Computing a Record Extension Analysis

Computing a Record Extension Analysis

To compute a Record Extension Analysis, press the Compute button located at the bottom of the editor once options have been set on the General tab. If the compute is successful, the user will receive a message that says Compute Complete. At this point, the user can begin to review the results.  Multiple Record Extension Analyses can be computed using the Compute Manager.  To open the Compute Manager, click the Analysis | Compute Manager menu option.  Then, select the analyses to be computed and press the Compute button.

Computational Methods

When using either the MOVE.1 or MOVE.3 (Bulletin 17C) computation methods, the following steps will be used to compute an extended record:

1. Compute summary statistics of the Secondary (i.e. shorter record) and Primary (i.e. longer record, X) data sets.
2. Compute the mean and variance estimators of the Secondary data set (i.e. shorter record).

3. Develop a linear relationship between the Primary (long record) and Secondary (short record) for the concurrent period.  This relationship is of the following type:

   $$//<![CDATA[ \begin{array}{l}\displaystyle \hat{y}_i=a+b\left(x_i-\bar{x}_e\right)\end{array} //]]>$$

   
   where $//<![CDATA[ \begin{array}{l}\hat{y}_i\end{array} //]]>$ = extended value for the Secondary data set (i.e. shorter record) for a given year or date, a = intercept, b = slope, $//<![CDATA[ \begin{array}{l}x_i\end{array} //]]>$ = value for the Primary data set (i.e. longer record) for the same year or date, and $//<![CDATA[ \begin{array}{l}\bar{x}_e\end{array} //]]>$ = mean of the values from the non-overlapping period of record for an effective record length, e.

4. Use the linear relationship to estimate values for the Secondary (short record) location for dates/times where only the Primary (long record) location contains values (i.e. non-concurrent period).

MOVE.3 (Bulletin 17C) Method Considerations

When using the MOVE.3 (Bulletin 17C) computation method, the effective record length (n_e, years) of the mean and variance estimators are calculated after step 2 within the list of steps above.  The estimator of the variance will always be the limiting factor (i.e. variance of the variance estimator > variance of the mean estimator).  Thus, the effective record length of the variance will be used as n_e.  ne is then used when developing a linear relationship between the Primary and Secondary data sets as well when extending the Secondary data set.

Record Extension Computational Considerations, Recommendations, and Requirements

The Record Extension Analysis will use all values within the concurrent time period when computing correlation coefficients, estimators for augmentation, and the extended record.  No values will be removed from consideration even though they may not have resulted from the same runoff event.

The following recommendations/requirements are checked during the computations:

• A minimum of 10 concurrent years between the Primary and Secondary data sets is required.  If the concurrent record length is less than 10 years, the compute will fail and an error message will be displayed.
• A correlation coefficient between the Primary and Secondary data sets of at least 0.8 is recommended (Bulletin 17C (England, et al., 2019).  If the correlation coefficient between the Primary and Secondary locations for the concurrent record is less than 0.8, a warning message will be shown.  However, the compute will proceed and results will be displayed.
• If the slope of the linear relationship (i.e. b parameter) is found to be equal to or less than 0, the compute will fail and an error message will be displayed.
• When using the MOVE.3 (Bulletin 17C) computation method, if the computed effective record length (n_e) is equal to or less than 1, the compute will fail and an error message will be displayed.