Smin Metrics for Water Availability

[glenncampagna]

• This issue will explain different methods for calculating Smin metrics and track decision-making in regards to implementation into Water Availability calculations #818

Methods

Original method explanations from https://github.com/HARPgroup/HARParchive/issues/818#issue-1710059974:

• Approximate method: Take the minimum Storage simulated during the year of the minimum low-flow simulated for the critical period of interest (e.g. 7, 30, 90 days)
  • This is the most time-efficient method, since it allows us to make this calculation from available data (since we already have L90_year and L30_year stored) and store a value for Smin_L90, Smin_L30, Smin_L7, etc,
  • This is the preferred method to start, since we can do it quickly AND it is the most conservative method, since Smin/CPL will be lower than the mean available if Smin is reached prior to the last day of the CPL and Storage increases after it reaches Smin
  • We can simple use sqldf, select min(Storage) from data where year = L90_year
    • Function created for this: fn_get_pd_min.R
• Near-Exact method : Iterate through flow/storage time series, and identify the days that comprise the min flow period, then take the minimum simulated Storage value during those N low-flow days.
  • Produces an exact value for Smin if Smin occurs on the final day of the low-flow period.
• Exact method: FInd the minimum Storage during the simulated low-flow period, and the number of days into that low-flow period that the minimum storage was reached, divide Smin/days_into_Lperiod.
  • This will only occur when inflows are above demand + minimum release from the impoundment, and/or MIF at the impoundment refill intake. Which will be a somewhat rare occurence in an actual drought.

Considerations

• The Exact method above does not consider the remainder of the low flow period after the minimum storage occurs, and therefore needs to be further understood before use
• The Smin values resulting from the approximate method will either be the same or less than the near-exact values, making the approximate method slightly more conservative in certain cases
• Approx. and near-exact Smin values will only be different if the minimum storage (in the low-flow year) occurs outside the low-flow period

Next Steps

• [ ] More in-depth comparison of approx. and near exact methods, quantify differences in Smin and time of the year it occurs see #1156
• [x] Improve Smin_methods.R
  • [x] Save runfile data for all impoundments to save time
  • [x] Enable working w/ non-impoundments (Smin = 0)
• [x] Investigate Vahydro Smin metrics (approx.) not matching those calculated in Smin_methods.R

[glenncampagna]

Smin discrepancy resolved

• The small differences between some Smin metrics from VAhydro and those calculated in Smin_methods.R were caused by the saving and re-reading of run file timeseries data
• The process used for saving the run files (instead of using fn_get_runfile) and then reading them locally to save time is not robust, which used write.zoo and fread