A spatial analysis and
modeling system of solve environmental problems
C. H. Vermillion, F. L. Stetina and J.
Hill
NASA (USA)
Background
Through
our economic and technological activity, we are now contributing to
significant global changes on the Earth within the span of a few human
generations. We have become a part of the Earth System and one of the
forces for Earth change.
Research holds the key to a deeper
understanding of the Earth as a integrated system of interacting
components, and of the consequences of global change for humanity. To
achieve this understanding, we need a new approach to Earth Studies- Earth
System Science-which builds upon the traditional disciplines, but promises
to provide a deeper understanding of the interactions that bind the
Earth's components into a unified, dynamical system. Fundamental to this
new approach is a view of the Earth System as a related set of interacting
processes operating on a wide range of spatial and temporal scales, rather
than as a collection of individual components. The goal of this new Earth
System Science is to obtain a scientific understanding of the entire Earth
System on a global scale by describing how its component parts and their
interactions have evolved, how the function and how the ma be expected to
continue to evolve on all time scales.
The challenge to Earth
System Science is to develop the capability to predict those changes that
will occur in the next decade to century, both naturally and in response
to human activity. Complimenting our innate curiosity Complimenting our
innate curiosity about our planet, the search for practical benefits to
improve the quality of human life continues to provide an important
motivation or Earth science. The problem is that the global changes cannot
readily be distinguished from the results of natural change on the same
time scale. We require a set of Earth Observations that will permit us to
disentangle the complex interactions among the Earth's components and to
document their effects over extended time periods. such observations will
allow us to establish casual relationships among the processes involved
and therefore to distinguish between the consequences of human economics
and technological activity, on the one hand, and the results of natural
ledge, we will then be able to take timely action to ensure an abundant
Earth for future generations.
we can begin to meet this challenge
today:
- Programs of global observations relevant to a number of Earth
system properties have already been carried out with great success.
- Global Vegetation Index
- Sea Surface Temperature
- Ocean Color
- Global Weather/Global Cloud Types
- Earth Radiation Budget
- Global Weather Experiment
Future Missions
- Ocean State/Currents
- Tropical Rain Measurements
- Earth Observing System.
- Information Systems specifically constructed to process
individual sets of global data are already in operation. New
developments in computing technology have now made feasible an advanced
information system to provide worldwide access to more extensive global
data to be obtained in the future, and the facilitate data analysis and
interpretation by the scientific community.
A new such
network-WETNET is discussed in another presentation during this
conference.
- A worldwide political awareness of the necessity for a
coordinated, international approach to the global study of the Earth has
been created, and cooperative research efforts b many nations across the
globe are underway.
To facilitate this cooperation, NASA has
developed a Spatial Analysis and Modeling System which allows easy
exchange of data; these software systems also allow the assembly of
information essential for effective decision-making for economic
development, emergency preparedness, and natural resources planning and
management. Thus the software which facilitates problem solving on local
and national levels also extends to regional and global scales without
design changes ad utilizing a realistic multidiscipline approach.
Introduction to the NASA spatial and modeling
system
The need and usefulness of spatial information is crucial in
all applications.
Experts estimate that almost 70 percent of all
information is location-based. Spatial data, thus, this a unique role in
the design and analysis of development policies and options. Since
verifiable information is scarce and expensive, it is important to have a
spatial information system which effectively handles all relevant spatial
data. It should allow for the comparison and integration of data from a
variety of satellite and ancillary sources, varying map scales, and fields
map scales, and fields of inguiry. Outputs of spatial data analyses may be
fed back to the system for further use in their applications, spatial
simulation models need to be integrated with the various data types. The
entire system complexity is exponentially increased by the varying data
types, scales, models and data structures. Expert systems are finally
needed to relate the rich fabric of the comprehensive spatial information
and modeling system to the needs of decision makers in various departments
and disciplines (discipline experts).
Spatial data comes from many
sources. Data sources include remote sensing systems, analog maps, digital
maps and ancillary ( textual ) data. One should not the efficacy remotely
sensed data in largely unmapped, rapid growing and/or inaccessible
regions. The launching of satellites and development of various imaging
sensors ( i. e., multispectral scanners, thermal radiometers, radar) have
added a wealth of additional economically acquired and qualitatively
improved spatial, spectral, and temporal data. Ancillary, non-spatial data
provide attributes of a spatial entity such as the function of a plant,
the permits issued, the population of a city or the type of materials in a
dump. Ancillary information can readily be processed by standard data base
management systems in current use, the spatial data are not so readily
handled.
