M.; ciations--odgin, concepts, mapping, and)

Guntenspergen, Glenn I_; Stearns, Forrest; application in the Lake States NationalYang, Zhao. 1991. Methods for analyzing Forests. St. Paul, MN: U.S. Department oftemporal changes in landscape pattern. Agriculture, Forest Service, North CentralIn: Turner, Monica G.; Gardener, Robert H., Forest Experiment Station.eds. Quantitative methods in landscapeecology. Ecol. Stud. 82. New York, NY: King, ,aLW. 1993. Consideration of scale andSpringer-Verlag: 173-198. hierarchy. In: Woodley, S.; Roy, J.; Francis,

G., eds. Ecological integrity and the manage-ECOMAP. 1993. National hierarchical frame- ment of ecosystems. Delray Beach, FL: St.

• work of ecological units. Washington, DC: Lucie Press. 220 p.U,S. Department of Agriculture, ForestService. KliJn, Frans; Udo de Haes, Helias _ 1994. A

hierarchical approach to ecosystems andFarrand, W.I_ 1982. Quaternary geology of its implications for ecological land classi-

Michigan. Ann Arbor, MI: State of Michigan flcation. Landscape Ecology. 9(2)" 89-104.. Department of Natural Resources, Geological

Survey, University of Michigan. 2 maps Li, Habin; Reynolds, James F. 1993. A newcontagion index to quantify spatial pat-

Forman, Richard T.T. 1995. Some general terns of landscapes. Landscape Ecology.principles of landscape and regional 8(3)" 155-162.ecology. Landscape Ecology. 10(3): 133-142.• Li, Habin; Reynolds, James F. 1994. A shnula-

Forman, Richard T.T.; Godron, Michael. 1986. tion experiment to quantify spatial het-Landscape ecology. New York, NY: John erogeneity in categorical maps. Ecology.Wiley and Sons, Inc. 619 p. 75(8): 2446-2455.

Frelich, Lee E. 1995. Old forest in the Lake Lorimer, Craig G. 1977. The presettlementStates today and before European settle- forest and natural disturbance cycle ofment. Natural Areas Journal. 15: 157-167. northeastern Maine. Ecology. 58: 139-148.

Frelich, Lee E.; Lorlmer, Craig G. 1991. Natu- Lorimer, Craig G. 1980. The use of landral disturbance regimes in hemlock- survey records in estimating presettle-hardwood forests of the upper Great Lakes ment fire frequency. In: Stokes, MA.;region. Ecological Monographs. 61(2): 145- Dieterich, J.H., eds. Proceedings of the Fire164. history workshop; 1980 October 20-24;

Tucson, AZ. Gen. Tech. Rep. GTR-RM-81.Gary, Margaret; McAfee, Robert, Jr.; Wolf, Fort Collins, CO: U.S. Department of Agricul-

Carol L., eds. 1974. Glossary of geology, ture, Forest Service, Rocky Mountain ForestWashington, DC: American Geological Insti- and Range Experiment Station. 142 p.

• tute. 805 p.17

Lor/mer, Craig C. 1984. Methodological Muller, Michael I_; Middleton, John. 1994. Aconsiderations in the analysis of forest Markov model of land-use change dynam-disturbance history. Canadian Journal of ics in the Niagara Region, Ontario,Forest Research. 15: 200-213. Canada. Landscape Ecology. 9(2): 151-157.

Luque, Sandra S.; Lathrop, Richard G.; Noss, Reed F. 1985. On characterizingBognar, John A. 1994. Temporal and spa- presettlement vegetation: how and why.tial changes in an area of the New Jersey Natural Areas Journal. 5: 5-19.Pine Barrens landscape. Landscape Ecol-ogy. 9(4}: 287-300. O_leill, I_V.; Drummel, J.l_; Gardner, I_H.;

Sugthara, G.; Jackson, B.; DeAngelis, D.L.;MacGarlgal, Key/n; Marks, Barbara J. 1993. Mflne, B.T.; Turner; M.G.; Zygmunt, B.;

FRAGSTATS: spatial analysis program for Chr/stensen, S.W.; Dale, V.H.; Graham, I_L.quantifying landscape structure. Corvallis, 1988. Indices of landscape pattern. Land-O1_ Oregon State University. 62 p. scape Ecology. 1(3}: 153-162.

