Here's a list of the people in our monitoring group, along
with the team they're from:
Garrett Marino—1
Martin Skelton—1
Alicia DeFrancesco—2
Jackie McConnell—2
Kristopher Tantillo—4
Allison Brown—5
Lisa Song—5
Monitoring group email list: m2008-tor-sensors@mit.edu.
Research topics (Nov. 6):
--What's being monitored on the Galapagos now—Garrett
--Monitoring technologies in use now and future possibilities—Kristopher
(plants/animals monitoring), Lisa (remote sensing), Martin (in general),
Alicia (marine monitoring)
--Recovery plans—Kristopher
--Parallel cases and how they're being monitored—Garrett (Great Barrier
Reef), Allison (Sea of Cortez, Birdlife International)
--monitoring volcanoes, pollution, etc.--Jackie
Final research topics (Nov. 16):
Jackie: air quality
Alicia: water quality
Kris: endangered species and recovery plants, invasive species
Allison: indicator species, plants
Marty: organization (logistics, working with Biopreserve/ecovillage, practicalities)
Garrett: cities and human impact on ecosystem
Lisa: soil pollution, indicator microbes, biogeochemical cycles
A list of questions we came up with (Nov. 6):
1. What will be monitored?
2. Who will collect/process the data? Will it go through the CDF?
3. Where will this data be stored?
4. If the monitoring system shows that something is wrong in the
ecosystem, who will take care of it?
5. Is it a good idea to make use of existing data (for example, to
get the data of water temperature, chlorophyll levels, etc. from existing
satellites, instead of trying to monitor every aspect by ourselves)?
6. The Charles Darwin Foundation is currently doing all kinds of
research/monitoring. What can we do that will be different/better? Will
we be trying to get the local people involved?
7. Is it acceptable to use unscientific methods of monitoring? For
example, fishermen know from personal experience the kinds of fish whose
numbers have declined in the past years, based on their catch. What if
we asked them and took polls? Should we consider their knowledge and data?
Questions asked on 11/19:
What are the acceptable limits for the things you are measuring?
Will the data be available to the public? Will people have to pay?
Will taking tree-ring samples requiring cutting down trees?
What do you want to see in a biopreserve subcommittee?
Where are you going to be monitoring?
Budget??? Cost??? Who will pay? Who will be paid? Who will profit?
How many trees are in the Galapagos?
WHAT IS BEING DONE NOW?
How many employees? Where will they be from?
What is the lifetime of the sensors? What maintaince will they require?
What will be the impact of the sensors? On the environment? On society?
How will you imcorporate sensors into the village?
How will you train people who work with the sensors? Are there associated
educational/job training benefits?
How much data will you collect?
How will you process this data? Who will do it?
How will you store it? How will it be transmitted? Are there issues of bandwith?
Information overload?
What *exact* data are you going to collect?
List of people I emailed to try to get more information:
Professor Geist, Idaho University. He did a lot of research on volcanoes
in the Galapagos.
dgeist@uidaho.edu. http://www.uidaho.edu/~dgeist/
Professor Nocera, Department of Chemistry, MIT. He gave a Terrascope lunchtime
lecture on nanosensors.
nocera@mit.edu. http://web.mit.edu/chemistry/dgn/www/index.html
Professor Sadoway, DMSE, MIT. He gave a short presentation on Nov. 19 about
flexible, paper-thin Li-ion batteries that he was working on developing.
We're hoping to get more information about possible applications to sensors.
dsadoway@mit.edu. http://web.mit.edu/dsadoway/www/
Remote Sensing
The following information comes from (Wilkie et al. 1996). I have
listed the page numbers from where I found the information.
Remote sensing measures without touching or directly
observing the object under study. Electromagnetic radiation (EMR) comes
in the form of sunlight, heat from warm-bodied organisms, or radar (artificial
and human-made). Remote sensors can record information almost instantaneously,
and over a large range of spatial scales (1-9).
