Remote sensing change detection tools for natural resource managers Understanding concepts& tradeoff
Remote sensing change detection tools for natural resource managers:
Understanding concepts and tradeoffs in the design of landscape monitoring projects
Robert E.Kennedy a ,?,Philip A.Townsend b ,John E.Gross c ,Warren B.Cohen a ,Paul Bolstad d ,Y.Q.Wang e ,Phyllis Adams f
a
USDA Forest Service,PNW Research Station,3200SW Jefferson Way,Corvallis,OR,97331,United States
b
University of Wisconsin-Madison,Department of Forest and Wildlife Ecology,1630,Linden Drive,Russell Labs,Madison,WI 53706,United States c
National Park Service,Of ?ce of Inventory,Monitoring,and Evaluation,1201Oakridge,Drive,Suite 150,Fort Collins,Colorado 80525-5596,United States d
University of Minnesota,Department of Forest Resources,301h Green Hall,1530,Cleveland Ave.N.,St.Paul,MN 55108-6112,United States e
University of Rhode Island,Department of Natural Resources Science,1Greenhouse Road,Kingston,RI 02881-0804,United States f
USDA Forest Service,620SW Main St,Suite 400,Portland,OR 97205,United States
a b s t r a c t
a r t i c l e i n f o Article history:
Received 7December 2007
Received in revised form 28May 2008Accepted 31July 2008Keywords:Monitoring
Change detection
Natural resource management Landscape ecology Protected areas
Remote sensing provides a broad view of landscapes and can be consistent through time,making it an important tool for monitoring and managing protected areas.An impediment to broader use of remote sensing
science for monitoring has been the need for resource managers to understand the specialized capabilities of an ever-expanding array of image sources and analysis techniques.Here,we provide guidelines that will enable land managers to more effectively collaborate with remote sensing scientists to develop and apply remote sensing science to achieve monitoring objectives.We ?rst describe fundamental characteristics of remotely sensed data and change detection analysis that affect the types and range of phenomena that can be https://www.wendangku.net/doc/f55204582.html,ing that background,we describe four general steps in natural resource remote sensing projects:image and reference data acquisition,pre-processing,analysis,and evaluation.We emphasize the practical considerations that arise in each of these steps.We articulate a four-phase process that guides natural resource and remote sensing specialists through a collaborative process to articulate goals,evaluate data and options for image processing,re ?ne or eliminate unrealistic paths,and assess the cost and utility of different options.
?2009Elsevier Inc.All rights reserved.
1.Introduction
Remote sensing science has become a critical and universal tool for natural resource managers and researchers in government agencies,conservation organizations,and industry (Gross et al.,2006;Philipson &Lindell,2003;Stow et al.,2004).The range of applications addressed in the papers of this special issue of Remote Sensing of Environment is testament to the growing use of remote sensing in natural resource management.For the resource manager,a particular attraction of satellite remote sensing technology is the ability to provide consistent measurements of landscape condition,allowing detection of both abrupt changes and slow trends over time.Detection and character-ization of change in key resource attributes allows resource managers to monitor landscape dynamics over large areas,including those areas where access is dif ?cult or hazardous,and facilitates extrapolation of expensive ground measurements or strategic deployment of more expensive resources for monitoring or management (Li et al.,2003;Schuck et al.,2003).In addition,long-term change detection results
can provide insight into the stressors and drivers of change,potentially allowing for management strategies targeted toward cause rather than simply the symptoms of the cause.
Despite their increased exposure to and appreciation of remote sensing,managers often must rely heavily on remote sensing specialists to design and implement monitoring programs based on change detection of remotely sensed data (Woodward et al.,2002).The authors'collective experience in monitoring projects has shown that success is the responsibility of both parties:the remote sensing scientists must understand the needs and the scienti ?c underpinnings of the managers'goals,and the managers must have or develop an understanding of the fundamental remote sensing issues that arise in remote sensing change detection and monitoring projects.The primary targets of this paper are natural resource managers or researchers who are considering remote sensing for monitoring resource attributes over time,and a fundamental goal is to provide them with enough information about the full arc of a remote sensing project to actively collaborate in designing successful monitoring projects.By doing so,we also hope to aid this audience in evaluating the case studies found in the other papers in this special issue.Despite our focus on educating natural resource managers,we emphasize that the dialog between managers and remote sensing specialists is bi-directional and iterative.
Remote Sensing of Environment 113(2009)1382–1396
?Corresponding author.Now at Department of Forest Science,Oregon State University,321Richardson Hall,Corvallis,OR 97331,United States.Tel.:+15417507498.
E-mail address:Robert.kennedy@https://www.wendangku.net/doc/f55204582.html, (R.E.Kennedy).0034-4257/$–see front matter ?2009Elsevier Inc.All rights reserved.doi:10.1016/j.rse.2008.07.018
Contents lists available at ScienceDirect
Remote Sensing of Environment
j ou r n a l h o m e pa g e :ww w.e l s e v i e r.c o m/l o c a t e /r s e
To discuss the full arc of a remote sensing study,we require the reader to have a basic understanding of a few key concepts in remote sensing change detection.The natural resource manager may consult the many excellent review papers(Cihlar,2000;Coppin et al.,2004; Lu et al.,2004;Mas,1999;Mouat et al.,1993;Yuan et al.,1998)and texts(Campbell,1996;Lillesand&Kiefer,2000;Lunetta&Elvidge, 1998;Richards,1993;Sabins,1997;Schott,1997;Schowengerdt,1997; Wulder&Franklin,2007)written on remote sensing in general and on change detection in particular.Despite the utility of these references, we?nd that the existing literature leaves two gaps.First,the natural resource manager will struggle to?nd references written for the non-specialist that also distill the key technical concepts needed to effectively make practical decisions about planned remote sensing projects.While we do not intend to be a simple review paper on basic remote sensing,our experience suggests that it is critical to highlight a few central concepts in remote sensing to lay the groundwork for later discussion.Second,most reviews focus on evaluating image types and analytical methods for change detection,but few review these issues in the context of long-term monitoring,particularly how decisions and constraints at all stages of a project can in?uence the types of monitoring goals that can be reached.To wisely distribute time and funds,a natural resource manager must be able to evaluate trade-offs among all of the components of the study before?nal plans are made. This paper represents our attempt to?ll these two gaps.
For simplicity of terminology,we refer in this paper to the“natural resource manager,”but in practice we consider our audience to be the broader group of scientists,managers,and agency of?cials who must bring remote sensing data into the realm of natural resource management.Because it is impossible in this paper to address each unique situation faced by natural resource managers and scientists, we have developed a set of broad resource attributes or indicators that encompass many speci?c issues faced by managers,scientists,and agency personnel(Table1).All subsequent tables will be structured around these attributes.Rather than being considered an exhaustive list,however,the attributes should be considered for their heuristic value in capturing the continuum of effects of different processes on landscapes.
The paper has three sections.The?rst describes underlying concepts in remote sensing and change detection that must be understood to effectively communicate with remote sensing specia-lists.The second section describes the steps involved in a typical remote sensing study designed for monitoring of natural resources, showing how the key concepts described in the?rst section are applied in practice.The third section provides a general framework of evaluation phases that should be considered before a remote sensing monitoring program begins.Throughout this paper,we use studies described in companion papers of this special issue to illustrate key concepts.
2.Key concepts
To appreciate the decisions that must be made in a remote sensing monitoring project,the natural resource manager must understand how sensors make measurements,how information is ascribed to those measurements,and how change is inferred from them.
The fundamental process in remote sensing is the measurement of electromagnetic energy to obtain useful information(Schott, 1997).That energy can originate from the sun or from a source associated with the sensor,such as a laser or radio emitter,or can be emitted directly from the material because of its temperature.Like human eyes,electronic sensors are designed to measure re?ected energy in discrete regions of the electromagnetic spectrum called “spectral bands.”Because the physical and chemical properties of a given material cause it to absorb,re?ect,and emit electromag-netic energy differentially in different parts of the electromagnetic spectrum,the relative amounts of energy measured in different spectral bands can be used to infer something about the character of the object being observed(Schott,1997;Verbyla,1995).For optical imagery,measurements made in each spectral band are arranged in regular grids of picture elements(pixels),and grids combined from different spectral bands create familiar color digital images.LIDAR data are provided as postings,at either regular or irregular intervals,but can be,and usually are aggregated to regular grid cells for interpretation,analysis and change detection.Depend-ing on the type of lidar(discrete return or waveform),data may be provided as elevations of one or several returns from each posting or as a continuous record or return intensity with height.Likewise, synthetic aperture radar(SAR)images are generally processed to regular grids,but originate as side-looking images recording the differences in travel times and return intensity of transmitted micro-wave signals.
Extracting information from a digital image begins with“spectral space”(which for our purposes includes SAR intensity or compar-able LIDAR measurements).Spectral space is the data space that can be visualized by plotting measured intensity of re?ected radiance in different spectral bands against each other(Lillesand&Kiefer,2000; Richards,1993).Fig.1illustrates this concept for a picture of a ?ower and green leaves.All objects that appear to be the same color in the digital image have pixels whose re?ectance values group together in the same region of spectral space.Thus,green leaves and reddish?ower bases occupy different regions of the spectral space de?ned by plotting the re?ectance values in the red versus the green bands.Once regions of spectral space are labeled“?ower”or “leaf,”all pixels that fall in that region of spectral space can be ascribed those labels.Note,however,that the observed spectral space depends not only on the object itself,but on the illumination source,and that consistency in illumination is needed to apply labels in spectral space.Similarly,the spectral space of an image of a landscape can be labeled with regions corresponding to labels such as forest,water,etc.
Labeling the regions of spectral space requires external informa-tion.In the case of the?ower in Fig.1,the external information is the observer's prior knowledge of the spatial and spectral properties of a
Table1
Common natural resource attributes or indicators that are the focus of monitoring programs,grouped into broad categories according to the process or threat in?uencing that attribute.
Resource attributes/Indicators Process of interest/Threat
Change in size or shape of patches of related cover types Vegetative expansion,in?lling,or encroachment,a erosion b
Change in width or character of narrow,linear features Visitor use of paths or roads,?ooding effects on stream vegetation c;dynamics of terrestrial and submerged near-shore aquatic vegetation d
Slow changes in cover type or species composition Succession,e competition,eutrophication,exotic species invasion f
Abrupt changes in state of cover Disturbance,human-mediated development,g,h
land management i
Slow changes in condition of a single cover type Climate-related changes in vegetative productivity,j slowly-spreading forest mortality caused by insect or diseases,k changes in moisture regime
Changes in timing or extent
of seasonal processes
Snow cover dynamics,vegetation phenology l
a Hudak and Wessman,1998,Harris et al.,2003.
b Allard,2003.
c Nagler et al.,2009-this issue.
d Wang et al.,2007.
e Hostert et al.,2003.
f Asner and Vitousek,2005.
g Goetz et al.,2009-this issue.
h Townsend et al.,2009-this issue.
i Huang et al.,2009-this issue.
j Skakun et al.,2003,Wulder et al.,2005.
k Nemani et al.,2009-this issue.
l Reed et al.,2009-this issue.1383
R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
?ower in a picture.In the case of an image of a landscape acquired by a satellite,external information is most commonly obtained from the observer's prior knowledge of the landscape,from actual descriptive measurements made at sample locations on the landscape,or from other imagery more detailed than that for which the spectral space labels are needed.Examples of such data would include airphoto-interpreted land cover type,?eld-measured species composition within 1-ha plots,or ?eld-measured estimates of forest basal area or cover-type areal proportions.More advanced approaches to obtain external data include the use of (sometimes complex)models of systems and/or system components (Peddle et al.,2007).Some model-based approaches can provide structural information that cannot be derived solely from spectral characteristics.Regardless of the source,without such reference data the measurements from a satellite image may be of limited utility to a natural resource manager.Thus,the acquisition of appropriate reference data is critical in any remote sensing study.
With appropriate reference data,several methods of labeling regions in spectral space are possible (Fassnacht et al.,2006;Fraser et al.,2009-this issue ).A common approach is discrete classi ?cation,where hard boundaries are drawn between discrete regions,resulting in a categorical map with discrete labels of land cover (Lillesand &Kiefer,2000).Another approach is to allow overlap between regions in spectral space,resulting in “fuzzy ”labels that retain some of the information about mixtures of components within a pixel (Foody,1996;Wang,1990).Alternatively,gradients within spectral space can be related to variables that vary continuously,such as the percent vegetative cover within a pixel or to proportions of spectrally pure cover types (Cohen et al.,2003).
The heart of change detection and monitoring is comparing the position of a pixel in spectral space at different points in time.Images are acquired of a landscape in different years or different seasons,and the spectral space of those images compared.If a pixel's spectral
values place it in a spectral region associated with one land cover type in one date and in another land cover type in another date,we could infer that a change has occurred on the ground for the area measured by that pixel.However,a variety of other effects could cause change in spectral values for pixels over time,and separating informative changes from non-informative types remains a central challenge in remote sensing change detection.Much like the case of a single spectral space,changes in spectral space can be described using categorical,fuzzy,and gradient-based techniques,with properties discussed in Section 3.3below.
In summary,the foundational process in most remote sensing change detection is quantifying and labeling changes in the spectral space represented by a given sensor.The types of change that can be detected,the ability to meaningfully label them,and the con ?dence in those labels all depend on the speci ?c choices made during several sequential steps in a change detection project.3.Steps in a remote sensing change detection study
Remote sensing change detection studies involve a series of sequential steps that are detailed extensively elsewhere (e.g.Cihlar,2000;Coops et al.,2007;Lunetta,1998;Schott,1997).For the natural resource manager,our goal here is to simplify these steps into four broad stages:data acquisition,preprocessing and/or enhancement,analysis,and evaluation.The better a manager understands how decisions in each stage affect the outcome of the study or project,the better he or she can guide those decisions.3.1.Data acquisition
The data acquired in this step are both image data and the reference data that will ultimately be used to label information in the image and to evaluate the ef ?cacy of
products.
Fig.1.An illustration of spectral space.a)A standard digital photo of a ?ower and green leaves taken with a handheld digital camera.b)The re ?ectance of green energy for that photo (e.g.the “green band ”of the image).c)The red band of the image.d)A plot of the intensity of red versus green band re ?ected energy for the images in b and c.The nearly-white parts of the ?ower petals are high in both red and green re ?ectance,placing them in a different part of spectral space from the reddish pixels from the base of the ?ower.Those pixels are fairly low in red re ?ectance (i.e.not near the top of the red axis),but even lower in green re ?ectance,making them appear dark red.
1384R.E.Kennedy et al./Remote Sensing of Environment 113(2009)1382–1396
Table2
Resource attributes and speci?c image characteristics that need to be considered when acquiring imagery to monitor the attributes.
Resource attribute(s)Image type Opportunities and challenges in tracking over time
Spatial Spectral Temporal Image quality
Change in size or shape of patches of related cover types Fine grain(IKONOS,
Quickbird,Airphoto)
Fine grain allows delineation of
shape;but detection of change
in shape requires strong
geometric integrity over time.a,b
Change information is mostly
tied to spatial,not spectral,
properties.a
Tasked-acquisition may allow
better control over image
timing,but historical archive
unpredictable.
