2986 Ecological Models | Remote Sensing

Caswell H (2000) Matrix Population Models: Construction, Analysis, and dynamics of large herbivores. Annual Review of Ecology and
Interpretation, 2nd edn. Sunderland: Sinauer Associates. Systematics 31: 367–393.
Coulson T, Gaillard JM, and Festa-Bianchet M (2005)Decomposing the Jones GP, Milicich MJ, Emslie MJ, and Lunow C (1999) Self-recruitment
variation in population growth into contributions from multiple in a coral reef fish population. Nature 402: 802–804.
demographic rates Journal of Animal Ecology 74: 789–801. Lande R, Engen E, and Saether BE (2003) Stochastic Population
Eriksson O and Ehrlen J (1992) Seed and microsite limitation of Dynamics in Ecology and Evolution. Oxford: Oxford University Press.
recruitment in plant populations. Oecologia 91: 360–364. Rodriguez SR, Ojeda FP, and Inestrosa NC (1993) Settlement of benthic
Fogerty MJ, Sissenwine MP, and Cohen EB (1991) Recruitment marine-invertebrates. Marine Ecology – Progress Series 97: 193–207.
variability and the dynamics of exploited marine populations. Trends Stearns SC (1992) The Evolution of Life Histories. Oxford: Oxford
in Ecology and Evolution 6: 241–246. University Press.
Gaillard JM, Festa-Bianchet M, Yoccoz NG, Loison A, and Toı̈go C Tilman D (1997) Community invasibility, recruitment limitation, and
(2000) Temporal variation in fitness components and population grassland biodiversity. Ecology 78: 81–92.

Remote Sensing
N Horning, American Museum of Natural History, New York, NY, USA
ª 2008 Elsevier B.V. All rights reserved.

Introduction Summary
Remotely Sensed Data Sets for Ecological Modeling Further Reading
Accuracy Assessment and Validation

Introduction images from airplanes and satellites to help identify or

better understand features on the Earth’s surface. In this
Remotely sensed data and methods are used extensi- article we will discuss a wide set of techniques, often
vely to provide data layers for ecological modeling. known by the alternative name of ‘Earth observation’
Understanding fundamental remote-sensing concepts (EO). We will only address EM remote sensing, so geo-
will help ecosystem modelers make informed decisions magnetic and acoustic remote-sensing techniques (sonar
regarding the utility and limitations of the broad spec- and seismic sounding) will not be covered.
trum of remotely sensed data and their derived products. A simple example of a remote-sensing instrument is a
The main focus of this article is to introduce photographic or digital camera. A camera records energy
remote-sensing science and associated data sets of in the form of light that is reflected from a surface to form
potential relevance to ecological modeling. The article an image. Most photographic cameras record visible
begins with an overview of fundamental remote-sensing light so that when we look at the photograph the image
concepts. This is followed by a section on how remo- resembles the feature that was photographed. More sophis-
tely sensed data can be used to derive a broad range ticated remote-sensing instruments are able to record
of biophysical data sets that are useful for terrestrial energy outside of the range of visible light. Data from
and aquatic ecological modeling. Several of the more remote-sensing instruments can be recorded as images or,
common data sets derived from remotely sensed in the case of lidar, a series of point data.
data are described with comments about how the
data sets are produced, their strengths, and limitations.
The article concludes with a summary of accuracy and Brief History of Remote Sensing
validation.
For our purposes we will begin the history of remote
sensing with the invention of the photographic camera in
the early nineteenth century. In the 1840s photographs
What Is Remote Sensing?
were taken from cameras secured to tethered balloons
In general terms, remote sensing is the science and prac- for purposes of topographic mapping. For the next 100
tice of acquiring information about an object without years or so camera technology improved but the major
actually coming into contact with it. In terms more appro- advances were in the platforms used to hold the camera
priate for our purposes remote sensing is a technology for systems. At first people experimented with platforms
sampling reflected and emitted electromagnetic (EM) such as kites, rockets, and even pigeons. A major step
radiation from the Earth’s terrestrial and aquatic ecosys- forward was made with the invention of the airplane and
tems and atmosphere. This is typically done by recording the next leap occurred when cameras could be mounted
Ecological Models | Remote Sensing 2987

on satellites, which provided a very stable and, of course, Spectral curves

high-altitude platform. Satellites also provide an ideal 100
platform for acquiring systematic data from around the Qu
artz
globe which has proved invaluable for large area ecolo-

