ANALYSIS OF TIME-SERIES SATELLITE IMAGES TO MONITOR

CHANGES IN NATURAL ENVIRONMENTS

Errol L. Montgomerya, Michael J. Roskoa, Janis Blainer-Fleminga, Kevin C. Horstmanb

Abstract: Spectral reflectance data from Earth-resource satellites have provided hydrogeologists with a new
technical tool for monitoring natural environments. Because the satellite instruments record reflected solar
radiation in digital format, the numerical data may be analyzed mathematically to identify and document
changes in areal extent of surface water and areal extent and density/vigor of vegetation using well-established
numerical methods. Graphs can be constructed to show changes in measured environmental conditions with
respect to time, or can be analyzed for changes over time in relations between environmental conditions and
other parameters, such as groundwater pumped, annual precipitation, or land use. Principal advantages of the
satellite image method for environmental monitoring include: a relatively long historic period beginning in 1985
for which information is available; the reliable numerical basis for the results; and the ability to assemble a
historic environmental baseline and provide on-going monitoring for large or remote areas where standard
monitoring methods may not be practical. Analyses of surface water and vegetation using satellite-recorded
spectral reflectance data are particularly well suited for project areas that cross political boundaries where data
collection may be difficult, or where little or no historic data exist.

The method requires processing and analysis of reflected solar radiation data obtained by Landsat 5 Thematic
Mapper, and Landsat 7 Enhanced Thematic Mapper Plus multispectral scanners. Materials on the Earth’s
surface absorb or reflect different parts of the spectrum depending on their physical and chemical properties and,
therefore, typically have characteristic spectral “signatures”. Using these signatures, environmental parameters
such as surface water and vegetation may be identified and analyzed. Because Landsat satellites 5 and 7 are in
sun-synchronous, near-polar orbits that pass the satellites over the same point on Earth every 16 days,
information obtained at regular intervals allowing comparison of spectral data for the same area at different
dates. Therefore, relatively continuous historic environmental parameters can be established for a location, and
change in those parameters can be measured. Results are readily exported to a geographic information system
for comparative analyses between environmental conditions and natural or man-caused factors that affect natural
environments.
a
Errol L. Montgomery & Associates, Inc.
1550 E. Prince Road
Tucson, Arizona 85719
(520) 881-4912
www.elmontgomery.com
b
Kevin C. Horstman P.G.
Geologist
P.O. Box 36084
Tucson, Arizona 85740
(520) 742-1083
[email protected]
 2

Introduction

Understanding and measuring natural systems have become increasingly important for detection and
prevention of damage to important ecosystems, and for restoration of ecosystems that have been damaged.
Analysis of time-series spectral reflectance data from satellites offers a way to characterize and detect change in
natural environments where standard monitoring methods may not be practical. Problems specifically associated
with monitoring natural systems in transboundary areas include: inability to obtain data, incomplete or
inadequate monitoring, and incompatible data resulting from use of different monitoring methods. For
monitoring areal extent of surface water and vegetation in transboundary areas, analysis of spectral reflectance
data from satellites offers a solution for these problems. In some cases, this method may be the only practical
way to monitor an area, or to obtain reliable post-1985 historic monitoring data.

Earth-Resource Satellites

Numerous civilian satellites and imaging systems are presently in orbit recording reflected and emitted
electromagnetic energy from Earth’s atmosphere and surface. The longest-operating earth-resource observation
satellite system is Landsat. Most Earth-resource satellites record reflected visible and infrared radiation in
several ranges or bands of the electromagnetic spectrum. To monitor local-scale changes in surface water and
vegetation, reflectance data must be obtained at suitable resolution and in appropriate spectral bands. To be able
to detect change, satellites must record reflectance for the same area at regular intervals. Landsat satellites 5 and
7 have these required characteristics, and reflectance data measured by these satellites can be used to monitor
changes in surface water and vegetation.

Landsat 5 is equipped with the advanced Thematic Mapper (TM) multi-spectral scanner. The TM
scanner is capable of recording three visible and three near- and mid-infrared spectral bands. Landsat 7 is
equipped with the Enhanced Thematic Mapper Plus (ETM) multi-spectral scanner, which records identical
spectral bands as the TM scanner on Landsat 5. Bands that are actively measured by the Landsat satellites and
used by this method are shown on the following illustration. Landsat satellites 5 and 7 are in sun-synchronous,
near-polar orbits that pass the satellites over the same point on Earth every 16 days. Landsat data are obtained at
regular intervals, allowing comparison of spectral values for the same area at different times. It is these regular
measurements for the same area that allow development of relatively continuous historic environmental
parameters for a location, and also allows detection of change with time for these parameters.
 3

Band 1 reflectance penetrates shallow water, making it useful for surface water studies; band 1 is also
sensitive to atmospheric haze. Haze includes the scattering effects of smoke, dust, and water vapor. Band 2 is
sensitive to turbidity and pollution in water bodies, and is moderately sensitive to vegetation type. Band 3
corresponds to a part of the spectrum that is strongly absorbed by chlorophyll found in green and dry grasses and
other plants, and is strongly reflected by moist soil. Band 4 corresponds to a part of the spectrum that is strongly
reflected by chlorophyll in plants. Bands 5 and 7 are sensitive to ferric iron and hydrous minerals, especially
clays.