What happens when EMR reaches an object (11-13)
· Transmitted—velocity of the energy is changed
as it passes through the object
· Absorbed—the energy is transferred to the
target, most of the time as heat
· Reflected—energy bounces back, in the same
form as before. This is responsible for color
· Scattered—the energy's direction is altered
randomly
· Reradiated—the energy is initially absorbed,
but then reflected back, usually as heat
All matter on Earth reacts in one of the above
ways when hit with EMR. The way they react is determined by pigmentation,
moisture, texture, and whether it's in sunlight or shadow (14).
Responses of water, soil, and vegetation (15-19)
· Water—clear water reflects wavelengths mostly
in the blue color region, and this gives water its characteristic color.
But clear water looks black in the reflected infrared region since it absorbs
radiation between 0.8 to 3.0 μm. Land also reflects in this range, but
the fact that water will appear black is a way of distinguishing the boundary
between land and water. Turbid water will reflect more EMR in the rest
of the visible range. To study water turbidity, quality, or depth, scientists
need sensors that gather reflectance data in the blue and green visible
spectrum. However, clouds, fog, and haze all interfere with remote-sensing
done with water
· Soils and rock—dry, tan and silty soils reflect
well in longer wavelengths. As the water content in soil increases, the
reflectance looks more and more like the response for water
· Vegetation—plants have a characteristic response
to EMR. They absorb well in the red and blue regions of the spectrum, while
reflecting strongly near the infrared section. In addition, the health
of plants can de determined by changes in reflectance. Chlorophyll levels
and water content differ according to a plant's species, its age, and its
health (green, yellow, or dying from disease or drought).
Thermal radiation (19-22)
· all objects warmer than absolute zero will
emit in the thermal infrared section. This allows us to detect ocean currents
(both location and direction) and water surface temperature over large areas
· thermal responses can be used to distinguish
animals within terrestrial environments. For example, this kind of remote
sensing is used to count walrus populations in Alaska and sandhill cranes
in Nebraska
· factors affecting thermal radiation response:
the angle of the sun (changes shadow patterns), rainfall and humidity,
cloud cover, and different stages in man-made differences. For example,
when monitoring slash-and-burn plots, the response from a field of growing
crops will look similar to the response of the vegetation-rich land around
the plot, but in a newly-burned place, the response will be characteristic
of soil and ash
Microwave sensing (22-25)
· these longer wavelengths (1mm-1m) are only
marginally affected by smoke, cloud cover, and snow. Large amounts of rain
will reflect wavelengths less than 3cm long, and this is used in weather
forecasts
· the energy comes from the sensor, and not
the sun.
· responses depend on the sensor's characteristics,
such as wavelength, polarization, or incidence angle. They also depend
on the target's size, shape, moisture, texture, and electrical conductivity
· smooth surfaces, such as leaves, scatter the
signal in all directions, so very little returns to the microwave sensor
(the antenna). Houses and other objects with sharp edges produce strong signals
· in addition, objects with higher electric
conductivity will reflect more. This includes metals and water, but excludes
soil, plants, and sediment-rich water
Information available from remote sensing imagery (26)
+ = available from that source
-- = not available from that source
|
Visible
|
IR
|
Feature
|
Blue
|
Green
|
Red
|
Reflected
|
Thermal
|
Radar
|
x, y position
|
+
|
+
|
+
|
+
|
+
|
+
|
size, shape
|
+
|
+
|
+
|
+
|
+
|
+
|
elevation
|
+
|
+
|
+
|
+
|
+
|
+
|
color
|
+
|
+
|
+
|
+
|
--
|
+
|
temperature
|
--
|
--
|
--
|
--
|
+
|
--
|
texture
|
+
|
+
|
+
|
+
|
+
|
+
|
vegetation—chlorophyll absorption
|
+
|
--
|
+
|
--
|
--
|
--
|
vegetation—biomass
|
--
|
--
|
--
|
+
|
--
|
--
|
vegetation—water content
|
--
|
--
|
--
|
+
|
--
|
--
|
soil moisture
|
--
|
--
|
--
|
+
|
+
|
+
|
fires
|
--
|
--
|
--
|
--
|
+
|
--
|
Applications of remote sensing imagery
I. Keeping maps up-to-date in a timely manner (222)
· cheaper and faster than aerial photography.