Sn
80

o
gical modeling.

w
Percent reflectance
By the 1940s instrument research was also becoming
60 Pinion Pine
increasingly sophisticated, pushing remote-sensing technol-
ogy beyond visible-spectrum photography into infrared
detection and radar systems. Leveraging this research, in 40 Dry grass
1972 the National Aeronautics and Space Administration
(NASA) began the Landsat program with the launch of the 20
Earth Resources Technology Satellite 1 (ERTS 1), which Asphalt

was later renamed Landsat 1. The Landsat program is now Water

0
the longest running program of satellite remote sensing 0.5 1.0 1.5 2.0 2.5
focused on EO. Following the launch of Landsat 1 other Wavelength (µm)
satellites were launched carrying different types of instru-
Figure 1 Spectral signatures for selected features. Materials
ments such as radar, lidar, and more precise optical sensors. on the Earth’s surface have unique spectral reflectance
Satellite remote sensing has evolved to the point where properties. This figure shows the spectral reflectance curves of
most environmental systems (hydrologic, atmospheric, eco- six common materials. Many of the methods used in remote
systems) now have dedicated satellite instruments recording sensing are designed to associate the spectral information
information to help us better monitor and manage Earth’s acquired by a sensor with the spectral qualities of features that
are to be identified.
environments, and providing valuable data for use in ecolo-
gical modeling. For example, NASA’s Earth Observing
System (EOS) is a mission that includes the acquisition of Earth’s surface is spectral reflectance: the ratio of the
satellite-based observations, science, and a data system to intensity of light reflected from a surface divided by
support the study of the land surface, biosphere, solid Earth, the intensity of incident light. Different features have
atmosphere, and oceans. One sensor in particular from the different spectral reflectance properties and we can use
EOS mission that offers a broad range of image products of this information to identify individual features. For
interest to ecologists is the Moderate Resolution Imaging example, white sand reflects most visible and near-infra-
Spectroradiometer (MODIS). Advances in satellite/instru- red light whereas green vegetation absorbs most red
ment packages will continue to be made providing more wavelengths and reflects most near-infrared wave-
precise and accurate data that can be used for ecological lengths. Figure 1 illustrates the spectral properties of
modeling. different materials.
Some remote-sensing instruments also provide infor-
Basic Concepts mation about how EM energy interacts with the surface of
a feature or within a three-dimensional feature such as a
It is necessary to understand a few basic remote-sensing forest. These will be discussed later in this article.
concepts before we begin discussing how remotely sensed
imagery can be used in ecological modeling.
What is an image?
EM spectrum The most familiar form of remotely sensed data is an
The electromagnetic spectrum (EMS) includes wave- image. An image is made up of individual elements that
lengths of EM radiation ranging from short-wavelength are arranged in a grid of rows and columns. These ele-
(high-frequency) gamma rays to long-wavelength (low- ments are called pixels. When zooming into an image,
frequency) radio waves. We focus on the region of the individual pixels can be seen (Figure 2).
spectrum starting in the ultraviolet and continuing In addition to rows and columns of pixels, images also
through the microwave wavelengths. Optical sensors are have layers. These layers are commonly referred to as
used to measure ultraviolet, visible, and infrared wave- ‘bands’ or ‘channels’. Throughout this article we use the
lengths, and microwave sensors are used for the term ‘band’ to refer to the layers in an image. These
microwave portion of the EMS. bands also correspond to different wavelengths of the
A fundamental physical principal that remote sensing EMS. Remote-sensing instruments vary in the number of
relies on is that different features on the Earth’s surface bands recorded, some only record a single band of data
interact with specific wavelengths of the EMS in differ- while others record hundreds. A convention has been
ent ways. When working with optical sensors the most established to use ‘hyperspectral’ to describe images with
important property used to identify features on the many bands (usually well over 100) and ‘multispectral’ for
2988 Ecological Models | Remote Sensing

Figure 2 Zooming to an individual pixel. The image in the left is printed at full resolution, it is a Landsat Enhanced Thematic Mapper
Plus (ETMþ) image acquired over Burlington, Vermont on 21 August 1999. The image in the top-right is a subset of this image that has
been magnified by a factor of 3. In the magnified image individual pixels (the square blocks that make up the image) can be seen. The
three black and white images in the bottom-right represent the three image bands that are used to create the color image. In this case
the red band is from the ETMþ band 4 (near-infrared), the green band is from ETMþ band 5 (mid-infrared), and the blue band is from
ETMþ band 3 (red). These three bands are combined to make the color image.