The orbit orientation and narrow
swath width of?ne grain imaging
satellites may require multiple
days to acquire image data for an
entire study area,which may
affect the effectiveness of
investigating time-sensitive
subjects on the ground.c
Change in width or character of narrow, linear features
Slow changes in cover type or species composition Fine grain(IKONOS,
Quickbird,Airphoto)
Useful when spatial texture
distinguishes cover types or
species,but limited spatial
extent may increase costs.
Broad physiognomic distinctions
between cover types possible,
but?ner distinction of species
and cover types compromised by
poor spectral depth.d
See above.Differences in view-angle and
shadowing introduce distortions
that affect interpretation of
cover and changes in cover over
time.e
Moderate grain,
multispectral
(Landsat,SPOT,
ASTER)
For many ecosystem types,slow
changes in cover occur over areas
larger than the grain of these
sensors,making them useful for
delineating bounds of affected
areas
Additional spectral depth of
short-wave infrared and
thermal bands can improve
separation among types,but
change in species composition
often impossible to track.
Historical archive of this type of
imagery among the longest
available and can be leveraged to
extract slow change information.f
Repeat interval is often
appropriate for changes that
occur over months or across
years,but relatively infrequent
overpasses can make matching
with seasonal or climatic
phenomena challenging.g
Consistent view angles aid in
change detection,but
unaccounted-for atmospheric
variations can introduce error;
cloudiness often a key
constraint.
Moderate grain,
hyperspectral
(AVIRIS)
Often used for tracking changes in
proportions of sub-pixel sized
components h;spatial extent
often smaller than multispectral
sensors.
The best chance for distinction
of species-composition,
although atmospheric
correction critical for detection
of subtle changes over time.
Tasked-acquisition may allow
better control over image
timing,but in practice can be
dif?cult to control.
Image quality typically high,
but geometric correction of
airborne platforms can be
challenging,and may introduce
more error than from
analogous satellite platforms.
Abrupt changes in state of cover Fine grain(IKONOS,
Quickbird,Airphoto)
Inference of land-use and land-
use change often possible
through direct image
interpretation,but automation
algorithms still in research phase;
small spatial extent may require
multiple images for large study
areas.Thus,costs may be high.
Poor spectral depth rarely a
hindrance because?ne spatial
resolution often allows
detection of disturbance or
development events.
Temporal depth of airphoto
archive(often many decades)
allows for detection of long-
term trends,but typically at a
fairly coarse temporal grain.
Image quality of historical
photos can sometimes reduce
con?dence in some land cover
labeling projects.
Moderate grain,
multispectral and
hyperspectral
(Landsat,SPOT,
Aster,AVIRIS)
Grain size a good compromise
that allows detection of many
disturbance type events across
large landscapes,although
unusable for some subtle types
of development or very small
disturbance events.i
Spectral depth allows detection
of many disturbance events
from spectral properties alone.j
Repeat interval generally
appropriate for most disturbance
types,although tracking of subtle
effects can be hampered by time-
of-season and cloud issues.Long
archive provides a useful baseline
for long-term monitoring.f,k,l
For most common disturbance
types,image quality suf?cient.
Clouds can obscure some
ephemeral disturbance events.m
Coarse grain
(MODIS,SPOT
VEGETATION)
Grain size appropriate for large
disturbances;subpixel
disturbances may be detectable
as proportional change.n,o
Spectral depth,particularly
thermal bands,can allow rapid
detection of?res.
Dense temporal record useful for
detecting lasting changes in land
cover at the sub-pixel scale.n
Cloud-screening and geometric
qualities of mosaicked images
can sometimes require temporal
smoothing to detect trends.n,q
Slow changes in condition of a single cover type Fine grain(IKONOS,
Quickbird,Airphoto)
Can be useful if process causes
noticeable changes in condition
(loss of vegetation,mortality)
in individual plants.d
Poor spectral resolution can
sometimes make detection of
subtle changes dif?cult d;spectral
distinction from background
likely dif?cult to automate,
forcing manual interpretation or
development of new methods
for automation.r
High cost of acquisitions may
make repeat imagery
untenable for capture of trends.
For long-term trends(many
decades),airphotos are the
only option,but shadowing and
view angle effects can make
even manual interpretation of
subtle change dif?cult.e
Moderate grain,
multispectral and
hyperspectral
(Landsat,SPOT,
ASTER,AVIRIS)
Many processes of interest
operate at spatial grain larger
than grain size of pixels,
making these sensors
especially useful.
Relative to?ne-grain sensors,
spectral depth of these sensors
improves spectrally-based
detection of changes in
condition,but subtle effects may
be dif?cult to discern spectrally
without hyperspectral imagery.s
Background noise can be
especially problematic because
signal of change is weak relative
to noise.
Long archive of some data
(Landsat)allows detection of
subtle effects over time.f If
effects are only manifested in a
narrow time of year(e.g.peak
biomass),however,lack of
control over timing of image
acquisition may introduce
noise.
Cloud effects are an issue,but
may be reduced if images over
many years are used to track
slow changes.j
Coarse grain
(MODIS,SPOT
VEGETATION)
Large grain and extent make
these sensors especially useful for
detection of change in vegetation
condition over very large areas.t
Coarse grain may make it dif?cult
to ascribe cause to changes within
pixels.
Spectral depth of coarse
grained sensors generally more
than suf?cient to capture slow
changes in vegetative cover.x
Temporal archive of AVHRR
data long enough to capture
trends,u but MODIS and SPOT
vegetation have records that
are currently too short to
capture long-term changes.
Ability to develop composite
cloud-free images allows for
capture of conditions at a
consistent point in the season
across years.
(continued on next page)
1385 R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
3.1.1.Image data acquisition
Rather than recreate lists found elsewhere of image sources or the broad categories of sensors(Kramer,1996;Lefsky&Cohen,2003; Sabins,1997),our goal is to describe the underlying considerations in image acquisition as they will speci?cally relate to the phases of decision-making in designing a remote sensing project(Section4of this paper).The four primary considerations are type,timing,quality, and cost of imagery.Table2lists the issues and challenges associated with using different image sources for each of the broad monitoring goals listed in Table1.
Radar and LIDAR imagery are not included in Table2,as they have not been used as widely for landscape change studies as have optical data,largely due to the comparable lack of availability of suitable data for land cover change detection until recently.However,SAR images have been used for a wide array of studies that are highly applicable to tracking changes in?ooding(Smith,1997;Townsend,2001),wetlands monitoring(Hess et al.,2003;Lang&Kasischke,2008;Wdowinski et al.,2008),for interferometric studies of geologic phenomena(Gens &VanGenderen,1996;Massonnet&Feigl,1998;Kaab et al.,2005),and to a lesser extent for landscape change studies(but see Quegan et al., 2000;Rignot&Vanzyl,1993).SAR imagery has been found to be especially useful for detection of change in urban areas(Dierking& Skriver,2002;Gamba et al.,2006;Henderson&Xia,1997;Ridd&Liu, 1998;Seto&Liu,2003).With increasing availability of airborne LIDAR data,more studies will likely use LIDAR to detect changes,especially in vegetation structure(Wulder et al.,2007a,b;Yu et al.,2006,2008)and topographic change(Woolard&Colby,2002;White&Wang,2003; Rosso et al.,2006).
Type of imagery refers to its spatial,temporal and spectral qualities,and re?ects the tradeoffs among these qualities in the design of sensors(Verbyla,1995).The spatial grain of a sensor is the area on the ground captured by a single sensor element,effectively the pixel size(although see Schott(1997)for a more detailed discussion), while the extent is the geographic scope of an image.The temporal grain is the frequency at which images of a given point on the Earth are acquired,and the temporal extent is the historical depth of that imagery.The spectral grain of a sensor relates to the width of the spectral bands in which it makes measurements,and the spectral extent to the breadth of the electromagnetic spectrum captured by all of the sensors.Generally,grain and extent in each domain are related: Finer-grain elements result in smaller extents.Tradeoffs across domains arise from engineering constraints.Spatial and spectral grain are opposed because the energy coming from a surface is?nite, and as that energy is divided into increasingly smaller pixels or narrower spectral bands,the signal strength falls(Schowengerdt, 1997).To maintain a signal above a critical threshold,one domain must be sacri?ced to facilitate?ner division of the other.In orbiting satellite systems,tradeoffs between spatial grain and temporal grain come about because larger pixels capture more of the Earth's surface at a time,allowing for more frequent overlap between images acquired on successive orbits and shorter repeat cycles for sensors with large pixels(Sabins,1987).The practical implication of these tradeoffs is that the natural resource manager may need to prioritize which domain is most relevant for a given monitoring goal of interest.
A key consideration driving many analytical and practical considera-tions in remote sensing studies is the relationship between the grain of the entities being mapped and the grain of the sensor(Woodcock& Strahler,1987).
Image timing and image quality must be chosen to minimize the in?uence of unwanted effects on spectral space,since such effects can obscure real change or produce the false appearance of change.Key issues to consider are phenological state of the landscape,sun angle,
Table2(continued)
Resource attribute(s)Image type Opportunities and challenges in tracking over time
Spatial Spectral Temporal Image quality
Changes in timing or extent of seasonal processes Coarse grain
(MODIS,SPOT
VEGETATION)
Broad extent allows detection
of regional trends in cyclic
processes;coarse grain size and
mosaicking make pixel-level
tracking of phenology dif?cult.v
Spectral depth suf?cient for
tracking phenology and snow
cover.w
Most products are composited
to near-weekly or bi-weekly
temporal grain,x which can
diminish precision of
estimates.
Most natural resource
managers will likely be
interested in using
automatically-produced maps
whose quality depends on
speci?c algorithms y.However,
case-speci?c maps can be
created from high-quality raw
data by remote sensing
specialists.z
a Zhang and Fraser,2007.
b Wang and Ellis,2005a,b.
c Wang et al.,2007.
d Lecki
e et al.,2004.
e Fensham et al.,2007,Fensham and Fairfax,2007.
f Kennedy et al.,2007b.
g Olthof et al.,2004.
h Asner et al.,2005.
i Cohen and Goward,2004.
j Huang et al.,2009-this issue.
k Kennedy et al.,2007a.
l Wang et al.,2009-this issue.
m Olthof et al.,2004.
n Potter et al.,2005.
o Zhan et al.,2002.
q Reed et al.,2009-this issue.
r Paci?ci et al.,2007.
s Asner and Heidebrecht,2002.
t Wessels et al.,2004.
u Myneni et al.,1998,Potter et al.,2005.
v White et al.,2005.
w Reed et al.,2009-this issue,Hall et al.,2002.
x Nemani et al.,2009-this issue.
y Cohen et al.,2006.
z Vikhamar and Solberg,2003.
1386R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
atmospheric condition,and geometric and radiometric quality of the imagery.These issues are described in greater depth in these key references(Coops et al.,2007;Yuan et al.,1998).Cost of imagery is an important consideration for most natural resource agencies,and is amply discussed in other references(Gross et al.,2006;Turner et al., 2003).Note that the greatest cost in many remote sensing studies is not the acquisition of imagery,but in the labor needed to process the imagery,derive information,and evaluate the results(Lunetta,1998).
3.1.2.Reference data acquisition
Reference data are independent sources of information that allow a remote sensing specialist to relate patterns in spectral space to real quantities or phenomena on the earth surface,or to validate or evaluate the products that come from such a process(Campbell, 1996).For example,?eld crews may make areal measurements on the ground of percent cover of different land cover types,and these land cover type proportions can be linked to the spectral space to build generalized rules that relate regions in spectral space to those land cover labels.If some data are withheld from the rule-making process, their measured land cover type proportions may also be used to evaluate how well the rules apply outside of the plots used to make them,providing a measure of the utility of the rules.Because the reference data affect the rules used to make maps as well as the ability to quantify their robustness,the quality and availability of reference data may drive the questions that can actually be addressed with remote sensing.This,in turn,makes assessment of reference data a critical step in the planning process.Table3lists the challenges and opportunities associated with using various reference data in support of the monitoring resource attributes listed in Table1.
Although not universal,reference data collected in a probabilistic statistical framework are commonly used to both train the classi?er and to assess classi?cation accuracy.The statistical framework provides rigor and credibility,and may be used to minimize bias and estimate variance, key to assessing data quality(Stehman,2000,2001).The statistically-based accuracy assessment consists of three primary components,a)a response design that describes how the“true”value for the ground
Table3
Resource attributes and the issues involved in collecting reference data to monitor these attributes.
Resource attribute(s)Reference data source Opportunities and challenges in tracking over time
Change in size or shape of patches of related cover types Airphotos Direct observation of patches or features often possible,but subtle changes
in shape or size may be dif?cult when comparing images from two different
acquisitions because of differences in sun or view angle,a or in phenological state.b
Change in width or character of
narrow,linear features Ground measurements Low ambiguity about species or feature type,but relatively low precision of
measurement of patch or linear feature metrics may diminish sensitivity to subtle
change over time.c Historical reference data may be dif?cult to co-locate.d
Slow changes in cover type or species composition Airphotos Subtle distinction of species type may be dif?cult.Quanti?cation of composition
may not be suf?cient for subtle change.e
Ground measurements Direct observation of land cover type or species usually reliable on the ground,but
co-location of plots and imagery often dif?cult,f and semantics of land cover or
species groupings may vary among observers or projects over time.Subtle
distinctions in cover or species type require many samples to resolve statistically,
which is often challenging with ground-based measurements.g
Abrupt changes in state of cover Indirect measurements(census data,
development data)Useful for validation of remotely-sensed measurements of development at broad spatial extents(county,state level).
Airphotos Often the best approach for quick and effective interpretation of abrupt
disturbance events h,i;historical data allow for statistically valid observation
of low-frequency disturbance events.
Ground measurements Field validation often must occur shortly after the event for?eld observers
to discern disturbance type;before-and after-?eld observations
of disturbance events often sparse.j
Repeat?xed-wing or helicopter over?ights Reliable and sometimes used for resource inventories(such as the USDA
Forest Service's Forest Health Monitoring program k),but expensive to implement.
Attention need be paid to geographic precision.f
Moderate grain sensors Landsat-type sensors can provide a measure of state or of changes in
broadly-de?ned land cover types for validation of coarse-grained sensors.l
Slow changes in condition of a single cover type Airphotos May be possible for changes that result in mortality,but often challenging for
more subtle measurements of vigor or health.n
Ground measurements Single date-direct measurements may allow discrimination of subtle changes in
condition,m but repeat measurements of plots are ideally needed to validate changes.
Changes in timing or extent of seasonal processes Repeat?xed-wing or helicopter over?ights See comments on same type above.
Indirect measurements(stream?ow data,etc.)May allow comparison of aggregated effects at watershed or basin scale,but
connection with remotely-sensed data may require mechanistic modeling. Airphotos Capture of processes dif?cult,but may be useful to model or to test estimates
of cover or phenological state at one point in time
Ground measurements Often the only means of capturing seasonal processes,but small spatial grain
of ground measurements and low number of samples make direct comparison
with remotely-sensed data challenging.Also timing of?eld data collection is
critical because of speed of change in processes,making?eld costs high.
a Wang et al.,2007.
b Goetz et al.,2003.
c Johansen et al.,2007.
d Kennedy et al.,2007a.
e Wulder et al.,2007a,b.
f Sánchez-Azofeifa et al.,2003.
g Congalton and Biging,1992.
h Cohen et al.,1998
i Jantz et al.,2005.
j Boutet and Weishampel,2003. k https://www.wendangku.net/doc/f55204582.html,/.
l DeFries et al.,2000.
m Leckie et al.,2004.
