DIGITIZATION OF A HISTORIC DATASET: THE WIESLANDER CALIFORNIA VEGETATION TYPE MAPPING PROJECT

2005·9.8 MB·English

by  Maggi Kelly

Madrono, Vol. 52, No. 3, pp. 191-201, 2005
DIGITIZATION OF A HISTORIC DATASET: THE WIESLANDER
CALIFORNIA VEGETATION TYPE MAPPING PROJECT
Maggi Kelly' and Barbara Allen-Diaz
Department of Environmental Sciences, Policy and Management, 137 Mulford
Hall, #31 14, U.C. Berkeley, Berkeley, CA 94720-31 14
Norma Kobzina
Marian Koshland Bioscience and Natural Resources Library, 2101 VLSB #6500,
U.C. Berkeley, Berkeley, CA 94720-6500
Abstract
In the 1920s and 1930s Albert Everett Wieslander and several others explored much of CaHfornia's
wilderness sampling vegetation, taking photographs, collecting specimens, and drawing detailed maps of
what they found. The collection is now known as the Wieslander Vegetation Type Mapping (VTM)
collection, and the entire survey encompassed nearly seventy million acres ofthe state, covering most of
the wild areas exclusive of the deserts and the larger agricultural areas. These data represent a valuable
resource for comparative and conservation ecology. We are digitizing the entire project, and making it
available to researchers and the public. The current "VTM Digitization" Project, as it is known, is
described in this article, and is a collaborative effort by teams at University of California Berkeley and
University ofCalifornia Davis, funded by the U.S. Forest Service and U.S. Department ofAgriculture's
Cooperative State Research, Education and Extension Service. The VTM Digitization Project aims to
digitize the entire VTM collection for use in ecological and geospatial analyses, and to facilitate access
and distribution of the data by researchers and interested parties. This article describes our efforts at
making the VTM plot data, maps and photographs digital, spatially referenced, readily available, and
web-accessible via an Internet-based Geographic Information System (webGIS) application. In addition,
we discuss potential uses for the data and caveats associated with its use, particularly spatial accuracy.
Key Words: VTM dataset, webGIS, digital database, California plant communities.
The 18th through early 20th century was a time "broadest coverage (Ohio to the west coast) and
of tremendous exploration of the landscape of the finest spatial resolution (one square mile) of any
United States, and the data generated by these sur- pre-settlement data source''.
veys include information about pre- and early set- Although there are numerous examples ofcoastal
tlement vegetation. The Public Land Survey Sys- maps of California dating from the 16"" century,
tem (PLS) which began in 1785 with the Land Or- surveying of the land from the Mexican border
dinance Act is probably the best known example northward did not begin until the early 19"' century
(Iverson 1988; Buisseret 1990; White and Mlad- (Heckrotte and Sweetkind 1999), and statewide
enoff 1994; Manies and Mladenoff 2000; Manies mapping endeavors began in earnest in the iTiid-19"'
et al. 2001), but there are others, including numer- century with statehood. The California PLS survey
ous 18th and 19th century land surveys, settlement began shortly after 1851, and the data collected in-
maps, and deeds (Siccama 1971; Russell 1981; Fos- cluded hydrography, vegetation, and natural re-
ter et al. 1998). While these efforts were land sur- sources, as well as cultural details (Buisseret 1990;
veys rather than botanical ones, valuable vegetation Heckrotte and Sweetkind 1999). Further statewide
data are contained in these collections, and in many efforts at mapping the natural landscape followed,
states, they provide the first statewide vegetation including an effort by the California Geological
map showing the natural environment before exten- Survey at statewide biological resource mapping in
sive changes associated with European settlement. the 1860s (Ertter 2000). The effort was abandoned
Many researchers are now interested in these col- in 1873 as a result ofseveral factors, including pol-
lections, recognizing their importance for historical itics, personnel, and poorly estimated costs.
ecological research, as well as to establish baselines Following these efforts, Albert Everett Wieslan-
for analyzing current patterns of land use change der and several others, with funding from the Forest
(Galatowitsch 1990; White and Mladenoff 1994; Service and other federal, state and county agen-
Minnich et al. 1995; He et al. 2000; Schulte et al. cies, began mapping California's vegetation in an
2002). As an example, Schulte and Mladenoff effort that was described lateras the most important
(2001) claim that the PLS records provide the and comprehensive botanical map of a large area
ever undertaken anywhere on the earth's surface
' Author for correspondence, email; mkelly@nature. (Jepson et al. 2000). The Wieslander crew explored
berkeley.edu much of California's wilderness sampling vegeta-
MADRONO
192 [Vol. 52
Table Some Potential Uses of the VTM Dataset.
1.
