SMILEY: A WEB-BASED REMOTE SENSING DATA MINING SYSTEM

Dr. William Perrizo, Longjun Chen, Dennis Amundson Computer Science
Department North Dakota State University Fargo, ND 58102 {Perrizo,
lchen, amundson}@plains.nodak.edu

ABSTRACT
Remote sensing imagery (RSI) data will be used by potentially
millions of users, provided that there exists a good way to access and
manipulate that data. Currently, the most common way to manipulate and
analyze RSI data is through a GIS system. Usually this method of the
GIS distribution (both the software and the data) is costly and slow.
Furthermore, the cost of owning RIS data is very high. These problems
may limit the application and usage of remote sensing data.  In this
paper, we describe a novel architecture to access, distribute, and
analyze the massive amount of RSI data in order to address the problems
described above. This architecture utilizes the most recent Internet
technology and provide a distributed multi-tier client/server
architecture for accessing and analyzing RSI data. SMILEY, a WWW-based
remote sensing imagery data mining system, provides a general interface
to access, display, and manipulate various sets of satellite imagery
data. Users, through any common WWW browser, can use virtually any
computer platform at any location to do RSI data analyzing.  The
distributed server architecture of SMILEY is scaleable so it can easily
be upgraded to handle an increasing workload thus achieving a high
degree of reliability.


1. INTRODUCTION

The Earth Observing System (EOS) [EOS93] is one of NASA's ongoing
space- and ground-based measurement systems to provide sustained
observations of the earth. The EOS consists of a series of
polar-orbiting and low-inclination satellites.  Each satellite bears
several sensors, for long-term global observations of the land surface,
biosphere, solid Earth, atmosphere, polar ice, and oceans.  The Earth
Observing System Data and Information System (EOSDIS) manages the data
sets derived from NASA's earth science research satellites and field
measurement programs [BAR91]. EOSDIS data is one kind of remote sensing
imagery (RSI) data. Other RSI data include radar data and remote
digital photography.  Initially, it was assumed that EOSDIS data would
be accessed by a few hundred primary EOS earth science investigators
and approximately 10,000 additional researchers to carry out basic
global-change research [VET95]. Since EOSDIS data belongs to the
public, non-research users such as educators, agribusinesses, community
planners, transportation users, and others who want to access it should
be able to do so. Thus, EOSDIS data could potentially be used by
millions of individuals, if a good method exists to access and
manipulate the data.  Currently, the common way to manipulate and
analyze EOSDIS data is through a commercial geographic information
system (GIS) system. A typical GIS system is a standalone software
package which the user purchases for geographical data manipulation.
The data sets, which the user wants to use, may be obtained either with
the purchased GIS system used or bought through third parties. These
methods of distributing the GIS software and data are costly and slow.
These problems may limit the application and usage of remote sensing
data.  In this paper, we provide a novel architecture for accessing,
distributing, and analyzing the massive amounts of remote sensing
imagery (RSI) data to address the problems described above. This
architecture utilizes the most recent Internet technology and make use
of a distributed multi-tier client/server architecture for flexible
accessing and analyzing of RSI data.  The World Wide Web (WWW) is an
attractive, easy to use vehicle that makes data available for
heterogeneous clients. At the present time, WWW browsers are available
for all major operating systems. Web browser provides us a good
platform for one to use and manipulate RSI data retrieved throuh the
WWW. Signature Mining & Interface Language for EOSDIS, Yet-another
(SMILEY) is a powerful online imagery analyzer and viewer.  Figure 1
shows a typical display of SMILEY.



Figure 1. SMILEY screen snapshot

SMILEY provides a general interface from the World Wide Web using
Internet technology to access, display, and manipulate various sets of
RSI data. Eventually, any computer platform from any location on the
Internet may be used to do imagery data analyzing. SMILEY uses the
client/server model to fully utilize the user's client machine's
processing power for achieving quick respond times. Users can also do
data mining operations based on individual pixel values in the image or
apply band-oriented functions.  They can also use many predefined
filters (transfer functions) to process the image being analyzed and
even define their own filter functions.