1387 R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
condition will be assigned and interpreted(what is“true on the ground”),b)a sampling design that describes how we will pick our speci?c?eld sampling locations,and c)analysis protocols that specify formulas and methods applied to the sampled reference data in estimating the value and accuracy of change(Strahler et al.,2006).
Ideal reference data are those that match imagery spatially and temporally,that measure a property that is thought to be detectable with the imagery,and that are designed to allow construction of the models required to label spectral space and to evaluate the robustness of?nal maps(Congalton&Green,1999).For the most part,such data rarely exist unless they were collected speci?cally for the purposes of remote sensing.Typical ecological measurements often do not capture the average conditions of an entire pixel,are not collected at the correct time,and are dif?cult to geographically link with the imagery(Kennedy et al.,2007a).The heterogeneity of the conditions within each reference plot is also critical for building and evaluating models with reference data(Fassnacht et al.,2006),as high classi?cation accuracies are harder to achieve in more heterogeneous environments(Smith et al.,2002),and the appropriate sampling for reference data is also affected by spatial autocorrelation and heterogeneity(Congalton&Green,1999;Friedl et al.,2000;Strahler et al.,2006).In addition to these challenges encountered with any remote sensing mapping project,challenges arise that are speci?c to the mapping of change.First,reference data ideally should be available for conditions before and after a change,which in practice can lead to validation using different data sources at different times (e.g.Huang et al.,2007).Second,even with similar data sources before and after a change,it is often dif?cult to completely replicate a given reference datum on the ground because of geolocational imprecision,and in high-resolution airphotos or imagery because of shadowing and view angle variations across years(Paine,1981;Wang et al.,2007).
In remote sensing,it is often preferable to collect many?eld plots of slightly lower quality or richness rather than collecting few plots that are rich in information.Strategies for sampling are well-covered elsewhere for both the general case(Cochran,1977;Congalton& Green,1999;Thompson,2002)and the special case of remote sensing change detection(Biging et al.,1998;Stehman,1999).Regardless of the particular method chosen,the geographic location of reference data should be determined before setting foot in the?eld or obtaining aerial photos,etc.,using a process that eliminates human bias in choosing plots.A common strategy is to use a higher-resolution remote sensing product to validate a coarser product(Cohen et al.,2001;Congalton& Green,1999;Lambin&Ehrlick,1997;White et al.,1996).Several papers in this special issue illustrate a range of approaches for reference data collection,from intensive?eld measurements of habitat condition at a relatively small number of plots,to simpler measurements of cover type at a larger number of plots(Nagler et al.,2009-this issue),to relatively quick GPS-linked?eld photos at an extremely large number of plots(Wang et al.,2009-this issue).
3.2.Image pre-processing
The goal of pre-processing is to ensure that each pixel faithfully records the same type of measurement at the same geographic location over time(Lunetta,1998).Preprocessing is especially critical in change studies because the detection of change assumes that the spectral properties of non-changed areas are stable,and inadequate pre-processing can increase error by causing false change in spectral space.(Coops et al.,2007;Lu et al.,2004;Lunetta,1998;Peddle et al., 2003;Schowengerdt,1997)Increasingly,pre-processing steps are becoming automated and resulting in free datasets of relatively high quality(Fraser et al.,2009-this issue;Masek et al.,2006).Note,how-ever,that each step in pre-processing alters the position or spectral properties of pixels in the imagery,and thus each step has the poten-tial to introduce error.
A?nal step often labeled pre-processing is image enhancement, which is the mathematical rotation,compression,or distortion of spectral space to accentuate desired features and suppress noise (Lillesand&Kiefer,2000).Many natural resource managers may be familiar with one type of enhancement known as vegetation indices,such as the normalized difference vegetation index(NDVI; Tucker,1979),but a wide range of enhancements are possible. Several papers in this special issue utilize derived indices as a key step in their process(Crabtree et al.,2009-this issue;Nagler et al., 2009-this issue;Nemani et al.,2009-this issue;Townsend et al., 2009-this issue).Note that image enhancement techniques do not create new information,but rather they highlight information present in the original spectral data.
In theory,if pre-processing has been perfectly successful,all changes in spectral value in a given pixel between two images can be ascribed to actual changes in the conditions of the surface represented by that pixel.In practice,no pre-processing steps account for all effects perfectly.Thus,some portion of the spectral change observed in a pixel over time is uninformative,and the analytical techniques in the next phase of the project must take this into account.
3.3.Extracting information
Once two or more images have been pre-processed and/or enhanced,many mathematical approaches are available to detect and label pixels that have or have not changed(Yuan et al.,1998). Despite the variety of methods,most change detection approaches contain a modeling(or functional algorithm)phase and a subtraction phase.The modeling phase refers to the development or implementa-tion of algorithms to infer meaning from spectral data,while subtraction refers to the process of comparing dates via image algebra or other methods.Key considerations are how the functional step treats spectral information,and whether the subtraction phase precedes or follows the modeling phase(Gong&Xu,2003;Yuan et al.,1998).Table4lists how various analytical techniques relate to the broad monitoring goals listed in Table1.
The algorithm phase can involve discrete,fuzzy or continuous methods.Discrete methods are attractive because changes are typically de?ned in terms of land cover classes that are familiar to natural resource managers(Wang et al.,2009-this issue)and that can be used directly in subsequent habitat fragmentation or similar analyses(Townsend et al.,2009-this issue).The primary drawbacks are that subtle changes of condition within a land cover class are missed,and that pixels near the spectral boundaries of classes are more likely to be incorrectly labeled as having changed,simply due to imperfect pre-processing,unless the change analysis is con-strained by available high resolution vector GIS data.Fuzzy methods acknowledge the potential confusion among classes in spectral space,and can be designed to capture subtle change within classes (Foody&Boyd,1999;Kennedy et al.,2007a).The fuzzy nature of these classes may be non-intuitive,however,and labeling change among a matrix of many overlapping classes may be untenable or non-informative in practice.Continuous-variable approaches allow for capture of subtle distinctions between two dates,but effort must be made to develop robust methods to de?ne what level of change is actually meaningful(Yuan et al.,1998).In addition,continuous-variable methods that simultaneously track several variables often must be collapsed into categorical variables to simply make sense of the change(Chen et al.,2003),potentially diminishing the advantage over strictly discrete methods.
If the models are?rst applied separately to the spectral space of each image to create two maps,then change is detected and labeled by comparing(differencing)those maps(Fig.2a;also Haertel et al., 2004).From the practical perspective of the land manager,taking this approach places a high premium on appropriate reference data tied temporally to each image,and less on costs associated with
1388R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
normalizing the spectral space of the two images(Yuan et al.,1998).A key challenge,however,is that errors in the maps from each image are compounded in the change detection map,limiting the maximum accuracy that can be achieved(Cohen&Fiorella,1998).If spectral space is?rst differenced(often through simple subtraction)and then an algorithm is applied to the spectral difference image,change is inferred from the spectral character of the spectral difference space (Fig.2b;also Lambin&Strahler,1994).The expected spectral difference for no-change is zero in all spectral bands,and change is detected as deviation from zero(although usually with a non-zero threshold to compensate for imperfect pre-processing).This approach is attractive in its explicit focus on the change,and avoids compounding errors in maps.It also allows for detection of change in any spectral direction(Lambin&Strahler,1994;Malila,1980),for detection of subtle effects like insect defoliation(Muchoney&Haack, 1994;Townsend et al.,2004),and for development of general models that can be applied across images from many years(Cohen et al., 2006).The challenge in using this approach is that radiometric normalization(e.g.,for atmospheric,phenological or BRDF differences [bidirectional re?ectance distribution function,Schaepman-Strub et al.,2006])must be very robust,and that reference data that speci?cally measure change(rather than just state)must be available. Moreover,the results can be confusing because the differencing step removes information about the origin or terminus of the pixel in spectral space(Cohen&Fiorella,1998).
Some important approaches combine or omit the differencing phase.One strategy begins with an existing land cover classi?cation map,and then uses image algebra to detect locations of change on the map.New classi?cation models are then applied only to label the changes,while the classi?cation labels from non-changed areas are simply carried forward(Fig.3;also Fraser et al.,this issue;Parmenter et al.,2003).For natural resource monitoring,this is attractive in allowing use of existing or familiar land cover maps while focusing on the change component of the spectral signal.However,it cannot be extended inde?nitely:cover classes in the original map are constantly degraded by change,reducing their spectral?delity over time and requiring eventual creation of a new land cover map. Another approach that does not include the differencing stage involves application of a model to the combined(stacked)spectral space from all of the component images(different years or dates)to infer information about change.Such an approach diminishes the need for robust normalization among images,but results can be
Table4
Resource attributes and the considerations involved in analytical change detection techniques to detect meaningful changes in them.
Resource attribute(s)Analytical techniques Opportunities and challenges in applying to change detection
Change in size or shape of patches of related cover types Segmentation or classi?cation and patch analysis
applied to two images,followed by subtraction
Direct measurement of changes in patch shape closely meets monitoring goal,a,b
but patch edge delineation may be dif?cult to reproduce over time.Also,patch
by patch observation over time is not a common technique,and summary
metrics of patch shapes,sizes,c etc.may obscure local-level issues.
Change in width or character of narrow, linear features Subtraction of images,identi?cation
of changes,followed by segmentation or
classi?cation and patch analysis
Focus on change may diminish false negatives relative to prior approach,but
requires that the change event be spectrally separable in the image data.
Labeling of the change may be dif?cult to automate if the shape characteristic
patches of change are ambiguous.
Abrupt changes in state of cover Time-series analysis of many years of
continuous-variable image data See comments in above cell.Loss of baseline conditions caused by differencing can make labeling the land cover change dif?cult.d
Discrete classi?cation of two images,followed by comparison of classi?ed maps Allows detection of phenomena more subtle than classi?ed approaches,and
use of time-series can reduce problems of variable image backgrounds and phenology.e,f,g Preprocessing steps are highly important,however,and
reference data to match each image are often impossible to?nd.f
Labeling of change is straightforward and radiometric pre-processing is of minor importance,h,i but errors in two single-date images are compounded.
Subtle effects are often dif?cult to detect.Reference data needed for both images, often forcing use of image-based reference.j
Slow changes in cover type or species composition Slow changes in land cover type can only be detected using discrete classi?cation if the interval between images is large.h Generally,continuous-variable methods are more appropriate.k
Slow changes in condition of a single cover type Two or more images subtracted,followed by
continuous-variable modeling of change
(regression,change vector analysis)
Focus on change may limit geographic scope needed to understand processes,l
but reference data that match beginning and end points are https://www.wendangku.net/doc/f55204582.html,beling
of change can be dif?cult because of loss of baseline.d
Continuous-variable models of sub-pixel
proportions(regression,spectral unmixing,
fuzzy-classi?cation)applied to two or more
images,followed by subtraction.
Proportional representation allows for detection of subtle effects,m but is also
more sensitive to variation in background re?ectance caused by year-to-year
variation in conditions during image acquisition.A high premium is placed on
accurate pre-processing,n and reference data must be robust and widespread
to allow building of statistical models.
Time-series analysis of many years of
continuous-variable image data or derived
(vegetation index)data
Detection of subtle trends more feasible than with any two-date approach,but
image pre-processing steps critical,including cloud and cloud-shadow screening,
and subtle change in sun angle or phenology may cause false positives.f
Changes in timing or extent of seasonal processes Allows detection of broad geographic and temporal patterns generally undetectable with two-date approaches.e Preprocessing steps(including cloud screening,image mosaicking, and trend smoothing)are critical to success of method and often challenging.
a Weisberg et al.,2007.
b Ellis et al.,2006.
c Li et al.,2003.
d Cohen and Fiorella,1998.
e Potter et al.,2005.
f Kennedy et al.,2007b.
g Huang et al.,2009-this issue.
h Wang et al.,2009-this issue.
i Viòa et al.,2007.
j Cohen et al.,1998,Kennedy et al.,2007a. k Dougherty et al.,2004.
l Lambin and Strahler,1994.
m Roberts et al.,1998.
n Yuan et al.,1998.
1389 R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
dif?cult to interpret and are generally applicable solely to the combined data space under study(Coppin et al.,2004;Fung&Siu, 2000).A second family of approaches using more than two dates of imagery seeks to identify temporal patterns or trajectories in the sequence of imagery(Garcia-Haro et al.,2001;Hostert et al.,2003; Huang et al.,2009-this issue;Kennedy et al.,2007b;Lawrence& Ripple,1999;Lu et al.,2003;Potter et al.,2005).These approaches are attractive because they capture overall temporal trends,but generally require robust radiometric normalization and may involve complex statistical analysis to infer change.As image processing and data storage capabilities improve,however,these approaches hold great promise in removing year-to-year variation from classi?cations of single date images,in detecting longer term processes than those typically captured,and in detecting more subtle processes than can be achieved through two-date change detection alone.
3.4.Evaluation and reporting
Monitoring may stimulate costly management responses.Erro-neous information may lead to inappropriate action,for example, remediation when it is unnecessary,or lack of action when interven-tion is needed(Ronnback et al.,2003).Therefore,information quality must be evaluated.In addition,the procedures used to create this information must be reported such that external parties can assess their results.3.4.1.Evaluation
Scientists have developed standard techniques for assessing map accuracy(Congalton&Green,1999;Gopal&Woodcock,1994).Error is typically quanti?ed statistically by comparing the map to independent reference data at a sample of locations in a landscape.When the map is categorical,the errors are reported as proportions accurately described within each class,often summarized across all classes in a table known as a contingency matrix and sometimes summarized on a per class basis(Wang et al.,2009-this issue;Fig.4).When the map is a continuous variable,the errors are reported as real numbers such as mean error,root mean square error,or other summary statistic. The actual agreement between the reference data and the map is a function of both the spatial accuracy of the two data sources,and the agreement in the labels assigned,but in practice the contribution of spatial error to the?nal agreement is dif?cult to disentangle.In all cases,large sample sizes improve estimation,but sometimes statistical approaches can be used to leverage small sample sizes (such as bootstrap or jackkni?ng procedures),allowing evaluation of accuracy when expensive?eld samples are sparse(Cohen et al., 2003).
To conduct a proper accuracy assessment,the independent data must be considered“truth,”in that they were collected without error (Congalton&Green,1999).In practice,reference data have errors in both location and in label,just as the map data do,and measurements not designed for remote sensing typically do not capture the average conditions of an entire pixel(Wulder et al.,2007a,b).Acknowledging that some error exists in reference data,values are often deemed true when they are known at substantially higher accuracy than the mapped values.When reference data are known to an accuracy level only moderately better than the map itself,the analysis is more appropriately considered an evaluation of agreement rather than a true accuracy
assessment.
Fig.3.An amalgam approach to change detection,where differencing is used to identify
only pixels that have changed,and single date mapping rules are applied only to those
changed
pixels.
Fig.4.Sample locations are often selected based on a randomized cluster method.
