Andrew W' Wood and Dennis P' Lettenmaier

1
Relative contribution of initial condition and
climate forecast error to seasonal hydrologic
forecast uncertainty
Andrew W. Wood and Dennis P. Lettenmaier
Univ. of Washington
Summary
Approach
Forecast uncertainty arises from uncertainty in
initial conditions and in boundary forcings,
among other sources. In the context of hydrologic
forecasting, the most important initial
conditions are the moisture states (mainly
snowpack and soil moisture), while the boundary
forcings are timeseries of climate variables such
as precipitation and temperature. Via a
retrospective analysis of six-month hydrologic
forecasts of snowpack, runoff and soil moisture
for 5 major river basins in the western U.S., we
estimated the relative contributions of
uncertainty in these two sources to forecast
uncertainty at different lead times, and for
different forecast initiation months. The
analysis is based on the comparison of the
results of Ensemble Streamflow Prediction (ESP
Twedt et al., 1977) forecasts with those of a
"reverse-ESP" approach. In the former, forecasts
are produced by coupling perfect initial
conditions with an ensemble of climate forecasts
derived from observed climate sequences whereas
for the latter, a perfect climate forecast is
coupled with an ensemble of initial conditions.
The climate sequences are taken from a
retrospective observation-based dataset, and the
initial conditions are simulated using this
forcing dataset and semi-distributed macroscale
hydrologic model.
Results for Basin Averages of Hydrologic Variables
In the western U.S., the hydrologic cycle
reflects (to varying degrees in different river
basins) the role of snowpack in storing
winter/spring precipitation for release as
runoff and soil moisture recharge in late spring
and summer.
Conclusions

• Generally
• uncertainty in initial snowpack and soil moisture
  dominate forecast uncertainty in winter and
  spring, and in summer, respectively, for lead
  times of 3-5 months while in autumn, initial
  soil moisture uncertainty influence is limited to
  1-2 months
• not surprisingly, boundary forcing uncertainty
  dominates forecast uncertainty at longer lead
  times.
• Variations by river basin and variable were
  notable, e.g.
• the CRB, CALI had longer IC influence on SWE in
  April than January, whereas the other basins had
  roughly equal IC influence for the two forecasts,
  mainly because of earlier melt
• for every forecast start date, IC influence on
  soil moisture was greater than met. forecast
  influence for at least 1 month, in contrast to
  runoff, which is influenced by ICs predominantly
  in spring, in most locations, but not at other
  times of the year.
• Implications for forecasting
• In spring and to a lesser extent, late winter,
  improvements in initial condition estimates will
  greatly reduce error in 1-5 lead forecasts of
  SWE, soil moisture and runoff, varying by
  location
• at other times of the year, particularly autumn,
  it is much less important to estimate ICs
  accurately than to forecast boundary conditions
  (meteorology) correctly.

Uncertainty in initial conditions that capture
snowpack levels at or near their peak, or soil
moisture levels at annual extremes (low or high)
tends to dominate forecast uncertainty for lead
times of 2-5 months. Uncertainty in (a)
precipitation and temperature forecasts spanning
periods when snowpack builds or melts, and (b)
precipitation forecasts during soil moisture
recharge periods, can dominate forecast
uncertainty for up to the entire forecast period,
depending on location and time of year. The
curves show, for each river basin and at four
times during the year, the extent to which SWE,
soil moisture and runoff forecasts are influenced
by IC error versus forecast meteorology error,
and hence would benefit from efforts to reduce
either.
Reference
Maurer, E.P., A.W. Wood, J.C. Adam, D.P.
Lettenmaier, and B. Nijssen, 2002, A Long-Term
Hydrologically-Based Data Set of Land Surface
Fluxes and States for the Conterminous United
States, J. Climate 15, 3237-3251. Twedt, T.M.,
J.C. Shaake, Jr., and E.L. Peck, 1977, National
Weather Service Extended Streamflow Prediction,
Proc. 45th Western Snow Conference, Albuquerque,
pp. 52-57, April. Liang, X., D. P. Lettenmaier,
E. F. Wood, and S. J. Burges, 1994, A Simple
hydrologically Based Model of Land Surface Water
and Energy Fluxes for GSMs, J. Geophys. Res.,
99(D7), 14,415-14,428.