This paper analyses the structure and functionality of SMILEY.
In section 2, we provide some background information about
SMILEY.  The structure of SMILEY version 1.1 is described in section 3.
The server and client data mining functions provided in the current
version of SMILEY are discussed in section 4. The conclusion along with
some suggested future enhancements of SMILEY are described in section 5.

2. BACKGROUND

2.1 Introduction to Remote Sensing Imagery

Of all data used in spatial data analyzing, one of the most
important is undoubtedly that provided by remote sensing.
Through the use of satellites, we now have a continuing program of data
acquisition for the entire world with time frames ranging from a couple
of weeks to a matter of hours. Very importantly, we also have access to
these remotely sensed images in digital form, allowing for rapid
integration of the results of remote sensing analysis to spatial data
analyzing tools like SMILEY.

Remote sensing can be defined as any process which gathers
information about an object, area, or phenomenon without
actually being in contact with it. Given this rather general
definition, the term remote sensing has come to be associated more
specifically with the gauging of interactions between earth surface
materials and electromagnetic (EM) energy.

Sensors can be divided into two broad groupspassive and active.
Passive sensors measure ambient levels of existing sources of
energy, while active ones provide their own source of energy. The
majority of remote sensing is done with passive sensors, for which the
sun is the major energy source. An example of this type of sensing is
airborne photographs which capture energy in the visible spectrum.
While aerial photography is still a major form of remote sensing, newer
solid technologies have extended capabilities for viewing in the
visible and near-infrared wavelengths to include longer wavelengths
(such as solar radiation) as well. By contrast, active sensors provide
their own source of energy.  The most familiar form of this is flash
photography. However, in environmental applications, the best example
of this type of sensing is radar.

When EM energy strikes a material, three types of interaction
can follow:  reflection, absorption, and/or transmission. For
remote sensing, our main concern is with the reflected portion since
that is what is usually returned to the sensor system. The exact amount
of reflected energy will vary, and will depend upon the nature of the
material and where in the EM spectrum our measurement is being taken.
If we look at the nature of this reflected component over a varied
range of wavelengths, we can characterize the measurement as a spectral
response pattern, which is sometimes called a signature.  A signature
is a description of the degree of which EM energy is reflected in
different regions of the spectrum. Finding distinctive spectrum
response patterns is key to most procedures for computer-assisted
interpretation of remotely sensed imagery. This task is non-trivial.
Rather, the analyst must interpret the right combination of spectral
bands along with the time of year at which distinctive patterns can be
found for each of the information classes of interest. A good tool is
needed to help correctly interpret the RSI data.

2.2   WWW: an Overview

As the popularity of the Internet increases, people are
becoming more aware of its potential. The World Wide Web (WWW,
or the web) is a product of the continuous search for innovative ways
of sharing information resources among heterogeneous machines and
platforms. The implementation of the Web follows a standard
client-server model. In this model, a user on a client machine executes
a program (i.e., a web browser) to connect to a remote server machine
(the web server) where a document store is located.  A typical Web
transaction takes place when the client submits a request to access a
particular document within the store, which the server usually complies
with by sending back a copy of the requested document.  Most documents
on the web are defined using a document markup language called
Hypertext Markup Language (HTML), which allows one to easily add
hyper-links to the text, which link to other HTML documents, as well as
to other types of media (such as graphic images). Web browsers are GUI
in nature. Because of these attributes, the WWW has quickly gained
great popularity among Internet users.  At the present time, web
browsers are implemented on most major operating systems.  The typical
transaction model of the web is shown in figure 2. The World Wide Web
is a client/server application. The server hosts the HTML content and
waits for a request by a client. When the server receives this request,
it will send the requested document back to the client. The client has
a WWW browser that is used to request, receive, and display the HTML
document. Previously, an HTML document consists of mostly static text
and/or image files which could be downloaded from the server when
requested by the client. With the help of Common Gateway Interface
(CGI) scripts, a mechanism that allows WWW servers to execute
special-purpose programs to handle specific requests, HTML documents
can now contain some dynamic information based on the clients special needs.