Cluster samples are selected across the landscape using some random process.A
number of points are then sampled in some distance-constrained manner near the
cluster center,either in random process(shown),or some systematic process(not
shown).Clustering reduces travel time among samples,thereby increasing the sample
size on a?xed budget.Cluster sampling is often an optimum tradeoff between the need
to seek independent samples,and increase the statistical power through higher sample
numbers.
Fig.2.Two means of conducting remote sensing based change detection.a)From two
separate spectral images,a mapping or classi?cation function is applied,resulting in
two separate maps.These maps are then compared through differencing or analogous
process to derive change.b)The spectral values of the two images are differenced
directly,and a mapping or classi?cation algorithm is applied to that different space.
1390R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
The balance between quality of reference data and number of samples must be considered carefully.Because larger sample sizes improve the precision of the estimate,it may be advisable to sacri ?ce some precision of measurement at any single reference plot in the interest of acquiring many more plots.This is particularly true in change detection studies,because areas that have changed typically occupy only a small portion of the landscape,and because there may be many different categories of possible change.Strati ?ed,cluster,and double-sampling methods may be particularly attractive approaches to distributing samples (Fig.5,also see Czaplewski &Patterson,2003;Kalkhan et al.,1998).These sampling strategies can increase the number of sample plots that can be collected for a given time or budget constraint,which is particularly important for monitoring projects where repeat visits across many years are planned.Strati ?cation and focus on those areas that have changed has been advocated to increase precision in the change estimate (Biging et al.,1998),as this often gives a more precise estimate,but requires useful strata be available.Changed areas are often 25%of the landscape or less,and without strati ?cation,these small regions may be under-sampled.While strati ?cation and clustering may substantially improve accuracy estimates and/or save time and money (Lohr,1999),many of these sampling strategies cannot be implemented without some knowledge of the spatial autocorrelation in the sampled variable,as most statistical accuracy estimates depend on an assumption of sample independence,and the estimates must be adjusted if samples are autocorrelated (Congalton,1998).
3.4.2.Reporting
Long-term monitoring will eventually rely on different sensors,training datasets,and analytical techniques.Accurate reporting of all phases of a project is thus critical to ensuring the long-term value of the data and the ability to evaluate prior results and infer change.
Data acquisition reporting should follow reporting requirements for non-imagery spatial and non-spatial data (Michener et al.,1997,FGDC:https://www.wendangku.net/doc/f55204582.html,/standards/standards.html ).Sources and disposition of data,including agreements on access and distribution,should be included in documentation.It may be important to include a discussion of the criteria used to choose imagery so that parallel criteria can be applied in the future.Part of this process is to document whether the spatial,spectral,or temporal characteristics of the imagery imposed constraints for the particular monitoring goals of the study.When ancillary spatial data are included as part of the project,they should be described from the perspective of how their spatial and temporal properties could affect the ?nal products.Documentation of reference data should be suf ?cient to allow future users to either recreate the data or re-visit a site.
Reporting on image pre-processing steps is critical because of the many image analysis steps involved in a typical remote sensing study.Documentation must be comprehensive enough to permit duplication of all steps,including the use of the same algorithms or models and parameters as well as discussions of why the methods were chosen.Errors associated with each model should be reported,noting that errors were caused by algorithm assumptions,by inaccurate reference data,and/or by spatial and temporal variation in imagery and datasets.
Analytical techniques for mapping and change detection also require detailed reporting.For projects that involve land cover class maps,particularly those speci ?c to a given site,documentation of steps used to build the classi ?cation must be provided to allow crosswalking between current and future land cover schemes.Legend design and cross walk procedures should follow an established approach (e.g.,Strahler et al.,2006).Error assessments conducted in the evaluation phase should include all raw data as well as the summary data used to evaluate overall performance.For all such data,the spatial and temporal grain of the analysis should be documented,especially if the analysis is conducted on the multi-pixel basis (for example,as average conditions across larger polygons).Also,it is important to evaluate whether the errors are equally distributed across the spatial extent of the study area,or whether different areas have different error properties (Fassnacht et al.,2006).
A part of reporting is archiving enough data to allow future investigators to re-evaluate or re-process the data.All raw imagery must be archived,using formats that are as transparent and generic as possible,as well as all models and reference data.Archiving of all intermediate products is not necessary,provided all information and algorithms needed to recreate those data are archived.If interpreta-tion of imagery (including photos)was conducted,then libraries of voucher specimens (type photos)should be included.4.Summary
The four phases of a remote sensing project described here are generally carried out in sequential order,but planning for such a study must consider all phases simultaneously.Each phase depends on prior phases,and decisions made early on can constrain options or inference later.Thus the entire arc of the study needs to be considered when managers are evaluating whether and how to include remote sensing in the monitoring of natural areas (Lunetta,1998).This is the subject of the next section.
5.Phases in the design of remote-sensing based monitoring projects This paper focuses on the remote sensing aspects of monitoring projects.Before initiating a project,the project manager must ?rst ensure two conditions exist.First,there must be an explicit process,with suf ?cient time,for collaborative development between natural resource and remote sensing specialists.Second,there must be a suf ?ciently clear and precise articulation of the monitoring (change detection)objectives.Fancy et al.(2008)emphasize the importance
of
Fig.5.A heuristic tool to illustrate the connections among image type,analytical techniques,analysis form,and the various monitoring goals outlined in Table 1.To use the ?gure,begin at the edge with one of the monitoring goals.The area de ?ned by that goal or the dashed/dotted lines indicates the domain of that goal in most common change detection studies.By following that domain in towards the center of the circle along the perpendicular,the other components most commonly associated with that goal are encountered.The shaded pie shape illustrates this for just one goal.
1391
R.E.Kennedy et al./Remote Sensing of Environment 113(2009)1382–1396
clear monitoring objectives.The monitoring objectives may be slightly revised during the collaborative development process,but inadequate speci?cation of objectives commonly leads to failure.
Ultimately,the appropriate strategy for extracting information on change will depend on the type of change being sought,the availability of appropriate imagery to detect that change,as well as the availability of reference observations to interpret and label the changes that are detected.Although in theory any combination of imagery,analytical technique,and reference data could be used in a natural resource monitoring study,in practice some combinations work more effectively and are found together more often in the literature.Fig.5shows how this reality can simplify the decisions that must be made during planning to monitor resource attributes listed in Table1and replicated in the outside ring of Fig.5.By traversing the concentric rings inward from any monitoring goal,the typical data types,reference data sources(e.g.“Airphoto data for reference”),and analytical techniques used to meet that type of goal are encountered. As an example,the pie-shaped shaded area in Fig.5shows that monitoring slow change in cover type typically requires airphoto data to develop reference information,could be analyzed at either the patch or the pixel level,could use either proportional or discrete descriptors of cover type,and likely would need moderate to high resolution imagery to carry out.Note that this?gure is intended to be suggestive rather than exhaustive;Turner et al.(2003),Kerr and Ostrovsky(2003)note ecological applications of similar sensors not identi?ed in Fig.5.
5.1.Phase1:Identify imagery appropriate to detect changes in resource attributes
5.1.1.Step1.Identify management or conservation attribute or indicator
In the initial phase of planning,the natural resource manager must identify the focal resource(sensu Fancy et al.,2008),key processes that act on the resource,and the resource attributes that are the focus of the monitoring.These are analogous to the Values/Threats/ Indicators paradigm of resource management(Hockings et al., 2006),but applied more broadly.The focal resource may range from a speci?c organism to an entire landscape or region,and may be biotic or abiotic.The processes that act on the focal resource may be external (e.g.hurricanes,?re,climate change,land cover conversion)or internal(e.g.succession of vegetation communities,eutrophication of water bodies).Such processes correspond to column2in Table1.
Key questions to address:What is the focal resource?What is(are) the process(es)of interest that act on that resource?What are the manifestations of that process on the resource attributes of primary interest?Is it critical that changes in the focal resource be detected everywhere they occur,or is a summary of an average effect useful? What are the management/conservation decisions in?uenced by detecting changes in the resource attributes?How quickly must changes be detected to implement appropriate management responses?
5.1.2.Step2.Identify potential imagery of appropriate grain and extent
In consultation with remote sensing specialists,use spatial, temporal,and spectral properties of the resource attributes to identify potential image sources.In all aspects of this phase,consider both the cover type(i.e.focal resource)of interest and the process that acts on it.Phinn et al.(2003)provided one framework for determining appropriate imagery.
Key spatial questions:What is the spatial grain needed to resolve the focal resource?What is the spatial grain of key process that acts on that resource?Is it necessary to capture the fate of individual organisms to capture changes in resource attributes,or can the behavior of many neighboring organisms at a larger grain size capture the necessary information?Over how large an area must change be tracked?Can a sample of images be used?
Key temporal questions:How fast do detectable changes in the focal resource occur?Do changes occur quickly in one place and then not recur for a long time(e.g.?re,?ood,etc.)or are changes a‘trend’that occurs slowly in the same place over time(e.g.successional changes, slow melting of glaciers,etc.)?Does the focal resource return to its prior state(in spectral terms)rapidly following the change,or do the spectral effects of the change persist?To capture resource changes with snapshots,what frequency of observations is required(con-sidering the pace of the process and the management activities that need to respond to it)?Are there certain windows of time when observations should or should not be obtained?Over what period must measurements occur to detect or track relevant changes in resource attributes?
Key spectral questions:Does the focal resource have a spectral quality that distinguishes it from its background?If not,is it related to some other resource or surface characteristic that is distinguishable? When the process acts on that resource,what changes in spectral quality are expected?Do those changes differ from ambient changes in spectral qualities of other areas unaffected by the process?Is there a sensor whose spectral measurements(grain and extent of spectral measurements)facilitate measurement of those spectral differences? If not,are there other related focal resources or associated resource attributes that have spectral properties that better match those of a given sensor?
5.1.3.Step3:Evaluate availability of potential imagery
Note cost and availability of both historic and future imagery, relative to spatial and temporal extent.If the imagery must be purchased,consider the costs needed to match its properties to the properties of existing data to which it will be compared.Resolve potential tradeoffs in spatial and temporal properties and availability, and explore whether modi?cations to monitoring objectives or a re-framing of questions could add alternative imagery types to the list.
5.2.Phase2:Estimate costs of pre-processing and analysis
With the assistance of remote sensing specialists,evaluate the pre-processing steps and analytical techniques that are required to detect meaningful changes in the resource attributes from conditions of no-change and of uninteresting change.
Key questions:What level of geometric processing is needed to align images to capture the spatial grain of the process or resource attribute of interest?Do the spectral changes associated with the resource attributes require normalization between images,and if so, what level of root-mean-square error is acceptable?Given the availability of imagery,are uninteresting changes(in the background, in the ambient vegetation,etc.)likely to be confused with spectral changes in resource attributes of interest?Do the changes of interest result in changes in cover type,or are they more closely associated with changes in the condition of the cover type?Should change information be detected and labeled with categorical variables,or is it necessary to capture change as a continuous variable?Do maps resulting from the change detection require additional processing(e.g. patch or pattern analysis,etc.)to provide useful information?How much labor/processing time and cost is likely associated with all of these steps?How much of the process can be automated?What level of expertise would be needed to carry it out?
At the end of this phase,each candidate set of imagery should have an associated set of potential processing and analytical steps associated with it,and a set of estimated costs for each step.
5.3.Phase3:Evaluate the availability and cost of appropriate reference data
Consider the full range of possible independent sources of information,including:?eld measurements,?ner-grained image
1392R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
data,ancillary geospatial information(vector or raster data),and expert knowledge.
Key questions:Do the reference data agree in spatial and temporal scope with the image source?If not,what potential error may be introduced when these data are used to train or evaluate change detection results?Do these data record quantities that can be related to the metrics resulting from the change detection techniques de?ned in Phase2?Can the precision and accuracy of the reference measure-ments be quanti?ed?What is the cost of acquiring these data?
At the end of this phase,the full suite of image,analysis,and reference data should be available to address the monitoring objective.This process may need to be repeated for other monitoring objectives that are to be addressed in a coordinated,parallel effort.
5.4.Phase4:Characterize performance of different options in terms of cost,con?dence in resulting maps,and the ultimate utility of those maps Each combination of image,pre-processing steps,analytical techniques,and reference data will likely produce a map or analytical result with different information content and different expected sources and magnitudes of error.The information content and errors may also vary for different focal resources.The goal in this?nal phase of planning is to evaluate the costs and bene?ts of the different options and select the approach that provides the best all-around bene?t.Most monitoring programs will want to simultaneously track as many resources and attributes as possible,and decisions on imagery sources and processing methods will necessarily require compromise that balances cost,imagery availability,and ability to detect changes in resource attributes of most interest.Practically speaking,a solution that meets80%of monitoring goals at a small cost will be selected over a very expensive solution that attempts to meet all goals.
This?nal decision is a type of cost–bene?t analysis,but it is important to recognize that resultant change detection maps will have essentially two levels of bene?t.The?rst level relates to how much the map itself can be trusted in the information it provides,which can be quanti?ed in a standard error analysis based on the reference data. This provides an important sense of how“good”the map is,and is often the criterion on which remote sensing specialists focus. However,the ultimate utility to a natural resource manager also depends on whether the information in the map is actually relevant and useful for management.A map that is90%accurate for a given attribute is still useless if that attribute has no management relevance, and conversely,a map that is60%accurate for a different attribute may be extremely useful for a manager,because the starting point on that attribute may be essentially zero(Czaplewski&Patterson,2003).
It is also important to recognize that remote sensing data may not be appropriate for many monitoring goals.This is particularly true when monitoring seeks to detect processes that result in little or no spectral change,for changes that require frequent,high-spatial resolution monitoring,or for subtle changes that occur within a background matrix of extreme variability.It may be more cost-effective to design a?eld-based sampling design,perhaps strati?ed in accordance with information from remotely sensed data.
6.Conclusions
Natural resource managers will increase the likelihood of meeting their monitoring goals with remote sensing by actively participating in the design and planning of a project.Remote sensing science can aid natural resource managers in understanding landscape dynamics over time,and the ultimate utility of derived maps can be strongly enhanced by matching the manager's expectations and needs with the available tools and techniques.In this paper,we developed a general framework that we have successfully used to collaboratively develop operational natural resource monitoring based on remotely sensed data.An understanding of the concepts and process articulated in this paper will help natural resource managers,and remote sensing scientists,productively engage in developing monitoring protocols.
We rely heavily on the concept of extracting change information from spectral space,but we emphasize that spectral space represents the more generic multivariate spaces derived from new sensor technologies and from other spatial data that describe landscapes. Other technologies will have different speci?c bene?ts,but ultimately the information source being tapped for information is variability in a data space.Increasingly,these data spaces involve a larger suite of environmental variables used to describe landscapes(Goetz et al., 2009-this issue;Ohmann&Gregory,2002).Because many of these ancillary data(elevation,average climate,etc.)are historically static, image data are often the most dynamic variables in the multivariate space used to track changes over time.Nevertheless,remote sensing change analyses are increasingly being incorporated into a data assimilation framework,i.e.the merger of available weather/climate, ocean,stream/lake,and ecosystems data with imagery and models to facilitate coordinated and operational analyses of environmental change(see https://www.wendangku.net/doc/f55204582.html,/).