Data type Dataset content Potential uses
rlUl LJdla. VII^IUd1llnLl1L1cfl^1l1l1\V/*C^ u^.ilnlcclill_\y/Jc>1icJ> cU\iT xV/Cf^^CciLFt'atLQtliUrl^ln rL^-rlmlndclrlf^^C u^.itL Ilcd^ltl^V/-^l
—scape or larger scales.
rioi iviaps iju id- ana ju-minuie Use in relocating plots for analysis ofdisturbance;
USGS topographic quad- e.g., historic forest stand structure in forests across a
rangles —gradient of Sudden Oak Death infestation.
Vegetation Maps 350 15- and 30-minute topo- Retrospective mapping to examine vegetation alliance
graphic quadrangles —changes.
Compare mapping methods in key areas such as Yo-
semite, examining e.g., average size ofpolygon, mini-
—mum polygon, and number ofdifferent mapping units.
Photographs 3100 black and white photo- Graphic examples of land-use change from stand to
graphs —landscape-scale.
Evidence ofdisturbance (e.g., fire, logging).
tion, taking photographs, collecting specimens and data, as well as some of the caveats in using the
drawing detailed maps (Wieslander 1935b; Colwell data.
1977). The natural vegetation mapping was carried
out by the U.S. Forest Service (USPS) California The Vegetation Type Mapping (VTM) Collection
Forest and Range Experiment Station in Berkeley The VTM collection consists offive components,
(Colwell 1977). Mapping continued using a proto-
bdcouelreindnegmveaWlpoorppleedddWb(ayWrieWsIiIle.asnBldayenrdtehre1n9,u6n1t;1i6lCmiotillwwlaeilsonlsht1ao9p7hp7ae)d.d feionrucerlduoodfnewmfholiorrciehsttiahcraenand1ed8s,ce0rn0iv0bierpdloonhtmesreena:trao1lu)nPddleottathield,astt2aa)tegaPttlhhoa-tt
The well-preserved collection is now known as the mVeagpestadteipoinctmianpgs,thseholwociantgiohnasnodfdprlaowtsn spaomlpylgeodn,s o3)f
Wieslander Vegetation Type Mapping (VTM) col-
forest types, and their associated species, and, 4)
lection, and the entire survey encompassed nearly
Landscape photographs and associated information
28 million ha, covering most of the state's natural about location and content of the photographs. The
areas exclusive of the deserts and the larger agri-
team also collected herbarium specimens for every
cultural areas. The collection represents a fascinat-
species recorded on the vegetation maps or in the
ing snapshot of California vegetation in the early sample plots (Ertter 2000). We have not begun in-
part of the last century, and represents a valuable
corporating the herbarium specimens into this dig-
raensdouortcheersfoirntfeorreesstteedrs,inetchoelongaitsutrsa,l leannvdirmoannmaegnetrso,f oituarl cowlelbecGtIioSn,sbyusttewme aarnedprtohveidJienpgsaonlinHkerbbeatrweieunm
early 20"^ century California, and land use changes
Specimen Catalog.
since then. For example, the collection has been
used in part as a basis for the statewide California
Plot Data
Gap Analysis maps produced decades later (Davis
et al. 1995; Walker 2000). There are approximately 18,000 plots statewide,
We are digitizing the collection, and making it concentrated primarily along the central and south-
available to researchers and the public. The current ern coastal ranges, and along the Sierra Nevada.
"VTM Digitization" Project is a collaborative ef- These plots were surveyed not only as a check on
fort by teams at University of California Berkeley the vegetation mapping, but also to survey the di-
(UCB) and University of California Davis (UCD), versity of California vegetation types for details
funded by the USFS, and U.S. Department of Ag- such as species composition, size and stand density
riculture Cooperative State Research, Education of trees and shrubs and depth of leaf litter These
and Extension Service (USDA CSREES). The sample plots were located across a gradient ofveg-
VTM
Digitization Project aims to digitize the entire etation types, and the historic records contain data
VTM
collection foruse in ecological and geospatial regarding tree stand structure (number per diameter
analyses, and to facilitate access and distribution of class), percent cover ofdominant overstory and un-
the data by researchers and interested parties. This derstory vegetation by species, soil type, parentma-
article describes our efforts at making the VTM terial, leaflitter, elevation, slope, aspect, parent ma-
collection digital, spatially referenced, readily terial, and other environmental variables.
available, and web-accessible via an Internet-based Details of the plot design are as follows (Wies-
Geographic Information System (webGIS) appli- lander 1935a; Wieslander 1935b). Each plot was
cation. In this paper we aim to: 1) describe the col- rectangular in shape with the longer axis running
lection, 2) alert readers to the digitization project, upslope. The plots were 800 m^ in size in forests,
3) describe the methods used to complete the dig- and 400 m- in scrub and chaparral communities.
itization, and 4) discuss the potential uses for the For example, the 400 m- plots were laid out using
!
2005] KELLY ET AL.: DIGITIZATION OF THE VTM DATASET 193
two chains (ca. 40 m) long and one-half chain (10 collection. A snapshot of the plot data is found in
m) wide, and divided into 100 squares (ca. 4 m-). Figure la.