While HTML and CGI were intended to add simple user
interactivity to WWW server sites, they are also an important
means of providing the user a simple GUI-like interface to a
distributed application without having to install additional software
into the users client machine. However, these tools are not enough to
handle complicated data- mining tasks. The limited layout control for
the GUI and the lack of continuous execution state on either the client
or server sides indicate a need for more sophisticated set of tools.

2.3 Java: a Brief Overview
Java is a new object-oriented programming language. Originally
developed by Sun Microsystems for hand-held electronic devices,
Java was engineered to be small and platform-independent, making it
well suited to apply to the Internet environment. Java itself is
object-oriented, implemented especially for network computing
applications. Java has many useful features. An important one is that
Java source code can be compiled into a machine-independent format,
which consists of the set virtual machine instructions and symbolic
data, commonly referred to as byte-code. Upon execution, a Java
interpreter acts as the virtual machine and will interpret this
byte-code. Most sophisticated web browsers have Java interpreters
embedded within them. Thus, Java byte-code can be downloaded along with
the HTML file to a web client machine and be executed locally within
the clients machine.

2.4 Application domain of SMILEY SMILEY has a very wide application
domain. It can be used in various scientific fields, such as precision
agriculture [BRA74], environmental protection, volcanic activity
analysis, Earth surface analysis, Earth science education, etc. Various
kinds of data mining can involve tools and techniques such as neural
networks, genetic algorithms, expert systems, database filters, etc.  -
Precision Agriculture A large store of land-process data is collected
from the LANDSAT series of earth observing satellites (part of EOS),
and is currently stored at the USGS EROS Data Center located in South
Dakota.  Agriculture scientists and farmers can use SMILEY to process
data, track site-specific farms, and monitor crop yields. SMILEY is
gaining visibility within many precision agriculture and agri-business
concerns.  - Education The SMILEY analyzer and satellite imagery viewer
will be used in several teaching and learning projects.  The Upper
Midwest Aerospace Consortium (a five-state effort to provide leadership
in the areas of Earth System Science) will be using it in its K-12
education component. In that component, school children and young
adults will be able to view and manipulate images showing various
places of interest.  They will be able to do a wide range of
interesting projects with the tool.  Some of these include: viewing
their farm for vegetation quality during the growing season, searching
for grasshopper infestations, watching the spring surface water
situation for flooding, monitoring wetlands drainage, as well as a
number of other projects. These possibilities are limited only by the
imagination of the users.

3. SMILEY ARCHITECTURE

SMILEY which stands for Signature Miner & interface Language
for EOSDIS, Yet- another and is an online tool for viewing and
analyzing remote sensing imagery data.  Using current Internet
distributed processing techniques (such as Java and the WWW), SMILEY is
being designed for the use of many types of remote users, who could be
operating on one of many types of different system platforms at any
place in the world.

In general, the SMILEY architecture extends the normal World
Wide Web architecture using the Java-enabled classic
client/server paradigm. SMILEY fully utilizes the advantages of
distributed computing, by taking advantage of the user's client
machine's processing power to achieve good response times. When a user
accesses the SMILEY- enabled web site using a Java capable web browser
(like Netscape or Microsoft's Internet Explorer), the browser will
download a number of SMILEY applets into the user's client machine. The
applets then contact the SMILEY server for a specific remote sensing
imagery data set. The SMILEY server retrieves the required data set
either from local disk or remotely from a SMILEY data server through a
dedicated network. Before transferring imagery data to the client, the
SMILEY server will cut out the snapshot needed by the client and
pre-format imagery data for easy client handling. The applets perform
image analysis and data mining functions with a user-selected view of
the image on the client host. This provides performance directly
relating to the processing power of the client machine.  Currently,
most PCs running in typical businesses location and at individuals
homes have the power and memory needed to run all functions that SMILEY
provides. The whole structure of SMILEY is shown in Figure 3.