Although the change detection framework described in this paper is likely relevant to the remote sensing portion of many monitoring projects,the resultant maps of change may be just the?rst step in a larger modeling or pattern analysis effort(Crabtree et al.,2009-this issue;Townsend et al.,2009-this issue;Goetz et al.,2009-this issue).A detailed consideration of pattern analysis or ecosystem modeling is beyond the scope of this paper,but the requirements for those(or similar)efforts may need to be part of the evaluation of overall utility of different remote sensing projects.Measurement information content (discrete vs.continuous variables)and the spatial and temporal grain will likely need to align with subsequent analyses.
In summary,remote sensing data are an increasingly important component of natural resource monitoring programs(Coppin et al., 2004;Gross et al.,2006;Wiens et al.,2009-this issue).The utility of remotely sensed data for monitoring is maximized by understanding the constraints and capabilities of the imagery and change detection techniques,relative to the monitoring objectives.This understanding is best achieved through a collaborative process that leverages the expertise of both natural resource specialists and remote sensing specialists throughout the entire planning and implementation process.
A careful consideration of the spatial,temporal,and spectral properties of focal resources and their alignment with imagery data will help determine the suitability of using remotely sensed imagery to effectively achieve monitoring objectives.
References
Allard,A.(2003).Detection of vegetation degradation on Swedish mountainous heaths at an early stage by image interpretation.Ambio,32,510?519.
Asner,G.P.,&Heidebrecht,K.B.(2002).Spectral unmixing of vegetation,soil and dry carbon cover in arid regions:Comparing multispectral and hyperspectral observa-tions.International Journal of Remote Sensing,23,3939?3958.
Asner,G.P.,&Vitousek,P.M.(2005).Remote analysis of biological invasion and biogeochemical change.Proceedings of the National Academy of Sciences of the United States of America,102,4383?4386.
Asner,G.P.,Elmore,A.J.,Hughes,R.F.,Warner,A.S.,&Vitousek,P.M.(2005).Ecosystem structure along bioclimatic gradients in Hawaii from imaging spectroscopy.Remote Sensing of Environment,96,497?508.
Biging,G.S.,Colby,D.R.,&Congalton,R.G.(1998).Sampling systems for change detection accuracy assessment.In R.S.Lunetta,&C.D.Elvidge(Eds.),Remote sensing change detection:Environmental monitoring methods and applications(pp.89?102).
Chelsea,MI:Ann Arbor Press.
Boutet,J.C.,&Weishampel,J.F.(2003).Spatial pattern analysis of pre-and post-hurricane forest canopy structure in North Carolina,https://www.wendangku.net/doc/f55204582.html,ndscape Ecology,18, 553?559.
Campbell,J.B.(1996).Introduction to remote sensing.New York,NY:The Guilford Press. Chen,J.,Gong,P.,He,C.,Pu,R.,&Shi,P.(2003).Land-use/land-cover change detection using improved change-vector analysis.Photogrammetric Engineering&Remote Sensing,69,369?379.
Cihlar,J.(2000).Land cover mapping of large areas from satellites:Status and research priorities.International Journal of Remote Sensing,21,1093?1114.
Cochran,W.G.(1977).Sampling techniques.New York:John Wiley&Sons.
1393
R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
Cohen,W.B.,&Fiorella,M.(1998).Comparison of methods for detecting conifer forest change with Thematic Mapper imagery.In R.S.Lunetta,&C.D.Elvidge(Eds.), Remote sensing change detection:Environmental monitoring methods and applica-tions(pp.89?102).Chelsea,MI:Ann Arbor Press.
Cohen,W.B.,Fiorella,M.,Gray,J.,Helmer,E.H.,&Anderson,K.(1998).An ef?cient and accurate method for mapping forest clearcuts in the Paci?c Northwest using Landsat imagery.Photogrammetric Engineering&Remote Sensing,64,293?300. Cohen,W.B.,Maiersperger,T.K.,Gower,S.T.,&Turner,D.P.(2003).An improved strategy for regression of biophysical variables and Landsat ETM+data.Remote Sensing of Environment,84,561?571.
Cohen,W.B.,&Goward,S.N.(2004).Landsat's role in ecological applications of remote sensing.Bioscience,54,535?545.
Cohen,W.B.,Maiersperger,T.K.,Spies,T.A.,&Oetter,D.R.(2001).Modeling forest cover attributes as continuous variables in a regional context with Thematic Mapper data.International Journal of Remote Sensing,22,2279?2310.
Cohen,W.B.,Maiersperger,T.K.,Turner,D.P.,Ritts,W.D.,P?ugmacher,D.,Kennedy,R.E., et al.(2006).MODIS land cover and LAI collection4product quality across nine sites in the western hemisphere.IEEE Transactions on Geoscience and Remote Sensing,44, 1843?1857.
Congalton,R.G.,&Biging,G.S.(1992).A pilot study evaluating ground reference data colletion efforts for use in forest inventory.Photogrammetric Engineering&Remote Sensing,58,1669?1671.
Congalton,R.G.(1998).Using spatial autocorrelation analysis to explore the errors in maps generated from remotely sensed data.Photogrammetric Engineering and Remote Sensing,54,593?600.
Congalton,R.G.,&Green,K.(1999).Assessing the accuracy of remotely sensed data: Principles and practices.Boca Raton:Lewis Publishers.
Coops,N.,Wulder,M.A.,&White,J.C.(2007).Identifying and describing forest disturbance and spatial pattern:Data selection issues and methodological implications.In M. A.Wulder,&S. E.Franklin(Eds.),Understanding forest disturbance and spatial pattern:Remote sensing and GIS approaches Boca Raton,FL: CRC Press,Taylor and Francis Group.
Coppin,P.,Jonckheere,I.,Nackaerts,K.,Muys,B.,&Lambin,E.(2004).Digital change detection methods in ecosystem monitoring:A review.International Journal of Remote Sensing,25,1565?1596.
Crabtree,R.L.,Potter,C.S.,Mullen,R.S.,Sheldon,J.W.,Huang,S.,&Harmsen,J.A.(2009).
A modeling and spatiotemporal analysis framework for monitoring environmental
change using NPP as an ecosystem indicator.Remote Sensing of Environment,113, 1486?1496(this issue).
Czaplewski,R.L.,&Patterson,P.L.(2003).Classi?cation accuracy for strati?cation with remotely sensed data.Forest Science,49,402?408.
Defries,R.S.,Hansen,M.C.,&Townshend,J.R.G.(2000).Global continuous?elds of vegetation characteristics:a linear mixture model applied to multi-year8km AVHRR data.International Journal of Remote Sensing,21,1389?1414.
Dierking,W.,&Skriver,H.(2002).Change detection for thematic mapping by means of airborne multitemporal polarimetric SAR imagery.IEEE Transactions on Geoscience and Remote Sensing,40,618?636.
Dougherty,M.,Dymond,R.L.,Goetz,S.,Jantz,C.A.,&Goulet,N.(2004).Evaluation of impervious surface estimates in a rapidly urbanizing watershed.Photogrammetric Engineering and Remote Sensing,70,1275?1284.
Ellis,E.C.,Wang,H.,Xiao,H.S.,Peng,K.,LIu,X.P.,Li,S.C.,Ouyang,H.,Cheng,X.,&Yang, L.Z.(2006).Measuring long-term ecological changes in densely populated landscapes using current and historical high resolution imagery.Remote Sensing of Environment,100,457?473.
Fancy,S.G.,Gross,J.E.,&Carter,S.L.(2008).Monitoring the condition of natural resources in U.S.National Parks.Environmental Monitoring and Assessment,151, 161?174.doi:10.1007/s10661-008-0257-y
Fassnacht,K.S.,Cohen,W.B.,&Spies,T.A.(2006).Key issues in making and using satellite-based maps in ecology:A primer.Forest Ecology and Management,222, 167?181.
Fensham,R.J.,Bray,S.G.,&Fairfax,R.J.(2007).Evaluation of aerial photography for predicting trends in structural attributes of Australian woodland including comparison with ground-based monitoring data.Journal of Environmental Manage-ment,83,392?401.
Fensham,R.J.,&Fairfax,R.J.(2007).Effect of photoscale,interpreter bias and land type on woody crown-cover estimates from aerial photography.Australian Journal of Botany,55,457?463.
Foody,G.M.(1996).Approaches for the production and evaluation of fuzzy land cover classi?cations from remotely-sensed data.International Journal of Remote Sensing, 17,1317?1340.
Foody,G.M.,&Boyd,D.S.(1999).Detection of partial land cover change associated with the migration of inner-class transitional zones.International Journal of Remote Sensing,20,2723?2740.
Fraser,R.H.,Olthof,I.,&Pouliot,D.(2009).Monitoring land cover change and ecological integrity in Canada's national parks.Remote Sensing of Environment,113,1397?1409 (this issue).
Friedl,M.A.,Woodcock,C.,Gobal,S.,Muchoney,D.,Strahler,A.J.,&Barker-Schaaf,C.
(2000).A note on procedures used for accuracy assessment in land cover maps derived from AVHRR data.International Journal of Remote Sensing,21,1073?1077. Fung,T.,&Siu,W.(2000).Environmental quality and its changes,an analysis using NDVI.International Journal of Remote Sensing,21,1011?1024.
Gamba,P.,Dell'Acqua,F.,&Lisini,G.(2006).Change detection of multitemporal SAR data in urban areas combining feature-based and pixel-based techniques.IEEE Transactions on Geoscience and Remote Sensing,44,2820?2827.
Gens,R.,&VanGenderen,J.L.(1996).SAR interferometry—Issues,techniques, applications.International Journal of Remote Sensing,17,1803?1835.Garcia-Haro,F.J.,Gilabert,M.A.,&Melia,J.(2001).Monitoring?re-affected areas using Thematic Mapper data.International Journal of Remote Sensing,22,533?549. Goetz,S.J.,Jantz,P.,&Jantz,C.A.(2009).Connectivity of core habitat in the northeastern United States:Parks and protected areas in a landscape context.Remote Sensing of Environment,113,1421?1429(this issue).
Goetz,S.J.,Wright,R.K.,Smith,A.J.,Zinecker,E.,&Schaub,E.(2003).IKONOS imagery for resource management:Tree cover,impervious surfaces,and riparian buffer analyses in the mid-Atlantic region.Remote Sensing of Environment,88,195?208. Gong,P.,&Xu,B.(2003).Remote sensing of forests over time.In M.A.Wulder,&S.E.
Franklin(Eds.),Remote sensing of forest environments:Concepts and case studies (pp.301?334).Boston:Kluwer Academic Publishers.
Gopal,S.,&Woodcock,C.E.(1994).Theory and methods for accuracy assessment of thematic maps using fuzzy sets.Photogrammetric Engineering&Remote Sensing,60,181?188. Gross,J.E.,Nemani,R.R.,Turner,W.,&Melton,F.(2006).Remote sensing for the national parks.Park Science,24,30?36.
Haertel,V.,Shimabukuro,Y.E.,&Almeida-Filho,R.(2004).Fraction images in multitemporal change detection.International Journal of Remote Sensing,25,5473?5489.
Hall,D.K.,Riggs,G.A.,Salomonson,V.V.,DiGirolamo,N.E.,&Bayr,K.J.(2002).MODIS snow-cover products.Remote Sensing of Environment,83,181?194.
Harris,A.T.,Asner,G.P.,&Miller,M.E.(2003).Changes in vegetation structure after long-term grazing in pinyon-juniper ecosystems:Integrating imaging spectroscopy and?eld studies.Ecosystems,6,368?383.
Henderson,F.M.,&Xia,Z.G.(1997).SAR applications in human settlement detection, population estimation and urban land use pattern analysis:A status report.IEEE Transactions on Geoscience and Remote Sensing,35,79?85.
Hess,L.L.,Melack,J.M.,Novo,E.,Barbosa,C.C.F.,&Gastil,M.(2003).Dual-season mapping of wetland inundation and vegetation for the central Amazon basin.Re-mote Sensing of Environment,87,404?428.
Hockings,M.,Stolton,S.,Leverington,F.,Dudley,N.,&Courrau,J.(2006).Evaluating effectiveness:A framework for assessing management effectiveness of protected areas, 2nd edition Gland,Switzerland:IUCN.
Hostert,P.,Roder,A.,&Hill,J.(2003).Coupling spectral unmixing and trend analysis for monitoring of long-term vegetation dynamics in Mediterranean rangelands.Re-mote Sensing of Environment,87,183?197.
Huang,C.,Goward,S.N.,Zhu,Z.,&Masek,J.G.(2009).Dynamics of national forests assessed using the Landsat record:Case studies in eastern U.S..Remote Sensing of Environment,113,1430?1442(this issue).
Huang,C.,Kim,S.,Altstatt,A.,Townshend,J.R.G.,Davis,P.,Song,K.,et al.(2007).Rapid loss of Paraguay's Atlantic forest and the status of protected areas—A Landsat assessment.Remote Sensing of Environment,106,460?466.
Hudak,A.T.,&Wessman,C.A.(1998).Textural analysis of historical aerial photography to characterize woody plant encroachment in South African savanna.Remote Sensing of Environment,66,317?330.
Jantz,P.,Goetz,S.,&Jantz,C.(2005).Urbanization and the loss of resource lands in the Chesapeake Bay watershed.Environmental Management,36,808?825. Johansen,K.,Phinn,S.,Dixon,I.,Douglas,M.,&Lowry,J.(2007).Comparison of image and rapid?eld assessments of riparian zone condition in Australian tropical savannas.Forest Ecology and Management,240,42?60.
Kaab,A.,Huggel,C.,Fischer,L.,Guex,S.,Paul,F.,Roer,I.,et al.(2005).Remote sensing of glacier-and permafrost-related hazards in high mountains:An overview.Natural Hazards and Earth System Sciences,5,527?554.
Kalkhan,M.A.,Reich,R.M.,&Stohlgren,T.J.(1998).Assessing the accuracy of Landsat Thematic Mapper classi?cation using double sampling.International Journal of Remote Sensing,19,2049?2060.
Kennedy,R.E.,Cohen,W.B.,Kirschbaum,A.A.,&Haunreiter,E.(2007a).Protocol for Landsat-based monitoring of landscape dynamics at North Coast and Cascades Network Parks.U.S.geological survey techniques and methods:USGS Biological Resources Division https://www.wendangku.net/doc/f55204582.html,/tm/2007/tm2g1/
Kennedy,R.E.,Cohen,W.B.,&Schroeder,T.A.(2007b).Trajectory-based change detection for automated characterization of forest disturbance dynamics.Remote Sensing of Environment,110,370?386.
Kerr,J.T.,&Ostrovsky,M.(2003).From space to species:Ecological applications for remote sensing.Trends in Ecology and Evolution,18,299?305.
Kramer,H.J.(1996).Observation of the earth and its environment:Survey of missions and sensors.Berlin:Springer.
Lambin,E.F.,&Ehrlick,D.(1997).Land-cover changes in sub-Saharan Africa(1982–1991):Application of a change index based on remotely sensed surface temperature and vegetation indices at a continental scale.Remote Sensing of Environment,61,181?200.
Lambin,E.F.,&Strahler,A.H.(1994).Change-vector analysis in multitemporal space:A tool to detect and categorize land-cover change processes using high temporal-resolution satellite data.Remote Sensing of Environment,48,231?244.