The term geographic information system (GIS), is often
used to describe these systems. Most geographic information systems. Most
geographic information systems, however, are designed for map data. The
most common data source for geographic information systems has been the
analog map. For this reason most of these systems are organized to
effective process analog map data and or link to spatial, simulation
models.
Hence, we use terminology Spatial Analysis ad Modeling
System (SAMS) for our system which includes remote sensing and image
analysis functions, simulation models as well as the mapping functions
ascribed to geographic information systems. It must be understood that all
the components: image analysis, modeling and information storage and
retrieval must effectively function as a system. SAMS tools are made
accessible by recent advances in the acquisition, interpretation and
synthesis of data.
Since data is the primary factor in determining
the structure and functions of varied spatial analysis systems, the
discussion below examines systems particularly with respect to spatial
data needs, SAMS must integrate the characteristics the two major types of
spatial systems: largely vector-based Geographic Information Systems
(GIS). The discussion below highlights the pertinent features of body
systems that are critical in the design the SAMS. The discussion
appropriately begins by listing the important attributes of GIS.
- Geographic Information and Image Processing Systems
GIS
systems have become used on s wide basis in recent ears. This has
occupancy because of advances in computing technology and GIS software.
Present systems provide:
- Map Digitizing and Editing: This includes changing scales
and projections. Joining separate map sheets and correcting for map
distortions.
- Ancillary data Entry and Management: The map features may
have attributes such as ownership, stream flow, etc. This allows one
to manage nonspatial data.
- Map production: This provides for output on plotters &
displays. Scale changes and various projections are supported.
- Analysis Function: Substantial analysis are supported such
as area, perimeter and distance calculations. Vehicle routing and
facilities sitting are also supported.
- Statistics: Statistical functions such as means, histograms
and multivariate analysis are supported.
- Data Management: This supports storage and retrieval of
spatial and associated data.
- Functions of SAMS
Various functions or capabilities that
need to be integrated for the development of an operational Spatial
Analysis and Modeling Systems (SAMS).
- Database Structures
The interaction of algorithms and
data structures in spatial systems is a complex issue and an important
area of research. Spatial databases tend to be vary large. The queries
made of the database consists of those common to other databases, but
also include queries that relate to distance, containment or
connected-to. There may be a large number of possible relationships
that can exist. The systems must effectively handle spatial
relationships. The data in a database are composed of entities and
relationships between the gentilities. The complexities in system
comes from the manner in which the relationships are represented.
Present systems use either tree network or relational schemes for
organizing databases. Relational Systems seems, the most promising. In
these systems seems, the most promising. In these systems,
relationships are given by tables. A query language is used to
describe relationships which are used in retrieval functions.
Most GIS's utilize vector data-structures. A GIS will have
algorithms suitable for the data structure. This means that most
systems have vector based algorithms. Image processing is very rich in
algorithm development, but has been oriented toward the raster data
structure.
Remotely sensed data re collected in raster format.
The output of complicated processing of pixel information might be the
input to SAMS. A problem is that currently the image processing
classification algorigthms are not accurate enough for input to a
SAMS. It is of course true that the SAMS can be used as a knowledge
base to create from the updating capabilities of remotely sensed data.
Significant benefits will derived from an integration of SAMS and
remote sensing and image processing systems.
- Vector-Raster Capabilities
SAMS effectively handles the
two major data formats amongst others: raster and vector data. The
data entities commonly utilized in spatial system are: points, lines
and polygons, and pixels. Points, lines and Polygons are readily
represented by sequences of x- pairs. These are known as vector data
and are easily processed b graphic output devices and digitizing table
input devices. This format also provides a compact way to store the
data. Vector format is used to encode a wealth of available data
including many geographic features, like political boundaries, that
cannot be remotely sensed. A vector is a straight line usually defined
by specifying the geographic ordinates of its endpoints. Many short
vectors can be laid end to define curves, or they can be arranged to
form closed polygons. Such vector groups are usually assigned a number
representing some attribute o the area enclosed by a polygon or the
path defined by a vector line tracing. Vector images such as contour
maps or highway maps consist of sets of named vector groups organized
for rapid access or storage efficiency. The resolution of a vector
image is defined by the accuracy of the vector endpoints as well as
the accuracy with which curved boundaries are approximated by straight
line segments. Vector data sets typically range in size from several
kilobytes up to a few megabytes per image.
