Filling Data Gaps by a Rank-data Distribution Method (R-D Method)

Presenter Information

Joon-Hee Lee

Location

ECC 303/305

Event Website

https://water.usu.edu/

Start Date

3-31-2008 4:00 PM

End Date

3-31-2008 4:15 PM

Comments

High frequency data are required to populate high frequency statistical water quality models, but the most common intervals between water quality samples are weekly or monthly and are irregular. The Rank-Data distribution method (R-D method) was developed based on the concept that time series of water resources data consist of data distributions and time series of the ranks of the data at the measurement times, and that the distribution of a full high frequency data set, including both observations and unknown values, is identical to the distribution of the observations. Cumulative Failure Probabilities (CFPs) of unknown values for dates with no observations were estimated by interpolating time series of the CFPs of the observations to create a daily time series of CFPs. This estimated time series of CFPs was then linked to the data distribution to obtain the flow time series. In tests, time series of mean daily flows from the R-D method were better estimation of time series of measured flows from the original daily data set than from simple interpolation between observations. These tests demonstrated the promise of generating time series of water quality or water quantity by combining the probabilistic results from Bayesian Networks to the CFP time series from the R-D method.