images with fewer (usually from three to a few dozen) Radiance at the sensor
bands. is a measure of the
electromagnetic
With most imagery the individual bands are used to radiation hitting the
Sun
record radiance values at different wavelengths. Radiance sensor’s detector
is a measurement of the intensity of EM energy. In other
words, the sensor is measuring the intensity of light when it
Incident light
hits the detector. The units for this measurement are typi- Reflected light
cally watts per steradian per square meter (W sr1 m2).
It is important to understand that optical sensors measure
radiance and not reflectance. Reflectance, which is the
ratio of reflected light over incident light, can be estimated Reflectance is the ratio of the
intensity of reflected radiation
using image-processing methods but the physical property over incident radiation
recorded by the sensor is radiance (Figure 3).
Target

Different platforms and orbits Figure 3 Reflectance and radiance. Remote-sensing detectors
For local and detailed information the airplane is still often measure radiance which is the energy of the radiation hitting the
detector. Reflectance, which must be calculated, is the ratio of
the platform of choice since it is possible to select which
the intensity of reflected radiation over incident radiation.
sensors should be mounted for a particular application
and it is possible to determine when to fly. Aircraft have
the ability to fly low to acquire imagery with a lot of detail. view of the Earth over a regular time period, for example,
For global and systematic coverage, satellites are the stan- every 16 days in the case of Landsat. It is important to note
dard remote-sensing platform. Most satellite orbits can that with a near-polar orbit the polar regions are not viewed
be classified as either geostationary or polar orbiting. from the satellite. For this reason, when people mention
Geostationary satellites orbit the Earth in the equatorial global remotely sensed data sets they often mean the data
plane with the same orbital period as the Earth so the sets are near-global. Polar and near-polar orbiting satellites
position remains fixed over a particular point on the Earth fly only several hundred kilometers above the Earth’s
and therefore it can continously view the same area. These surface.
satellites are commonly used to monitor the weather but are
too far from the Earth’s surface (38 500 km) for detailed Passive versus active remote sensing
environmental monitoring. More common for Earth remote Remote-sensing instruments are often categorized as having
sensing is a near-polar orbit that provides a near-global either active or passive sensors. An active sensor generates
Ecological Models | Remote Sensing 2989

Table 1 Active and passive satellite-based remote-sensing instruments

Sensor name Type Wavelength range Resolution (m)

IKONOS Optical 450–900 nm 1–4

SPOT5a Optical 500–1750 nm 2.5–10
IRS-P6 – LISS-4 Optical 520–860 nm 5.8
ALOS AVNIR-2 Optical 420–500 nm 10
ASTER Optical 520–11 650 nm 15–60
Landsat ETMþ Optical 450–2350 nm 15–30
MODIS Optical 459–14 385 nm 250–1000
AVHRR Optical 580–12 500 nm 1000
ENVISAT Radar 5.7 cm (C band) 25
RADARSAT-1 Radar 5.7 cm (C band) 10–100
RADARSAT-2 Radar 5.6 cm (C band) 3–100
ALOS – PALSAR Radar 23.5 cm (L band) 10–100