In both types of plots, the dominant species within
each milacre was recorded. When less than 50% of Plot Maps
the square was vegetative cover, ground surface The VTM plot maps show the locations of all
characteristics such as bare ground, rock outcrops the individual plots surveyed by the original VTM
or tree trunks were noted. A summary of the crews. Hollow circles of about 3.5 mm in diameter
squares within the rectangle plot was provided, not- depicting the location of the plots were stamped in
ing the average height of the dominant species (to red ink on USGS topographic maps (editions of
gtrheeanteearretshtan0.510ftcomr 1i5ncdmb)h. Awtitthhiens1a0memtiomfe,eittreheesr c1u8t93i-nt1o92s0e,ctiroenpsr,inmteodunitnetdheon19c3a0nsv)ast,haatnhdadfolbdeeedn,
side of the center-line were tallied by species and to facilitate use in the field. This allowed for re-
diameter class. In both types of plots, additional peated folding along the seams, without loss of
information such as slope, soil character and year mapped information. Unfortunately, the resulting
oflast burn was recorded (Wieslander 1935a; Wies- maps were not dimensionally stable, and 80 years
lander 1935b). of use, temperature extremes, and other factors
All plot data was stored on paperdata sheets, and have warped many of the plot maps. This has re-
individual plots were numbered according to U.S. quired us to take steps in the georeferencing pro-
Geological Survey (USGS) topographic quadrangle cess to reduce these errors, described below. The
map name, quad section number and plot number. plot map collection comprises about 150, 15-mi-
There are about 150 original maps remaining in the nute (1:62,500 scale) and 30-minute (1:125,000
Fig. 1. Plot and vegetation map data examples: a) scanned plot maps and plot data sheets, b) a sample vegetation
map, and c) a sample of the database return for one plot.
MADRONO
194 [Vol. 52
printed in color on 15- and 30-minute USGS to-
pographic quadrangles, and simplified, uncolored
blue-line print sheets. Some areas of the state have
"zoomed-in" vegetation maps drawn on 6- and
7.5-minute USGS quads. The major vegetation
types are shown in different colors and separated
by ink lines (Colwell 1977). These are further sub-
divided on the map into pure and mixed stands,
with notation on species composition. The teams
used a protocol designed to be useful forengineers,
foresters, and managers, and attempted to group
vegetation types by their fire hazard characteristics,
uses, or economic importance.
There are about 330 of these detailed maps in
the collection, covering about 16 million ha. Some
were published, and some remain unpublished. Fig-
ure lb shows an example ofa vegetation map from
the San Mateo area. The plotdata sheets, plotmaps,
and vegetation maps are currently curated by Dr.
Allen-Diaz in the Environmental Sciences, Policy
and Management Department in the College of
Natural Resources at UCB.
y/// VegetationMap
Counties Photographs and Associated Data
Fig. 2. Inventory of VTM plot maps in California. We There are approximately 3100 black and white
have produced inventory maps for other parts ofthe col-
lection. These can be found at: http://vtm.berkeley.edu/ paphportooxgirmaapthesl(y9.1200X c1o3l.o6rctmo)pofgrroamphi1c92m0a-p1s9.41M,aannyd
of the photographs are keyed to USGS topograph-
scale) USGS quadrangles, primarily concentrated ical maps. For these photographs, the location of
along the central and southern coastal ranges, and the place where the photos were taken is written in
along the Sierra Nevada. Figure 2 displays our in- red pen on the maps, with an arrow marking the
ventory of the plot maps and Figure la shows an vantage point and view of the photo. Some of the
example of the plot maps and data sheets. photographs are "panorama" style images, but
most are focused on a stand (see examples in Fig.
Vegetation Maps 3). The photographs document the typical and atyp-
The vegetation type maps were mapped in the ical subtype, wider species, timber standconditions,
field by VTM crews, directly upon 15-minute (1: range of variation, and consequences of land use
62,500-scale) topographic quadrangles by direct Tanhde cmualitinvaptihoont,oggrraapzhinegr,wlaosggiAnlgb,ermtiWniiensglaanndder.firAe.
observation, and "sketching from ridges, peaks,
second "series" was done by Richard C. Wilson, a
and other vantage points", and supplemented by UCB School of Forestry graduate. Several other
sample plots (Wieslander 1935b). With average
cwoenedkistiotnos,coitmptloeotkeatthweo-fimealdnwcorrekw ffroromasi1x5-tmoieniugthet piphaotteodgrianphtehresp,roijnecclt.udTinhgeCM.arRiaaynmKoonsdhlCalnard,Bpiaorstciic--
quad of about 6,070 ha (Wieslander 1935a). Plant ence and Natural Resources Library at tVheTUMniver-
communities were mapped to a minimum of 16 ha sity of California, Berkeley houses the Pho-
(Colwell 1977). Dominant plant species were tographs Collection.
mapped, while understory vegetation information
was collected in the sample plots. The vegetation Methods and Results to Date
mapping scheme, since it was done in the field via Initial Preparation
overlooks and remote vantage points was by ne-
cessity driven by overstory species recognition (i.e., In order to create a complete and linked database
"the dominant vegetation visible externally" accessible to the public and to researchers, we
(Wieslander 1935a)). The VTM method used two needed to make all parts of the collection digital,
vegetation classification concepts: they mapped index each component by its spatial location, and
"mosaic types" which are complex vegetation con- develop a web tool for users to query, view, and
ditions that result from fire or other disturbances, download data from the database. These are de-
or pure and mixed stand conditions which they as- scribed in turn here. Similarefforts have been made
sociated with natural plant associations (Wieslander in other states (e.g. Iowa and Wisconsin) with in-
1935a). teresting results (Anderson 1996; Schulte et al.
