eLearning

he reports between 10 and 100 kg N ha-1
y-1
, only
the repeated soil sampling in the harvested Walker
Branch sites had a strong experimental design. The
report of N accretion 120 kg N ha-1
y-1
came from
the Australia chronosequence, in which resampling
by the same investigators showed no net accretion
of N.
We note that the set of studies with the most
surprising results from strong experimental designs
all involve mixed hardwood forests. The intensive,
replicated, repeated sampling of the postharvest
hardwood sites in Michigan found amazing disap-
pearances of N, at the same time that a similar study
in harvested hardwood sites in Tennessee found
surprisingly high rates of N accretion. An undis-
turbed watershed in the Coweeta Hydrologic lab
found an unexplained, high rate of N loss, whereas
a similar undisturbed watershed in nearby eastern
Tennessee found no significant loss. The design and
description of each of these projects warrants rela-
tively high confidence in the estimated rates of N
change; either some real (and unexpected) changes
are occurring or experimental errors crept into the
studies in unexpected and undocumented ways.
We have three general recommendations and
conclusions. The first is simply that researchers who
deal with changes in soils over time (in chronose-
quences or multiple, repeated sampling of single
sites over time) should use an appropriate statistical
method for analyzing the functional relationship
between forest N content and time. A large portion
of the studies we reviewed treated different periods
of time as though each time period was a different
categorical treatment, ignoring the information that
may result from knowing that 10 years comes be-
fore 20 years, which in turn comes before 30 years.
A variety of time series approaches may be used to
evaluate soil change (such as linear regression,
curve-fitting, or orthogonal contrasts), all of which
would be more powerful at detecting time trends
than simple analyses of variance with time periods
treated as separate, independent treatments.
Our second recommendation is that we simply
have too few studies that have documented
changes in forest N content with rigor and preci-
sion. We recognize that it would be almost impos-
sible to obtain funds for a long-term project with
the simple goal of determining changes in forest N
content, so we recommend that projects that are
well funded for other goals consider attempting to
document forest N contents with high precision as a
part of larger objectives. Heroic efforts have been
invested in attempting to detect changes in mineral
soil N in watershed #5 at Hubbard Brook, but high
spatial variation in soils (and in rock contents) will
preclude precise tracking of changes over time that
are less than approximately 700–800 kg N/ha
(Johnson 1995). In contrast, the relatively uniform
soils under the loblolly pine plantations in the Cal-
houn Experimental Forest allowed 95% confidence
in changes of 250 kg N/ha or more. More effort
should be invested in testing ideas of N accretion in
sites with uniform, easily sampled soils. Studies that
examine changes in soils with uneven thicknesses
of horizons, changing bulk densities, and variable
content of coarse fragments need to use great cau-
tion and clear methods (such as Johnson 1995) to
have any hope of measuring even moderate rates of
change in soil nutrient pools.
Our last conclusion is that we think occult N
inputs are unlikely, based on the quality of the
available evidence. However, we suggest that this
“null” conclusion be considered as fairly tenuous,
owing to the paucity of well-designed studies to
address this fundamental question about the N cy-
cle and forest ecosystems.
ACKNOWLEDGMENTS
This review benefited substantially from discussions
with our colleagues, especially Martin Christ, Dan
Richter, Jennifer Knoepp, Chuck Rhoades, and
Dick Fisher. Support for this work came fromMcIn-
tire-Stennis appropriations to Colorado State Uni-
versity, the National Science Foundation (Grant
DEB 9816006 and DEB97-08521, Korea University,
and the Bonanza Creek Long-Term Ecological Re-
search Project funded at the University of Alaska by
the National Science Foundation