its own signal which is subsequently measured when the EMS. In a lidar system a single light pulse can reflect
reflected back by the Earth’s surface. A passive sensor mea- off of several features in vertical space such as different
sures solar energy that is either reflected or emitted from layers in a forest. A single emitted pulse will result in a
features on the Earth’s surface. Table 1 lists a number of wave or series of returned pulses that are recorded by the
different active and passive instruments mounted on satellite detector. These return pulses can be recorded as a wave
platforms. (full-waveform lidar) or in discrete pieces that correspond
Although most passive sensors operate in the visible to the peaks in the returned signal. A number of different
and infrared portions of the EMS, there are also some types of lidar systems have been developed but most pro-
passive microwave sensors in use that measure a number vide the capability to record the first and last return of the
of parameters such as wind speed, atmospheric and sea light pulse. The returns correspond to the top of an object
surface temperature, soil moisture, rainfall, and atmos- (i.e., top of a tree canopy) and the base substrate that the
pheric water vapor. object is resting on (i.e., the ground). This is ideal for
An advantage of passive sensors is that most rely on the measuring the height of features such as trees or buildings.
Sun’s energy to illuminate the target and therefore do not Radar systems behave differently from optical systems
need their own energy source so in general they are simpler with respect to how they interact with materials. The signal
instruments. A limitation for most passive optical sensors is from most radar systems can penetrate well into a forest
that they require daylight to operate, although there are canopy, and radar systems with especially long wavelengths
some sensors that record nighttime lights and clouds at (e.g., P-band systems) can even penetrate dry ground.
night and others that record energy emitted from the Although lidar is probably best know for its capability
Earth’s surface. Since most of these sensors operate in the of acquiring digital elevation data which can be used to
visible and infrared wavelengths, they are adversely describe topography, these systems show a lot of promise
affected by weather and cloud cover. Lastly, since sunlight to be able to directly measure vegetation structure char-
is primarily reflected from the top of a feature, such as a acteristics. Lidar can make direct measurements of
forest, it is not possible to ‘see’ under a canopy to measure vegetation structure and provide vertical information
vegetation structure. To obtain this kind of information it is about structure that is largely missing in data collected
necessary to use active sensors. using passive remote-sensing instruments which only
Active sensors, such as radar and lidar emit their own record light reflected from the top of a canopy.
energy to illuminate a target and are comprised of a signal
generator and receiver. They measure the strength of the
returned signal and the time delay between when the What qualities determine what can be identified
instrument emits the energy and when it receives in an image?
the returned pulse. These two types of information are There are different characteristics that affect the detail that
used to describe vegetation structure. Radar is an acronym can be resolved (seen) in a digital image. These are tradi-
tionally referred to as the four types of image resolution.
for ‘radio detection and ranging’. Radar systems operate in
Most people think of ‘resolution’ as being synonymous
the long-wavelength microwave portion of the EMS and
with spatial resolution but other ‘resolution’ terms used in
thus are largely unaffected by clouds and rain. They can be
the formal literature are as follows:
considered all-weather systems. Lidar is an acronym for
‘light detection and ranging’ and these systems use lasers Spatial resolution. This is often simply referred to as ‘reso-
that emit light in the visible and near-infrared portions of lution’ and is the size of a pixel (smallest discrete scene
2990 Ecological Models | Remote Sensing

