It is evident from the above that proper sampling of the marine environment 
is an important step in monitoring and impact assessment. For example, even 
though analytical methods for estimating certain environmental parameters 
may be highly accurate and precise, if the sample being analyzed is not 
representative, the data resulting from the analysis is relatively worthless. 
A sampling program for any environmental monitoring or baseline study 
must consider explicitly the following items: a) the number of samples 
required; b) sampling frequency; c) parameters to be measured; and 
d) sampling locations. These items are premised on some accepted definition of 
the level of perturbation or impact which is ecologically significant. It is 
recognized that a complete environmental assessment program encompasses a 
relatively comprehensive characterization (physical, chemical, biological) of a 
system, and includes determination of the potential impacts of pollutants or 
environmental changes on human health and ecological systems. Lucas (8) and 
Eberhart (5) have provided a review of some of the difficulties in assessing 
impacts and have proposed some models as bases for taking and analyzing 
environmental data. It is our belief that programs with more limited objectives 
of characterizing existing conditions or identifying previously defined impacts 
or changes can be developed as subsets of larger environmental assessment and 
monitoring programs. It is our objective to briefly describe some aspects of the 
design and analysis of such experiments to answer some specific questions on 
sampling with particular reference to ichthyoplankton. 
Most biotic elements of the environment are highly variable and 
everchanging. They must be sampled with sufficient intensity to determine the 
course of such changes in time and space. Empirical evidence to date 
concerning ichthyoplankton, as well as juvenile and adult fishes, suggests that 
several years of sampling may be necessary to detect reasonable changes in 
these populations. 
STATISTICAL METHODOLOGY 
There have been two general kinds of statistical methodology applied to 
environmental impact analysis and monitoring. They are linear model analysis 
and time series analysis. 
Time series analysis will not be considered in detail in this report. However, 
it will be briefly mentioned. In time-series analysis, the correlation of a 
response variable to past observations is taken into account in the formulation 
of a statistical model. Statistical time-series analysis has been treated in several 
excellent books including: Anderson (1), Box and Jenkins (2), and Nelson (9). 
The treatment of the methodology and time-series is comprehensive in these 
references, and the reader is referred to them for details. Our limited 
338 