Lang,M.W.,&Kasischke,E.S.(2008).Using C-band synthetic aperture radar data to monitor forested wetland hydrology in Maryland's coastal plain,USA.IEEE Transactions on Geoscience and Remote Sensing,46,535?546.
Lawrence,R.,&Ripple,W.J.(1999).Calculating change curves for multitemporal satellite imagery:Mount St.Helens1980–1995.Remote Sensing of Environment,67,309?319. Leckie,D.G.,Jay,C.,Gougeon,F.A.,Sturrock,R.N.,&Paradine,D.(2004).Detection and assessment of trees with Phellinus weirii(laminated root rot)using high resolution multi-spectral imagery.International Journal of Remote Sensing,25,793?818. Lefsky,M.A.,&Cohen,W.B.(2003).Selection of remotely sensed data.In M.A.Wulder, &S.E.Franklin(Eds.),Remote sensing of forest environments(pp.13?46).Boston: Kluwer Academic Publishers.
Li,Y.,Liao,Q.F.,Li,X.,Liao,S.D.,Chi,G.B.,&Peng,S.L.(2003).Towards an operational system for regional-scale rice yield estimation using a time-series of Radarsat ScanSAR images.International Journal of Remote Sensing,24,4207?4220.
1394R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
Lillesand,T.M.,&Kiefer,R.W.(2000).Remote sensing and image interpretion.New York: John Wiley&Sons,Inc.
Lohr,S.L.(1999).Sampling:Design and analysis.New York:Duxbury Press450pp. Lu,D.,Mausel,P.,Brondizio,E.,&Moran,E.(2004).Change detection techniques.In-ternational Journal of Remote Sensing,25,2365?2407.
Lu,H.,Raupach,M.R.,McVicar,T.R.,&Barrett,D.J.(2003).Decomposition of vegetation cover into woody and herbaceous components using AVHRR NDVI time series.
Remote Sensing of Environment,86,1?18.
Lunetta,R.S.(1998).Applications,project formulation,and analytical approach.In R.S.
Lunetta,&C.D.Elvidge(Eds.),Remote sensing change detection(pp.1?19).Chelsea, Michigan:Ann Arbor Press.
Lunetta,R.S.,&Elvidge, C. D.(Eds.).(1998).Remote sensing change detection: Environmental monitoring methods and applications Chelsea,MI:Ann Arbor Press. Malila,W.A.(1980).Change vector analysis:An approach for detecting forest changes with Landsat.In P.G.Burroff,&D.B.Morrison(Eds.),West LafayettePurdue https://www.wendangku.net/doc/f55204582.html,b.
App.Remote Sens.(pp.326?335).
Mas,J.F.(1999).Monitoring land-cover changes:A comparison of change detection techniques.International Journal of Remote Sensing,20,139?152.
Masek,J.G.,Vermote,E.F.,Saleous,N.E.,Wolfe,R.,Hall,F.G.,Huemmrich,K.F.,et al.
(2006).A Landsat surface re?ectance dataset for North America,1990–2000.IEEE Geoscience and Remote Sensing Letters,3,68?72.
Massonnet,D.,&Feigl,K.L.(1998).Radar interferometry and its application to changes in the earth's surface.Reviews of Geophysics,36,441?500.
Michener,W.K.,Brunt,J.W.,Helly,J.J.,Kirchner,T. B.,&Stafford,S.G.(1997).
Nongeospatial metadata for the ecological sciences.Ecological Applications,7, 330?342.
Mouat,D.A.,Mahin,G.G.,&Lancaster,J.(1993).Remote sensing techniques in the analysis of change detection.Geocarto International,2,39?49.
Muchoney, D.M.,&Haack, B.N.(1994).Change detection for monitoring forest defoliation.Photogrammetric Engineering&Remote Sensing,60,1243?1251. Myneni,R.B.,Tucker,C.J.,Asrar,G.,&Keeling,C.D.(1998).Interannual variations in satellite-sensed vegetation index data from1981to1991.Journal of Geophysical Research-Atmospheres,103,6145?6160.
Nagler,P.L.,Glenn,E.P.,&Hinojosa-Huerta,O.(2009).Synthesis of ground and remote sensing methods for monitoring ecosystem functions in the Colorado River delta, Mexico.Remote Sensing of Environment,113,1473?1485(this issue).
Nemani,R.R.,Hashimoto,H.,Votava,P.,Melton,F.,White,M.,&Wang,W.(2009).Monitoring and forecasting ecosystem dynamics using the Terrestrial Observation and Prediction System(TOPS).Remote Sensing of Environment,113,1497?1509(this issue). Ohmann,J.L.,&Gregory,M.J.(2002).Predictive mapping of forest composition and structure with direct gradient analysis and nearest-neighbor imputation in coastal Oregon,U.S.A.Canadian Journal of Forest Research,32,724?741.
Olthof,I.,King,D.J.,&Lautenschlager,R.A.(2004).Mapping deciduous forest ice storm damage using Landsat and environmental data.Remote Sensing of Environment,89, 484?496.
Paci?ci,F.,Del Frate,F.,Solimini,C.,&Emery,W.J.(2007).An innovative neural-net method to detect temporal changes in high-resolution optical satellite imagery.
Transactions on Geoscience and Remote Sensing,45,2940?2952.
Paine,D.P.(1981).Aerial photography and image interpretation for resource management.
New York:John Wiley&Sons.
Parmenter,A.W.,Hansen,A.,Kennedy,R.E.,Cohen,W.B.,Langner,U.,Lawrence,R.,et al.
(2003).Greater Yellowstone land cover change.Ecological Applications,13,687?703. Peddle,D.R.,Johnson,R.L.,Cihlar,J.,Leblanc,S.G.,Chen,J.M.,&Hall,F.G.(2007).
Physically based inversion modeling for unsupervised cluster labeling,independent forest classi?cation,and LAI estimation using MFM-5-Scale.Canadian Journal of Remote Sensing,33,214?225.
Peddle,D.R.,Teillet,P.M.,&Wulder,M.A.(2003).Radiometric image processing.
In M.A.Wulder,&S.E.Franklin(Eds.),Remote sensing of forest environments Boston:Kluwer Academic Publishers.
Philipson,P.,&Lindell,T.(2003).Can coral reefs be monitored from space?Ambio,32, 586?593.
Phinn,S.R.,Stow,D.A.,Franklin,J.,Mertes,L.A.K.,&Michaelsen,J.(2003).Remotely sensed data for ecosystem analyses:Combining hierarchy theory and scene models.
Environmental Management,31,429?441.
Potter,C.,Tan,P.N.,Kumar,V.,Kucharik,C.,Klooster,S.,Genovese,V.,et al.(2005).
Recent history of large-scale ecosystem disturbances in North America derived from the AVHRR satellite record.Ecosystems,8,808?824.
Quegan,S.,Le Toan,T.,Yu,J.J.,Ribbes,F.,&Floury,N.(2000).Multitemporal ERS SAR analysis applied to forest mapping.IEEE Transactions on Geoscience and Remote Sensing,38,741?753.
Reed,B.,Budde,M.,Spencer,P.,&Miller,A.(2009).Integration of MODIS-derived metrics to assess interannual variability in snowpack,lake ice,and NDVI in southwest Alaska.Remote Sensing of Environment,113,1443?1452(this issue).
Richards,J.A.(1993).Remote sensing digital image analysis:An introduction.Berlin: Springer-Verlag.
Ridd,M.K.,&Liu,J.J.(1998).A comparison of four algorithms for change detection in an urban environment.Remote Sensing of Environment,63,95?100.
Rignot,E.J.M.,&Vanzyl,J.J.(1993).Change detection techniques for Ers-1Sar data.
IEEE Transactions on Geoscience and Remote Sensing,31,896?906.
Roberts,D.A.,Batista,G.T.,Pereira,J.L.G.,Waller,E.K.,&Nelson,B.W.(1998).Change identi?cation using multitemporal spectral mixture analysis:Applications in eastern Amazonia.In R.S.Lunetta&C.D.Elvidge(Eds.),Remote sensing change detection:Environmental monitoring methods and applications(pp.137–161).
Chelsea,MI:Ann Arbor Press.
Ronnback, B.I.,Nordberg,M.L.,Olsson, A.,&Ostman, A.(2003).Evaluation of environmental monitoring strategies.Ambio,32,495?501.Rosso,P.H.,Ustin,S.L.,&Hastings,A.(2006).Use of lidar to study changes associated with Spartina invasion in San Francisco Bay marshes.Remote Sensing of Environment,100,295?306.
Sabins,F.(1987).Remote sensing:Principles and interpretation.New York:W.H.Freeman and Company.
Sabins,F.(1997).Remote sensing:Principles and interpretation.New York:W.H.Freeman and Company.
Sánchez-Azofeifa,Kachmar,M.,Kalácska,M.,&Hamilton,S.(2003).Experiences in?eld data collection.In M.A.Wulder&S.E.Franklin(Eds.),Remote sening of forest environments.Boston:Kluwer Academic Publishers.
Schaepman-Strub,G.,Schaepman,M.E.,Painter,T.H.,Dangel,S.,&Martonchik,J.V.
(2006,07-15).Re?ectance quantities in optical remote sensing—De?nitions and case studies.Remote Sensing of Environment,103(1),27?42.
Schott,J.R.(1997).Remote sensing:The image chain approach.New York:Oxford University Press.
Schowengerdt,R.A.(1997).Remote sensing:Models and methods for image processing.
San Diego:Academic Press.
Schuck,A.,Paivinen,R.,Hame,T.,Van Brusselen,J.,Kennedy,P.,&Folving,S.(2003).
Compilation of a European forest map from Portugal to the Ural Mountains based on earth observation data and forest statistics.Forest Policy and Economics,5, 187?202.
Seto,K.C.,&Liu,W.G.(2003).Comparing ARTMAP neural network with the maximum-likelihood classi?er for detecting urban change.Photogrammetric Engineering and Remote Sensing,69,981?990.
Skakun,R.S.,Wulder,M.A.,&Franklin,S.E.(2003).Sensitivity of the thematic mapper enhanced wetness difference index to detect mountain pine beetle red-attack damage.Remote Sensing of Environment,86,433?443.
Smith,L. C.(1997).Satellite remote sensing of river inundation area,stage,and discharge:A review.Hydrological Processes,11,1427?1439.
Smith,J.J.,Wickham,J.D.,Stehman,S.V.,&Yang,L.(2002).Impact of patch size and land-cover heterogeneity on thematic image classi?cation accuracy.Photogram-metric Engineering&Remote Sensing,68,65?70.
Stehman,S.V.(1999).Basic probability sampling designs for thematic map accuracy assessment.International Journal of Remote Sensing,20,2423?2441.
Stehman,S.V.(2000).Practical implications of design-based sampling inference for thematic map accuracy assessment.Remote Sensing of Environment,72,35?45. Stehman,S.V.(2001).Statistical rigor and practical utility in thematic map accuracy assessment.Photogrammetric Engineering and Remote Sensing,67,727?734. Stow,D.A.,Hope,A.,McGuire,D.,Verbyla,D.,Gamon,J.,Huemmrich,F.,et al.(2004).
Remote sensing of vegetation and land-cover change in Arctic tundra ecosystems.
Remote Sensing of Environment,89,281?308.
Strahler,A.H.,Boschetti,L.,Foody,G.M.,Friedl,M.A.,Hansen,M.C.,Herold,M.,et al.
(2006).Global land cover validation:Recommendations for evaluation and accuracy assessment of global land cover maps.GOFC-GOLD Report No.25.Luxembourg: European Communities.
Thompson,S.K.(2002).Sampling.New York:John Wiley&Sons,Inc.
Townsend,P.A.(2001).Mapping seasonal?ooding in forested wetlands using multi-temporal radarsat SAR.Photogrammetric Engineering and Remote Sensing,67,857?864. Townsend,P.A.,Eshleman,K.N.,&Welcker,C.(2004).Remote sensing of gypsy moth defoliation to assess variations in stream nitrogen concentrations.Ecological Applications,14,504?516.
Townsend,P.A.,Lookingbill,T.R.,Kingdon,C.C.,&Gardner,R.H.(2009).Spatial pattern analysis for monitoring protected areas.Remote Sensing of Environment,113, 1410?1420(this issue).
Tucker,C.J.(1979).Red and photographic infrared linear combinations for monitoring vegetation.Remote Sensing of Environment,8,127?150.
Turner,W.,Spector,S.,Gardiner,N.,Fladeland,M.,Sterling,E.,&Steininger,M.(2003).
Remote sensing for biodiversity science and conservation.Trends in Ecology& Evolution,18,306?314.
Verbyla,D.(1995).Satellite remote sensing of natural resources.Boca Raton,FL:CRC Press,Inc.
Vikhamar,D.,&Solberg,R.(2003).Snow-cover mapping in forests by constrained linear spectral unmixing of MODIS data.Remote Sensing of Environment,88,309?323. Viòa,A.,Bearer,S.,Chen,X.,He,G.,Linderman,M.,An,L.,Zhang,H.,Ouyang,Z.,&Liu,J.
(2007).Temporal changes in giant panda habitat connectivity across boundaries of Wolong nature reserve,China.Ecological Applications,17,1019?1030.
Wang, F.(1990).Fuzzy supervised classi?cation of remote sensing images.IEEE Transactions on Geoscience and Remote Sensing,28,194?201.
Wang,H.,&Ellis,E.C.(2005a).Spatial accuracy of orthorecti?ed IKONOS imagery and historical aerial photographs across?ve sites in China.International Journal of Remote Sensing,26,1893?1911.
Wang,H.Q.,&Ellis,E.C.(2005b).Image misregistration error in change measurements.
Photogrammetric Engineering&Remote Sensing,71,1037?1044.
Wang,Y.Q.,Mitchell,B.R.,Nugranad-Marzilli,J.,Bonynge,G.,Zhou,Y.,&Shriver,G.
(2009).Remote sensing of land-cover change and landscape context of the national parks:A case study of the Northeast Temperate Network.Remote Sensing of Environment,113,1453?1461(this issue).
Wang,Y.Q.,Traber,M.,Milstead,B.,&Stevers,S.(2007).Terrestrial and submerged aquatic vegetation mapping in Fire Island National Seashore using high spatial resolution remote sensing data.Marine Geodesy,30,77?95.
Wdowinski,S.,Kim,S.W.,Amelung,F.,Dixon,T.H.,Miralles-Wilhelm,F.,&Sonenshein, R.(2008).Space-based detection of wetlands'surface water level changes from L-band SAR interferometry.Remote Sensing of Environment,112,681?696. Weisberg,P.J.,Lingua,E.,&Pillai,R.B.(2007).Spatial patterns of pinyon-juniper woodland expansion in central Nevada.Rangeland Ecology&Management,60, 115?124.
1395
R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
Wessels,K.J.,Defries,R.S.,Dempewolf,J.,Anderson,L.O.,Hansen,A.J.,Powell,S.L.,& Moran,E.F.(2004).Mapping regional land cover with MODIS data for biological conservation:Examples from the Greater Yellowstone Ecosystem,USA and Para State,Brazil.Remote Sensing of Environment,92,67?83.