element and image display unit) in ground dimensions. In sets that are useful for terrestrial and aquatic ecological
most cases an image’s resolution is labeled with a single modeling. In this section several of the more common
number, such as 30 m, which represents the length of a data sets derived from remotely sensed data will be
side of a square pixel if it were projected onto the Earth’s described with comments about how the data sets are
surface. If the pixel were rectangular (not very common any produced, their strengths, and limitations. This is not an
more), then the length and width of the pixel would be exhaustive list but it highlights the diversity of data
provided. derived from remote sensing that can be integrated into
Spectral characteristics. This includes bandwidth, band ecological modeling.
placement, and the number of bands. Spectral bandwidth,
or spectral resolution as it is often called, refers to the range
of wavelengths that are detected in a particular image band.
Land Cover
This is effectively a measure of how precisely an image
band measures a portion of the EMS. Band placement Land cover data are available in image and vector formats
defines the portion of the EMS that is used for a particular with individual types of vegetation assigned to discrete
image band. For example, one band might detect blue classes. For example, in an image format each vegetation
wavelengths and another band might detect thermal wave- type would be assigned a unique numeric value and in a
lengths along the EMS. The properties of the features one vector format each polygon would have attribute infor-
is interested in sensing indicate which bands are important. mation that would describe the type of land cover in that
The last spectral variable is the number of bands. The more polygon. These data are available with a wide range of
bands that are available the more precisely spectral proper- thematic (classification scheme) and spatial (spatial reso-
ties of a feature can be measured. lution) detail.
Acquisition dynamics . This has two components. The first The specific classification scheme used for a particular
is the minimum time a particular feature can be recorded land cover data set can be as simple as forest/nonforest
twice, often called the repeat frequency of temporal reso- classes or as detailed as a species-level map. One impor-
lution. Some sensors with a very wide field of view can tant point related to thematic detail is that the more
acquire multiple images of the same area in the same day classes that are used, the lower the per-class accuracy
whereas some sensors have a repeat frequency of several will be. In other words, the classes in a forest/nonforest
weeks. It should also be reiterated that most remote-sen- map will be more accurate than the individual classes in a
sing satellites have a near-polar orbit and are not able to species-level map. The spatial detail in a land cover data
acquire imagery at the poles since their orbit does not go set is usually a direct result of the type of remotely sensed
over these areas. The other component is the timing of the data on which the classification was based. Using aerial
acquisitions. Dynamic features such as deciduous forests photography or high-resolution satellite imagery indivi-
and events such as flooding often have an optimum time dual tree crowns can be discerned allowing improved
for which they should be imaged. For example, the identi- capabilities for mapping species-level information.
fication of deciduous vegetation is aided by acquiring For the most part land cover maps are created using
imagery during leaf-on and during leaf-off periods. data from optical sensors. One area where radar sensors
Sensitivity of the sensor . This is defined by the dynamic excel is in mapping wetlands and water under forests,
range of the sensor as well as the range of digital numbers such as in flooded forests.
that can be used to represent the pixel values. Sensors have Land cover data sets can be created using manual
lower limits below which a signal is not registered and upper and/or automated methods. The basic principle of land
limits above which the sensor saturates and is unable to cover classification is to translate the pixel values in a
measure increases in radiance. The detail that can be mea- satellite image into meaningful land cover categories.
sured between these extremes is determined by the range This is often accomplished using automated procedures,
between the minimum and maximum digital numbers per- in which a computer algorithm is used to assign individual
mitted for a particular data type. For example, Landsat TM pixels or groups of pixels to one of the valid land cover
data values can range from 0 to 255, whereas IKONOS categories. The classification process can also be accom-
values range form 0 to 2048. This potential range of values plished using visual interpretation methods where the
is often referred to as quantization or radiometric resolution. interpreter uses visual cues such as tone, texture, shape,
pattern, and relationship to other objects to identify and
group similar land cover types. In general, the human brain
Remotely Sensed Data Sets for Ecological is better at interpreting the spatial characteristics in an
Modeling image and automated algorithms are better suited for pro-
cessing spectral (the many image bands) information.
Remote sensing provides instruments and methods that There are dozens of classification methods in use but
can be used to derive a broad range of biophysical data there is not a single ‘best’ approach.
Ecological Models | Remote Sensing 2991

One of the possible limitations of classified land cover Measurements of these characteristics are based on the fact
data is that the information is discrete instead of contin- that reflectance, transmittance, and scattering of energy in
uous. One way around this is to create a ‘continuous a canopy is greatly affected by the structure of the vegeta-
fields’ image data set for selected types of vegetation. In tion and how the vegetation components (leaves, branches,
this data set each pixel value represents the percentage of trunk) interact with the spectrum of energy being used by a
that pixel covered by a particular land cover type. For particular remote-sensing instrument.
example, in a broadleaf tree continuous fields data set a Vegetation indices have been used extensively for
pixel value of 65 would mean that 65% of that pixel is global studies to monitor changes in vegetation health
covered by broadleaf tree species. In addition to different and cover and have been effective in mapping droughts,
types of land cover it is also possible to create a contin- desertification, phenology, net primary productivity, and
uous fields data set for imperviousness. This is called an deforestation around the world. The most common vege-
impervious surface data set and it is being used increas- tation index, the ‘normalized difference vegetation index’
ingly in ecological modeling particularly when it is (NDVI), is based on the principle that healthy green
necessary to quantify water runoff. vegetation absorbs most of the incident red wavelengths
of light and reflects most of the near-infrared wave-
Landscape metrics lengths. The formula for NDVI is
Once a land cover map has been produced it is sometimes
NDVI ¼ ðNIR – redÞ=ðNIR þ redÞ
desirable to quantify different aspects of patterns in the
landscape using assorted landscape metrics for use in eco- where NIR is the radiance value from the near-infrared
logical models. These metrics provide an objective way to band and red is the radiance value from the red band.
describe patterns commonly described using subjective Two other common vegetation indices that use a simi-
terms such as ‘highly fragmented’, ‘small patches’, and lar principle as NDVI are the ‘soil adjusted vegetation
‘heterogeneous landscape’. Using software tools it is easy index’ (SAVI), which was developed to reduce the effect
to create these metrics. Some common metrics include of background material (i.e., soil, sand, snow) and the
‘enhanced vegetation index’ (EVI) which is less sensitive
1. Landscape composition to atmospheric scattering effects. Vegetation index data
• Proportion – area of one cover type compared to
the total area
sets are usually available as temporal composites, such as
10-day or monthly. In a composite product the ‘best’
• Richness – number of different patch types index values from the composite period are provided.
• Evenness – relative abundance of different patch
types
Using this approach it is possible to reduce the negative
effects of clouds and haze.
• Diversity – composite measure of richness and
evenness
Vegetation structure and biomass data sets are often
created using data acquired from radar and lidar sensors.
2. Spatial configuration Although radar has been used to measure vegetation
• Patch size and shape properties such as biomass, leaf area index, and forest
• Connectivity of patches structure most of this has been experimental so data are
• Dispersed or clumped patches somewhat limited. This is an area of active research and as
• Setting with respect to neighboring patches new instruments are developed, operational methods
using radar instruments may be available in the not too
Although landscape metrics can be of great value, caution distant future.
must be exercised when using these metrics. Many of the Commercial lidar instruments are available for
metrics are very sensitive to scale and extent of the study area mounting in airplanes that can quickly provide vegetation
so comparisons across time and space must be done with care. height information and this can be correlated to tree volume
and biomass using allometric tables. Unfortunately, using
these instruments is expensive and it is often not feasible to
cover large areas. Research using airborne and satellite lidar
Vegetation Characteristics instruments to measure vegetation structure directly is
underway and early results look promising.
In addition to land cover, there are several vegetation
characteristics that can be measured using passive and
active remote-sensing instruments. These include
Topography
• phenology,
• primary productivity, Elevation data sets and their derived products (Table 2)
• vegetation health
vegetation and vigor, and are used extensively in ecological modeling. Digital ele-
• structure. vation data are available as digitized points or contour
2992 Ecological Models | Remote Sensing

