Governor Bond Lake TMDL
Final Report 27 September 2002
6.0 MARGIN OF SAFETY
6.1 METHOD FOR CALCULATING MARGIN OF SAFETY (MOS)
The MOS is an additional factor included in the TMDL to account for scientific uncertainties, growth, and
others such that applicable water quality standards/guidelines are achieved and maintained. The MOS
can be included implicitly in the calculations of the WLA and LA or can be expressed explicitly as a
separate value.
For the Governor Bond Lake TMDL, the MOS includes both implicit and explicit determination.
Conservative input values (examples) were chosen for modeling purposes in order to implicitly include a
MOS. The BATHTUB model calculated a measure of potential model error (coefficient of variation).
This error term was used in combination with the coefficient of variation for percent load reductions in
order to meet target water quality goals. The summation of these error terms was used to determine an
additional explicit MOS. The coefficient of variation is a measure of variation in numbers relative to the
mean value and can be expressed as either a fraction or percent of the mean.
6.2 RATIONALE FOR MOS
Potential sources of error are inherent in measured data, default values chosen for modeling, and model
calculation procedures. The first two error sources are included in the implicit MOS, while the last error
source is included in the explicit MOS.
1. Measured flow data had several potential errors; flows were determined based on staff gage depth
of water flow, cross-section geometry, and float method stage-discharge relationships. Changing
channel morphology during the growing season is highly likely. Stage-discharge relationships
were measured at the end of the monitoring season, and are therefore only approximate for the
monitored season. Bending of staff gages, limitations of float method, and changing channel
morphology all contribute to measured flow errors.
2. Measured concentrations were from single grab samples, often following a precipitation event in
order to capture high flow transport of pollutants. These samples may not accurately describe
total loads during high flow conditions or base flow conditions. If samples happened to miss the
peak loading times, loads may be under predicted. If samples were taken only during peak
concentrations, loads may be over predicted.
3. Suspended sediment samples do not often characterize the entire water column, and the
measurements often miss heavier particle fractions because they settle out before the subsample
can be drawn out for laboratory analysis.
4. GIS analysis was used to minimize calculation errors, however, watershed characteristic data is
based on several data sources, each containing inherent errors (e.g., soil survey polygons, soil
survey k factors, land use type and area, etc.). A difference in calculated areas, when one data set
is re-projected to be consistent with other data sets, is another source of error.
5. Model calculations use various regression and decay functions. Each submodel has errors
associated with it. These are tabulated during the modeling process for an overall coefficient of
variation.