White,J.D.,Ryan,K.C.,Key,C.C.,&Running,S.W.(1996).Remote sensing of forest?re severity and vegetation recovery.International Journal of Wildland Fire,6,125?136. White,S.A.,&Wang,Y.(2003).Utilizing DEMs derived from LIDAR data to analyze morphologic change in the North Carolina coastline.Remote Sensing of Environment, 85,39?47.
White,M.A.,Hoffman,F.M.,Hargrove,W.W.,&Nemani,R.R.(2005).A global framework for monitoring phenological responses to climate change.Geophysical Research Letters,32,L04705.doi:10.1029/02004GL021961
Wiens,J.A.,Sutter,R.D.,Anderson,M.,&Blanchard,J.(2009).Selecting and conserving lands for biodiversity:the role of remote sensing.Remote Sensing of Environment, 113,1370?1381(this issue).
Woodcock,C.E.,&Strahler,A.H.(1987).The factor of scale in remote sensing.Remote Sensing of Environment,21,311?332.
Woodward,A.,Acker,S.A.,&Hoffman,R.(2002).Use of remote sensing for long-term ecological monitoring in the North Coast and Cascades Network:Summary of a workshop.https://www.wendangku.net/doc/f55204582.html,/olympic/research/https://www.wendangku.net/doc/f55204582.html, Dept.
of the Interior US Geological Survey.
Woolard,J.W.,&Colby,J.D.(2002).Spatial characterization,resolution,and volumetric change of coastal dunes using airborne LIDAR:Cape Hatteras,North Carolina.
Geomorphology,48,269?287.
Wulder,M.A.,&Franklin,S.E.(Eds.).(2007).Understanding forest disturbance and spatial pattern:remote sensing and GIS approaches Boca Raton,FL:Taylor and Francis Group,LLC.Wulder,M.A.,Han,T.,White,J.C.,Sweda,T.,&Tsuzuki,H.(2007a).Integrating pro?ling LIDAR with Landsat data for regional boreal forest canopy attribute estimation and change characterization.Remote Sensing of Environment,110,123?137.
Wulder,M. A.,Skakun,R.S.,Dymond, C. C.,Kurz,W. A.,&White,J. C.(2005).
Characterization of the diminishing accuracy in detecting forest insect damage over time.Canadian Journal of Remote Sensing,31,421?431.
Wulder,M.A.,White,J.C.,Magnussen,S.,&McDonald,S.(2007b).Validation of a large area land cover product using purpose-acquired airborne video.Remote Sensing of Environment,106,480?491.
Yu,X.,Hyyppa,J.,Kaartinen,H.,Maltamo,M.,&Hyyppa,H.(2008).Obtaining plotwise mean height and volume growth in boreal forests using multi-temporal laser surveys and various change detection techniques.International Journal of Remote Sensing,29,1367?1386.
Yu,X.W.,Hyyppa,J.,Kukko,A.,Maltamo,M.,&Kaartinen,H.(2006).Change detection techniques for canopy height growth measurements using airborne laser scanner data.Photogrammetric Engineering and Remote Sensing,72,1339?1348.
Yuan,D.,Elvidge,C.D.,&Lunetta,R.S.(1998).Survey of multiespectral methods for land cover change analysis.In R.S.Lunetta,&C.D.Elvidge(Eds.),Remote sensing change detection(pp.21?40).Chelsea,Michigan:Ann Arbor Press.
Zhan,X.,Sohlberg,R.A.,Townshend,J.R.G.,DiMiceli,C.,Carroll,M.L.,Eastman,J.C., Hansen,M.C.,&DeFries,R.S.(2002).Detection of land cover changes using MODIS 250m data.Remote Sensing of Environment,83,336?350.
Zhang,C.S.,&Fraser,C.S.(2007).Automated registration of high-resolution satellite images.Photogrammetric Record,22,75?87.
1396R.E.Kennedy et al./Remote Sensing of Environment113(2009)1382–1396
初中定语从句专项讲解与练习
定语从句（初中） 在复合句中，修饰某一名词或代词的从句叫定语从句。被修饰的名词或代词叫先行词，定语从句一般放在先行词的后面。 二、定语从句的关系词 引导定语从句的关系词有关系代词和关系副词，常见的关系代词包括that，which，who（宾格whom，所有格whose）等，关系副词包括where，when，why等。关系代词和关系副词放在先行词及定语从句之间起连接作用，同时又作定语从句的重要成分。 三、定语从句的分类 根据定语从句与先行词的关系，定语从句可分为限制性定语从句及非限制性定语从句。限制性定语从句紧跟先行词，主句与从句不用逗号分开，从句不可省去，非限制性定语从句主句与从句之间有逗号分开，起补充说明作用，如省去，意思仍完整。 四、关系代词的用法 1. that 既可以用于指人，也可以用于指物。在从句中作主语或宾语。例如： Mary likes music that is quiet and gentle.（that作主语） The coat （that）I put on the desk is blue.（that作宾语） ； 用于指物，在句中作主语或宾语。例如： The building which stands near the train station is a supermarket.（作主语） The film （which）we saw last night was wonderful. （作宾语） ，whom用于指人，who 用作主语，whom用作宾语。在口语中，有时可用who代替whom。例如： The girl who often helps me with my English is from England.（作主语） Who is the teacher （whom）Li Ming is talking to（作宾语） 注意： 1）当关系代词在定语从句中充当宾语时，who、that、which可省略，但介词在关系代词前时，只能用“介词+which/whom”结构（此时关系代词不能用that代替）。例如：This is the house (which/that) we lived in last year.(可省) This is the house in which we lived last year.(不可省，关系代词不可用that) - Please tell me (whom/that) you borrowed the English novel from.(可省) Please tell me from whom you borrowed the English novel.(不可省，关系代词不可用that) 2）关系词只能用that的情况： a. 先行词被序数词或形容词最高级所修饰，或本身是序数词、基数词、形容词最高级时，只能用that，而不用which.例如： He was the first person that passed the exam. This is the most beautiful place that I have been to . b.被修饰的先行词为all，any，much，many，everything，anything，none，the one 等不定代词时，只能用that，而不用which.例如： Is there anything that you want to buy in the shop= I didn't understand the words all that he said. c.先行词被the only，the very，the same，the last，little，few ，no，just等词修饰时，只能用that，而不用which.例如： 】 This is the same bike that I lost.这就是我丢的那辆自行车。 d. 先行词里同时含有人或物时，只能用that，而不用which.例如： I can remember well the persons and some pictures that I saw in the room. e.以who或which引导的特殊疑问句，为避免重复，只能用that.例如： Who is the girl that is crying 正在哭泣的那个女孩是谁
定语从句讲解关系代词的用法
定语从句讲解关系代词 的用法 文件排版存档编号：[UYTR-OUPT28-KBNTL98-UYNN208]
U n i t 1《s c h o o l l i f e 》 Grammar （1） 定语从句(AttributiveClause) Ⅰ根据初中所学知识，请用红笔标出下列表格中的定语 定语从句的定 义及其 作用： 定语从句是又称形容 词性从 句,在句子中 起定语作用,修饰一个名词或代词,有时也可修饰一个句子.受定语从句修饰的词叫先行词.定语从句的作用和作定语的形容词、介词词组、分词词组相似，有时可以相互转换，例如：金发女孩可译作ablondegirl ，agirlwithblondehair 或agirlwhohasblondehair 。定语从句通常由关系代词 that/which/who/whom/which/as 或关系副词when/where/why 引导，这些词既指代主句中要说明的名词或代词,又充当从句中的某个句子成分。定语从句可分为：限制性定语从句和非限制性定语从句。 定语从句一般都紧跟在它所修饰名词后面，所以如果在名词或代词后面出现一个从 句，根据它与前面名词或代词的逻辑关系来判断是否是定语从句。 找出下列句中的定语从句；分析定语从句三要素 1. Youaretherightmanwhomwearelookingfor. 2. I’vespentallthemoneythatwasgivenbymyparents. 3. Iwillneverforgettheday whenIjoinedtheparty. 4. Thisisthefactorywherethemachinesaremade. 1 2关系词 3 relativepron.(assub.,obj.,pred.) (that 指人或物/which 指物/who(m)指人/whose) relativeadv.(asadverbial) Ⅲ定语从句的必备三要素
体验商务英语综合教程中文双语对照版
体验商务英语综合教程3 第二版 双语对照版 Unit1 Made in Europe 欧洲制造 Almost every fashion label outside the top super-luxury brands is either already manufacturing in Asia or 5 thinking of it. Coach, the US leather goods maker, is a classic example. Over the past five years, it has lifted all its gross margins by manufacturing solely in low-cost markets. In March 2002 it closed its factory in Lares, Puerto Rico, its last company-owned plant, and outsources all its products. 除了顶级奢侈品牌外几乎所有的时尚品牌都已经在亚洲生产，或者正在考虑这么做。美国的皮革商品制造商蔻驰（Coach）就是一个经典的例子。在过去的五年中，它通过仅在低成本市场生产来提升毛利率。在2002年的3月，它关闭了在波多黎各拉雷斯的最后一间公司所属工厂，将所有产品全部外包。Burberry has many Asian licensing arrangements.In 2000 it decided to renew Sanyo's Japanese licence for ten years. This means that almost half of Burberry's sales at retail value will continue to be produced under license in Asia. At the same time however, Japanese consumers prefer the group's European-made products. 巴宝莉（Burberry）在亚洲持有许多许可授权安排。2000年它决定给日本三洋公司的特许授权延长十年。这意味着按零售价计算巴宝莉几乎一半的销售额将是亚洲授权生产的。但是同时，日本的消费者却偏好于该集团在欧洲生产的产
体验商务英语综合教程2 教案
外语系教案 第次课学时：授课时间：第周
Context: Unit 1 Title: Introductions The tone of a business relationship can be set by an initial introduction. It is important to make a good impression right from the first handshake. When meeting businesspeople for the first time, is it better to be formal or informal? If in doubt, advise students to adopt a more formal approach. Here are some points to remember when making business introductions in English-speaking Western countries: a.Introduce businesspeople in order of professional rank –the person of highest authority is introduced to others in the group in descending order, depending on their professional position. b.When possible, stand up when introductions are being made. c.If clients are present, they should be introduced first. d.The same and title of the person being introduced is followed by the name and title of the other person. PROCEDURES Lesson 1 Starting up Ss listen to four businesspeople and match the speakers to their business cards. Vocabulary 1: Job titles Ss list word as job titles or departments. Then Ss talk about their jobs or studies. Vocabulary 2: Nationalities Ss match countries and nationalities. Reading: Describing people This reading section can be completed in two parts. Ss can start preparatory work on the article about Phil Knight, the founder and CEO of Nike, and complete Exercise A. Lesson 2
初中定语从句复习课教案
初中英语语法定语从句复习课教案 教学目标：1.学生能掌握关系代词和关系副词的正确使用。 2.学生能正确理解整个句子的意思 教学重难点：定语从句中引导词（who, whom, whose，that , which，when，where, why）的正确使用教学过程： 例子导入： The girl is my sister. The girl is standing under the tree. The girl （who is standing under the tree）is my sister. 先行词定语从句 一.定义：在复合句中，用来修饰某一名词或代词的从句叫做定语从句。被定语从句修饰的名词、代词叫先 行词。定语从句放在先行词之后。 I like the music that I can dance to. 先行词（物）↘引导词（that指代the music） She is a girl ( who has long hair.) 先行词（人）↘引导词（who指代a girl） 二.引导词：关系代词：who, whom, whose，that , which（表人、表物，作主语，宾语） 关系副词：when，where, why, （表时间、地点、原因，作状语） （1）先行词表人时可用who，that或whom 分点练习：① He is a boy（______ is confident.） (The boy is confident.) 主语 ②He is the teacher for _____ you are waiting. (介词提前) ③ He is the teacher ( ______ you are waiting for.) (You are waiting for the teacher) 宾语 归纳总结：当先行词是人：①引导词在句中作主语，引导词用who，that
体验商务英语综合教程--Unit-4-答案
Unit 4 Advertising Part I Business Vocabulary Directions: There are 20 incomplete sentences in this part. For each sentence there are four choices marked A, B, C and D. Choose the ONE that best completes the sentence. Then mark the corresponding letter on the Answer Sheet with a single line through the center. This part totals 20 points, one point for each sentence. C1 Outdoor advertising is one of the fastest growing _______________ in the market. A markets B sections C segments D sectors D2 The world of outdoor advertising billboards, transport and ‘street furniture’is ______ about $18 billion a year, just 6% of all the worl d’s spending on advertising. A worthwhile B worthy C valued D worth C3 The soaring costs of TV are ______________ clients to consider alternatives. A making B driving C prompting D letting A4 BMW ran a ‘teasers’ campaign in Britain on bus shelters. A exclusively B largely C greatly D inclusively C5 Placing an ad on a bus shelter for two weeks ________________ at about ￡90. A works on B works away C works out D calculates D6 We are facing a ________________ with our market share. What are we going to do about it? A promotion B sale C order D crisis A7 Focus, a large advertising agency based in Paris, has a reputation for creating imaginative and ____________ campaigns. A effective B efficient C effect D efficacious C8 Focus now needs to ________________ potential clients that it still has plenty of creative ideas to offer. A ensure B assure C convince D persuade B9 Focus has been asked to _________________ ideas for advertising campaigns to managements of the companies concerned. A offer B present C supply D furnish
初中定语从句讲解及练习
初中定语从句讲解及练 习 -CAL-FENGHAI-(2020YEAR-YICAI)_JINGBIAN