Plus, remote sensing works over a larger scale
· used by the US Forest Service
II. Mapping habitat suitability (222-228)
1. General facts
· one flight would cost $275, verses $9000 for
a single flight using aerial photography
· remote sensing allows the user to set the
resolution at will
2. The piping plover and the least tern
· in Nebraska, data was gathered on the habitat
available for the piping plover and the least tern along 400km of the Platte
river
· the total data was gathered on 2.4 hours of
tape
· the impacts of dams and irrigation on water
levels were studied to find the range of water levels which would allow
these birds to thrive.
3. Sandhill cranes on the Platte River, Nebraska
· the cranes roost during the night, on wide,
bare patches of the river channels
· differences in thermal reflectance between
the cranes and the water allowed for identification of the cranes
· the river was split into ten segments, the
number of roosts per segment counted
· several roosts were randomly selected from
each segment, and the density of cranes per roost was calculated
· errors could have resulted from: distortion
from the plane banking around river bends, and missed roosts due to poor
navigation
III. Monitoring Floods (228-229)
· in April 1998 the Darling River in Australia
flooded
· satellites were used to determine the flood's
progress
· using thermal reflectance at night was more
accurate than daytime measurements
· during the day lakes and flooded land have temperatures
similar to those of the land
· at night the land cools faster, and water temperatures
remain much higher
· in the end, the estimation of the area of land
flooded by the river was accurate to plus or minus 25%
Remote sensing advantages (30-32)
· no interference with what you are sensing
· data is gathered instantaneously, repeat gatherings
can be done on a scale of hours to years. This is good for mapping change
· there are many kinds of sensors with varying
spatial and spectral resolutions
· our eyes can detect only hundreds or thousands
of different color hues on various backgrounds, but multispectral remote
sensors can detect much subtler differences. Digital image processing can
cluster together all the pixels with similar spectral responses, which isolates
important information and decreases the number of colors needed for a visually
useful map. For example, this kind of technology was used to map Adelie penguin
nests (which includes the penguins, guano, soil, and debris) against the Antarctica
landscape. This allowed scientists to find previously undiscovered nesting
places. (36)
More examples (32-33)
· speed, dispersion rate, area covered of oil
spills
· how often certain areas are affected by floods
· development of phytoplankton blooms and how
they move
· individual trees can be identified, and their
health observed by aerial videography
· human population density
Detail
· resolution is defined as a measure of the smallest
feature that can be identified spatially, spectrally, temporally, or radiometrically
from the background
· the resolution depends on the field-of-view
of the sensor and the contrast between the object of interest and its background
(38)
1. Spatial resolution (38-46)
· is the smallest size of terrestrial features
that can be distinguished from the background, or whose size can be measured
· in aerial photography, this depends on the
film's grain (smaller grain = higher resolution) and the photographic scale
(which depends on the camera's focal length and the altitude of the plane)
· some aerial photography is taken from space.
The Large Format Camera (LFC) can achieve a resolution of 20m
· Detectability--ability of a sensor to sense
the presence or absence of an object
· recognizability--our ability to name the object
· an object can be detected but remain unrecognizable.
