


2.3.1 LAKE ASSESSMENT, THE FIRST STEP
To complete the task of lake assessment, there are implicit components that must be addressed.
Often these are violated. For example, an assessment might not have the luxury of a data collection
effort because abundant data are already available. Or else there are insufficient resources to acquire
more data. (In this example, one can substitute <observation' for <data') In this case all of the practical
questions leading to the formation of a database are skipped and the last step  data analysis  is all that
is left for completion. If there are abundant appropriate data this is the ideal situation for lake
assessment. If there are not abundant data, this is called <reality'.
The process of collecting needed observations is important and needs to planned as carefully as
possible. There are two extremes for this situation, a thorough statisticallysound design versus a simple
minimal design specifically for the question or problem at hand.
2.3.2 SAMPLING DESIGN
As with tailwater assessment the first step in this process is clearly identifying the problem or
question. This will allow definition of the objectives. The objectives are an important reference during
the entire process of sampling design. Rational consideration of the sampling requirements necessary to
accomplish the objectives will facilitate early identification of needed resources, or if the objectives are
even possible.. or not.
The design procedure involves answering several questions:
1. What needs to be observed (measured)?
2. How often and where?
3. How many replicates?
An objective way to answer these is to apply statistical methods to preexisting knowledge of
the lake or reservoir. Fortunately there are abundant resources for a statisticallysound sampling
design. Recent published works include Montgomery (1997), Green (1979), Gilbert (1987), and
Gaugush (1987 and 1993). These are specialized for sampling design and build on knowledge of basic
statistical concepts. For those basic concepts other standard references include Neter and Wasserman
(1974), Steele and Torrie (1980) or Sokal and Rohlf (1969). Gaugush, in particular, addresses the
requirements for reservoir water quality investigations and has published (Gaugush 1993) software to
aid in the design. All of these approaches incorporate statistical considerations of intended objectives;
trend or prediction?, statistical confidence, QA and QC considerations. Pragmatically, these statistical
approaches tend to assume that some minimal standard (for example, a confidence limit) exists that
must be exceeded in the results of the assessment. Such limits of effort are nearly always present. In
these treatments of statistical sampling design there are dominant issues of meeting statistical
2.32


Integrated Publishing, Inc. 