NWP2StreamflowBench

Error propagation from NWP precipitation to streamflow. Daily basin-mean precipitation from each forecast model drives the calibrated dCREST model (dCREST-CAMELS v1.0 parameters) on CAMELS-US catchments from a common analysis-forced state; simulated discharge is scored against USGS daily observations with the Nash–Sutcliffe efficiency (NSE), per lead day and per NOAA climate region.

Loading…

Leaderboard

Median score across basins at selected lead days. Reference is the same model driven by the gridMET analysis instead of a forecast — the ceiling any precipitation forecast can reach. Skill horizon: last lead day whose median score exceeds 0 (better than the observed mean).

Streamflow skill vs lead time

Median across basins; band = 25th–75th percentile. Dashed grey line: analysis-forced reference.

Streamflow KGE vs lead time

Kling–Gupta efficiency for the same runs: combines correlation, variability ratio and volume bias.

Precipitation skill vs lead time

Median Pearson correlation of basin-mean daily forecast precipitation with the gridMET analysis over the same days — the input error that the hydrograph inherits.

Skill by climate region

Median streamflow score vs lead day in each of the nine NOAA climate regions (basin counts in panel titles).

Basin map

Per-basin score for one model and lead day. Diverging scale centred on 0 (clipped to −1 … 1); grey = not scored.

Table view: median score by region and lead day

Download: streamflow metrics (per basin, lead, model) · precipitation metrics · reference run · summary.json