Land Use/Land Cover Change

Detection with RS (Ch 12)

Natural and human-made features/systems can be
monitored with remote sensing (and in situ) to detect and
understand stable or dynamic processes and examine
impacts of global change

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Questions to think about:
• What is remotely sensed change detection?
• Why is image preprocessing so important for
change detection?
• What are some common change detection
algorithms, and what are their advantages and
disadvantages?
• How do you interpret change detection
images?
 What is change detection?
• The use of remotely sensed data to map the
spatial extent, magnitude and variability of
changes on the land surface through time;
• Examples:
– Vegetation dynamics
– Urban growth
– Industrial development (e.g., gas wells, mines)
– Coastal zone changes
– Climate change effects (e.g., sea ice)
OUTLINE
Land Cover and Land use Changes
• The current rates, extents, and intensities of land cover
and land use change (LCLUC) are far greater than ever
before.
• LCLUC drive unprecedented changes in ecosystems and
environmental processes at local, regional, and global
scales.
• Remote sensing can provide:
• broad, precise, impartial, and easily available information for quantifying the location, extent, and
variability of LCLU change; (inventory, monitoring, assessments)
• the causes and processes of LCLU change; (human, climate variab.+change)
• and the responses to and consequences of LCLU change. (society, ecosystem)

• How are remote sensing data products used to:

• map the extent of major land use types e.g. urban, forest, and agricultural land?
• detect changes in these landuse types?
• evaluate the environmental impacts (biol-biod-Ecol-geo-phys_H2O-chem-C-soc-econ-health)
resulting from these land use changes?
Ch 24 Advanced Remote Sensing
The general steps used to
perform digital change
detection of remotely
sensed data.
1. State the Nature of the
Problem
2. Significant considerations for
change detection
3. Remote sensing data
processing steps to extract
change information
4. Perform accuracy
assessment
5. Answer questions/distribute
results/report/product
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Thematic Attribute(s) or Indicator(s) of Interest

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 - Scale,
watershed, globe
- Coverage in
space and time

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 - Traffic and vegetation
time scales
- Project needs (system,
environmental
organization needs)

e.g. NLCD
Discrete and soft

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
• Remote Sensors
• Environment
• Time of day
• Anniversary dates
• Look angle
• Similar spatial
resolution
• Georectification
• Multi-spectral/sensor
• 8-12 bits

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
a,c) Zero percent cloud cover on Landsat 8 scene Path 16/Row 37 on May 14, 2013
and May 17, 2014, respectively. These data are excellent for change detection
purposes.

b,d) 30% cloud cover on May 30, 2013 and massive 80% cloud cover on July 20,
2014. The use of these images for change detection would yield poor results (imagery
courtesy of NASA).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Phenological cycles of a) sugarbeets, and b) cotton grown in the California San Joaquin Valley. Landsat MSS
images were obtained over a 12-month period. The color composite images (MSS RGB = bands 4, 2, 1) of two
fields were extracted and placed below the crop calendar information.

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Annual phenological cycle of cattails and waterlilies in Par Pond, SC. Cattails
are present throughout the year while waterlilies emerge in April and disappear in
November. The best time to acquire imagery to discriminate between cattails and
waterlilies is from mid-April to mid-May. Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Comparison with the 1976 aerial photography reveals
substantial residential land development.

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Dichotomous key used to identify progressive
stages of residential development.
Such development in Denver, CO,
normally begins by clearing the terrain
of vegetation prior to subdivision.
In many geographic areas, however,
some natural vegetation may be kept as
landscaping. The absence or existence
of natural vegetation dramatically
affects the range of signatures that a
parcel of land undergoes as it
progresses from natural vegetation (1)
to fully-landscaped residential housing
(10 or 17).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
a) The stages of residential
development described
previously applied to an area
in Beaufort County, SC, in
2007.
b) The stages of residential
development for the same
geographic area in 2011.
Many of the parcels in 2011
are subdivided, have access
to paved roads, buildings are
present, and the property is
fully landscaped (i.e.,
residential development Stage
10). Each building with an
icon has a Beaufort County
residential address
(imagery courtesy of Beaufort
County GIS Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Obscuration considerations
a) Buildings, roads and other
terrain may be obscured from
view by tall dense tree canopy
when imagery is collected at
Nadir in leaf-on conditions. In
this example the image is
centered on the driveway of an
obscured single-family
residence.
b-e) The same residential area
viewed from four different
cardinal directions. The view
from the South to the North (b)
provides the most detailed
information about the residential
home (imagery courtesy of
Beaufort County GIS
Department). Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
The general steps used to perform digital change detection
of remotely sensed data.
1. State the Nature of the Problem
2. Significant Considerations for change detection
3. Remote sensing data processing steps to extract change information
4. Perform accuracy assessment
5. Answer questions/distribute results/report/product

