12 To do list

12.1 Characterize sampling error in pool estimates

Current methods aggregate samples to mean monthly values to calculate pools. However, this masks sampling variability among sites and may give an unrealistic sense of prediction error for load estimates. One alternative would be to estimate UBL and DBL pools for each individual sample, rather than aggregating to a mean prior to estimating pools. This would allow the generation of prediction intervals and the ability to show plots that accurately reflect variability in estimates (e.g. as boxplots or line plots with confidence intervals).

12.2 Compare pools & loads

Characterize loads as a proportion of the pool. Which time steps are relevant for comparison?

12.3 Account for negative flows at FB

Currently, negative flows at the FB outlet are removed before calculating loads. These could be accounted for by generating loadest models for both positive and negative flows and subtracting.

12.4 UPRR bridge load analysis

Constituent exchanges between Gilbert and Gunnison Bays at the UPRR bridge are currently unestimated. Ideally these would be estimated from high frequency discharge measurements, but these are not currently available. Alternatively, it may be possible to estimate positive and negative loads based on existing paired field samples and discharge measurements.