Application of Streamflow Reconstruction to Water Resources Management - Dendroclimatology - page 251

8.4 Challenges
8.4.1 High Flows
Attempts to reconstruct annual streamflow totals from tree-ring indices by regres-
sion methods often run up against the obstacle of heteroscedastic errors, in which
the variance of the regression errors increases with the magnitude of the estimated
flows (e.g., Meko and Graybill 1995 ; Mekoetal. 2001 ) . This is especially true
the smaller the basin and the more arid the climate. To satisfy regression assump-
tions, one solution is log transformation of the flows before regression analysis.
The reconstruction is then in log units of flow, and in such units the residuals
may not be heteroscedastic. But water resources planners do not compute their
water budgets in log flow units, and when reconstructions are back-transformed
to original units, the heteroscedasticity returns. The problem essentially is that for
basins with 'flashy' flow regimes (e.g., rapid runoff from heavy storms in arid
regions), tree-ring data are unlikely to accurately distinguish magnitudes of high
flows (Meko et al. 1995 ) . The runoff events may simply occur too rapidly to
leave a strong signature in soil moisture in the root zone of the trees, and in tree
growth. At the same time, high flows are of great importance in water resources
planning because they often contribute the pulses of runoff that refill storage
reservoirs.
8.4.2 Seasonality
Cool season precipitation and snowmelt contribute proportionally more water to
streamflow than warm season precipitation, especially for large watersheds. Reasons
include the higher evapotranspiration losses and requirements for soil moisture
recharge in the summer season, as well as the generally spotty nature of summer
convective storms. Improvement of the seasonal resolution of precipitation or soil
moisture signal from tree rings is therefore likely to yield more accurate streamflow
reconstructions. Separate measurements of earlywood width and latewood width
have proved useful in various regions for seasonal resolution of precipitation vari-
ations (e.g., Lodewick 1930 ; Meko and Baisan 2001 ; Cleaveland et al. 2003 ) . A
different issue related to seasonality concerns streamflow reconstructions in areas
where the climate signal in tree growth and the main climatic contribution to annual
flow do not correspond as well as they do across most of the western United
States. In the Pacific Northwest, the main source of runoff is winter precipitation,
while low-elevation moisture-sensitive trees are more tuned to summer precipitation
(Gedalof et al. 2004 ) . In the eastern United States, species such as bald cypress also
tend to be more sensitive to summer conditions, making reconstructions of summer
low flows a more appropriate hydrologic variable for reconstruction than annual
flows (Cleaveland 2000 ) .