Pixels or gird data
are image data and are normally colleted by remote sensing systems
(i.e., multispectral scanners, radar, video cameras). These data are
called raster data and are readily processed by raster based image
area into a regular grid whose resolution many vary from cells sizes
of several square kilometers down to less than one square meter each.
Each grid square is represented b a number that signifies some
characteristics of the imaged piece of land. The collection of numbers
from all grid squares forms a rectangular array are known as a digital
image. Raster images can have multiple channels where each channel has
a separate array of numbers, with their sampling grids assumed to be
perfectly matched. Raster data sets are typically quite large, with
image sizes in excess of 24 megabytes becoming increasingly
commonplace.
The raster and vector representations of spatial
data have both their strengths and weaknesses such that neither data
type can adequately replace the other. Raster data are easily acquired
by remote sensing techniques and are available in prolific quantity on
a near real-time basis. The raster format has the advantage that the
data are stored in the computer in a manner that preserves spatial
geometry. Neighboring pixels are neighbours on the earth's surface.
This format is convenient for man tasks, especiall image analysis and
some simulation tasks. Raster images, however, require large amounts
of storage and cannot reliably resolved objects that are smaller than
the image grid cell size. Vector images, on the other hand, can afford
a very high spatial resolution due to the greater storage efficiency
of the format. Most GIS systems process vector data and most image
analysis systems process raster data.
- Raster-Vector Conversions
The potential of merging the
two technologies, GIS and remote sensing, has been recognized for
several years. From the GIS side, the remote sensing data represent
an important source of data, which are easier to input, relatively
up-to-date, and can be available on a continuous basis. This helps
to release the major bottleneck of developing a large-scale GIS: The
problem of digitization and inputting data into the computer. From
the remote side, GIS represents an excellent approach to handling
and analyzing the large amount of spatial data. In addition, the
conventional map data contained in most GIS provide another
dimension of information which remote sensing data lack, such as
county or census tract boundaries and socio-economic data.
Despite its great potential, research and development in the
integration of these two technologies is still limited due to the
many technical as well as theoretical problems involved. Among the
technical problems are: different computer hardware and software,
different data storage formats and different procedures for handling
and analyzing the data. These technical problems are further
complicated b the accuracy of raster-vector conversions, conversions
between different data structures, and the reliability of the map
overlay and interpolation procedure must be addressed before
accurate interpretation of the results from the integrated GIS is
made.
- Vector and Raster Processing Systems
Integration of
vector and raster displays involves more than simply drawing vectors
on the raster display device. The software must keep track of the
data residing on the screen. Physically the screen tracking data
must be centrally located and readily accessible to all application
programs. A library of data management routines must also be provide
to assist applications in navigating and updating the complex
tracking structures.
Many of the standard vector graphics
systems define textual equivalents of graphic device commands called
graphics metacode. These metacode instructions allow drawings to be
saved in ordinary text files and later redrawn using a standard
metacode interpreter program. As such, the metacode description of a
drawing contains all information necessary to perform display screen
tracking of vector images. To accommodate raster image tracking, the
metacode could be augmented with special image description code
which contain the information necessary to regenerate the image on
the screen. Such augmented metacode would provide a uniform data
structure capable of tracking both raster and vector data in a given
screen presentation. An added advantage to such a metacode based
tracking system is that the textual metacodes are human readable
and, thus easy to debug.
- Simulation Models
The running of predictive models is one
of the main attractions offered by SAMS. These models are often quite
complex, each representing ma man-years of development, and will usually
be garnered from a wide variety of sources. Each research site of
interest will generally have a different format for the required data.
The data format will often fundamentally affect the processing strategy
of the modeling software. When importing a new model from an outside
site, it is preferable to modify the in-house data to fit the models
requirements rather than modifying the model to fit the data. The most
widespread of such data format conversions will occur between raster and
vector format data.