Information for each band is recorded for a square area on the Earth’s surface 30 meters on a side, and is
referred to as a picture element or “pixel”. Amount of radiation detected for each pixel in each spectral band is
recorded as a unitless number that represents a fraction of the sensor range; sets of pixels for a specific time and
area are referred to as a “scene” or an “image”. Although colors are commonly assigned to band data to allow
visual presentation, numeric values assigned for each measured spectral band allow quantitative evaluation of
the data.
 4

Method Concepts

Earth objects absorb or reflect different parts of the spectrum, and as a result, different materials
typically have characteristic spectral “signatures”. Surface water and individual plant species have unique
spectral signatures and may be characterized using spectral data obtained from the TM and ETM scanners.
Because the satellite recording instruments report reflectance in digital format, images may be analyzed
mathematically for areal extent of surface water, and areal extent and density/vigor of vegetation using well-
established numerical methods (Huete, 1988; Qi et al., 1994; Avery and Berlin, 1992).

Several processing steps are required to provide quantitative and temporally consistent results for area of
surface water, and for areal extent and density/vigor of vegetation. These steps include: 1) obtaining satellite
image data and support information for selected calendar time periods; 2) geo-referencing the image pixels to a
common local map coordinate system so that reflectance values can be referenced to the same geographic
location for different satellite images; 3) correcting for variation in geometric relations for sun angle and satellite
position; 4) correcting for variations in atmospheric haze; and 5) calculating area of surface water, area of
vegetation, and/or health and abundance of vegetation.

To effectively monitor an area, surface water and vegetation must be distinguishable from other
materials on the Earth’s surface. Because water and vegetation have relatively distinctive spectral reflectance
signatures, equations have been developed to identify these materials (Huete, 1988; Qi et al., 1994; Avery and
Berlin, 1992). For example, areal extent of surface water can be estimated from satellite image data using ratios
of reflectance in infrared and blue-green parts of the spectrum. Landsat TM bands 4, 5, and 7 are strongly
absorbed by water, and bands 1 and 2 are strongly reflected by water. Ratios of bands 1 or 2 to bands 4, 5, or 7
are larger for water than for most other common materials on the surface of the Earth and, therefore, can be used
to estimate area covered by water.

Similarly, the difference in reflectance of plant chlorophyll between bands 3 and 4 is large; chlorophyll
strongly absorbs wavelengths detected by band 3 and strongly reflects wavelengths detected by band 4. No
other common Earth materials produce a comparable large difference in amount of reflected energy in these
bands. Comparison of amounts of reflectance between bands 3 and 4 is commonly used to estimate areal extent
and density/vigor of vegetation from satellite image data.

Examples

Example results for the satellite image method are taken from a closed-drainage “salar” basin with salt
flats and saline lakes on the basin floor. The example basin is located in the Atacama Desert region in the Andes
Mountains of northern Chile. Many salar basins in the region contain large amounts of groundwater stored in
basin-fill alluvial and fractured rock aquifers; these aquifers are commonly in hydraulic communication with
springs and saline lakes on the basin floors. The aquifers in salar basins have been developed as groundwater
sources for municipal and industrial use. However, because of the relative scarcity of surface water in this
extremely arid region, springs and lakes tend to be critical habitats for many fish, bird, and animal species,
including three species of flamingos that feed in shallow saline lakes.

Mitigation of groundwater drawdown impact from pumping of wells in the environmentally important
basin floor areas has become a focus of attention and regulatory concern. Therefore, monitoring of surface water
and vegetation to detect potential impacts has become a required part of management plans for development in
these areas. Because the salar basin floors are commonly large and remote, Chilean regulatory agencies have
accepted the satellite image monitoring method as a useful technique for detection of potential past and future
impacts.
 5

Results from Baseline Satellite Image Monitoring

The following examples represent environmental baseline conditions for Laguna Tebinquiche, one of
several saline lakes located in the north part of the Salar de Atacama Basin, Chile. Large amounts of
groundwater pumping have not yet occurred in the north part of the basin. Variability demonstrated in the
following examples is natural and results chiefly from seasonal variation in temperature, precipitation, and
evapotranspiration.

Areal Extent of Surface Water: Processed images showing location of surface water for Laguna Tebinquiche
during summer and winter of 1985 are shown below.