The mapped products include map sheets over- 2002). We refer to some of the lessons learned in
2005] KELLY ET AL.: DIGITIZATION OF THE VTM DATASET 195
Fig. 3. Representative photographs from the original VTM collection (all photogra|)lis taken by A. E. Wieslander.
and descriptions from photograph captions): (a) In Mill Creek State Park, Del Norte County, showing coastal redwood
and R. D. Garver, taken I I/I/1940; b) Snow Lake in Tuolomne County, showing whitebark pine and associated ground
cover, taken 8/1/1936; c) In San Mateo County, showing east arm ofPilarcitos Lake, with Douglas fir, coast live oak,
madrone and California bay woodland and coastal sagebrush, taken 7/24/1932; d) Looking southwest down Cuyuma
Creek in Santa Barbara County, showing coast live oak woodland and Salvia leiicophylla: taken 4/5/1936; and e)
Panorama looking north and northeast showing Long Valley Peak and Oak Valley in San Diego County, showingcoast
live oak woodland with scattered Jeffrey pine and Engelmann oak woodland on upper left slope, taken 4/4/1931.
MADRONO
196 [Vol. 52
those projects, and others utilizing historic vegeta- (ESRI 2004). Average RMSE for the process was
tion data. around 60 m. We have performed some preliminary
examination of this error, and our results suggest
VTPlMotcodlalteacbtaisoen.foMrarnesyearrecshea(rBcrhaedrbsurhyave197u4s;edMitnh-e mthiantianggeeorfrobrasceontmraipbutiisoonn;ebcarsiteimcaalpsfacmtaodreinbdeeftoerre-
nich et al. 1995; Vayssieres et al. 2000; Franklin et 1911 have larger RMSE values; and a second im-
hala.v2e00m4a;deTayploortrio2n0s04obf),thaendVTsoMmeploofttdhaetsaedpiegoitpalle. peovrattainotnscohnatvreibluetsisonRMisSeEle.vation: maps on lower el-
Allen-Diaz produced the largest database contain-
ing about 4,000 plots, as part of a statewide oak Vegetation maps. The vegetation maps are also
woodland classification effort (Allen et al. 1989). being scanned and georeferenced using the same
We collected existing electronic versions of VTM protocol as the plot maps. The vegetation maps
plot data and merged them into a standardized da- have not all been cut into sections, and can be
tabase; now we are in the process of entering all scanned and georeferenced in their entirety. Asso-
remaining plot data into this common database ciated metadata from each vegetation map is col-
(Fig. Ic). Once completed, each plot will have a lected at the time of scanning. This component of
unique identifier, based on its quad, the section of the project is being completed by Jim Thorne and
the quad, and its unique number in that section. Jeff Kennedy at UCD.
This identifier will geographically locate the plot, Photograph collection. The Library Photograph-
and link the plot location with the database for fu- ic Services is digitizing the photograph collection.
ture analysis. Each photograph is scanned, and information from
Plot maps. All plot maps have been scanned at its captions has been entered into an Access data-
600 dpi, one cut segment at a time. Uncut, scanned base, which will be available on the Library Web.
versions of each USGS topographic map of the Once the project is complete, users will be able to
same edition and reprint were searched for in sev- search by keyword, genus and species, and quad
eral California map libraries and spatial data clear- name. Data from the caption include a brief de-
inghouses. To date, we have found more than half scription of the location and subject of the photo-
of the maps in an uncut state in either the UCB graph including relevant genus and species, and
Map Library, the California State University Chico quad name. The photographer, date of the photo-
Meriam Library California Historic Topographic graph, and occasionally township and range are in-
Map Collection, the Los Angeles County Historical cluded. In addition, researchers will be able to
Topographic Map Collection, or the map library at search for the photographs using the webGIS tool
described below.
the University of California Santa Barbara.