Figure 3. SMILEY (V1.1) system architecture

There are two kinds of main components which constitute SMILEY. One set
of components are passive in nature (stores of data and procedures),
while the rest are active.  The two passive components are:  - Data
sets of satellite imagery (data store).  This data can consist of TM
data sets or other type images, stored both locally (with respect to
other components) as well as remotely (being accessible through a
broadband network). Because the volume of the data store can be so
massive in nature (one uncompressed band of a TM scene requires 44MB of
storage, and normally one TM scene consists of 7 bands), it is likely
that most data will be stored away from the active components. The data
is stored in a hierarchical fashion, in descending order of
popularity.  This allows hot (i.e.  heavily used) data to be stored
relatively close to the active components for quick access. Data that
is relatively cool in nature would probably be stored at a distant
location, such as an archive host accessible via a high-speed network
or a near-line jukebox of persistent storage.  - Set of stored
procedures (proc store).  The main functionality of SMILEY is
implemented as a set of procedures written in Java. These, along with
other additional miscellaneous documents (like on-line help pages) and
procedures, are stored in the proc store.  The active components are:
- Remote WWW browser (or web browser).  It is expected that the remote
user will have on his/her local system a web browser that is capable of
executing the applets composing SMILEY (from the proc store). At the
present time, most graphical web browsers have an embedded Java
interpreter to allow seamless integration of applet execution with
normal web browsing.  - WWW server (or Web server).  A WWW server is
mainly used for transferring hypertext documents (written in HTML) from
a central store to a WWW browser.  The server may also be used to
transfer other types of documents (such as graphic files) and for
execution of procedures (typically called CGI programs) to accomplish
simple tasks (such as authentication, or dynamic file creation). SMILEY
uses the web server for transferring data from the procedure store to
the web browser at the remote site.  When a remote user invokes SMILEY,
the applets that compose its functionality will be downloaded to the
browser via the web server.  - SMILEY server The purpose of the SMILEY
server is to catalog the items composing the data store and to transfer
to remotely executing applets the necessary remote sensing data
requested by them.  Many times, items within the data store are not
formatted as the applets would like them. As a result, the SMILEY
server needs additional functionality to perform simple
graphics-oriented tasks (such as clipping and rotating of images,
filtering out one or more bands of the original images, etc.)  The
complexity of this server varies directly with the complexity of the
data store.

4. SMILEY SERVER AND CLIENT FUNCTIONALITIES

SMILEY is programmed using Java for its primary language. As
discussed in section 2.2, Java is an object-oriented language.
It is platform-independent and designed for purposes of network
computing.

Figure 4. SMILEY (V1.1) software structure

A diagram representing the general functionality of SMILEY is
shown in Figure 4.  The functionality can be divided into two
parts: client-side processes and server-side processes. The clients
main functional purpose is to do image data mining processing and
result visualization; the servers main purpose is to process user
registration information and to process image data requests initiated
from client processes. At present, most client machines have very
powerful processing abilities.  To take advantage of this fact, SMILEY
distributes its data mining calculation and visualization processing
onto the client, in order to achieve faster response times, and to
decrease the load on the server.  4.1 SMILEY Server Processes The
SMILEY servers are responsible for accepting client requests,
maintaining image data indices, retrieving and pre-formatting image
data, and sending image data back to the requesting client. The
functionality of the SMILEY server can be divided into three parts:
request handling and dispatching, data indices managing, and data
retrieving and pre- formatting. These functions do not necessarily need
to be implemented on the same machine.  Dividing up the server
functionality gives more flexibility for the server to cope with an
increasing workload.

4.1.1 SMILEY Server The SMILEY server uses the concurrent server
technique; it can serve an increased workload with better efficiency
than one single server can. The SMILEY server acts as a task
dispatcher, as shown in Figure 5.  When the SMILEY server receives a
client request, it will check the request against its internal dispatch
table. When it finds the appropriate data server or other server (such
as index server for browsing request) to answer this request, it will
dispatch the task to this server and make itself available to handle
next request.

4.1.2 SMILEY Index Server
The SMILEY index server is a special server that serves the
clients browsing requests. In the SMILEY index server, the
server keeps all data sets indexed in a hierarchical tree structure as
shown in Figure 6.