Pixels with TM band 4/2 ratio in the range from 0 to 0.75 are highlighted in blue and represent surface
water area. It is clear from inspection of the images that surface water is less abundant in the summer than in the
winter, when evaporation rates are much smaller. Reflectance data was processed mathematically to determine
total areal extent of surface water. Computed total area of surface water for the summer 1985 image is about
564,000 square meters (m2); and for the winter 1985 image is about 1,751,000 m2.

By combining time-series data sets for the Laguna Tebinquiche area, trends and ranges can be
established. After determining normal ranges for environmental parameters, results can be analyzed
mathematically to determine impacts from external factors, such as groundwater pumping. Water surface area of
Laguna Tebinquiche for the period 1985 to 1998 is shown below. These results indicate that areal extent of
surface water is influenced by seasonal differences. Evaporation potential is about 2 to 3 times larger in the
summer than in the winter. Vegetation growth in the summer is much larger than in the winter and may also
result in decreased areal extent of shallow surface water in the summer due to increased plant transpiration.
 6

3,0 00,0 00
SUMME R
20 -Jan -89
Area of s urface water

Range of surface water area

2,5 00,0 00 WINTER

27-Jun-94 Area of surface water
20-Jan- 89 Area of sur face water

Range of surface water area

16-Jul-95
07-Jul-86 27-Jun-94 21-Jul-97
2,0 00,0 00
24-Jun-93 02-Jul-96
04-Jul-85
21-Jun-89

1,5 00,0 00

1,0 00,0 00
31-Jan-93

20-Jan-89
25-Jan-85 29-Jan-98

5 00,0 00

0
19 84 198 5 1 986 198 7 1 988 198 9 1 990 199 1 19 92 199 3 19 94 199 5 1 996 199 7 1 998
YEAR
SEASONAL VARIATION IN AREA OF SURFACE WATER, 1985 THROUGH 1998
ERROL L. MONTGOMERY & ASSOCIATES, INC.
 7

Areal Extent of Vegetation: The images below have been processed to show areal extent for
vegetation for Laguna Tebinquiche during summer and winter, 1985.

Pixels with vegetation are shown in yellow and green, with green representing increased density or vigor
of vegetation. The magnitude of the index value for a given pixel is also a function of the abundance of
chlorophyll in the vegetation, and is a function of both density and health of the vegetation. Computed total area
of vegetation for the summer 1985 image is about 3,472,000 m2; and for the winter image is about 2,516,000 m2.

Seasonal variation in areal extent of vegetation for the Laguna Tebinquiche area during the period 1985
to 1998 is shown on the graph below. Inspection of results for years 1985 and 1993 suggests that for most single
years, areal extent of vegetation is larger in the summer than in the winter, due chiefly to decreased plant growth
during the colder winter months. However, inspection of long-term trend compared to precipitation suggests
that areal extent of vegetation is generally correlated to annual precipitation.
 8

10 ,000 ,000 200

20-Jan-89 Area where vegetation index is larger than 0.11

for satellite image obtained during summer
27-Jun-9 4 Area where vegetation index is larger than 0.11
for satellite image obtained during winter
Annual precipitation
8 ,000 ,000 160

ANNUAL PRECIPITATION, MILLIMETERS

6 ,000 ,000 29-Ja n-9 8 120

4 ,000 ,000 80
25-Jan-85
0 7-Jul-86

0 4-Jul-85 21 -J un-89
31-Ja n-9 3
2 0-Jan-89
2 ,000 ,000 2 1-Jul-97 40
27-Jun-94
24-Jun-93 02 -J ul-9 6
16 -J ul-9 5

Re po rted
p recip itation
= 0 mm
0 0
1984 1 985 1 986 19 87 198 8 198 9 1 990 1 991 19 92 199 3 1994 1 995 1 996 19 97 199 8
YEAR
SEASONAL VARIATION OF AREA OF VEGETATION AND COMPARISON TO PRECIPITATION,
1985 THROUGH 1998
ERROL L. MONTGOMERY & ASSOCIAT ES, INC.

Conclusions

Spectral reflectance data from Earth-resource satellites have provided hydrogeologists with a new
technical tool for use in hydrologic monitoring of natural environments. Monitoring natural systems in
transboundary areas may be difficult due to inability to obtain data in large and remote areas, to incomplete or
inadequate monitoring, and/or to incompatible data resulting from use of different monitoring methods.
Analysis of time-series spectral reflectance data from satellites offers a method for characterizing and detecting
change in natural environments where standard monitoring methods may be not be practical.

REFERENCES

Avery, T.E., and Berlin, G.L. 1992. Fundamentals of Remote Sensing and Airphoto Interpretation, 5th edn.
Prentice-Hall, Inc., New Jersey.

Huete, A.R., 1988. A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment 25,
295-309.

Qi, J, Chehbouni A, Huete A.R., Kerr Y.H., and Sorooshian S. 1994. A modified soil adjusted vegetation index.
Remote Sensing of Environment 48, 119-126.