Forests and Water: A State of the Art Review for Colorado

• Year: 2003
• Title: Forests and Water: A State-of-the-Art Review for Colorado
• Author: Colorado State University
• Description: Colorado State University 2003
• Publications - Doc Type: Other

downstream? (3) What change in flow is detectable? and
(4) Can changes in forest cover cause changes in flows
on larger basins, even if the change in flow is not statis-
tically detectable? The answers to these questions have
important implications with respect to the larger -scale
changes in runoff that are of most interest to resource
managers and the public.
The answer to the first question — how much of a basin
must be treated in order to detect a change in runoff — is
relatively well known. A 1982 review of paired catch-
ment experiments concluded that at least 15 -20% of a
forested basin must be treated within a short time period
in order to detect a change in runoff (Bosch and Hewlett,
1982). Troendle and Leaf (1980) noted that 20 -30% of a
watershed must be treated to detect a statistically signifi-
cant change in flow, and this value has been supported
by the magnitude of flow changes observed in paired
watershed experiments in the Rocky Mountains (Table
2.1). At Coon Creek in south - central Wyoming there was
a detectable change in annual water yield as a result of
harvest and road - building on 24% of the watershed (Tro-
endle et al., 2001). Small increases in annual water yields
were observed by removing 31 -33% of the trees in two
small watersheds in northern Arizona (Baker, 1986).
Harvesting 36% of the North Fork of Deadhorse Creek
caused a significant change in runoff at the sub - watershed
scale, but a change in water yields was not detectable at
the main weir because the treatment only removed about
5% of the forest canopy above this weir. For ponderosa
pine forests in the Black Hills of South Dakota, 25 -30%
of the forest canopy may have to be removed in order to
obtain a detectable increase in annual water yields. This
higher threshold is probably due to the drier conditions
and relative absence of a strong snowmelt peak.
The second question is whether a flow increase at an
upstream location will be delivered to a downstream res-
ervoir, diversion intake, or other location of interest. The
general principle is that streams are relatively efficient
in conveying water downstream, and this is particularly
true in headwater forested areas (e.g., Troendle, 1982).
However, as streams flow from the mountains and onto
the plains streams may become losing rather than gain-
ing. This means that groundwater is no longer flowing to
the streams ( "gaining "), but streamflow is seeping out of
the channel to support the local groundwater table ( "los-
ing"). Hence some of an increase in streamflow due to
forest management may be transferred to the adjacent
groundwater, particularly in larger basins. The availabil-
ity of the resultant increase in groundwater for human
use will vary from basin to basin. Riparian vegetation
19
and surface evaporation also can reduce the amount of
streamflow that is delivered downstream.
The proportion of a change in streamflow that is trans-
mitted to a specific location can vary seasonally and
from year to year. During spring snowmelt and early
summer streams are more likely to be gaining, while
later in the summer groundwater levels tend to drop
and streams may shift from gaining to losing (e.g., Graf,
1997). Losses to riparian vegetation and surface evapo-
ration generally will be much less in spring than later in
the summer. Thus an increase in spring snowmelt gener-
ally should be more readily translated downstream than
an increase in low flows. Some of the increases in peak
flows may be lost to bank storage and overbank flows,
and this will attenuate high flows in the downstream di-
rection. The extent to which an increase in flow is trans-
lated downstream will depend on the conditions within
that basin, but generally one would expect greater losses
with increasing spatial scale and increasing distance
from the mountain front.
The third issue is the detectability of a given change.
The ability to detect change will vary according to the
accuracy of the discharge measurements, the magnitude
of the imposed change, the variability of the data, the
length of the record prior to and after treatment, the cer-
tainty with which we want to detect change (i.e., level of
significance), and the certainty of detecting change when
in fact there is a change (i.e., power) ( Loftis et al., 2001).
For small research catchments discharge is typically
measured at carefully- designed weirs with impermeable
cutoff walls, and in these situations the uncertainty in
discharge is generally believed to be on the order of only
2 -3 %. In contrast, the uncertainty of a single flow mea-
surement in a natural channel is generally believed to
be around 5% for moderate flows (Kennedy, 1983) and
substantially larger at higher flows (Dickinson, 1967).
The extrapolation from individual flow measurements
to seasonal or annual water yields may involve a further
degradation in accuracy because of the greater error in
measuring the highest flows, the uncertainty in the re-
lationship between stage and discharge, and the greater
potential for missing data (Kennedy, 1983). Hence the
uncertainty of flow measurements at a typical stream
gauging station is probably closer to 10 %, or several
times greater than in a research setting, and this greater
uncertainty limits our ability to detect a change in runoff
due to a given change in vegetation or land use.
The use of paired- watershed design greatly increases
our ability to detect change (Wicht, 1967; Loftis et al.,
2001), as the data from a control watershed can be used