Figure 5. SMILEY server task dispatch diagram


Figure 6. Data index hierarchy tree

All supported remote sensing data types are represented as top-level
branches within the tree. Under these, all available imagery data sets
are stored in separate sub-tree.  With this organize of data, combined
with SMILEY clients index browser rendering function, the user is able
to find a particular imagery set quite easily. This structure is
flexible in nature, which makes it easy to organize data within the
server.

4.1.3 SMILEY Data Server One SMILEY data server is designed to serve
one specific imagery data type.  Because each remote sensing data has a
different format and structure, and furthermore, each data set in the
same data type has different parameters, procedures serving each
specific data type are designed to increase flexibility, scalability,
and maintainability.  When a SMILEY data server receives a request from
the SMILEY server, it will analyze the request and retrieve the header
file of the required data set. From the header file, the data server
can retrieve the necessary information needed to handle this data
request. After analyzing the header file, the SMILEY data server then
retrieves the imagery data needed to fulfil the request (either the
full iamge or snapshot), and then formats it into what the SMILEY
client requires. The resulting image is then transferred to the SMILEY
client that requests the data.

4.1.4 Typical SMILEY Server Processing Scenario Summary When the SMILEY
server receives a data request from a SMILEY client, the SMILEY server
will parse that request. If the request is a browsing request, the
request is the dispatched to the SMILEY index server. Otherwise, the
request is dispatched to a corresponding SMILEY data server.  When the
SMILEY index server is activated, it will parse the request, retrieve
the indices which the SMILEY system currently provides, reformat the
index entries or viewing purpose, and transfer them back to the SMILEY
client.  When a SMILEY data server receives a request, it will first
parse the request. If the request is for a full-screen request (which
is used to display a overview of the whole image set), the data server
will reformat imagery data according to the user screen size and
transfer it back to client after doing some additional formatting. If
the request is for a snapshot request, the data server just retrieve
the detailed data and transfer it back to client.

4.2 SMILEY Client Processes

4.2.1 Image Data Browsing & Selection

SMILEY version 1.1 supports TM, SPOT, AVHRR and digital photography
(TIFF) data set formats. There can be many imagery data sets for each
format taken for the same or different places at different times. This
process enables the user to select a data type and browse the entire
data store.  As shown in Figure 7, users can click to see information
about the data set, such as image size, its geo-location, the date the
image was taken, etc. This information helps the user select the data
set that he is interested in.

Figure 7. Image browsing and selection screen

4.2.2 Snapshot Selection & Creation Many imagery data sets are very
large in size. For example, a typical TM image scene covers
approximately 110 miles by 110 miles of area and consists of
approximately 6200 by 6200 pixels with a resolution of 28.5 meters per
pixel. This size is too big for individual processing needs, since most
users will only care about their own farms or towns.  Based on the
position of a sliding window, which the user moves, SMILEY will make a
snapshot of the image scene before transferring it back to the client.
Not only does this process help save on network bandwidth and reduce
response time, but the processing requirements of the client are also
greatly reduced. A screen dump of this process is shown in Figure 8.

Figure 8. Snapshot selection and creation screen

A typical snapshot is 250 by 250 pixels in size, which covers an area
of about 56 square kilometers using TM data. All the data mining and
visualization functions are performed on the snapshot. Thus, the main
requests that the client sends to the SMILEY server are snapshot data
requests. When the user selects a snapshot area, the client will send
the data request to the server, the server will then process the
request, pass it along to the data server. After retrieving the
snapshot data, the SMILEY server sends the requested data back to the
client.  4.2.3 Transfer (image filter) Functions Transfer functions are
a specialized set of image filtering/visualization functions.  The
SMILEY system predefines several often-used image filters: linear
filter, band filter, hybrid filter, etc.  SMILEY also provides a
user-defined filter interface, so that the user can visually define his
own filter. By applying filter functions to the image snapshot, the
user can perform interactive data mining on this image and visualize
the results. The user can also use transfer functions to find data
signatures before applying band-oriented or pixel- oriented data mining
functions and visually see those results.

