DWR_3066022

• Title: WESTERN DAM ENGINEERING NEWSLETTER, VOLUME 5, ISSUE 1 MAY 2017
• Author/Source: AECOM
• Keywords: EMERGENCY RESPONSE, EROSION, CALIBRATION, HYDROLOGIC MODELING, EMBANKMENT DAMS, SEEPAGE, SINKHOLES
• Document Type - Reference Library: Research, Thesis, Technical Publications
• Document Date: 5/1/2017
• Year: 2017
• Team/Office: Dam Safety

Western Dam Engineering
Technical Note

May 2017

17
(using the full record of gage data), etc. Other
methodologies, like those associated with ungaged
watersheds (e.g., regional regression equations to
predict runoff rates), could also be used; however,
observed and measured data are preferable,
where available.
 Pseudo-Calibrations – Independent data are based
on both observed (i.e., gage) and predicted (i.e.,
ungaged) data and could include the
aforementioned independent gage data as well as
runoff rates predicted from local and regional
regression equations, correlations with
neighboring gaged watersheds, etc.

Figure 7. Gaged Discharge Time Series Used for Model
Calibration and Validation.
Model Validation
Figure 8 presents a flow chart of a typical model runoff
result validation process. This process is similar to that
of the calibration process with the exception of the
data used to perform the validation, which is
independent of that used as part of the calibration
process.
One can see that the calibration, validation, and overall
verification process is iterative in nature—the model is
calibrated for one condition or scenario and then
initially verified prior to being evaluated for additional
conditions and scenarios and further verified. If you
cross your fingers and toes while holding a four-leaf
clover, you may only have to go through the whole
process once, but more likely than not, a series of
refinements and adjustments will be required to
produce a model that is fully verified across a suite of
conditions and scenarios. Just remember, verification
is valid when validation is verified!
But what constitutes an adequately verified model?
Model accuracy evaluations are performed and
provide a basis for satisfactory agreement between
observed or predicted data and modeled result data.

Figure 8. Hydrologic Model Validation Process.
Model Accuracy Evaluations
Model accuracy is frequently evaluated by statistical
comparison of measured or predicted data and
modeled results [5]:
1. Percent Bias Coefficient (𝐵𝑝),
𝐵𝑝(%)=100 ∙∑(𝑁𝑖−𝐵𝑖
𝑁𝑖
)𝑛
𝑖=1
Where: 𝑛= number of pairs of the observed
and modeled variables; 𝑁𝑖= observed data;
and 𝐵𝑖= modeled value. The 𝐵𝑝 is expressed as
a percentage and describes the tendency of
the modeled data to be greater or smaller than
the observed data. The corresponding
accuracy classification is:

𝐵𝑝 ≤ ±10 Very good
±10 < 𝐵𝑝 ≤ ±15 Good
±15 < 𝐵𝑝 ≤ ±25 Satisfactory
𝐵𝑝 ≥ ±25 Unsatisfactory