Decision support experiments and evaluations using seasonal to ...

The U.S. Climate Change Science Program Chapter 2
46
logic forecasting community’s intuition about
the current levels of hydrologic forecast skill using
long-lead climate forecasts generated from
various sources. The analysis first underscored
the conclusions that, depending on the season,
knowledge of initial hydrologic conditions conveys
substantial forecast skill. A second finding
was that the additional skill available from incorporating
current (at the time) long-lead climate
model forecasts into hydrologic prediction
is limited when all years are considered, but can
improve streamflow forecasts relative to climatological
ESP forecasts in extreme ENSO years.
If performance in all years is considered, the
skill of current climate forecasts (particularly
of precipitation) is inadequate to provide readily
extracted hydrologic-forecast skill at monthly
to seasonal lead times. This result is consistent
with findings for North American climate
predictability (Saha et al., 2006). During El
Niño years, however, the climate forecasts have
Figure 2.16 CPC objective consolidation forecast made in June
2007 (lead 1 month) for precipitation and temperature for the
three month period Aug-Sep-Oct 2007. Figure obtained from
.
adequate skill for temperatures, and mixed skill
for precipitation, so that hydrologic forecasts
for some seasons and some basins (especially
California, the Pacific Northwest and the Great
Basin) provide measurable improvements over
the ESP alternative.
The authors of the Wood et al. (2005) assessment
concluded that “climate model forecasts
presently suffer from a general lack of skill,
[but] there may be locations, times of year and
conditions (e.g., during El Niño or La Niña)
for which they improve hydrologic forecasts
relative to ESP”. However, their conclusion
was that improvements to hydrologic forecasts
based on other forms of climate forecasts, e.g.,
statistical or hybrid methods that are not completely
reliant on a single climate model, may
prove more useful in the near term in situations
where alternative approaches yield better
forecast skill than that which currently exists
in climate models.
2.3 CLIMATE DATA AND
FORECAST PRODUCTS
2.3.1 A Sampling of Seasonal-to-
Interannual Climate Forecast
Products of Interest to Water
Resource Managers
At SI lead times, a wide array of dynamical prediction
products exist. A representative sample
of SI climate forecast products is listed in Appendix
A.1. The current dynamical prediction
scheme used by NCEP, for example, is a system
of models comprising individual models of the
oceans, global atmosphere and continental land
surfaces. These models were developed and
originally run for operational forecast purposes
in an uncoupled, sequential mode, an example
of which is the so-called “Tier 2” framework
in which the ocean model runs first, producing
ocean surface boundary conditions that are
prescribed as inputs for subsequent atmospheric
model runs. Since 2004, a “Tier 1” scheme was
introduced in which the models, together called
the Coupled Forecast System (CFS) (Saha et
al., 2006), were fully coupled to allow dynamic
exchanges of moisture and energy across the
interfaces of the model components.