Maclean, Gordon, A. 1994. Inventory of deer Pastor, John; Broschart, Michael. 1990. Thehabitat in Michigan's Upper Peninsula spatial pattern of a northern conifer-utilizing Landsat thematic mapper data. hardwood landscape. Landscape Ecology.Final report for Michigan Department of 4(1): 55-68.Natural Resources, _sing, MI.

Peterson, W.L. 1986. Late WisconsinanMartin, Helen M. 1957. Map of the surface glacial history of Northeastern Wisconsin

formations of the Northern Peninsula of and Western Upper Michigan. USGS Bull.Michigan. Publ. 49. Michigan Department of 1652. Washington, DC: U.S. Geological

• Conservation, Geological Survey Division. 1 Survey. United States Government Printing•map. Office. 14 p.

McNab, W. Henry. 1996. Classification of Price, David L. 1994. An ecological study of•local- and landscape-scale ecological the composition, structure and distur-types in the Southern Appalachian Moun- bance regimes of the pre-European settle-tains. Env/ronmental Monitoring & Assess- ment forests of western Chippewament. 39:215-229. County, Michigan. East Lansing, MI: Michi-

gan State University, Department of For-McNab, W. Henry; Avers, Peter E., eds. 1994. estry. 214 p. M.S. thesis.

Ecological subregions of the UnitedStates: section descriptions. WO-WSA-5. Rowe, J. Start. 1991. Forests as landscape

• Prepared in cooperation with R_gional ecosystems: implications for theirCompilers and the ECOMAP Team of the regionalization and classification. In:Forest Service. Symposium: Ecological land classification:

applications to identify the productive poten-Mflne, Bruce T. 1991. Lessons from applying tial of southern forests; 1991 January 7-9;

fractal models to landscape patterns. In: Charlotte, NC. Ashevflle, NC: U.S. Depart-.Turner, Monica G.; Gardener, Robert H., eds. ment of Agriculture, Forest Service, South-

Quantitative methods in landscape ecology, eastern Forest Experiment Station. 149 p.Ecol. Stud. 82. New York, NY: Springer-Verlag: 199-235. R_we, J. Stan. 1996. Land classification and

ecosystem classification. EnvironmentalMNFI. 1994. Pre-European settlement land- Monitoring & Assessment. 39:11-20.

scape of Michigan. Lansing, MI: MichiganNatural Features Inventory. Santer, Richard _ 1977. Michigan: heart of

the Great Lakes. Dubuque, IA: Kendall/Nature Conservancy. 1993. IAmdscape con- Hunt Publishing Co. 364 p.

servation--the next big step. News fromthe Michigan Chapter of The Nature Conser-vancy, Autumn.

18

Simpson, John W.; Boerner, Ralph E.J.; Uhlig, Peter W.C.; Jordan, James K. 1996. ADeMers, Michael N.; Berns, Leslie IL; spatial hierarchical framework for the co-Artigas, Francisco J.; Silva, Alejandra. 1994. management of ecosystems in Canada andForty-eight years of landscape change on the United States for the Upper Greattwo contiguous Ohio landscapes. Land- Lakes Region. Environmental Monitoring &scape Ecology. 9(4): 261-270. Assessment. 39: 59-73.

