International Archives of Photogrammetry and Remote Sensing. Voi. XXXII Part 7C2, UNISPACE III, Vienna 1999
10
I5PR5
UNISPACE III - ISPRS Workshop on
“Resource Mapping from Space”
9:00 am -12:00 pm, 22 July 1999, VIC Room B
Vienna, Austria
ISPRS
1979; Lees and Ritman 1990; Olsson 1989; Skidmore et al.
1997; Zhou and Garner 1990). The difference between image
processing (of remotely sensed data) and analysis using
geographic information systems becomes blurred when
remotely sensed and other geographically registered data are
merged in a (raster) GIS data model. There are many common
processing techniques including geometric correction,
cartographic output, and classification methods.
One processing technique that incorporates remote sensing and
is executed in a GIS, is the expert system. Expert knowledge
from farmers, agronomists, soil scientists, foresters, indigenous
people or others interested in land resource management is an
important (and sometimes the only) source of information.
Participatory techniques have been adapted to extract
information through interviews and discussion groups,
particularly for land use planning. Expert systems can
incorporate such knowledge into a GIS, and then automatically
map areas and features of interest. An example is the Land
Classification and Mapping Expert System (LCMES)
(Skidmore et al., 1997). Forest soils were mapped using terrain
parameters such as slope and aspect, remote sensing, and a
knowledge base generated by interviewing soil scientists.
Interestingly, there was no difference between the accuracy of
the expert system map and a map prepared by a soil scientist.
Expert systems retain the robustness of the scientific method,
but allow knowledge and other ‘soft’ data to be used in analysis,
as well as mapping more rapidly and with low'er input of labour.
Despite these successes with expert systems, as w'ell as the rapid
growth of GIUS described above, satellite remote sensing has
played a relatively minor role in the commercial success of GIS.
There are a number of reasons for the poor uptake of remote
sensing, but five appear to dominate. Firstly, the coarse spatial
resolution of satellite images is fine for regional scale studies
but most GIS analysis is at the local scale. In addition a limited
suite of wavelength bands have been available, and these have
usually been rather broad. Secondly, the primary data input into
GIS has been conventional cartographic maps in the form of
vector data layers, which are expensive to convert to a digital
(vector) format. The new high spatial resolution imagery
described in the previous section will probably become the
primary data model, as well as providing the primary data
source. Thirdly, there is a mismatch between the vector data
structure (based on points, lines and polygons) used in GIS, and
the raster (pixel) based data structure used in remote sensing.
There are difficulties in converting between these two data
models. In addition, users have been conditioned to accept line
work on maps as representing a boundary', when in fact most
boundaries of natural resources features (such as soils,
vegetation species mixes, or wildlife habitat suitability') are
gradients. ‘Cartographic conditioning’ results in most users
delineating homogeneous areas (polygons) on maps, instead of
mapping variables as a continuum, such as the density' of a
species, or the concentration of soil nutrients. The fourth reason
for the slow uptake of remote sensing in GIS is poor
accessibility to data due to the expense of higher resolution
imagery, poor delivery systems for the data, or data being in the
wrong format. Tills problem is particularly acute for
environmental monitoring where budgets are often tight.
Fifthly, there is a lack of expertise to extract information
available in remotely sensed images.
Key issues in the field of integrating GIS and remote sensing
are scale and accuracy. Scale of the imagery, or the map data
with which the imageiy is being fused, may be mismatched to
each other. The scale may alos be inappropriate to the task.
Detailed local information may be viewed as noise in a regional
scale, whilst important local detail may be lost when working at
regional scales. The second key issue is accuracy'. It is still rare
to find maps, or other spatial products, with a quantitative
statement of accuracy’, even though accuracy assessment
techniques are by now well known.
CONCLUSIONS
In this presentation, it is shown that remote sensing has not
penetrated the natural resource management arena as deeply,
nor as quickly, as predicted 20 years ago. This is despite rapid
changes to a service economy, and associated demand for
information, including spatial information. In less developed
countries. GIS and remote sensing offer regional and local
mapping and monitoring, which may be used to assist in
sustainable development. Growth of GIS has remained very
high for the last 10 years, but the remote sensing industry' has
not capitalised on this market, and a number of reasons for this
are discussed. Some new remote sensing systems may offer
solutions, including high spatial resolution imageiy, high
spectral resolution imagery and to a lesser extent radar.
REFERENCES
Aspinall, R. and N. Veitch (1993). “Habitat mapping from
satellite imagert and wildlife survey using a Bayesian
modeling procedure in a GIS.” Photogrammetric
Engineering and Remote Sensing 59(4): 537-543.
Britannica. E. (1989). The New Encyclopaedia Brittanica.
Chicago. Encyclopaedia Britannica.
Burrough, P. A. (1993). Modelling land resource scenarios with
field data, remote sensing and process models in GIS
is easy. NARGIS 93, Darwin, AGPS: Canberra.
Campbell, H. and I. Masser (1992). GIS in British Local
Government: An Overview of Take-Up and
Implementation. Proceedings of EGIS '92, Munich,
EGIS Foundation.
Drucker, P. (1991). “The new' productivity challenge.” Harvard
Business Review! November-December).
Fairall, J. (1995). ESR1 still ahead says report. Spatial Business.
1: 1.
Green, R. (1990). “Geographic information systems in Europe.”
Cartographic Journal 27: 40-42.
Hoffer, R. M. (1975). Natural resource mapping in mountainous
terrain by computer analysis of ERTS-1 satellite data,
Purdue University: Indiana.