Table 2 Products commonly derived from digital elevation resolution. People have experimented using data col-
models (DEM) lected from optical remote-sensing instruments to map
Slope steepness soil moisture but results have been mixed.
Slope aspect
Hillshade and perspective views
Viewshed/line of sight Surface Temperature and Precipitation
Topographic features Although temperature and precipitation data are routi-
Ridges nely collected using satellite-based instruments, data sets
Peaks created using meteorological station data are still often
Channels
Pits
preferred for ecological modeling. In some cases data sets
Passes derived from station data are improved by integrating
Plateaus data collected from satellite-based instruments.
Hydrologic parameters Satellite-based rainfall estimates are made using pas-
Flow direction sive microwave, radar, and optical instruments. Rainfall-
Flow accumulation estimate data sets derived from satellite remote sensing
Predicted watercourses are often too coarse (>4 km) for many ecological model-
Watershed boundaries
ing tasks.
There are many satellite-based instruments that can
measure surface temperature including some that pro-
lines, triangulated irregular networks (TIN), and as duce a global daily and 8-day composite data set
gridded surfaces or images. available with a 1 km resolution. Finer resolution data
When using elevation data it is important to know can be acquired using satellite sensors. However, they
what the elevation values represent. These values can do not have as frequent repeat cycles as those producing
represent the surface of the bare Earth, the surface of 1 km products.
the features on the Earth (i.e., top of the canopy), or
somewhere in between.
Atmospheric Properties (Clouds)
A relatively new topographic data set is the Shuttle
Radar Topographic Mission (SRTM) ‘digital elevation Of all the remote-sensing-derived atmosphere products
model’ (DEM). The resolution of this data set is 30 m (Table 3), the cloud mask is arguably the most frequently
and it covers land areas between 56 south latitude used for ecological modeling since it provides a measure
and 60 north latitude. It has been degraded to 90 m of cloud cover. For the MODIS cloud mask product a
for most areas of the world outside of the United States. clear-sky confidence level (high confident clear, probably
The data for the SRTM DEM were collected using an clear, undecided, cloudy) is assigned to each pixel. This
interferometric radar instrument mounted on the Space daily data set is available with 250 m and 1 km spatial
Shuttle. Over forested areas the elevation value provided resolutions.
by the SRTM data set represents a point somewhere
roughly half-way between the ground surface and the
Oceans
top of the canopy. The exact point depends on the struc-
tural characteristics of the forest stand. In the marine environment, remote sensing provides data
Lidar is increasing being used to collect elevation data for a wide variety of environmental variables (Table 4).
with vertical accuracy on the order of centimeters. Lidar Optical remote-sensing methods are commonly used to
instruments are flown on aircraft and are routinely used for map the ocean–land interface and coral reefs. Methods
monitoring coastal areas. The accuracy and speed of lidar
elevation data collection is unmatched by other methods.
Table 3 Information about the atmosphere derived from
remote sensing

