kelleythompson
commented
2 years ago A small script was written to facilitate collection of transit time sample points in gage data. For a reach, upstream and downstream gage data were plotted on the same figures, and subsequent mouse clicks between the flow and time of a peak of valley in the upstream gage and the apparent same peak or valley in the downstream gage. The script recorded the average flow, transit time, and other info like WD, reach, and year. A total of 3131 observation pairs were collected in the following amounts: WD17: R2: Y2018: 36 Y2017: 136 Y2019: 103 Y2015: 108 R3: Y2018: 26 Y2017: 123 Y2019: 74 Y2015: 137 R4: Y2018: 30 Y2017: 186 Y2019: 98 Y2015: 108 R5: Y2018: 31 Y2017: 128 Y2019: 94 Y2015: 76 R6: Y2018: 5 Y2017: 46 Y2019: 64 Y2015: 63 R7: Y2018: 24 Y2017: 30 Y2019: 46 Y2015: 69 WD171 Y2020: 4 Y2016: 4 Y2015: 1 WD172 Y2020: 191 WD67: R2: Y2018: 17 Y2017: 17 Y2019: 21 Y2016: 26 Y2015: 21 Y2020: 24 Y2014: 13 Y2013: 34 Y2012: 23 Y2011: 17 R3: Y2018: 17 Y2017: 38 Y2019: 58 Y2016: 37 Y2015: 53 Y2020: 34 Y2014: 31 Y2013: 18 Y2012: 17 Y2011: 32 R4: Y2018: 20 Y2017: 39 Y2019: 50 Y2016: 23 Y2015: 12 Y2020: 11 Y2014: 18 Y2013: 14 Y2012: 3 Y2011: 21
Excel was used to develop power based trendlines between flow and celerity.
In WD17 in particular, celerity values above about 2000cfs seemed to go down. Relationships with dispersion and it was found that dispersion could not be causing this degree of change at least based on the dispersion rates presented by Livingston (which will continue to be used at least until calibration). The only other explanation is that at rates above some threshold - ie around 2000cfs - some amount of flow could be overtopping the banks and going into overbank flow. Overbank flow would have significantly lower velocity, and this slower water could be skewing the observed peaks. Therefore, for WD17, it was reasoned that the celerity relationship should be based on flow rates below the overbank flow, and maximum celerity then limited by a single value that would often take effect over about 3000cfs. This was enabled in the code; hence the small code change..
Celerity (ie wave velocity) and dispersion should inherently be a function of flow. Livingston had a single celerity relationship for WD17 based on C=a*Q^b. For WD67, Livingston had used a different celerity form with different values by Reach, but the celerity values apparently didn't work for calibration and a single, non-flow dependent celerity and dispersion were used for calibration in WD67 - so we really don't have a starting celerity relationship from Livingston for WD67. We've also added some small tributaries in the model (ie timpas creek, crooked arroyo, horse creek, and gageby creek) - primarily to convey aug station releases - that also do not have a celerity relationship. Therefore, just prior to calibration, celerity values were revised for individual reaches in WD17, WD67, and WD171-174 by visually measuring transit time in gage data by looking at sharp peaks and valleys that can be observed in upstream and downstream hourly gage measurements.