Figure 9. An sample screen of the transfer function interface

4.2.4 Band Mining Functions Band mining functions are used to do data
mining based on band data. As discussed in Section 2, bands of the TM
image are selected to maximize their capabilities for detecting and
monitoring different types of Earth resources. The user can use
predefined formulas to explore a variety of information contained in
the image snapshot.  These functions also provide an interface for the
user to enter his own formulas to generate a virtual band and visualize
it. As shown in Figure 10, the VI formula (G-B) uses TM band 4 as G
(Green) and band 7 as B (Blue) to represent the vegetation index in the
region. After calculation, SMILEY can visualize the results.

Figure 10. An example screen of band mining functions

4.2.5 Pixel Mining Functions Pixel mining functions are used to do data
mining to detect some specific digital signatures within the image.
These functions provide the major support of the data mining
capabilities of SMILEY. The signatures allow the user to detect
specific characteristics from the image snapshot which could represent
some phenomenon in the land being sensed.  To define a signature, the
user selects a specific value, as well as a tolerance, for each band
constituting the image.  After the user specifies the signature, SMILEY
will find the pixels that match the signature, and visualize the pixels
after performing some additional operators (such as merge or isolate).
SMILEY will also calculate various statistics of the mining operations,
such as percentage of pixels within the image snapshot that matched the
signature. Figure 11 shows an example of the pixel mining function
screen snapshot in SMILEY.

Figure 11. Pixel-oriented data mining screen in SMILEY

4.2.6 Geo-statistical Estimation Functions Any type of remote sensed
measurement has some resolution due to the hardwares physical
limitations and the large area usually under consideration. Some
statistical procedures [REK87] can help to achieve improved and more
efficient measurement analysis. The Kriging algorithm is a type of
statistical method used to predict unknown points based on the known
measurements using the regionalized variable theory. It is basically an
interpolation process using moving-weighted averages. With kriging, the
resolution of spatial data can be enhanced. SMILEY Version 1.1
implemented the semi- variance function as shown in figure 12.


Figure 12. SMILEY geo-statistical implementation

4.3 Scalability and Reliability Consideration
	Currently, SMILEY (Ver. 1.1) is implemented at North Dakota
	State University (NDSU) on a UltraSparc workstation running
Solaris. All server components (the web server, SMILEY server, SMILEY
directory server, and SMILEY data server), the data store, and the proc
store are implemented on the same system. When client requests increase
in number, or the data store increases in size, the performance of the
server will degrade. To solve the performance limitation, a distributed
server architecture can be used.  When handling a large WWW load, the
processing will be distributed over a number of low-end systems,
interconnected by a high-speed dedicated network, as shown in Figure
13. This architecture will also provide higher reliability than a
single server architecture could by using redundant servers.


Figure 13. Distributed architecture of SMILEY

5. CONCLUSION
In this paper, we present the system architecture of SMILEY, a
WWW-based remote sensing imagery data mining system that is
currently available on the Internet (http://smiley.cs.ndsu.nodak.edu).
SMILEY uses current Internet techniques such as Java and the World Wide
Web to explore a novel method in accessing, distributing and
manipulating massive remote sensing imagery data. It achieves platform
independence and easy of use. The client/server model is used in SMILEY
to achieve quick response times.  The current implementation of SMILEY
is numbered version 1.1. In this version, the implementation is a Java
applet which can only be used on the Internet. The data sets to be
analyzed are all stored online in the SMILEY server. The storage
capacity of the server and the variety of the data is limited by the
server that we are using. The geo-reference system in the current
SMILEY system is also pre-mature, which will limit the usage of SMILEY
for the time being.  As third party GIS systems evolve, the
functionality provided by them will also increase. To utilize the power
of these systems and to avoid re-inventing the wheel, SMILEY will need
to provide a two-way interfaces with other commonly used GIS systems,
such as ARC INFO/VIEW, MAP INFO, or IDRISI. These GIS systems have
already provided a very good geo-reference system which can be used by
SMILEY. What GIS systems lack is a rich collection of data mining
tools, which is exactly what SMILEY provides. With the interface,
SMILEY can take advantage of the functionality of existing GIS systems.
Users will be benefited with this type of interaction between SMILEY
and other common GIS systems.