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– LiDARgrammetric change detection
– Post-Classification Comparison Change Detection
– Spectral change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
The logic of analog
visualization on a
computer screen using
two or three dates of
registered remote
sensor data placed in
the RGB memory
banks. Analysts can use:
a) individual original bands,
and/or
b) multiple-date
transformations of the
original data such as the
Normalized Difference
Vegetation Index (NDVI)

(imagery courtesy of Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Richland County GIS Boston: Pearson Education, Inc., 623 p.

Department).
The use of analog
visualization to detect
changes in high spatial
resolution imagery
collected in 2004 and
2007 of an area near
Ballentine, SC

(imagery courtesy of
Richland County GIS
Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Use of analog visualization
to detect changes

a,b) Multiple-date images of

Lake Mead, NV, collected by
Landsat ETM+ and ASTER.
c) Analog visualization change
detection using ASTER band 3
(0.76 – 0.86 μm) and ETM+
band 4 (0.75 – 0.90 μm)

(imagery courtesy of NASA). Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Output from the Esri, Inc., ChangeMatters® used to identify change on Mt. St. Helens from 1975 to 2010.
a) A 1975 Landsat Multispectral Scanner (MSS) image (RGB = bands 4,2,1).
b) A 2010 Landsat Thematic Mapper image (RGB = bands 4, 3, 2).
c) A 2010 NDVI image was placed in the green graphics memory and a 1975 NDVI image was placed in the red and
blue graphics memory. Unchanged areas appear in shades of grey because the earlier and later images should
have approximately the same NDVI values. The NDVI change image highlights areas with vegetation decrease
in magenta and vegetation increase in light green
(images courtesy of USGS and NASA).
http://changematters.esri.com/compare

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– LiDARgrammetric change detection
– Post-Classification Comparison Change Detection
– Spectral change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
 The logic of image algebra change detection
using image differencing or band ratioing

(imagery courtesy of Richland County GIS Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
FIGURE 12–13c The logic of image algebra change detection using image differencing or band ratioing (imagery courtesy of
Richland County GIS Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Image Differencing
Change Detection
a) Landsat ETM+ imagery of a portion of
Lake Mead in Nevada obtained on May
3, 2000.
b) ASTER data of Lake Mead obtained on
April 19, 2003.
c) The histogram of a change image
produced by subtracting the ETM+
2000 band 4 (0.75 – 0.90 μm) data
from the ASTER 2003 band 3 (0.76 –
0.86 μm).
d) Map showing the change as a function
of the two thresholds identified in the
change image histogram

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
(original imagery courtesy of NASA Earth Boston: Pearson Education, Inc., 623 p.

Observatory).
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– LiDARgrammetric change detection
– Post-Classification Comparison Change Detection
– Spectral change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
Multi-Date Composite
Image Change
Detection
The logic of performing change
detection using multiple dates of
imagery placed in a layerstack (a).
The layerstack can then be analyzed
using:
b) traditional unsupervised or
supervised classification
techniques, or
c) by using Principal Components
Analysis (PCA)

(imagery courtesy of Richland Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
County GIS Department). Boston: Pearson Education, Inc., 623 p.
Principal components
derived from a multiple-date
dataset consisting of
Landsat ETM+ and ASTER
imagery. Principal
component 2 contains
change information. The
first three principal
components were placed in
various memory banks to
highlight more subtle
changes,

(original imagery courtesy

of NASA).
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Multiple-date composite image
change detection using PCA
images that were created from a
multiple-date composite dataset
consisting of Landsat ETM+ and
ASTER data. Analog visualization
was used to highlight the information
content of the various principal
component images. Principal
components 2 and 3 contain much of
the change information