Slms, Richard ,_; Corns, Ian G.W.; Kllnka, Urban, Dean L.; O'Neill, Robert V.; Shugart,Karel. 1996. Introduction--global to local: Herman H., Jr. 1987. Landscape ecology, aecological land classification. Environmen- hierarchical perspective can help scien-tal Monitoring & Assessment. 39: 1-10. tists understand spatial patterns. Bio-

, science. 37: 119-127.Smith, Marie-Louise; Carpenter, Constance.

i996. Application of the USDA Forest USGS. 1990. In: Anderson, James I_, et aL,Service national hierarchical framework eds. A land use and land cover classiflca-of ecological units at the sub-regional tion system for use with remote sensor

, level: the New England-New York example, data. Reston, V_: U.S. Geological Survey.Environmental Monitoring & Assessment.39: 187-i98. Veatch, J.O. 1928. Reconstruction of forest

cover based on soil maps. Quarterly BulIe-SOLEC. In Press. Impacts of changing land tin of the Michigan Agricultural Experiment

use. In: Thorp, Steve; Rivers, Ray; Pebbles, Station. I0:116-126.Victoria, eds. Proceedings of the 1996 Stateof the Lakes ecosystem conference; 1996 Veatch, J.O. 1959. Presettlement forests ofNovember; Windsor, Ontario. EPA 905-D-96- Michigan. East Lansing, MI: Department of

• 00 l e. Windsor, Ontario: Environment Natural Resources, Michigan State Univer-Canada and U.S. Environmental Protection sity. Map.

.. Agency.Whitney, Gordon G. 1986. Relation of

Spurr, S.H.; Barnes, Burton V., eds. 1980. Michigan's presettlement pine forests toForest ecology. 3d ed. New York, NY: John substrate and disturbance history. Ecol-Wfley and Sons. 687 p. ogy. 67: 1548-1559.

Turner, Monica G.; Ruscher, C. Lynn. 1988.Changes in landscape patterns in Georgia,USlL Landscape Ecology. 1(4): 241-251.

19

------._._______--________

°

APPENDIX A:

Formulae used to calculate landscape, class, and relative indices

Landscape Indices:1. Number of Patches = N(lJ)

2. Total Area = SUM a(iJ) = A (ha)

3. Mean Patch Size (MPS) = SUM a(iJ)/N(ij)

4. _est Patch Index (LPI) = max a(iJ)/A * I00

5. Patch Density (PD) = n(iJ)/A * I0000 * I00 (# per I00 ha)

6. Diversity {SI-IDI)= -SUM (P(i) * log P(i))Class Indices:

1. Number of Patches = N(J)

2. Class Area = SUM a(J) = CA (ha)

3. Mean Patch Size (MPS) = SUM a(J)/N(i)

4. Landscape Simfliarity Index (LSIM) = SUM a(j)/A * I00, or CA/A * I00

5. Largest Patch Index (LPI) = max a(J)/CA * 100

6. Patch Density (PD) = n_])/CA * I0000 * I00 (# per I00 ha)

7. Patch Size Coefficient of Variation (PSC_ = patch size standard deviation / mean patch size

8. Mean Shape Index (MSI) = SUM (p(J)/2sqrt(pie*aU))) / N(J)

a(iJ} = area of patch iJ

p(iJ) = perimeter of patch lJ

P(i} = proportion of patch type I = a(iJ)/A

0

2O

Table l.--IJst of cover classes and numeral codes for historic land cover map developed by MN_ (1994)GIX) Surveyors' notes of 1840"s; legend is based on expanded Michigan Resowce Informatk)n

System (MIRIS) legend of land use cover classes

I

Category • 1st order 2nd order 3rd order 4th order Land cover class description

Palustrlne62 Non-forestedWetland

, .. 622 Emergentmarsh�meadow�prairie6221 Eemergentmarsh

, 6222 Great lakes marsh6223 Interdunalwetland

" 6224 Wet meadow6225 Inlandsalt marsh6226 Llakeplainprairie6227 Inlandwet prairie6228 Intermittentwetland

