Soil moisture from ERA5: the next initial-condition upgrade
Soil moisture in the live pipeline comes from GFS, which can lag reality after fast-moving wet fronts. Switching to ERA5 reanalysis would help. Here is the trade and where it sits in the queue.
Any WRF pipeline that initialises soil moisture from a parent global model inherits a structural weakness: the soil state lags reality after fast-moving wet fronts. GFS, which we initialise from, runs a lower-fidelity land surface model at 13 km, and its soil moisture field evolves slowly over many forecast cycles. After a fast-moving wet front, GFS soil moisture can lag observed soil moisture by 24 to 72 hours. While that lag persists, the model believes the surface is drier than it really is, and partitions too much of the incoming radiation into sensible heat flux. The downstream effect on a soaring forecast: trigger temperature comes out too low, trigger time too early, `wstar_ms` slightly too strong.
This is independent of which land surface scheme you run downstream (the Noah post covers the scheme itself). Even an upgrade to Noah-MP would not fix the inherited soil-moisture lag. The fix has to come from the initial condition, and that is what this post is about.
The standard fix is to replace GFS soil moisture at initialisation with a more responsive source. ERA5 is the obvious candidate: a reanalysis that assimilates observed precipitation with about a one-day latency, producing a soil moisture field that responds to recent wet events much faster than a free-running forecast model can. Other research-grade products exist (SMAP satellite soil moisture, various land data assimilation systems) but ERA5 is the one with the cleanest production pipeline.
What is in the way of just doing it. Three things. First, the engineering: ERA5 has its own access pattern (Climate Data Store API, different file format, different latency profile than NOMADS). Adding it as an alternative initial-condition source for soil moisture means a parallel ingest path that has to be reliable enough to run unattended four times a day. Second, the validation: the size of the win in our specific UK configuration is something we want to measure rather than estimate, and the validation pipeline is a precondition for that measurement. Third, the prioritisation: a soil-moisture fix is downstream of basic validation work, and the basic validation work is downstream of having the engineering capacity to build it.
What we do today instead. Nothing specific. The pipeline runs on GFS soil moisture and inherits whatever lag GFS has. On settled weather the lag is invisible. On rapidly-changing weather just after wet fronts, the bias is real and the forecast is more uncertain than its self-confidence suggests.
What we tell users. If you notice forecasts that feel too optimistic immediately after a wet front (early trigger, strong forecast thermals, low cloudbase mismatch), the soil moisture initial condition is the most likely culprit. This is a known structural issue in any WRF pipeline that takes soil state from a parent global model without an independent correction.
ERA5 soil moisture initialisation is on the roadmap behind validation. When it lands, this post gets rewritten with the comparison numbers. Until then, the Noah land surface post describes the shipping scheme.