(original imagery courtesy of

NASA).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– Photo- LiDAR- grammetric change detection
– Post-Classification Comparison Change Detection
– Spectral change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
a) Date 1 aerial photography with the building footprint extracted and outlined in yellow. The footprint
of the same building derived from the Date 2 photography is displayed inJensen,
red. J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
b) Date 2 aerial photograph with the building footprint extracted displayed in red Pearson Education, Inc., 623 p.
Boston:
(aerial photography courtesy of Pictometry International, Inc.).
a) LiDAR-derived building footprints for 33 cities
and towns in metropolitan Boston, MA. The
buildings are so dense in most areas that they
appear dark gray at this small scale.
b) An enlargement of a small part of the study area.
Such LiDAR-derived information obtained on Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
multiple dates is used routinely to monitor Boston: Pearson Education, Inc., 623 p.
change
(source of data: www.mass.gov/mgis/massgis.htm).
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– Photo- LiDAR- grammetric change detection
– Post-Classification Comparison Change Detection
– Spectral change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
The logic of performing post-
classification comparison change
detection using multiple dates of
imagery classified using either a per-
pixel or object-based image analysis
(OBIA) classification and a change
detection matrix tailored to meet the
change detection information
requirements
(imagery courtesy of Richland County GIS
Department).
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
ab) Rectified Landsat Thematic
Mapper data of an area centered
on Fort Moultrie, SC, obtained
on November 9, 1982, and
December 19, 1988 (RGB =
bands 4, 3, 2).

cd) Classification maps of the Fort

Moultrie, SC, study area
produced from the November 9,
1982, and December 19, 1988,
Landsat TM data. Some barren
land is included in the
Developed/Exposed Land
category
(original imagery courtesy of NASA).
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
a) Change detection map of the
Fort Moultrie, SC, study areas
derived from analysis of
November 11, 1982, and
December 19, 1988, Landsat
Thematic Mapper data.
The natures of the change
classes selected for display are
summarized in
b) the “From-To” Change
Detection Matrix

(based on Jensen et al., 1993a).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Change detection map of the Fort Moultrie, SC, study areas derived
from analysis of November 11, 1982, and December 19, 1988, Landsat
Thematic Mapper data. Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
The natures of the change classes selected for
display are summarized in the “From-To”
Change Detection Matrix (based on Jensen et al.,
1993a).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Post-classification
comparison can be
performed using per-pixel
or object-based image
analysis classification. This
is an example of OBIA
segmentation. Selected
object-based segmentation
of the Bluffton, SC,
Pictometry International,
2007 and 2011 natural color
aerial photography at (a,b)
40 scale, and (c,d) 120 scale
(imagery courtesy of
Beaufort County GIS
Department). Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Post-classification change
detection of Bluffton, SC, based
on 2007 and 2011 Pictometry,
Inc., natural color aerial
photography and object-based
image analysis.
a) Classification of 2007
photography (rasterized).
b) Classification of 2011
photography (rasterized).
c) Union of 2007 and 2011
classification maps using GIS
map overlay function with
selected “from–to” classes
highlighted.

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Select the Most Appropriate
Change Detection Algorithm
• “Change/no change” information
– Binary Change Detection
• Analog on screen
• Image algebra (band differencing and ratioing)
• Multi-date composite image change detection (e.g.
classification, PCA)
• Thematic “From-to” change information (many)
– Photo- LiDAR- grammetric change detection
– Post-Classification Comparison Change Detection
– Spectral Change Vector Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
– Visual on screen detection
 Schematic
diagram of the
Spectral Change
Vector Analysis
change detection
method

Possible change sector codes for a pixel measured in three bands on two dates.

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Diagram of change detection
using an ancillary data source as
Date 1

(imagery courtesy of Richland County GIS

Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Diagram of change detection
using a binary change mask
applied to Date 2