Aerosol loading and size distribution over oceans

Soil Moisture Aerosol content and optical thickness over land
Water vapor (precipitable water)
Soil moisture is a much sought after data set for ecological
Cloud optical thickness
modeling. However, data that meet the needs of a parti- Cloud top temperature and height
cular application are often not available because existing Cloud locations
data spatial resolutions are too coarse, available data do Cloud particle phase
not cover the area of interest, or data do not exist for Cloud particle radius
Total-column ozone content
the required timeframe. Global soil moisture data col-
Atmospheric wind and temperature profiles
lected from microwave sensors is available with a 25 km
Ecological Models | Remote Sensing 2993

Table 4 Information about marine ecosystems derived from Accuracy Assessment and Validation
remote sensing

Sea surface temperature When using data sets derived from remote-sensing meth-
Ocean color (productivity) ods, it is important to understand the level of accuracy
Coral reef mapping associated with a particular product. Accuracy figures can
Ocean surface topography refer to the accuracy of the calculated value when com-
Oil slick detection and mapping
Ocean circulation pared to the actual biophysical value or it can refer to
Wind speed and direction the positional accuracy. In some cases both figures are
Fluorescence provided. Accuracy statistics should be distributed with
the data set. Unfortunately, this is not always the case. In
some cases, accuracy statistics for a data set simply do not
used for these applications are similar to those used for exist.
terrestrial land cover mapping except there is greater use For data sets that represent categorized data, the sta-
of the short-wavelength blue bands for coral reef mapping tistics usually provide per-class and overall accuracy
since those wavelengths are better able to penetrate into information. This is common practice with land cover
the water to provide more information on features several data sets. For data sets such as elevation, values are
meters under the water surface. given for horizontal and vertical accuracy. These values
Two other common global marine data sets are sea are usually given as a probability of being within a speci-
surface temperature, which is derived using methods fied distance. Other data sets, such as those derived from
similar to those used for land temperature, and ocean MODIS data are validated by a team of scientists and in
color which uses optical imagery to determine levels of some cases the validation effort is incomplete or ongoing.
phytoplankton in the water. Both of these data sets are When using these data sets, it is important to research the
acquired on a daily basis. most current information available about the data set’s
accuracy. This information is often available on the
Internet.
Different methods for reporting accuracy exist
Inland Water (Streams, Lakes, Wetlands, etc.) and this is an active research area. For example, new
Inland water features include wetlands, streams, and accuracy methods are being developed that provide
lakes. A mix of optical and microwave remotely sensed information about the spatial distribution of the error.
data is used to measure and monitor a number of para- Methods are also being developed using fuzzy statistics
meters (Table 5) related to inland waters although many to indicate the severity of the error instead of using the
of these measurements are only available at local or traditional approach of noting a value as either correct or
regional scales and many of the methods require signifi- incorrect.
cant field work to correlate actual values with what is
recorded by the remote-sensing instrument.
Mapping the extent of inland water features is done
using a mix of optical and microwave instruments. For Summary
example, radar is an ideal technology for locating and
mapping standing water, even if it occurs under a forest Remotely sensed and derived data are an invaluable
canopy such as in a flooded forest. Optical sensors are asset for ecological modeling. These data provide
used to measure temperature and identify aquatic broad area and repetitive coverage that is impractical
vegetation. to gather using field methods. This is a very dynamic
field and as sensors and processing tools improve,
these data will continue to become more precise and
Table 5 Information about inland water ecosystems
accurate.
derived from remote sensing
This article provides a brief overview of remote sen-
Water body and wetland mapping sing and the types of data that can be derived to support
Flooded forest mapping ecological modeling. The intended audience for this
Water surface elevation
article is the consumer, not producer of remote sensing
Water depth
Turbidity/secchi depth products and it is not a comprehensive treatise on remote
Water temperature sensing. More in-depth information about remote sen-
Aquatic and wetland vegetation mapping sing and its application to ecological modeling can be
Riparian buffer mapping found in the section titled ‘Further reading’ for this
Flow rates
article.
2994 Ecotoxicology | Reproductive Toxicity