These uncut maps are used in the georeferencing Website development. We have developed a web-
process. Georeferencing is the process whereby a site to update interested parties on our progress.
map or set of maps is referenced to true world co- The website is currently available at: http://
ordinates through the collection of "tie points" vtm.berkeley.edu/. Here you can find program pro-
(Jensen 1996). The tie points are used to position gress, inventory maps and other information, in-
two images coincident with each other through a cluding our "webGIS" application that unites all
geometric transformation that translates the location aspects of the project. WebGIS is a new term that
of each tie point on the historic map to that found refers to websites that unite two components: (1)
on the modern reference map. A root mean squared GIS database storage and maintenance and (2) In-
error (RMSE) is then calculated to estimate overall ternet accessibility. Although not yet widely used
accuracy of the transformation. Others have at- in natural resource management, such systems are
VTM
tempted this process with the maps. For ex- a promising option for entering and storing hetero-
ample, Walker (2000) used modern satellite imag- geneous datasets, indexed by location, and making
ery as the reference map in the georeferencing pro- them widely available in a visual, dynamic and in-
cess. In this case, we first registered the historic teractive format (Kearns et al. 2003). When fin-
uncut maps of the same vintage as the VTM cut ished, the website will allow users to view all data
maps to modern maps (1:24,000-scale USGS Dig- associated with the project in mapped form. Users
ital Raster Graphic digital images ofmodern USGS will be able to turn on and off different "layers"
quadrangles) of a known projection and coordinate ofdata (e.g., modern vegetation, elevation, orVTM
system using stable tie points such as roads and plot map quads), zoom in and out, viewing increas-
peaks. We used between eight and 16 tie points per ing detail as they zoom in, search for particular
map. Next, the uncut scanned VTM maps were areas of interest by entering quad name informa-
georeferenced to the georeferenced historic uncut tion, or geographic coordinates, view photographs
maps using common map features as the tie points. tied to the location, query the plot database, create
We used a minimum of six tie points per segment. their own maps, and finally, request the data itself
We used first order polynomial transformations for for research purposes.
each step. All georeferencing used Erdas Imagine Figure 4 shows plot maps, and plot locations ov-
8.7 software (Leica 2004), and ArcGIS software erlayed on modern topographical data. Once the en-
2005] KELLY ET AL.: DIGITIZATION OF THE VTM DATASET 197
Fig. 4. VTM plots in the San Francisco Bay area. The original, scanned and georeferenced plot maps from the VTM
collection are superimposed on a modern elevation map. Red circles depict locations of VTM plots.
tire collection has been digitized, we will be able able time thinking about the uses of these data in
to provide this kind ofseamless digital view for the the future; and claimed that ''the data not only
entire collection. serve as a basis for mapping sites in the field but
also are suitable for scientific study later'' (Wies-
Discussion lander 1935a). He outlined some potential uses for
the data in Madrono in 1935 that still hold true
Potential Users and Uses of the Data today. He wrote the data might provide: 1) a partial
There are many potential users and uses for the explanation ofthe present distribution ofvegetation
data in this collection. Wieslander spent consider- types and dominant species; 2) a better understand-
MADRONO
198 [Vol. 52
ing of vegetation changes that have occurred in the Diego County area (Taylor 2004b). These studies
past, those now in progress, or those to be expected that relocate VTM plot data and compare plant
to occur in the future; 3) further contributions to abundance and distribution measures with current
our knowledge of the values of certain plants and data might be more successful in certain vegetation
vegetational types as indicators of particular soil types; for example, areas with trees or with less
and climatic conditions; and 4) suggestions for fu- diversity might be more amenable to comparative
ture investigations and also a foundation for further analyses. Efforts over larger scales also should be
research. He also provided explanations of change encouraged, as successful relocation of individual
that might be controversial today: "the most strik- plots can be problematic (Keeley 2004).
ing and significant of such changes are those rep- Finally, the vegetation map portion of the VTM
resenting a progressive deterioration from higher collection has also been put to use. Walker (2000)
and more valuable to lower and less valuable types compared the VTM vegetation maps and other his-
of vegetation as a result of such land abuse as de- toric sources of vegetation distribution in Yosemite
structive logging, accidental and willful summer National Park with modern sources ofmapped veg-
fires, the practice of annual burning in many foot- etation data to examine decadal changes across the
hill and mountain localities, and excessivegrazing" Park. Such retrospective mapping projects are not
(all quotes from Wieslander [1935b]). without challenges: the original inclination of the
The VTM methods and resulting data lend them- VTM design was forestry-based, with a vegetation
selves to modern ecological research, and to date classification scheme developed for the project and
the data have been used in three broad areas of designed for further forestry applications. In addi-
inquiry. First, the data have been used to classify tion, the vegetation mapping was done via remote
vegetation communities in the state (Jensen 1947; vantage points. These choices necessarily resulted
Griffin and Critchfield 1972; Allen et al. 1989; Al- in large minimum mapping units and a classifica-
len et al. 1991; Allen-Diaz and Holzman 1991). tion scheme based on overstory canopy dominance.
The detailed collection and recording of floristic Modern ecologist wishing to examine vegetation
and environmental detail at the plot scale allows for changes through these maps should be cognizant of
a posteriori determination of vegetation types by these facts. Indeed, these kinds of considerations
researchers (Walker 2000), making the data more are also necessary in comparing different contem-
useable and flexible for modern ecologists. This ap- porary vegetation datasets (e.g.. Manual ofCalifor-
proach was used by Allen-Diaz et al. (1989) in cre- nia Vegetation Classification and the USFS's
ating a comprehensive classification scheme for CalVeg dataset (Thorne et al. 2004).