623 Mudflats6231 Marlflats

612 Shrub-dominatedwetland6121 Bog6122 Alder/willow/Bogbirchthicket6123 Bottonbush/DogwoodNVi!lowswamp6124 Patternedpeatland6125 Muskeg

91 LandscapeComplex911 Wooded dune/swalecomplex

4 ForestedWetlands41 Hardwood/Conifer- Hardwoodspredominating

414 Lowlandhardwood(broadleaf)4141 Ash4142 Elm4143 Silver/Redmaple4144 Cottonwood4145 Balsampoplar4146 Aspen4147 White birch4148 blackwillow

42 Hardwood/Conifer-Coniferspredominating423 Lowlandconifer

• .- 4231 Cedar4232 blackspruce4233 Tamarack4234 Balsamfir/Whitespruce4235 Balsamfir

, 4236 Jackpine, 4237 Hemlock

4238 White pine

LacUstrine and Rlverine51 MajorRiver52 Lake or Pond53 Great Lakes

(table I continued on next page)

21

i ___.._ _ _--_- _mr_.m = =-.. - ,i ,--, - -, _-- .- .=___

li

(tab/e 1 .oont/nued)

Category let order 2nd order 3rd order 4th order Land cover class description_.

Terrestrial3 Non-Forested(grassland- savannas)

31 Herbaceous- UplandGrassland32 Shrub- ShrubSavanna33 Tree Savanna

331 Lakeplainoak opening332 Oak barrens333 Pinebarrens334 Oak/pinebarrens335 Buroak savanna

• 336 Oakopening4 Forested

411 Northernhardwoods4111 Sugarmaple, Beech4119 Beech,Hemlock

412 Centralhardwoods4121 Beech,Sugar maple,

Basswood,Redoak4122 Whiteoak, Hickory,Blackoak4123 Blackoak, Whiteoak4124 Pin/Blackoak

413 Aspen/Whitebirch421 Pine

' 4211 Whitepine4212 Red pine4213 Jackpine4215 Red pine/Jackpine4216 Red pine/Whitepine4217 Whitepine/Whiteoak '4218 Red pine/Oak4219 White pinelBeechlMaple

422 Other uplandconifer4221 White spruce4223 FirlSprucelCedar4226 Hemlock4227 Hemlock/Whitepine

• 4228 Hemlock/Sugarmaple• .

4229 HemloddYellow birch. Open, LltUe/NoVegetation

72 Beach,Riverbank73 Open SandDune74 ExposedBedrock

741 Alvar742 Bedrockglade743 Sinkhole744 Limestoneledge/Outcrop745 Sandstoneledge/Outcrop746 Igneous-met.ledgelOutcrop

(table I continued on next page)

22

(table I conttnue_

Category 1st order 2nd order 3rd order 4th order Land cover class description

Natural -DistUrbances/

Cultural Feature92 Wildfire93 Windthrow94 Beaver Flooding95 Great LakesLevelChange96 CulturalFeatures97 Jack PineThicket

Wetland Grade I IntactD DegradedE EliminatedM Manipulated

23

Table 2.DLIst of cover classes and numer_al codes for present land cover map (Maclean 1994) through classifi-cation of 1991 Thematic Mapper satellite imagery

Category Cover type name Cover class Numerical code

Non-ConiferousUrban 1

Non-vegetative •2Agricultural-cropland 3Herbaceousopenland 4

Shrubland 5Northernhardwood 6

Oak 7Aspen/birch 8

Lowlandhardwoods 9Dry hardwood/conifermix 10Wet hardwood/conifermix 11

Wetlands 12Water 13

ConiferousPines Red pine 14

Jackpine 15White pine 16

Other(mixedpine) 17Tamarack 18Hemlock <70% crownclosure 19

>70% crownclosure 20. Blackspruce <70% crownclosure 21

>70% crownclosure 22White spruce <70% crownclosure 23

>70% crownclosure 24Balsamfir <70% crownclosure 25

>70% crownclosure 26White cedar <70% crownclosure 27

>70% crownclosure 28Mixedconifer <70% crownclosure 29

>70% crownclosure 30

24

L

Table 3.--List of 24 Cross-llsted cover classes used to compare historic and present land cover maps