(imagery courtesy of Richland County GIS

Department).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
a) Panchromatic orthophoto of Sullivan’s
Island, SC, obtained on July 1, 1988,
prior to Hurricane Hugo. The data
were rectified to State Plane
Coordinates and resampled to 0.3 ×
0.3 m spatial resolution.
b) Panchromatic orthophoto obtained on
October 5, 1989, after Hurricane
Hugo.
c) Change information overlaid on
October 5, 1989, post-Hurricane Hugo
aerial photograph, Sullivan’s Island,
SC. Completely destroyed houses are
outlined in red. Partially destroyed
houses are outlined in yellow. A red
arrow indicates the direction of houses
removed from their foundations.
Three beachfront management setback
lines are shown in cyan (base, 20 year,
40 year). Areas of beach erosion are
depicted as black lines. Areas of beach Jensen, J. R., 2016, Introductory Digital Image
accretion caused by Hurricane Hugo Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
are shown as dashed black lines.
 Visual on-screen
change detection
a) Panchromatic orthophoto of Sullivan’s
Island, SC, obtained on July 1, 1988,
prior to Hurricane Hugo. The data were
rectified to State Plane Coordinates and
resampled to 0.3 × 0.3 m spatial
resolution.
b) Panchromatic orthophoto obtained on
October 5, 1989, after Hurricane Hugo.
c) Change information overlaid on
October 5, 1989, post-Hurricane Hugo
aerial photograph, Sullivan’s Island,
SC. Completely destroyed houses are
outlined in red. Partially destroyed
houses are outlined in yellow. A red
arrow indicates the direction of houses
removed from their foundations. Three
beachfront management setback lines
are shown in cyan (base, 20 year, 40
year). Areas of beach erosion are
depicted as black lines. Areas of beach
accretion caused by Hurricane Hugo Jensen, J. R., 2016, Introductory Digital Image
are shown as dashed black lines. Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
a) GOES-12 image of Hurricane Katrina
making landfall on August 29, 2005,
overlaid on MODIS true-color
imagery (image courtesy of USGS
Coastal and Marine Geology
Program).
b) Vertical high spatial resolution aerial
photograph of the Grand Casino in
Gulfport, MS, on August 29, 2005
(image courtesy of USGS Coastal and
Marine Geology Program).
c) Low oblique aerial photograph
documenting the location of a part of
the Grand Casino now lying on
Highway 90 (photography courtesy of
NOAA Remote Sensing Division).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
A portion of the Aral Sea shoreline
recorded by the Landsat Multispectral
Scanner (MSS) and Landsat Enhanced
Thematic Mapper Plus (ETM+) in 1973,
1987, and 2000

(imagery courtesy of NASA Earth Observatory).

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Atmospheric Correction for Change Detection

• Necessary for:
– Data transformations
• ratio, NDVI,
– Biophysical variables over time,
– Consistency in change–no change variables
– Field data extension thru time
• Unnecessary for:
– Individually classified imagery
– Multidate imagery within a classification (normalize…)
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
The general steps used to
perform digital change
detection of remotely sensed
data.
1. State the Nature of the Problem
2. Significant Considerations for
change detection
3. Remote sensing data processing
steps to extract change
information
4. Perform accuracy assessment
5. Answer questions/distribute
results/report/product
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Perform accuracy assessment

1. Assessing the Accuracy of the Individual

Maps used in a Change Detection Study
2. Assessing the Accuracy of a “From-to”
Change Detection Map
3. Assessing the accuracy of a binary Change
Detection map

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
Assessing the Accuracy of the Individual Maps
used in a Change Detection Study

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
2. Accuracy of a “From-To” Change Detection Map

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
3. Assessing Accuracy of a Binary Change
Detection Map

Jensen, J. R., 2016, Introductory Digital Image

Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
The general steps used to
perform digital change
detection of remotely sensed
data.
1. State the Nature of the Problem
2. Significant Considerations for
change detection
3. Remote sensing data processing
steps to extract change
information
4. Perform accuracy assessment
5. Answer questions/distribute
results/report/product
Jensen, J. R., 2016, Introductory Digital Image
Processing: A Remote Sensing Perspective,
Boston: Pearson Education, Inc., 623 p.
 Change Detection Issues
Things to keep in mind for all of the
algorithms
• Atmospheric effects – may need to correct for
atmospheric differences (and sun angle differences,
etc.)!
• Spatial error – must accurately register all of the
image dates
• Must consider other causes of change besides the
ones you are interested in (e.g. wet vs. dry years, etc.)
 Summary – Change Detection
• Change detection is a very useful application of
remote sensing—used a lot (= good job skill) at many
different scales
• Various methods and data have strengths and
weaknesses…choice depends on the application and
expected results
• Important to control or minimize changes that don’t
relate to the change you are studying (e.g. sensors,
atmospheric effects)
 Tucson – Find the Changes!
January 8, 1983 ± February 9, 2015

0 10 20 Miles

Changes in Land Cover Jan 8, 1983 and Feb 9, 2015

http://www.snre.arizona.edu/ Color composites (SWIR, NIR, Green wavebands) of Landsat Satellite Imagery (30m resolution). Data from USGS