23. Large-Scale Water Manipulations
water migration to control edge effects. While this
can be an effective and appropriate solution, it can
also generate experimental artifacts for studies of
large plants. Root barriers may reduce the "true"
rooting volume and lead to unintended changes in
root/shoot relationships of large trees. In experiments without barriers, the issue of water migration
between plots and root growth beyond plot boundaries is a serious potential problem. Hanson et al.
(1995, 1998) addressed this issue by choosing very
large plots with predefined buffer zones to allow
for potential overlap of water or roots. During the
course of their experiment, incidental treefalls
within treated plots showed that the effective rooting area under a tree was highly correlated with
each tree's crown spread. This observation suggests
that a conservative approach for edge effects would
be to allow a minimum of two mean crown widths
between water manipulation plots. However, this
"rule-of-thumb" should not substitute for good surveys of the water distribution patterns throughout
any large-scale water manipulation study.
Statistical Replication
Because of costs and logistics associated with
stand- or catchment-level manipulation experiments, true replication is often very limited or lacking in large-scale water manipulation experiments.
This lack of replication leads to the unfortunate
consequence of researchers having to resort to subsampling within treated areas or "pseudoreplication" (Eberhardt and Thomas 1991). Recognizing
pseudoreplication in any large-scale manipulation
study design is critical (Hurlbert 1984). Without
this recognition, simple differences in site characteristics might be overlooked and be totally confounded with the interpretation of the treatment
response.
The practical and scientific need for large contiguous albeit unreplicated water manipulation
structures must be preceded by serious evaluation
of the uniformity of the pretreatment site variables (i.e., soil type and vegetation distribution,
slope, etc.). Although unreplicated experimental
designs are clearly not ideal, pseudoreplication
has been recognized as a reasonable limitation
and an acceptable approach when costly experimental designs are being initiated for answering
349
scientific questions that can not be addressed at
smaller scales (Eberhardt and Thomas 1991).
Eberhardt and Thomas (1991) recommend that
unreplicated experiments be supported by adequate sampling of site environmental parameters
(including climatic conditions), comparable ambient areas, and pretreatment observation of critical variables.
Conclusions
Large-scale water manipulation experiments represent the only logical choice for scientific
questions demanding an understanding of integrated responses of plants and soils in a natural
microclimate. Large-scale manipulations were defined as field experiments that are of sufficient size
and complexity to handle questions of individual
plant response, interplant interactions, as well as
stand-level carbon, water, and nutrient cycling responses. Because large-scale manipulations are intended to retain all features of natural forest stands
or ecosystems except those being manipulated (i.e.,
precipitation inputs), extreme care must go into the
planning and design of the experimental approach
so that normal physical, chemical, and climatic features of the forests or ecosystems are retained.
A commitment to multi-year maintenance and
operation of large-scale water manipulation facilities should be demonstrated during the planning
and design of such studies. Maintaining large-scale
manipulations over multiple years minimizes the
initial costs of construction to levels that are comparable to more traditional experimental designs.
Multi-year studies also provide an opportunity to
see the impacts of a given manipulation against natural year-to-year variations in weather, thus expanding the scope of the planned manipulations.
Finally, it is important to understand that sustained
treatments on large-scale water manipulation studies may be the only way that we can detect responses that are normally manifested only slowly
in a natural ecosystem (e.g., changes in stored carbohydrates or nutrient cycling processes).
Acknowledgments This research is sponsored by
the Program for Ecosystem Research, Environmental Sciences Division, Office of Health and Envi-
water migration to control edge effects. While this
can be an effective and appropriate solution, it can
also generate experimental artifacts for studies of
large plants. Root barriers may reduce the "true"
rooting volume and lead to unintended changes in
root/shoot relationships of large trees. In experiments without barriers, the issue of water migration
between plots and root growth beyond plot boundaries is a serious potential problem. Hanson et al.
(1995, 1998) addressed this issue by choosing very
large plots with predefined buffer zones to allow
for potential overlap of water or roots. During the
course of their experiment, incidental treefalls
within treated plots showed that the effective rooting area under a tree was highly correlated with
each tree's crown spread. This observation suggests
that a conservative approach for edge effects would
be to allow a minimum of two mean crown widths
between water manipulation plots. However, this
"rule-of-thumb" should not substitute for good surveys of the water distribution patterns throughout
any large-scale water manipulation study.
Statistical Replication
Because of costs and logistics associated with
stand- or catchment-level manipulation experiments, true replication is often very limited or lacking in large-scale water manipulation experiments.
This lack of replication leads to the unfortunate
consequence of researchers having to resort to subsampling within treated areas or "pseudoreplication" (Eberhardt and Thomas 1991). Recognizing
pseudoreplication in any large-scale manipulation
study design is critical (Hurlbert 1984). Without
this recognition, simple differences in site characteristics might be overlooked and be totally confounded with the interpretation of the treatment
response.
The practical and scientific need for large contiguous albeit unreplicated water manipulation
structures must be preceded by serious evaluation
of the uniformity of the pretreatment site variables (i.e., soil type and vegetation distribution,
slope, etc.). Although unreplicated experimental
designs are clearly not ideal, pseudoreplication
has been recognized as a reasonable limitation
and an acceptable approach when costly experimental designs are being initiated for answering
349
scientific questions that can not be addressed at
smaller scales (Eberhardt and Thomas 1991).
Eberhardt and Thomas (1991) recommend that
unreplicated experiments be supported by adequate sampling of site environmental parameters
(including climatic conditions), comparable ambient areas, and pretreatment observation of critical variables.
Conclusions
Large-scale water manipulation experiments represent the only logical choice for scientific
questions demanding an understanding of integrated responses of plants and soils in a natural
microclimate. Large-scale manipulations were defined as field experiments that are of sufficient size
and complexity to handle questions of individual
plant response, interplant interactions, as well as
stand-level carbon, water, and nutrient cycling responses. Because large-scale manipulations are intended to retain all features of natural forest stands
or ecosystems except those being manipulated (i.e.,
precipitation inputs), extreme care must go into the
planning and design of the experimental approach
so that normal physical, chemical, and climatic features of the forests or ecosystems are retained.
A commitment to multi-year maintenance and
operation of large-scale water manipulation facilities should be demonstrated during the planning
and design of such studies. Maintaining large-scale
manipulations over multiple years minimizes the
initial costs of construction to levels that are comparable to more traditional experimental designs.
Multi-year studies also provide an opportunity to
see the impacts of a given manipulation against natural year-to-year variations in weather, thus expanding the scope of the planned manipulations.
Finally, it is important to understand that sustained
treatments on large-scale water manipulation studies may be the only way that we can detect responses that are normally manifested only slowly
in a natural ecosystem (e.g., changes in stored carbohydrates or nutrient cycling processes).
Acknowledgments This research is sponsored by
the Program for Ecosystem Research, Environmental Sciences Division, Office of Health and Envi-