See also: Ecological Models, Optimization; Monitoring, Lillesand TM, Kiefer RW, and Chipman JW (2004) Remote Imaging and
Image Interpretation, 5th edn. New York: Wiley.
Observations, and Remote Sensing – Global Dimensions;
Maune DF (2001) Digital Elevation Model Technologies and
Sensitivity, Calibration, Validation, Verification. Applications: The DEM Users Manual. Bethesda, MD: American
Society of Photogrammetric Engineering and Remote Sensing.
Rencz AN (ed.) (1999) Manual of Remote Sensing, Vol. 3: Remote
Sensing for the Earth Sciences: Manual of Remote Sensing, 3rd edn.
Further Reading Hoboken, NJ: Wiley.
Schowengerdt RA (1997) Remote Sensing: Models and Methods for
Congalton RG and Green K (1999) Assessing the Accuracy of Remotely Image Processing. San Diago, CA: Academic Press.
Sensed Data: Principles and Practices. Boca Raton, FL: Lewis. Skidmore A (ed.) (2002) Environmental Modeling with GIS and Remote
Gergel SE and Turner MG (eds.) (2002) Learning Landscape Ecology: Sensing. New York: Taylor and Francis.
A Practical Guide to Concepts and Techniques. New York: Springer. Turner W, Spector S, Gardiner N, et al. (2003) Remote sensing for
Henderson FM and Lewis AJ (eds.) (1998) Manual of Remote Sensing, biodiversity science and conservation. Trends in Ecology and
Vol. 2: Principles and Applications of Imaging Radar: Manual of Evolution 18(6): 306–314.
Remote Sensing, 3rd edn. Hoboken, NJ: Wiley. Ustin SL (ed.) (2004) Manual of remote sensing, Vol. 4: Remote Sensing
Jensen JR (2000) Remote Sensing of the Environment: An Earth for Natural Resource Management and Environmental Monitoring,
Resource Perspective. Saddle River, NJ: Prentice-Hall. 3rd edn. Hoboken, NJ: Wiley.
Jenson JR (2004) Introductory Digital Image Processing: A Remote Wilkie DS and Finn JT (1996) Remote Sensing Imagery for Natural
Sensing Perspective. Saddle River, NJ: Prentice-Hall. Resources Monitoring: A Guide for First-Time Users. New York:
Kerr JT and Ostrovsky M (2003) From space to species: Ecological Columbia University Press.
applications for remote sensing. Trends in Ecology and Evolution
18(6): 299–305.

Reproduction See Mating Systems

Reproductive Toxicity
L V Tannenbaum
Published by Elsevier B.V.

Introduction Looking to the Future

Scope of the Issue Does Reproductive Toxicity Really Need to Be
Standard Approach to Reproductive Toxicity Evaluated?
Assessment – Hazard Quotients Summary
Other Approaches to Reproductive Toxicity Further Reading
Assessment

Introduction this encyclopedia’s arrangement. Of nearly 70 subtopical

entries under the larger heading of ‘ecotoxicology’, just
Although chemicals released to the environment have the four involve specific types of illness or disease; aside from
potential to tax any number of biological processes in ‘reproductive toxicity’, there are only ‘carcinogenic
nonhuman species, the case can be made that reproduc- effects’, ‘sublethal effects’, and ‘teratogenesis’ that are
tive effects are the ones that need to be most seriously discussed.
addressed. Whereas the public’s perception of reproduc-
tive toxicity may primarily or only reflect matters of
societal relevance (e.g., hunting or fishing as recreational
pastimes may all of a sudden be jeopardized; the desire to Scope of the Issue
have in abundant number, a species whose image is on a
state emblem), a stark biological reality should heighten A proper review of reproductive toxicity within an eco-
the attention given to reproductive toxicity, more so toxicological context should draw out a distinction
than any other chemically caused ill-effect – if reproduc- between concerns over large geographical expanses and
tion is being compromised in a species, extinction, in a those that pertain to discrete and relatively small area
specific locality or beyond, may ensue. The argument locations. Examples of the former would be the US’s
can be made that the primacy of reproductive toxicity as Gulf States region, or the Eastern deciduous forest; an
an ecotoxicological endpoint of concern is reflected in example of the latter would be a 5 ac hazardous waste