California's hardwood rangelands, which in turn We are interested in using the collection to com-
was used in the manual of California Vegetation pare current stand structure and composition with
(Sawyer and Keeler-Wolf 1995). The data have also 70-80-year-old stand structure in areas affected by
been used to validate modern models of vegetation the new forest disease called "Sudden Oak Death"
composition (Vayssieres et al. 2000; Franklin (Rizzo and Garbelotto 2003). We hope to under-
2002). These efforts assume a static vegetation stand why there is such a patchy distribution of
composition over decades at a large scale (e.g.. infection in forests with Sudden Oak Death. We
Franklin (2002) examined a 3880 km- study area). intend to look at interrelationships between landuse
Secondly, the decades-old collection ofplot data changes, climate changes, pest invasions, stand
has also been used in reconstructing Californiaveg- growth, and infection, and model change in plant
etation conditions in the early 20"' century in order community composition and structure. This project
to examine changes due to disturbances such as fire is one example of a multi-scaled approach to his-
and disease, or to examine local trajectories ofland torical ecology: it utilizes plot level information on
cover change. Bradbury (1974) was perhaps the current vegetation community structure across a
earliest published example ofthis. He examined 40 gradient of sites with infected trees, and historical
years of change in San Diego County vegetation, vegetation data at a larger-scale to examinepossible
and found little change in the area, most of which explanations for the patchiness ofthe disease state-
was clearly disturbance-related (Bradbury 1974; wide.
Franklin 2002). In contrast, other researchers found
significant changes in species composition. For ex- Spatial Accuracy Considerations
VTM
ample, Minnich et al. (1995) used data to
document significant shifts in species composition Re-locating plots from historic maps is recog-
and stand density in the San Bernardino Mountains nized by many to be a challenge (for some the
of California, and discussed the role of fire in greatest challenge) facing modern ecologists using
changing the forests in California conifer forests. these data. Several researchers have discussed ca-
VTM
Taylor (2004) relocated 78 plots in San Diego veats associated with use of historical ecological
County, and discovered that while overall shrub data, and recommended caution in using older col-
cover loss was less than reported in a comparable lections. Indeed, the use of PLS data for historical
area (Riverside County), there was a shift in com- ecology has been controversial because ofpossible
position in several shrub communities in the San inconsistencies caused by surveyor variability (Gal-
2005] KELLY ET AL.: DIGITIZATION OF THE VTM DATASET 199
atowitsch 1990). While Schulte and Mladenoff equivalent to 110 m on the ground when using a
(2001) and Siccama (1971) found that surveyors 15-minute (1:63,500-scale) map. In addition, the
were consistent in recording distance and direction georeferencing process contributes an additional 60
to witness trees, they also note that surveyors were m error (based on the average RMSE for the pro-
not consistent in describing species and diameter cess, reported in Erdas Imagine). Finally, we need
characteristics, possibly due to seasonality of sam- to include the error contributed by the original
pling (Schulte and Mladenoff 2001). In addition, uses map itself, which is about 30 m (USGS
several researchers note variability among survey- 1947). We have combined these sources to generate
ors across species, diameter, size and location an overall error of about 200 m per plot, which is
(Bourdo 1956; Galatowitsch 1990; Manies et al. similar to that reported by Walker (2000) using a
2001; Schulte et al. 2002). Overall, however, Man- different georeferencing method. This amount is in
ies and Mladenoff (2000) found the PLS method excess of any error in locating the original plot on
VTM
accurately estimated relative species composition a quad by the crew in the field. The overall
and the order of dominance of land cover types at error might be lessened through use of the slope
broad spatial scales. and aspect values recorded in the plot data in com-
These experiences are applicable to the use of parison with topographic data from a modern Dig-
the VTM collection, although surveyor inconsisten- ital Elevation Model (Taylor 2004a).
cies might be less than in the PLS collections as a
result ofa smaller project, more modern equipment, Conclusions
and perhaps a better-trained staff. Indeed, Minnich
California's landscape is complex, varied and ex-
ewtitahli.n(a1919050)mdersacdriiubsebcyonrfeifdeernecneceitnolfoicxaetdinfgeatpulroetss tremely dynamic, and in the 20th century has ex-
perienced numerous agents of large-scale change.
such as roads, and the location of prominent trees
Fire, urban and exurban development and expan-
included in the original dataset (Minnich et al. sion, harvesting, invasive species, and new forest
p1l9o9t5,)t.oIennasdudrietitohna,t tthheeyy htoaodkltohcraeteedretphleicpaltoetss.atWaelakc-h diseases such as Sudden Oak Death combine to al-
ter the environment in ways that could not be imag-
er (2000) found that while the overall fidelity of ined 80 years ago when Weislander was working.
species composition mapping was good, the spatial
Yet, the legacy that he and his crews have left us,
accuracy of the plots (and of vegetation polygons)
a detailed and accurate picture of California vege-
varied with topography, and spatial error increased
tation of that time, help us to understand change,
in areas of high terrain (Walker 2000). Franklin
(2002) found discrepancies between the VTM data and better manage it for the future. Clearly, the
and the more modern USPS Forest Inventory and ability to relocate plots with sufficient precision and
accuracy for detailed historical analysis will remain
Assessment (FIA) plots over a large area, but pro- an important challenge in historical ecology, but we
poses that this might be a result of differing sam-
pling schemes between the VTM and FIA plots, not contend that researchers using historical data have
found a range of approaches that address this and
a result of spatial accuracy.
