Pre-registration: Snow-gauge undercatch, the within-catchment sensitivity of streamflow to snowfall fraction, and its inheritance by deep-learning streamflow models (p305 A′, v1)

How much of the observed within-catchment sensitivity of runoff ratio to snowfall fraction survives a physically based snow-undercatch correction? And does training deep-learning models on uncorrected gauge-based precipitation change the streamflow response they project for +2 °C warming? Deposited before any undercatch-corrected quantity, blocked-CV model run on observed flow, or warming counterfactual on such models was computed; results already seen are disclosed in §0.