MNFI Code" TM code2 Class name MNFI Code _ TMcode: Class name

not applicable 1 Urban 6221 1272 2 Non-vegetated 6222 1273 2 6223 1274 2 6224 12

741 2 6227 12742 2 6228 12743 2 6231 12744 2 51 13 Water745 2 52 13746 2 54 13

notapplicable 3 Agriculutralcropland 4212 14 Red pine31 4 Herbaceousopenland 4211 15 Jackpine32 5 Shrubland 4238 15

4111 6 Northernhardwoods 333 16 White pine4119 6 334 16332 7 Oak 4213 16

4121 7 4236 164122 7 4215 17 Mixedpine4123 7 4216 174124 7 4217 17413 8 Aspen/Birch 4218 17•41 9 Lowlandhardwoods 4219 17414 9 4233 18 Tamarack4141 9 4226 19 Hemlock4142 9 4227 194143 9 4228 194144 9 4229 194145 9 4237 19

. 4146 9 4232 21 Blackspruce4147 9 4242 214148 9 4221 23 White spruce

notapplicable 10 Dry hardwood-conifermix 42 25 Balsamfir42 11 Wet hardwood-conifermix 4234 25

6121 12 Wetlands 4235 256122 12 4231 27 Whitecedar.6124 12 423 296125 12 4223 29 Mixed conifer

• .

_Michlgan Natural Features Inventory (MNFI 1994) code from historic cover class legend (table I).='[hematlc Mapper code from present cover class legend (Maclean 1994, table 2).

25

'---.,-..,-_,__,__,__

f

29

°

80

The UnitedStates Departmentof Agriculture(USDA) prohibitsdiscriminationin itsprogramsonthe basisof race, color,nationalodgin, sex,religion,age, disability,politicalbeliefs,and maritalor familialstatus. (Not all prohibitedbases applytoallprograms.) Personswithdisabilitieswho requirealternativemeansof oommunicationof programinformation(braille,largeprint,audiotape,etc.) shouldcontactthe USDA

•Office ofCommunioationsat (202) 720-2791.

Tofile a cornplaint,writethe Secretaryof Agriculture,U.S. Dq)anrnent ofAgriculture,Washington,13(320250, or call 1-800-245-6340 (voice),or 202-720-1127 (TDD).USDA is an equal employmentopportunityemployer.

o

' _lB_ Pdnted on recydable paper.41

g_'U.S. GOVERNMENT PRINTING OFFICE: 1997-557-052

Sflbernagel, Janet; Chen, Jiquan; Gale, Margaret R.; Pregitzer, Kurt S.;Probst, John.

1997. An interpretation of landscape structure from historic andpresent land cover data in the Eastern Upper Peninsula of Michi-gan. Gen. Tech. Rep. NC-192. St. Paul, MN: U.S. Department ofAgriculture, Forest Service, North Central Forest Experiment Station.30 p.

Compares historic and present landscape structure among fourlandtype association groups in Upper Michigan. Provides an example

; of a landtype association framework for assessing landscape composi-•tion and pattern.

KEY WORDS: Landtype association framework, landscape composi-tion, landscape pattern.

Our job at the North Central Forest Experiment Station is discovering andcreating new knowledge and technology in the field of natural resources andconveying this information to the people who can use it. As a new generationof forests emerges in our region, managers are confronted with two uniquechallenges: (1) Dealing with the great diversity in composition, quality, andownership of the forests, and (2) Reconciling the conflicting demands of thepeople who use them. Helping the forest manager meet these challengeswhile protecting the environment is what research at North Central is allabout.

' NORTHCENTRAL_

EHPERII11ENTSTATION