A less optimistic view is presented by Keeley make the dataset valuable for ecological research.
(2004). He found considerable spatial variability in For example, in areas of little change, the floristic
detail of the plot data can support vegetation com-
coastal sage scrub and chaparral communities in munity classification, or validate new models of
Southern California. Consequently, he maintains
VTM species distribution. In areas ofchange, researchers
that the accuracy and precision to which plots
have pursued landscape and larger scale analysis of
pinercfhoarpamrrapllomti-gbhyt-pbleotrelaoncaaltyesdisis nooft scuoffmimciuennitttyo historical data, combining several plots and exam-
change. In other words, when using VTM data on ining changes across watersheds, regions, or even
states.
a plot-by-plot basis in chaparral communities, any The kind of data repository we are building will
change found might be an artifact of errors in plot
provide valuable data and tools for those interested
re-location, and not indicative of real change. This
in California's changing vegetation through the
is a reasonable argument for chaparral communi- 20th century and beyond. We hope to have the data
ties, but many authors contend that in forested com- accessible to researchers by Fall 2005 at
munities the problem of relocation might not be as
vtm.berkeley.edu.
severe due to the existence of persistent trees, as
described above.
Acknowledgments
Plot relocation still remains an important chal-
lenge in ecological research with these data. We The authors would like to acknowledge the following
estimate that our georeferencing technique gener- people who have contributed to the production ofthe da-
ates plot locations with a combined error ofaround tabase. Included are numerous students entering data,
200 m on a 15-minute (1:63,500-scale) quad. The aKnend-iDcahvieUSehdaaa,riTifrmomDothheertEyn,viArnonnmHeunbtearl.SLcaiuenrceens,McPGoleiec,y
plot markings on the original maps constitute the and Management Department at UC Berkeley; Brian
largest component ofthis error. These plots markers Thomas from the UC Berkeley College of Natural Re-
mm
are 3.5 in size, generating circles with radii sources; Beth Weil from the Marian Koshland Bioscience
MADRONO
200 [Vol. 52
and Natural Resources Library at UC Berkeley; and Jeff records (1839-1866) for northern Wisconsin at mul-
Kennedy and Jim Thorne from UC Davis. Without all tiple scales. Journal of Biogeography 27:1031-1042.
these people, this collection could not have been brought Heckrotte, W. and J. SwEETKiND. 1999. California 49:
to life again. We are grateful for the insightful comments Forty-nine maps ofCaliforniafrom the sixteenthcen-
fromthreereviewers; incorporationofthesereviewsgreat- tury to the present. Occasional PaperNo. 6, The Cal-
ly strengthened this paper. ifornia Map Society, San Francisco, CA.
Iverson, L. R. 1988. Land-use changes in Illinois, USA:
Literature Cited the influence of landscape attributes on current and
historic land use. Landscape Ecology 2:45-61.
Allen, B. H., R. R. Evett, B. A. Holzman, and A. J. Jensen, H. A. 1947. A system for classifying vegetation
Martin. 1989. Rangeland cover type description for in California. California Fish and Game 33:199-266.
California hardwood rangelands. Contract . 1996. Introductory digital image processing: a
#8CA63912, California Department of Forest and remote sensing perspective, Second Edition. Prentice
Fire Protection: Forest and Rangeland Assessment Hall, Upper Saddle River, NJ.
Program, Sacramento, CA. Jepson, W. L., R. Beidleman, andB. Ertter. 2000. Willis
, B. a. Holzman, and R. R. Evett. 1991. A clas- LinnJepson's "Mapping inForestBotany". Madroiio
sification system for California's hardwood range- 47:269-272.
lands. Hilgardia 59:1-45. Kearns, F R., M. Kelly, and K. A. Tuxen. 2003. Ev-
Allen-Diaz, B. H. and B. A. Holzman. 1991. Blue oak erything happens somewhere: using webGIS as atool
communities in California. Madroiio 38:80-95. for sustainable natural resource management. Fron-
Anderson, P. F. 1996. GIS research to digitize maps of tiers in Ecology and the Environment 1:541-548.
Iowa 1832-1859 vegetation from general land office Keeley, J. E. 2004. VTM plots as evidence of historical
townshipplatmaps, IowaState University, Ames,lA. change: goldmine or landmine? Madroiio 51:372-
Bourdo, E. a. 1956. A review ofthe general land office 378.
survey and ofits use in quantitative studies offormer Leica. 2004. Erdas Imagine 8.7 software. Leica Geosys-
forests. Ecology 37:754-768. tems. Inc. Heerbrugg, Switzerland.