Data in the spatial data bank will be
acquired from a wide variety of sources, all offering very little choice
in the format of their data. To ensure consistency in the data bank, all
data should be converted to an uniform standard format that is strictly
enforced throughout the data archive. The definition of such a standard
format poses demanding requirements in the data storage strategy. The
standard format must handle raster and vector data types independently,
as the fidelity of either data type is still served by permanent
conversion to the other type. The data management software must be
capable of converting between data types as required to suit each
application. Provisions must also be made to handle non-image data such
as surveyour's notes or tables of toxic agent half-lives.
The
working spatial data bank may contain hundreds of separate images. To
keep track of such a large volume of data, a management system must be
included to support queries on data bank coverage based on the requested
locations or time frames. New image can be ingested into the data bank
by simply identifying them to the data bank manager, which would also
maintain a complete account of the history of each image in the data
bank. Separate data sources within the data bank would be integrated
into the configuration requested for a given application. This involves
automatic registration of images having disparate resolutions and data
types. Such a scheme for automatic data integration could be greatly
enhanced b the enormous storage capacity of laser disk media. Offering
the possibility of maintaining the entire data bank in a constantly
on-line mode.
- Expert Systems
Expert systems are needed to integrate
image processing and GIS's. Both systems tools for integration and
utilization in an effective interactive mode with human operators.
Effective vision systems in the future will have expert system shells in
the implementation. In additation. In addition, the effective analysis
of spatial data requires discipline experts. This knowledge must be
available to the system if effective results are to be achieved. Expert
system tools are needed to incorporate and extract expert knowledge from
discipline experts. The overall system will necessitate expert system
tools in system building.
Expert systems are generally defined
to be programs that perform intellectual tasks. A system of this type
can then give advice to make decisions based on the expert knowledge
available to it. Expert systems have been successfully applied to such
areas as medical diagnosis, geological exploration, petroleum
production, circuit design and computer vision. Expert systems have an
appeal over other forms of analysis because they can explain and justify
their results. This is done be translating the rules and assertions used
to draw a conclusion into a line of reasoning. This allows the personnel
developing and evaluating the system to medical diagnosis, geological
exploration, petroleum production, constantly monitor and evaluate its
performance. A system must have good communication capabilities so that
experts through simple dialogues can examine the reasoning and improve
the system as needed. The advantages of developing an expert system is
that such a system can be: 1) continually improved as more knowledge is
obtained, 2) reproduced easily for other machines, and 3) combined with
other expert systems to build a single system with increased
capabilities.
Expert systems have matured to the point that the
can be applied to spatial analysis problems. Tools are available as a
basis for developing these systems. The variables in our spatial system
are so complex and interwoven that it is unlikely that any other
approach will predict the behavior of these systems.
Expert
systems are frequently constructed from a class of programming languages
commonly referred to as rule-based production system.
A
rule-based system can be (used to integrate the information contained in
maps, aerial images, and demographic data bases with the textual
informations produced b various spatial models. The incorporation of
symbolic processing capabilities with the vector and raster processing
capabilities will be an aspect of this project. These capabilities will
be important to enhance the capabilities of the SAMS. These capabilities
will be utilized interactively with human operators and with image
processing operators as the reliability of the vision operations
warrant.
- Applications Software Modules
Numerous applications
specific software packages have been developed b scientists at GSFC,
government agencies, and universities. These include the following:
- Atmospheric and Oceanic Software
There are 3 major
programs for this area. These are GEMPAK, SEAPAK, and the
international TOVS processing package (ITPP) GEMPAK, and SEAPAK are
products of NASA, and read data from a variety of sources, performs
analysis and image display. They output pots in monitors such as that
on the IIS systems or standard graphics terminals. GEMPAK and SEAPAK
also can be run under the NASA developed Transportable Applications
Executive (TAE). TAE provides the use with a friendly interface to
applications programs. For example, under TAE, GEMPAK, and SEAPAK can
be easily used on a menudriven basis, and the user can conveniently
and interactively choose a variety of options.
GEMPAK is meant
for atmospheric and meteorological applications. It can process,
analyze and display data. It can perform objective analysis using the
Barnes algorithm and derive standard meteorological parameters such as
winds from appropriate data. It can evaluate and plot pressure
temperature profiles for sounding data, and can draw Stuve, Skew-T,
Log P, or vers T graphs. Geopotential heights, potential temperature
and equivalent potential temperature ma be obtained and displayed.
