Configuring Medium Range Ensemble Members

[JordanLaserGit]

The National Water Model article describes the medium range members as:

While starting from the same set of initial conditions as member 1, members 2–7 ingest time-lagged forcing data from successively older forecast cycles of the GFS. These six members extend to 8.5 days, creating an overall seven member equal-weighted medium range ensemble forecast with an 8.5 day forecast horizon

Mostly to demonstrate performance benchmarks, we are initially aiming to replicate this configuration in the NRDS #202.

As we build this system, we could add ensemble members that use the non-lagged forcings but are configured with different physics. For example, we could run member 8 with Muskingum and member 9 with diffusive wave for example. In my limited experience evaluating ensembled weather models, it's common for ensemble member to be designed such that the envelope of predictions encloses the observations. So we may deliberately choose members that have differing physics that may lead to competing biases.