初中定语从句讲解及练习 定义：在复合句中，修饰某一名词或代词的从句叫做定语从句。 如：1) The man who lives next to us is a policeman. 2) You must do everything that I do. 上面两句中的man和everything是定语从句所修饰的词，叫先行词，定语从句放在先行词的后面。 引导定语从句的词有关系代词that, which, who(宾格who, 所有格whose)和关系副词where, when、why 关系词常有三个作用：1、引导定语从句 2、代替先行词 3、在定语从句中担当一个成分 二、关系代词引导的定语从句 1.who指人，在从句中做主语 (1)The boys who are playing football are from Class One. (2)Yesterday I helped an old man who lost his way. 2. whom指人，在定语从句中充当宾语，常可省略。 (1) Mr. Liu is the person (whom) you talked about. 注意：关系代词whom在口语和非正式语体中常用who代替，可省略。 (3) The man who/whom you met just now is my friend. 3. which指物，在定语从句中做主语或者宾语，做宾语时可省略 (1) Football is a game which is liked by most boys. ( which 在句子中做主语) (2) This is the pen (which) he bought yesterday. ( which 在句子中做宾语) 4. that指人时，相当于who或者whom；指物时，相当于which。 在宾语从句中做主语或者宾语，做宾语时可省略。 (5) The people that/who come to visit the city are all here. (在句子中做主语) (6) Where is the man that/whom I saw this morning (在句子中做宾语) 5. whose通常指人，也可指物，在定语从句中做定语 (1) he has a friend whose father is a doctor. (2) i once lived in a house whose roof has fallen in. whose指物时，常用以下结构来代替 (3) the classroom whose door is broken will soon be repaired. (4) the classroom the door of which is broken will soon be repaired. (5) do you like the book whose cover is yellow? (6) do you like the book the color of which is yellow? 三、关系代词在定语从句中做介词宾语时，从句常由介词+关系代词引导 (1) the school (that/which) he once studied in is very famous. (2) the school in which he once studied is very famous. (3) tomorrow i will bring here a magazine (that/which) you asked for. (4) tomorrow i will bring here a magazine for which you asked. (5) we'll go to hear the famous singer (whom/that/who) we have often talked about. (6) we'll go to hear the famous singer about whom we have often talked. 注意：1. 含有介词的动词短语一般不拆开使用，如：look for, look after, take care of等 (1) this is the watch which/that i am looking for. (t) (2) this is the watch for which i am looking. (f)
体验商务英语视听说答案
体验商务英语视听说答案【篇一：体验商务英语 4 综合教程】lass=txt> 一．课程基本信息课程编号：0142524 课程类别：选修 总学时：36 课程简介：商务英语课程作为翻译专业的一门专业先选课程，主要目标是培养学生掌握国际贸易的基础知识、基本技能，能独立从事一般的对外贸易业务工作。具备听说读写译的基本技能，语音语调正确、语法概念清楚，能用英语较熟练地从事外事接待、外贸业务洽谈的口、笔译工作。在培养学生英语语言能力的同时让学生了解和熟悉各种商务情景和商务活动，掌握相关的商务及商务文化知识，并使他们能够把所学的知识运用到各种商务活动中。二．教材简介 教材名称：体验商务英语综合教程4 教材编者：david cotton;david falvey; simon kent; 《体验商务英 语》改编组出版社：高等教育出版社 教材情况简介：本教材话题紧跟国际经济发展形势，循序渐进地训练学生用英语进行调研分析、归纳总结和使用正确语体作书面或口头表述的能力。既可以帮助在校生了解真实的商务环境和话题，学习地道的商务英语；也可以帮助从事各种经济活动的商务人员通过语言技能综合训，较快地提高语言能力。将国际商务活动引入课堂，体验真实的商务世界。角色扮演和案例学习将体验式学习引向深入，教学设计严谨，为体验式学习打好基础。教学资源丰富，为体验式教学提供有力支持。
三．课程教学内容 教学重点和难点 1. 重点：掌握各种商务活动情景对话中的语言要点及专业词汇 2. 难点：由于缺乏实战经验，学生难以理解不断涌现的商务方面的新知识和商务活动的实战环节。 教学内容、目标和学时分配 教学内容教学目标课时分配unit1communication 掌握 4 unit2international marketing 掌握4 unit3building relationships 掌握 4 unit 4 success 掌握 4 unit 5 job satisfaction 掌握 4 unit 6 risk 了解0 unit 7 e-commerce 掌握4 revision unit a unit 8team building 了解0 unit 9raising finance 了解0 unit 10 customer service 了解0 unit 11 crisis management 了解0 unit 12 management styles 掌握4 unit 13 takeovers and mergers 了解0 unit 14 the future of business 了解0 revision unit b 四．课程各教学环节的基本要求 课堂讲授：要求学生在课堂上就国际商务中的各种场景进行对话和听力练习，并通过看录像、vcd 及听录音和mp3 等多媒体手段提高学生的商务英语听说能力。作业：练习章节语言要点及专业词汇，布置角色扮演任务。mp3 或网络资源听力练习。 课程设计：本课程注重把语言技能的训练和专业知识有机结合起来。除了课堂讲解，有些练习属于开放式的，要求学生理论联系实际，认真独立地思考问题、深入探究问题、最终解决问题。在这一过程中学生的表达能力同时得以锻炼。4）测试： 测试方法：测试；闭卷；口试（其中平时成绩占30% ；测试成绩占70%） 五．教学参考书
体验商务英语综合教程1-词汇
体验商务英语综合教程1-词汇
Unit 1 a cc ountant 会计会计师 ad agency (advertising agency) 广告代理 advertising n.广告业广告 advertise v. advertisement n. advertising永远是不可数名词。 有三个意思： 1、指整个广告业； 2、指广告活动（比如说某公司正在筹备做一个广告）； 3、指所有广告的总称（例：All the advertising here in this chamber is brilliant. 这个屋子里的所有广告都很有创意。） 而advertisement则在第一个意思中是可数名词，第二个是不可数名词。 有两个意思： 1、一则广告，某一个具体的广告； 2、做一则广告或某一个具体广告的动作（例：A company plans to do an advertising which will do advertisement 10 times. 一个公司计划做一个广告，这个广告准备播出十次。） commercial 商业广告 agency 代理处行销处代理中介 agent n./v./adj. airway s航空公司 airline Aloha [?'l?u?, ɑ:'l?uhɑ:]欢迎再见 a nalyst 分析师 analyze v. Argentina 阿根廷 Beaverton 比佛顿 Brazil 巴西 Brazilian 巴西人巴西的巴西人的 Buenos Aires 布宜诺斯艾利斯 cashier 出纳 CEO（chief executive officer）首席执行官 officer一般指军官，或者有指挥权利的官员，比如警官，船长。 official一般指公务员，官方。 conference 会议 meeting consultant 顾问 adviser consulting n.咨询adj.顾问的资讯的 consult 查阅商量向请教当顾问 e-mail 电子邮件 executive [iɡ'zekjutiv] n.管理人员经理adj.经营管理的 finance 财政金融 financial analyst 金融分析师
体验商务英语视听说 profile summary
Unit 1: Steve Morgan, a good dynamic salesman, is on the way to a job interview on Bateman Retail Technologies. But he is very nervous because his IT skills are not good enough to get the job. Jim, the security camera, wants to help him. He thinks that Steve has not prepared well for the interview as Steve does even not know what sort of products do the company sell. When Steve is thinking about the interview, Jim tells him that the first impression is very important for interview and encourages him to be confident. In the waiting room, Jim encourages Steve to chat with the lady called Jennifer who is the head of personnel but is mistaken as the one of candidates by Steve. They talk about many things like the expansion of the company. Steve tells her that he is taking part in an evening course to improve his IT skills. He also demonstrates his opinion about the on-line selling that no customer care, no customer loyalty. And few days later, with the help of Jim, Steve gets the job and successfully makes a deal with Jack Smedley. Unit 2: Hilary Baker is under a lot of pressure because there will be a big presentation about a web directory for inventors, called The Great ID,tomorrow. It is an important presentation for them because if Pablo Duarte, the European chairman, likes their idea, they will receive a great number of money invested by him and other investor. Because of the nervousness, Hilary decides to go back to the office to have another look of the material. Jackie goes with her and help her. Back to the office, they begin to rehearse the presentation. Hilary is too tense to finish the presentation. However, with the encouragement and help of Jackie,Hilary becomes better and better. While they are rehearsing, Simon, the partner of Hilary, comes back to the office with the same reason as Hilary. And then they go through the presentation together. Their efforts have been paid. The presentation is very successful. Both Pablo and Pik Sen are satisfied with their presentation. They also exchange present for each other and Pik Sen and Simon will go to the opera together. Unit 3: Jonathan and Katherine, the two senior directors of Preston Valley Racecourse, are talking about the problem of the racecourse which has lost money for the two years and is nearly bankrupt. Katherine suggests to ask the consultant for help. But Jonathan opposes to do so for worrying that the consultant wants to turn the racecourse into a theme park. Warren ,the owner of the racecourse, is 1.2 million dollars in debt and he has to put a proposal in front of the bank manager on Friday. Unluckily, he failed to get some suggestions from Jonathan and Kathrine. Jonathan wants to accept Warren’s suggestion of opening a riding school which would take too much time to bring in the profit they need and they also need to invest extra money but Warren doesn’t have another money to put in. Finally, as the time is limited, they have no choice but to consult with the expert. Therefore, they consult with Jean-Piere Dubois who is the best in the business. After the visit of the racecourse, Jean-Piere gives two suggestions. One is to sell part of the land and then use the money to relaunch the business. The other one is to sell the racecourse to the National Racecourse Association, which the future of the racecourse is guaranteed but Warren would not be the owner of the racecourse any longer. Katherine agrees on the first suggestion while Jonathan supports the second one. Warren asks them to think about it more cautiously and then make a right choice. After a heated discussion, Warren makes a decision to choose the second suggestion. One week later, Katherine don’t want to work for a big cooperation and leaves the racecourse for France to
初中定语从句详细讲解
定语从句 定语从句是一个重点语法项目，而且也是各种考试中考查的重点，要牢固掌握好关系代词和关系副词的基本用法及特殊用法，并且学会利用相关知识来作出判断，准确解答相关试题。 知识详单 何为定语从句：在复合句中充当定语的从句叫定语从句。 定语从句的作用相当于形容词，用来修饰主句中的某一名词、代词或整个主句。 先行词：被定语从句修饰的词，定语从句一般紧跟在它所修饰的先行词之后。 关系词：在先行词和定语从句之间起连接作用的词叫关系词。 关系词有关系代词和关系副词两种。 知识点1关系词
知识点2关系代词的用法
知识点3关系副词的用法
知识点4定语从句的注意事项
考点突破 考点1考察关系代词的用法 1.(哈尔滨中 考)Everyonehashisaiminlife.However,youcan'tgetfatononemouthful.Startwiththeeasi estthing______youcancontrol A.who B.that C.which 【解析】选B。先行词thing是物，且由最高级修饰，关系代词用that, 2.(绥化中考)Thestorybook_________youlentmeisveryinteresting. A.which B.it C.what 【解析】选A。先行词为Thestorybook，指物，关系词which在从句中作lent的宾语。故选A, 3.(龙东中考)Shirleyisthegirl_______taughtmehowtouseWe-chat(微信). A.whom B.which C.who 【解析】选C,thegirl作先行词，表示人，关系词在从句中作主语，故关系代词用who, 4.(咸宁中考)-Haveyouseenthedocumentarynamed ABiteofChina(舌尖上的中国)) -Yeah!It'sthemostfunnyone__________Ihaveeverseen. A.that B.what C.which D.where [解析]选A。先行词one前有最高级修饰，故关系词用that, 考点2考察关系副词的用法 5.(枣庄中考)Heisunlikelytofindtheschool________hetaught50yearsago. A.where B.when C.how D.why
体验商务英语综合教程3-课文翻译(全)
第一单元 欧洲制造 除顶级奢侈品牌外几乎所有的时尚品牌都或者已经在亚洲生产，或者正在考虑这样做。美国皮具制造商蔻驰（Coach）是一个典型的例子。在过去的五年中，它通过完全在低成本市场生产已经提高了毛利率。2002年3月它关闭了在波多黎各拉雷斯的工厂（公司拥有的最后一家工厂），将所有产品全部外包。 巴宝莉（Burberry）在亚洲有许多特许授权安排，2000年它决定给日本三洋公司的特许授权延长十年。 这意味着按零售价计算巴宝莉几乎一半的销售额将是亚洲授权生产的。但是同时，日本的顾客却偏爱该集团欧洲生产的产品。 为了应对这种对巴宝莉在亚洲工厂所生产产品的需求，三洋公司在东京银座开设旗舰店，出售从欧洲进口的巴宝莉产品。 在《金融时报》的采访中，许多企业高管表示，消费者认为顶级的奢侈品牌来自欧洲，在亚洲尤为如此。古琦（Gucci）的多米尼克·德索尔说：“无论如何，亚洲的消费者只相信：奢侈品来自欧洲，而且一定是那里制造的最好。” 古琦的控股公司（Pinault Printemps Redoute）的首席执行官塞格·温伯格说，公司不会将古琦的生产线移到海外。然而一些业内人士认识到，就算对豪华奢侈品牌而言，这种变化也将来临。普拉达（Prada）的首席执行官帕特里齐奥·埃特里说：“‘意大利制造’的标签很重要，但我们真正提供的是一种风格，风格是文化的表现”。因此，他认识到高品质的时尚产品并非总是要在意大利生产。 欧洲工商管理学院市场营销系的Amitava Chattopadhyay教授说：“品牌就是消费者心中的一系列联想，其中之一就是原产地。对于奢侈品，品牌的作用至关重要。破坏它是一种弥天大罪。没有哪个品牌经理愿意将产地和品牌形象之间的关系搞错。” 第三单元 活儿脏，点子棒 SOL清洁公司是欧洲北部最令人向往的公司之一，走进它的总部SOL城，你会感觉就像走进了一个商业广场。它坐落在赫尔辛基市中心一家翻新过的电影制片厂里，里面色彩炫烂、气氛喧闹，彰显着非凡的创造力。墙壁刷上了明亮的红色、白色和黄色；员工在大厅里来去行走，不时用黄色的手提电话交谈。丽莎·乔洛南11年前在家族拥有的150年工业帝国的基础上开发了SOL清洁服务。SOL的竞争公式有五个关键成分。 很少人会梦想成为一个清洁工。但是，这并不意味着清洁工不能在工作中找到满足。乔洛南认为，满足的关键是乐趣和个人自由。SOL的清洁工穿着红色和黄色的连身衣裤，强化了公司的乐观形象。SOL的标志是一张黄色快乐的脸，它出现在所有的东西上，从鲜艳的外套到公司的预算报告。自由意味着废除企业传统文化中所有的条条框框。在SOL没有头衔或秘书，没有个人办公室或工作时间表。公司取消了所有的特权和身份符号。 SOL的培训计划包括七个模块，每个历时四个月，最后是严格的考试。当然，擦桌子
体验商务英语综合教程3 unit2 case study
Good afternoon, everyone. Welcome to my presentation today. My presentation will be divided into two parts. At first, I’ll talk about the problems about our company .Then I’ll show you my ideas about the problems. If you have any questions, I will be very glad to answer you after my presentation. Ok, first, let’s talk about the problems. As we all known, our market share has declined by 25% in the last two years. Brand loyalty, price, copycat’products, Brand image, account for the big loss, and these are also what we need to devote ourselves to solving. Now, let’s move on to the solutions, this will be also the main part of my presentation. As far as I’m concerned, copycat’s products, brand image, these two are the problems demanding prompt solution. We can see the chart from the report, most of our former market share was taken away by the top five European coffee brands, our competitors. So a new strong competitive product is needed. For example, we can bring out an instant coffee, which aims for office workers. Or we can bring out a series of food related to coffee under our brand. Besides , stretching our brand is also a good idea to be considered, that means we can allow selected manufacturers of coffee equipment (cafetieres, percolators, coffee machines) to use our brand on
- 体验商务英语综合教程--Unit-4-答案
- 体验商务英语综合教程第二版unit
- 体验商务英语综合教程3 中文 双语对照版
- 体验商务英语综合教程3(第二版)答案Unit2Travel
- 体验商务英语综合教程中文双语对照版
- 体验商务英语综合教程三(第二版)答案Unit 1 Brands
- 体验商务英语综合教程3 unit2 case study
- 体验商务英语综合教程1第二版 unit1
- 体验商务英语综合教程第二版unit
- Unit 5 Stress体验商务英语综合教程2
- 体验商务英语综合教程5 answers to unite 1&2
- 体验商务英语综合教程4考试参考
- 体验商务英语综合教程3(第二版)答案
- 体验商务英语综合教程答案
- 体验商务英语综合教程第二版unit
- 体验商务英语综合教程3-课文翻译(全)
- 体验商务英语综合教程3(第二版)答案
- 体验商务英语 综合教程2
- [新版]体验商务英语综合教程1unit1
- 体验商务英语综合教程4翻译