Variables can also be contoured, and wind barbs and streamlines can
also be drawn. Derivatives of data can also be obtained.
These
are done in a interactive and menudriven manner, and the user can
conveniently and flexibly choose options. GEMPAK can also run on the
existing system and is compatible with IIS and a variety of DEC
terminals.
SEAPAK is similar to GEMPAK, put is oriented
towards oceanic and coastal applications. Image data such as that from
AVHRR and one Coastal Zone Color Scanner (CZCS) a rocessed, analsed
and displaed. VPAK and also work in conjunction with me IIS image
processing system. Parameters such as sea surface temperatures,
oceanic primary production and sedimentation can be measured.
Dynamical and biological processes can be inferred. Physical processes
such as currents, eddies and instabilities can be studies.
The
ITPP is a product of the National Oceanic and Atmospheric
Administration (NOAA) and the Cooperative Institute for Meteorological
Satellite Studies (CIMSS) and retrieves geophysical parameters from
the TOVS, data. The TOVS instrument consist of measurement in 27
spectral channels in the visible, infrared and microwave regions to
sound the atmosphere. As such atmospheric temperatures, and water
vapor content at selected altitudes can be retrieved b 'inverting' the
radiative transfer equation. In addition, total ozone amount are also
available. These results can be input to GEMPAK for analysis and
display. The availability of temperatures allow for a quantitative
atmosphere. Additional applications software will also be written for
TOVS analysis.
- Land Analysis
The NASA-developed Land Analysis System
(LAS) software will be provided. As with GEMPAK and SEAPAK, this
software can also run under TAE and consists of a comprehensive set of
routines to display and analyze LANDSAT and SPOT data. Currently, the
software is being enhanced to work in conjunction with a geographic
information system. LAS is currently also able to work compatibly with
the IIS system.
LAS can perform a wide variety of data
manipulation, mathematical and geometrical function on images. These
include all commonly required analysis functions.
- Mathematical and Statistical Packages
A numerical
analysis packaged used by US is the International Mathematical and
Statistical Package (IMSL). This package includes a subroutine library
containing all of he commonly needed numerical and statistical
subroutines. Examples are spectral analysis, least squares fitting,
numerical quadruture, solution of differential equations, correlative
analysis and interpolation. The user can write a driver which reads
data and calls the subroutine package to perform the required
functions.
There are also packages which are more interactive
and useroriented. An example is the Interactive Data Language (IDL),
which also can perform some image analysis and graphics.
- Graphics
There are many comprehensive commercial and
government graphics packages that are available. This software provide
for a wide variety of displays, including simple graphs,
three-dimensional drawings, animation and contouring. It is proposed
that, at a minimum, some user-oriented, interactive package such as
IDL be implemented. IDL is also compatible with IIS system. IDL can
also perform much numerical analysis and image processing. If needed,
more sophisticated graphics packages can be obtained and installed.
The latter require more software development. It may be best to
implement both types of capability.
- Data Base Management Systems (DBMS)
It is important to
have data base management and archiving capability to handle all the
data. Among commercial systems that are popular are the products from
Ingress, Oracle and DEC. These are relational databases and provide
for automated catalogs and storage of data for convenient usage.
- SAMS data flow
It is clear from the above discussion that
a flexible spatial data analysis system should include five main
components:
- A data input subsystem for collecting data maps, images, textural
sources, DCPS and other data sources;
- A data storage and retrieval subsystem for organizing and quickly
retrieving the data;
- A data manipulation subsystem model that allows user-selected data
to be aggregated and modified, and;
- A data reporting subsystem for output manipulated data in map,
tabular or image from (Marble 1984);
- Simulation model subsystems for predictive purposes. The data
input subsystem will have digitizing capabilities that will allow maps
to be input using an X-Y unitizing table. The digitizing software will
allow maps of various projections and scales to be mathematicall tied
to the digitizer so that point, line, and polgon data can be manually
traced and entered into existing data bank. The software will also
have facilities for correcting digitizing formats such as the U.S.
