687 
:vers, 
closer 
1 informa- 
i applied 
Dresen- 
surround- 
2StS of 
Earth 
earth resources data. Some recent developments in extractive processing 
methodology and hardware and in user applications model development are 
explored in this paper. 
Techniques for spectral-temporal-spatial processing were developed to 
permit more reliable recognition of classes of vegetation and other terrain 
categories with distinctive time-varying spectral-spatial signatures. 
Examples of wetlands mapping and land resource inventory are presented and 
discussed. 
assors 
d be 
To perform extractive processing rapidly to meet user timeliness require 
ments, a high speed parallel digital special purpose processor MIDAS (Multi 
variate Interactive Digital Analysis System) is being developed by ERIM under 
NASA funding. This low-cost, prototype processor will implement maximum 
d help 
nents to 
sensors 
rocessors, 
register 
ayed to 
ocessing 
data, 
puts from 
most 
likelihood classification, with versatile data processing, at rates comparable 
to the data collection rate of ERTS making a regional processing center concept 
realistic. The design philosophy of this processor is discussed. 
Last, an example of a user application model developed to predict the yearly 
production of mallard ducks, an important migratory waterfowl, from remote 
sensing and ancillary data, is described. 
EARTH RESOURCES INFORMATION SYSTEMS 
remote 
reased 
tential of 
ge detec- 
e signa- 
the ERTS 
re elements 
of this 
interested. 
With the launch of the Earth Resources Technology Satellite (ERTS-1) 
in July 1972, the attention of many scientists focused on the application of 
data from the multispectral scanner (MSS) to earth resources problems in the 
United States and throughout the world. For those of us involved in the design 
of prototype earth resources information systems to assist in resource 
management, ERTS provided a synoptic, periodic, broad scale look at the earth’s 
surface attainable only with great difficulty from previously available air 
craft sensors. 
While data from aircraft sensors will still be sought and used by 
resource managers because of its relatively high spatial resolution, ERTS-like 
he exist- 
because 
ause of 
with 
. required, 
considerable 
ull 
ief icial 
data from satellites is also definitely here to stay. ERTS, because of its 
wide area coverage, has renewed the interest of resource managers in operational 
earth resource information systems which can supply them information on a 
routine basis. For systems designers, this is a challenge for three reasons: 
1) the data rate from earth resources sensor systems increased by over an order 
of magnitude (to 10 10 data vectors per week) with the launch of ERTS-1, 
creating problems of data acquisition and processing; 2) the demands for timely 
information from an earth resources system call for throughput requirements at 
least two orders of magnitude greater than can be obtained with currently 
implemented algorithms on general-purpose digital computers; 3) what information 
¡mote sensing 
i method- 
l to users 
value of 
is easily extracted from the data by current processors is often not digestible 
to users. An enzymatic user model must be included to make the system complete.