Limited by Javas security features, the current SMILEY system
cannot access any data stored locally on a client system (i.e.,
the data which is stored in a users hard disk or CD-ROM). This may
limit somewhat the usage of the SMILEY system. Two techniques are under
review to solve this problem: a standalone SMILEY system and a plug-in
add-on to SMILEY. Both techniques will solve the local disk-accessing
problem; but both will have pros and cons. The stand-alone technique
will lose all advantages of the applet version, such as easy
maintenance, upgrading, and centralized control. The plug-in technique
will keep the all applet advantages but will bring in multiple binary
executable code problems. When considering what the main motivation was
for developing the current SMILEY system (to explore a novel
architecture to access, distribute and manipulating remote sensing
imagery data over Internet), the plug-in technique seems likely to prevail.

REFERENCE:

[BAR91] B. Barkstrom, A preliminary EOSDIS user model, NASA Langley
Research Center, 1991

[BRA74] E. J. Brach, Measurement of agriculture crops by remote
spectral techniques, Remote Sensing Earth Resources, Vol. 3, 1974

[BUR80] Burgess, T. M., and R. Webster, Optimal interpolation and
isarithmic mapping of soil properties. I. The semivariogram and
punctual krigging.  Journal Soil Sci. 31:315-331, 1980

[DAV77] David, M., Geostatistical ore reserve Est. Elsevier Publ, NY, 1977

[DEE75] Deering, D. W., Rouse, J. W., Haas, R. H., and Schell, J. A.,
Measuring forage production of grazing units from Landsat MSS data.
Proceedings of 10th International Symposium on Remote Sensing of
Environment, II, 1169-1178, 1975

[ELE92] F. J. Eley, R. Granger, and L. Martin, Soil moisture: Modeling
and monitoring for regional planning, National Hydrology Research
Center, Saskatoon, Saskatchewan, 1992

[ELK94] M. Elkington, R. Meyer, and G. McConaughy, Defining the
architectural development of EOSDIS to facilitate extension to a wider
data information system, Hughes Applied Info System, EOSDIS Core Syst.
Proj Tech. Paper 194-00131. Apr. 1994

[EOS93] EOS Ref. Handbook, National Aeronautics and Space
Administration, Earth Science Support Office, Washington, DC, Mar. 1993

[HUE88] Huete, A. R., A soil-adjusted vegetation index. Remote Sensing
and the Environment 25, 53-70, 1988

[JAC83] Jackson, R. D., Spectral indices in n-space. Remote Sensing and
the Environment 13, 409-421, 1983

[JOU78] Journel, A. G. and Ch. J. Huijbregets, Mining Geostatistics,
Academic Press, New York, 1978

[JUE97] P. Juell, Implementation Issues of NDSU WWW server, 5th Wksp on
Enabling Tech:  Infrastructure for Collaborative Enterprises, Stanford,
Cal., Jun. 1996

[KRI66] Krige, D. G., Two-dimensional weighted moving average trend
surface for ore evaluation. Journal S. Afr. Inst. Min. Metall.
66:13-38, 1966

[REK87] R. E. Knighton and R. J. Wagenet, Geo-statistical Estimation of
Spatial Structure, Continuum Water Resources Institute, Vol.1 Jul.,
1987

[RON97] Ronald J. E., Idrisi Users Guide, Version 3.0, Clark Labs for
Cartographic Technology and Geographic Analysis, June 1997

[ROU73] Rouse, J. W. Jr., Haas, R. H., Schell, J. A. and Deering, D.
W., Monitoring vegetation systems in the Great Plains with ERTS. Earth
Resources Technology Satellite-1 Symposium, Goddard Space Flight
Center, Washington D.C., 309-317, 1973

[TML96] EROS data center, Thematic Mapper Landsat Data, available on
Internet at URL : http://edcwww.cr.usgs.gov/glis/hyper/guide/landsat_tm

[VET95] R. Vetter, et al, Accessing Earth System Science Data and
Applications Through High-Bandwidth Networks, IEEE Journal on Selected
Areas in Comm, V13:5, June 1995

Partially supported by grants, NSF No. OSR-9553368 and DARPA No.
DAAH04-96-1-0329.