Geological Survey Digital Line Graph (DLG) or that obtained from other
mapping system such as Intergraph or Arc-Info will be capable of entry
via phone lines or magnetic tapes. The system will allow
geometrically-corrected digital images obtained from aircraft,
satellites and digitized photographs to be entered via entered via
magnetic tape and textural data can be entered via the keyboard, phone
line, or magnetic tapes.
The data storage and retrieval
subsystem will provide a means of string data in an organized and
efficient manner for later retrieval and manipulations. Retrieval
options will include quicklook capabilities for on-one maps, images, and
text which allow these data to be input into other system ad externals
programs for further analysis.
To be an effective management and
planning tool, a spatial data system must provide user friendly and
efficient means of selectively retrieving data. A typical query of the
data bank would be: "locate within a user-defined polygon all soils of a
particular type within five miles of a specific highway and not
currently under cultivation." This request would require boolean,
polygon, and proximity queries of the data bank. Data query options to
be included in the proposed system will include boolean operations,
distance, and promixity calculations, area measurements, and polygon,
point, and line retrieval. These operations can be done for area defined
by irregular polygons, circles and rectangles.
The data
reporting subsystem will provide a means to outputting processed data in
a clear and concise manner. Data output will include textual data in
report form of formatted for input into other external statistical or
graphics programs, business graphics such as bar-charts, line-graphs and
pie-charts, and maps output to graphics terminals or pen plotters.
Typical map generalization routines for coordinate thinning, edge
matching and projection changes will be available.
The analysis
subsystem will have comprehensive functions for georeferencing,
classifying , distance calculations, statistical analysing,, and image
processing functions. The modeling subsystem will have transportation
and plume models. The INGRES data manager runs on many types of
computers. It is the data base system chosen by U.S. Federal Emergecy
Management Agency (FEMA) as a part of the (IEMIS) system. INGRES has the
ability to take in text, properly formatted, and transform it to data
base from whence to data may be retrieved using the INGRES query
language. The SAMS is modeled on the IEMIS software.
- Hardware and Software
SAMS is a software package which was
developed on Digital Equipment Corporation's (DEC) VAX line of
computers, using DEC's proprietary operating software VMS. It is written
in the FORTRAN language and consists of hundreds of thousands of lines
of program code. Color, high-resolution graphics terminals are needed to
output the images to a screen. These terminals should be Tektronix
terminals or SAMS is a software package which was developed on Digital
Equipment Corporations' (DEC) VAX line of computers, using DEC's
proprietary operating software WMS. It is written in the FORTRAN
language and consists of hundreds of thousands of lines of program code.
Color, high-resolution graphics terminals are needed to output the
images to a screen. These terminals should be Tektronix terminals or
SAMS is a software package which was developed on Digital Equipment
Corporation's (DEC) VAX line of computers, using DEC's proprietary
operating software VMS. It is written in the FORTRAN language and
consists of hundred of thousands of line program code. Color
high-resolution graphics terminals are needed to output the images
emulations thereof. PC-based terminals are available as output devices.
Digitizers, plotters, cameras and other peripheral devices can be linked
to the system.
- Summary of NASA - SAMS
The Spatial Analysis and Modeling
System has evolved over a number of years from two important software
packages-NASA's direct satellite reception and analysis systems as
exemplified by the ground processing system developed for
SPARRSO-Bangladesh, also the Regional Severe Storm Warning System
developed for the Fiji Island Meteorological Service and the US Federal
Emergency Management System Integrated Emergency Management Information
System developed to facilitate U.S. disaster response and planning.
Thus, the system is presently suited for applications in emergency
management. Simulation models available include plume generation for
nuclear power plants (adaptable for chemical plume plotting), evacuation
time estimation for point-source and regional disasters), siren
propagation model TYAN cyclone track model and projected models
including dam break, storm surge, hazard mitigation models, etc. The
need for such a system has already been expressed by numerous agencies
and industries. The occurrence of the Bhopal explosion in the early
1980s and the series of spills on the Rhine River years have created a
latent demand for cost-effective industrial accident management systems.
The spatial analysis system can serve both as a simulation and learning
tool, and as an actual dents. The base capabilities available from the
system and the projected capabilities of the near future form a
combination that could allow the use of the system, with modifications,
for planning for economic development, coastal management, resource
management and areawide planning.
|