Bradbury, D. 1974. Vegetation history of the Ramona Manies, K. L. and D. J. Mladenoff. 2000. Testing meth-
quadrangle, San Diego County, California (1931- ods to produce landscape-scale presettlement vege-
1972). Ph.D. dissertation. University of California, tation maps from the U.S. public land surveyrecords.
Los Angeles, CA. Landscape Ecology 15:741-754.
BuissaEgReEs:T,iDn.te(repdr.e)t.in1g99N0.orFtrhomAmseeraicchaanrtshitsotosrayteltlihtreouimg-h larg,e-scale s,uArNveDyoEr. vVa.riNaboiRlDitHyEiiMn.th2e00h1i.stoArsisccfsosriensgt
maps. University ofChicago Press, Chicago, IL. dataofthe original U.S. public land survey. Canadian
CoLWELL, W. L. 1977. The status of vegetation mapping Journal of Forest Research 31:1719-1730.
in California today. Pp. 195-223 in M. G. Barbour MiNNicH, R. A., M. G. Barbour, J. H. Burk, and R. F.
Davisna,inad.F.JJ.Woh,MnajPW.oirAl.e(yeSdts&i.)n.eS,oTneDrs.r,esMSt.aricaSrltamovemensgt,eot,Mat.CioAIn..BoofrCcahleifrotr,- FcCooennrisnfeaerurv.afto1ir9oe9ns5t.sBiSooilfxotgytyhey9e:aSr9as0n2o-f9B1ec4rh.naanrgdeinion CMaoluinftoarinnisa.n
and a. D. Hollanger. 1995. Gap analysis ofthe ac- Rizzo, D. M. AND M. Garbelotto. 2003. Sudden oak
tual vegetation of California: 1. The Southwestern death: endangering California and Oregon foresteco-
Region. Madrono 42:40-78.
systems. Frontiers in Ecology and the Environment
Ertter, B. 2000. Our undiscovered heritage: Past and fu-
1:197-204.
ture prospects for species-level botanical inventory. Russell, E. W. B. 1981. Vegetation ofNorthern New Jer-
Madrofio 47:237-252.
sey before European settlement. American Midland
ESRI. 2004. Environmental Systems Research Institute,
Naturahst 105:1-12.
Isneca.rcAhrIcnGstIiStutes:ofRtewdalrea.ndsE,nvCiAr.onmental Systems Re- Sawyer, J. and T. Keeler-Wolf. 1995. A Manual ofCal-
Foster, D. R., G. Motzkin, and B. Slater. 1998. Land- ifornia Vegetation. California Native Plant Society
Press, Sacramento, CA.
use history as long-term broad-scale disturbance: re-
gsiyosntaelmsfor1e:s9t6-d1y1n9a.mics in central New England. Eco- ScHULUTSE,PuLb.liAc.LaannddDS.uJr.veMylardeecnorodfsf:.T2h0e0i1r.uTseheanodrigliinmail-
Franklin, J. 2002. Enhancing a regional vegetation map tations in reconstructing presettlement vegetation.
with predictive models of dominant plant species in Journal ofForestry 99:5-10.
chaparral. Applied Vegetation Science 5:133-146. , D. J. Mladenoff, and E. V. Nordheim. 2002.
70,yCe.arLs.iCnoualtseourt,hearnndCSa.liJ.foRrenyi.a2c0h0a4p.arCrahlancgoemomvue-r cQuoannstiinta(tUi.vSe.Ac.l)aslsainfdicsactaipoen:omfaapphiisntgorfiocrensotrsthaterrnegWiiosn--
nity related to fire history. Journal ofVegetation Sci- al scales. Canadian Journal of Forest Research 32:
ence 15:701-710. 1616-1638.
Galatowitsch, S. M. 1990. Using the original land sur- SiccAMA, T. G. 1971. Presettlement and present forest
vey notes to reconstruct presettlement landscapes in vegetation in Northern Vermont with special refer- j
the American West. Great Basin Naturalist 50:181- ence to Chittenden County. American Midland Nat-
191. uralist 85:153-172.
Griffin, J. R. and W. B. Critchfield. 1972. The distri- Taylor, R. S. 2004a. National Park Service. Personal |
bution offoresttreesinCalifornia. PSW-82,U.S.D.A. Communication.
Forest Service, Pacific Southwest Forest and Range . 2004b. Changes in coastal sage scrub composi-
Experiment Station, Berkeley, CA. tionandstructureover70yearsinanurbanizingland-
He, H. S., D. J. Mladenoff, A. Sickley, and G. G. Gun- scape. Pp 500-501 in Abstracts of the 89"^ Annual
tenspergen. 2000. GIS interpolations ofwitness tree Meeting of the Ecological Society of America, Au-