696 
the decision model used by the USDI Fisheries and Wildlife Service. The 
model shown is, at present, conceptual and represents the work of FWS 
personnel. ERIM and FWS are working jointly (under NASA funding from 
ERTS and Skylab programs) to develop remote sensing inputs to this model. 
Figure 3 shows at the left, inputs to the user model. Astrisks indicate 
those types of information obtained from low altitude aircraft-borne observers 
(supplemented by ground observations), and data potentially derivable from 
remote sensing techniques. The goal of the user model is to estimate the fall 
population of mature and immature birds. Old birds are estimated by low 
altitude aerial survey and ground counts of nesting pairs on a sampled basis. 
Estimates of summer mortality are also made from ground observations and past 
experience. The number of breeding pairs, coupled with the May and June pond 
numbers are used to estimate the number of new ducks. This information is 
augmented by the number of broods obtained on a sampled basis from ground 
survey. 
Hunting regulations are defined based on the population size, the 
estimate of harvesting of ducks in Canada, and the carrying capacity of the 
habitat. Remote sensing has an impact in assessing capacity of habitat, 
especially in the assessment of quantities and distribution of natural 
vegetation. 
As an example of the annual production equation, Figure 4 shows a linear 
equation in several variables which FWS personnel have derived to predict new 
production of mallard ducks. The coefficients of the model are derived on 
the basis of historical experience. If similar models can be constructed 
for other species, then the general user-decision model of Figure 4 can 
become more operational. 
DATA 
Numbe 
by 
Pere 
Remò 
May 
Inde 
Numt 
Broc 
*Obt 
**0b 
ADVANCES IN PROCESSING EQUIPMENT 
My assessment of current multispectral processing equipment systems to 
meet the needs of keeping pace with multispectral data acquisition which 
occurs typically at rates of hundreds of kilopixels/sec or megapixels/sec is 
given below: 
1. Sensor capability exceeds processing capability by large factors. 
2. Digital and analog implementations of present techniques will not 
keep pace with the needs of most operational-prototype informa 
tion systems. 
3. Multiple digital computer approach is probably too costly 
4. Implementation of special purpose parallel processing with improved 
techniques appears promising from both throughput and cost aspects 
but requires development. 
Despite work on faster techniques, presently available conventionally 
organized digital computers are too slow (with current algorithms) by 
orders of magnitude so that even many computers per sensor are still