ESP - A Few Words of Caution
In using ESP products and information, it is very important to recognize that all of the variability in the streamflow traces comes from the precipitation and temperature traces. In reality, uncertainty is introduced from a whole host of sources.
The ultimate challenge is to appropriately consider all sources of uncertainty in
developing accurate ensembles from which accurate statements and effective decisions can be made.
Currently, the state of the operational science is limited to the uncertainty introduced by future
temperature and precipitation. In other words, the errors (uncertainty) introduced by observations,
model states, the modeling system, model parameters, and human input are not currently considered.
Is that at big deal? If you’re only interested in longer range forecasts and
with parameters aggregated over time, it probably isn’t a big deal. The uncertainty associated
with future weather pretty much dwarfs everything else. As one moves towards the near term (hours to
weeks), the dominance of the weather forcing is much less. Additionally, if you’re interested
in specifics of flow (instantaneous minimums and maximums), the uncertainty of the modeling system
and its parameters can play a significant role. Work is currently underway to address this short-coming,
but it is not ready. In the meantime, users are advised that the system is likely to underestimate the
uncertainty in near-term (a few weeks) and critical level forecasts.
In addition to considering only a portion of the total variability, one must remember
that the ESP analysis is based on a limited number of samples (30 to 50 years of data). As such the
precision is limited. The uncertainty in the uncertainties is very real and users are advised to
consider the values derived as estimates. Work is currently underway to develop and provide verification
and validation information that will allow users to assess the ability of the system to provide accurate
information for each location.
Regulated versus Unregulated Streamflow Forecasts
All results provided on the CNRFC AHPS and AHPS / ESP Trace Analysis website are for
“full natural streamflow” unless otherwise specifically stated. Actual measured can and will
be different if there are reservoirs and/or diversions upstream from the location of interest. The vast
majority of river gages in the Western U.S. have some sort of regulation or diversion upstream. Until
the NWS has the operational ability to simulate the future operation of these reservoirs and diversions
in an ensemble environment, they cannot be reasonably reflected in our products. Certainly, this limits
the utility of the forecasts for many customers. The NWS is actively working on solutions, but we are
likely several years away as of 2005.
|