Land cover mapping is a new protocol that we are introducing this year
in Forest Watch. Given the power of image processing in the field of environmental
science, the numerous Forest Watch sites that are participating in different
areas of New England, and the educational importance of the activity, we
think that it will make a great addition to the program. Students and teachers
have also expressed a great deal of enthusiasm and interest in remote sensing
and Multispec
in the past, so it seems like an appropriate direction to lead them.
What is Land Cover Mapping?
Land cover mapping is the assessment of different land cover types over
a certain geographic extent. There are many ways of doing this. Field workers
can go into the field and actually take an inventory of species and draw
a map showing what species occur in what areas. This is very time consuming
and expensive. Aerial photographs can be used to trace areas of certain
species, or damage classes of trees. This method has been used extensively
in the past. It is much more cost effective and faster than field workers,
but it still has its limitations. In recent years, we have had the benefit
of satellite data and computers to help us better map the extent of various
land cover types. Now, we can get larger areas mapped in less time with
more consistency. With the help of image processing software such as Multispec
we can learn much more about what types of land cover make up the area
that we live and how that land cover has changed over time.
How does it work?
When aerial photo interpreters separate land cover types, they look
through a stereoscope and trace onto a clear mylar that is placed over
the aerial photo. They trace around areas of similar land cover types so
that when they separate the mylar from the photo, they have a sheet with
many different polygons. Each polygon represents a different type of cover,
which they label.
Multispec and other image processing software programs use a system
that does the same type of analysis. When you open a Landsat image, you
see it as a mixture of many different colors that represent different land
cover types such as trees, grass, water, soil, and urban areas. If you
zoom into the image, you begin to see individual pixels. In the case of
the Landsat 7 multispectral images that you will be using, the pixels will
each be 30 meter squares. Each of these pixels has a value that describes
that 30 meter piece of the earth. When you do a classification, Multispec
analyzes each of these pixels and groups pixels together that have similar
spectral properties. When a 10-class classification is done, Multispec
assigns each pixel to one of 10 different groups, depending on what group
best represents the value of each pixel. Multispec can output the results
in the form of a color coded image. The new image would have only 10 colors,
with each color representing an area with similar spectral properties to
other areas with that same color. Multipec can also output text results
such as the percentage of area represented by each group. Figure 1 shows
a Landsat 7 ETM+ false color composite (bands 4,3,2) of Durham, NH. Figure
2 illustrates a 10-class unsupervised classification of the same area,
done with Multispec software. Multispec has separated the image into 10
different colors and assigned an arbitrary color to each cluster.
When the new image is produced, it needs to be compared to the original.
Multispec has done the hard work by separating the types of land cover,
but now you have to tell it what groups are what. If you know that just
by looking at the original image, you can label that group with the appropriate
name. The best way of determining the classification is to go into the
field and check each type of classification. When you have determined what
land cover method is represented by each cluster, you must have a method
of labelling each of the land cover types. The method that we will show
you is called the MUC classification system. This is a standard classification
method that is used throughout the world and provides unique number values
for each type of land cover. It is important to have a standard methodology
so that the results can be compared to other areas.
As you can see, this can be a very valuable tool and is a much faster
method than going out in the field or using the aerial photo methods. It
is quite accurate and you can achieve higher levels of accuracy with better
resolution. Any number of classes may be used as well. Ground checking
is important and although it takes some time to go to each area, it is
much faster than having to go to each spot on the image and map it yourself.
Applications
Land cover mapping is a valuable tool that allows us to see the patterns
that exist across the New England landscape. The patterns that we see are
the products of many years of natural and human influences. The classification
process not only separates the land cover into different classes, but also
can show what percentage of the image is covered by each class. Knowing
how much and what types of cover make up the landscape allows scientists
to predict insect infestations, assess the amounts of different types of
timber, and assess wildlife habitat. Having this data gives people the
power to make decisions about what areas can be utilized in some way and
which areas need to be protected. These are only a few of the many applications
that land cover mapping provides and we hope that students will come up
with their own ideas while having a lot of fun and learning about their
environment.
We presented this activity at the Land Cover Classification workshop
this past summer, as well as the spring workshops and teachers have really
shown great enthusiasm. One of the questions that came up was, "What is
the minimum number of classes that the classifications need to be useful."
Using the Durham Landsat image, we found that a six class image was effective
in delineating classes with enough accuracy. That also makes it easier
for students to correctly field check the classes. Figure 3 shows a six
class image of the Durham, NH subset. Figure 4 illustrates a closer view
of Durham where you can easily pick out boundaries between each of the
six classes. The unsupervised classification was edited to show labels
and more realistic colors for each class.
One great advantage to this classification method is that once the land
cover is determined for a particular area near your school, it can be applied
to the other areas of the image with the same color. In this way, it takes
a small amount of checking to get an accurate assessment of the different
land cover types within the larger image. It would be helpful for students
to find the areas on the image where they live and determine what type
of land cover is present in those areas. When that information is matched
with what others have seen in that same class, it will allow a more accurate
assessment of what type of cover that class represents.
The methodology is the basically the same as that used by the GLOBE
program, so we urge you to check out the information that they have about
the process. If you would like more information about land cover mapping,
we would be glad to provide that to you. Also, if you have comments or
questions about the activity please ask one of us at UNH. We would also
like you to share any ideas about other types of materials, information,
or activities that you would like to see.
Requirements:
Teachers or individuals who would like to participate in this part of
Forest
Watch will need to be trained in the procedures of classification.
If you have experience in the classification process and would like to
participate, we still need to go over the procedures of submitting data
and let you know what products are needed. The following information
gives a general guideline of the procedures and expected products from
each school that participates.
Methods:
Products:
* Optional protocol
please feel free to contact Mike Gagnon
at 603-862-4113 or email this site.