252
dry and occult deposition. Lovett et al. (1992) also
found that estimates of sulfur (S) deposition at Hubbard Brook were significantly higher using the
catchment mass balance method compared to
throughfall or inferential measurements. Similarly,
Shepard et al. (1989) found that estimates of dry
deposition inputs of S provided by a catchment
level budget were significantly higher than estimates by regressions applied to wet-only deposition
or models of atmospheric concentrations of S coupled to deposition velocities. Thus, although
throughfall measures of dry deposition have the advantage of being relatively inexpensive and easy,
they may underestimate true deposition. Although
it is not clear which measurement technique is more
accurate, the catchment budget can provide either
verification of an independent measure or else evidence that a particular measurement approach is in
error.
Stream Outputs
The classic small watershed approach to ecosystem
budgets ideally uses long-term data from gauged
streams as the major output measurement (Likens
1985). Techniques for sampling streams for representative seasonal and annual estimates of hydrology and nutrient flux can be found in Likens et al.
(1977) and Stottlemyer (1987). A detailed discussion of stream gaging sites and types of stations can
be found in Reinhart and Pierce (1964). Methods
for accurate and precise hydrologic sampling are
necessarily site- and climate-specific, and thus few
set rules apply. Flumes might be needed for steep
slopes where high sediment yields are common,
while V-notch weirs are more accurate in systems
without high sediment loads. While some sites use
proportional, automatic samplers for collecting
stream water, other sites such as Hubbard Brook
have found that weekly grab samples are equally as
accurate in their ecosystems (W. Martin, personal
communication). A detailed analysis of sampling
frequency requirements in different catchment
types for accurate representation can be found in
Semkin et al. (1994).
Multiple years of data are critical to reduce the
chance that input-output budgets are biased by hydrologic and climatic variability (Williams and Melack 1997a), as the retention of nutrients within the
ecosystem can change with variations in the hydroKate Lajtha
logic (Lajtha et al. 1995, Seely et al. 1998) or temperature regime (Mitchell et al. 1996). However,
even long-term data cannot accurately determine
climatic links to variations in stream water chemistry, and thus a paired, or replicated, watershed approach is a preferred goal if manipulative experiments are to be performed. Clearly, finding or
measuring a control, or reference, watershed is not
always possible, although most long-term ecological research (LTER) sites with watersheds make
some attempt at replication. Experiments involving
effects of forest harvest, sewage sludge disposal, or
experiments within the stream ecosystems at the
Hubbard Brook Experimental Forest provide some
of the best examples of this approach. More often,
a single control and a single manipulated watershed
are measured, limiting true statistical power (Chorover et al. 1994) although differences might be
large enough to be convincing even without much
replication (e.g. Kahl et al. 1993). Paired-catchment
data can be augmented with pre- and postdisturbance comparisons, but there is always the possibility that interannual variability could be falsely interpreted as an effect of the manipulation (Williams
and Melack 1997b). Local "control" watersheds
cannot, of course, serve as controls for the analysis
of regional disturbances, such as climate change or
increased anthropogenic deposition of nutrients or
pollutants. In these cases, large regional assessments can be performed (e.g., Stoddard 1994) or
else annual discharge-concentration relationships
can be separated into trajectories of time, discharge,
and ion concentration to determine processes that
regulate streamwater concentrations (Williams and
Melack 1997a). Due to funding and time constraints, however, most watershed studies are performed without much replication or over long time
spans, and thus cannot reliably address biogeochemical trends.