For example, small circles in the plains of Sudan may be interpreted as patches
of shrubs or as nomad camps
· a general rule: identification requires three
times the resolution needed for detection, and analysis requires up to ten
times the resolution needed for identification
· the resolution of a sensor needs to be less
than half of the smallest dimension of the object to be monitored. This minimizes
(though it does not completely erase) the edge effect (when the reflectance
comes from a mixture of features rather than one feature on particular)
· higher resolution is not always better. Pick
the resolution that is the most appropriate for the purpose
Scale
|
Identifiable Features
|
1:500
|
Individual tree size, plant species identification,
uses of buildings and industries
|
1:5,000
|
Volume of timber, wetland boundaries, small tributary
outlines, transportation systems, property boundaries
|
1:50,000
|
Outlines the areas of evergreen and deciduous trees,
flow direction of water, shoreline outlines, finds major transportation routes,
measures agricultural land area
|
1:500,000
|
Regional vegetation and land use classification
|
1:5,000,000
|
Major rivers, continental vegetation zones, continental
cloud cover
|
2. Spectral resolution (46-47)
· refers to the size and number of wavelength
regions that the sensor can detect accurately
· black-and-white aerial photography film can
detect all reflected blues, greens, and reds
· other sensors such as the Landsat thematic
mapping sensor (Lansat TM) has a more fine spectral resolution, and in one
band records radiation between 0.63 and 0.68 μm
· depending on what's being monitored, you need
to choose the bands of a sensor system that maximizes the spectral contrast
between the object and its background
3. Radiometric resolution (47-48)
· measure of the sensitivity of a sensor to differences
in the radiant flux reflected from objects on the ground
· to increase the resolution, the difference
between 2 brightness levels should be bigger than the system's noise level.
This way, brightness level difference won't be mistaken for errors or vice
versa
4. Temporal resolution (48-50)
· a measure of how often a place can be revisited
with the sensor system
· good for measuring changes: used to track floods,
progress of growing season in the Sahel
· can succeed where other kinds of resolutions
fail. For example, two grain crops may look the same spectrally, and are
indistinguishable most of the time. However, if they flower at different
times, temporal resolution will be able to tell the difference between the
crops. At times like this, temporal sensors can reduce costs by decreasing
the need for higher spatial, spectral, or radiometric resolutions
5. Resolution trade-offs (51-53)
· higher resolutions mean more data, but the
data is not always useful
· for example, a picture of the Mona Lisa shown
with 16 gray tones looks almost exactly the same as the same picture with
256 gray tones. The human eye sees almost no difference between the two
· more resolution can mean extra noise. For example,
when measuring the area farmland verses urban areas, it's not necessary to
see every patch of vegetation on the land. This might lead to mistakes, as
lawns or gardens may be taken for farmland
· cost and the time needs to analyze the data
rises greatly with increased resolution
· for example, an inventory of all the wetlands
in Massachusetts (using aerial photography) would cost more than $8 million
and 200,000 man-hours at the scale of 1:5,000. The same inventory could be
done, without loss in accuracy, by using a 1:12,000 scale, costing $1 million
and 30,000 hours of work
· the following graph shows complete coverage
of Massachusetts using different scales
Scale
|
Number of aerial photogrphs needed
|
1:12,000
|
11,556
|
1:25,000
|
2,889
|
1:60,000
|
300
|
Different sensor systems (54-57)
1. Satellite imagery
· can be bought from companies or government
agencies (ex: US Geological Survey)
· available in digital, prints (both black-and-white
and color), or transparencies
· good for broad ranges (at least 10,000 km2),
areas needing repetitive coverage
· targets objects larger than the image's spatial
resolution
· gives worldwide coverage without government or
security limitations over most of the world (64)
2. Large format aerial photography
· can be bought from national or state cartographic
centers
· topographic and land-use maps
· used for small areas (les than 1000km2) that
need one-time mapping
· for smaller features (less than 10m in diameter)
3. Aerial 35mm photography
· usually used by individual researchers who
rent a plane and take pictures through an open window with their own 35mm
cameras
· used for areas less than 5km2
· good for applications requiring frequent, repetitive
monitoring
· targets objects less than 5m in diameter
4. Aerial videography
· usually used by individual researchers who
rent a plane
· the video camera is mounted on the window and
controlled by the researcher