Because stream water chemistry integrates many
processes over a large spatial scale and often over
a large and poorly understood time scale, it is difficult and often misleading to directly relate stream
water chemistry to processes within the forest
stand. However, most watershed budgets have attempted to attribute differences in inputs and outputs to retention processes within the terrestrial
ecosystem. Denitrification of N leached from the
terrestrial ecosystem can be significant, variable,
and high, depending on the presence and charac-
dry and occult deposition. Lovett et al. (1992) also
found that estimates of sulfur (S) deposition at Hubbard Brook were significantly higher using the
catchment mass balance method compared to
throughfall or inferential measurements. Similarly,
Shepard et al. (1989) found that estimates of dry
deposition inputs of S provided by a catchment
level budget were significantly higher than estimates by regressions applied to wet-only deposition
or models of atmospheric concentrations of S coupled to deposition velocities. Thus, although
throughfall measures of dry deposition have the advantage of being relatively inexpensive and easy,
they may underestimate true deposition. Although
it is not clear which measurement technique is more
accurate, the catchment budget can provide either
verification of an independent measure or else evidence that a particular measurement approach is in
error.
Stream Outputs
The classic small watershed approach to ecosystem
budgets ideally uses long-term data from gauged
streams as the major output measurement (Likens
1985). Techniques for sampling streams for representative seasonal and annual estimates of hydrology and nutrient flux can be found in Likens et al.
(1977) and Stottlemyer (1987). A detailed discussion of stream gaging sites and types of stations can
be found in Reinhart and Pierce (1964). Methods
for accurate and precise hydrologic sampling are
necessarily site- and climate-specific, and thus few
set rules apply. Flumes might be needed for steep
slopes where high sediment yields are common,
while V-notch weirs are more accurate in systems
without high sediment loads. While some sites use
proportional, automatic samplers for collecting
stream water, other sites such as Hubbard Brook
have found that weekly grab samples are equally as
accurate in their ecosystems (W. Martin, personal
communication). A detailed analysis of sampling
frequency requirements in different catchment
types for accurate representation can be found in
Semkin et al. (1994).
Multiple years of data are critical to reduce the
chance that input-output budgets are biased by hydrologic and climatic variability (Williams and Melack 1997a), as the retention of nutrients within the
ecosystem can change with variations in the hydroKate Lajtha
logic (Lajtha et al. 1995, Seely et al. 1998) or temperature regime (Mitchell et al. 1996). However,
even long-term data cannot accurately determine
climatic links to variations in stream water chemistry, and thus a paired, or replicated, watershed approach is a preferred goal if manipulative experiments are to be performed. Clearly, finding or
measuring a control, or reference, watershed is not
always possible, although most long-term ecological research (LTER) sites with watersheds make
some attempt at replication. Experiments involving
effects of forest harvest, sewage sludge disposal, or
experiments within the stream ecosystems at the
Hubbard Brook Experimental Forest provide some
of the best examples of this approach. More often,
a single control and a single manipulated watershed
are measured, limiting true statistical power (Chorover et al. 1994) although differences might be
large enough to be convincing even without much
replication (e.g. Kahl et al. 1993). Paired-catchment
data can be augmented with pre- and postdisturbance comparisons, but there is always the possibility that interannual variability could be falsely interpreted as an effect of the manipulation (Williams
and Melack 1997b). Local "control" watersheds
cannot, of course, serve as controls for the analysis
of regional disturbances, such as climate change or
increased anthropogenic deposition of nutrients or
pollutants. In these cases, large regional assessments can be performed (e.g., Stoddard 1994) or
else annual discharge-concentration relationships
can be separated into trajectories of time, discharge,
and ion concentration to determine processes that
regulate streamwater concentrations (Williams and
Melack 1997a). Due to funding and time constraints, however, most watershed studies are performed without much replication or over long time
spans, and thus cannot reliably address biogeochemical trends.
Because stream water chemistry integrates many
processes over a large spatial scale and often over
a large and poorly understood time scale, it is difficult and often misleading to directly relate stream
water chemistry to processes within the forest
stand. However, most watershed budgets have attempted to attribute differences in inputs and outputs to retention processes within the terrestrial
ecosystem. Denitrification of N leached from the
terrestrial ecosystem can be significant, variable,
and high, depending on the presence and charac-