· used for areas less than 50km2
· good for applications requiring frequent, repetitive
monitoring
· targets features bigger than 25m in diameter
5. Radar
· from planes or satellites
· EMR signal is in the microwave region, does
not involve the sun
· good for places with year-round cloud cover
· for features bigger than 50m in diameter
6. Remote sensing platforms (61-63)
· tripods and cranes can raise sensors from
several meters to dozens of meters above ground
· planes, helicopters, rockets, and hot air balloons
go up to 500,000 meters above sea level
· low-orbiting platforms (e.g. space shuttles)
go up to 200-950 km high
· geosynchronous sensors orbit high over the earth,
at an altitude of 36,000 km. They remain over the equator, in synch with the
Earth's rotation, and thus gather data from the same location. There are
5 placed around the equator, and they've been gathering global meteorological
data since 1976
Format
|
Advantages
|
Disadvantages
|
Satellite imagery
|
--spectral range is wide (UV-IR)
--available both digitally and as photos
--wide range of detectors
--consistent sun angle and perspective
--good for comparing different scales and wavelengths
|
--low spatial resolution compared to other sensors
--expensive to start
--high training costs
|
Large format aerial photography
|
--good spatial resolution
--little geographic rectification needed
--simple operation
--analysis equipment is cheap
|
--small spectral range film
--digital analysis available only by scanning prints
--hard to interpret large volumes of data
--expensive to analyze data for areas bigger than 1000km2
|
Aerial 35mm photography
|
--very cheap for sampling (small plane rentals
are $100-$200/hr)
--high resolution (1:500 or higher)
--easy to operate the equipment
--analysis equipment is cheap
|
--hard to connect the individual photos together
--digital analysis possible only by scanning the prints
--hard to find places that develop color infrared film
--photographic film has a small spectral range
--hard to interpret large areas of data
|
Aerial videography
|
--cheapest system for small to mid-range monitoring
--can see the images while they're being monitored
--visual and digital data
--easy to use equipment
--spectral range from visible to near-IR
--you can talk into the video camera and comment as the monitoring occurs
|
--low spatial resolution
--need lots of light for image acquisition
--to estimate the areas of features obtained, you need to account for
spatial distortion
|
Radar
|
--data is obtained regardless of cloud cover or
atmospheric conditions
|
--expensive aircraft equipment
--satellite systems are still in testing stage
--low spatial resolution
--analysis methods not well developed
--only gives information on terrain texture and water content
|
Pitfalls of remote sensing (254-255)
· all results and their accuracy depend on the
user's knowledge of the land's history and the spectral characteristics
of the target area
· the use of computers and other technology gives
the impression that monitoring is an automated process. However, the user
must understand the characteristic reflectance of soil, plants, water, and
the distortions caused by the sun, landscape, and atmosphere to make an accurate
classification of land-use
· cloud cover can limit or prevent data gathering
· shadows caused by low sun angles and topographic
relief (occurs in high altitudes during the winter, and occurs world wide
in the early morning when remote sensing satellites pass over the area) will
decrease a sensor's ability to distinguish features
· places with no roads or human activity makes
it hard to gain accurate geometric corrections of remote sensing data
· keep in mind that landscapes change with the
weather, the seasons, and with human or natural intervention
· there is always a trade-off between land area
covered and spatial imagery. Cost is also a factor, meaning that the data
gained may be less-than-ideal
· it's often wasteful to purchase the most up-to-date
and expensive equipment. This kind of technology requires expensive and long
training before they can be used properly. Hardware and software can be
upgraded when needed; it may be better to start with less expensive equipment
Conclusion and thoughts
- use remote sensing to monitor floods, fires, other natural disasters
- use it to determine land usage (agricultural land verses cities, farms
verses wild vegetation, factories in cities, etc.)
- weather, water temperature, soil moisture, chlorophyll absorption
- population density of animals that live above ground, that are big
enough to be distinguishable from the air
- data from global satellites can be bought from agencies such as the
US Geological Survey, since the data alreadye exists; this is a lot cheaper
than launching satellites for the sole purpose of monitoring the Galapagos
- buy current land-use and topographic maps from agencies. Then continue
to monitor to observe changes
- aerial videography is cheap. It can be used to monitor changes
Back