JAWRA - Ag Conservation

• Project Type: General OWC
• Title: Designing Impact Assessments for Evaluating Ecological Effects of Agricultural Conservation Practices on Streams
• Date: 12/15/2009
• Water Conservation - Doc Type: Reports

DESIGNING IMPACT ASSESSMENTS FOR EVALUATING ECOLOGICAL EFFECTS OF AGRICULTURAL CONSERVATION PRACTICES ON STREAMS
experimental design (Galeone, 1999). Control streams
are necessary to isolate the effect of conservation
practices from other factors (Downes et al., 2002).
Control streams are similar to laboratory controls in
that both serve as the untreated treatment for the
experiment (Hulbert, 1984). Yet, control streams dif-
fer from laboratory controls in that environmental
conditions of control streams cannot be regulated to
reduce the influence of confounding factors, but envi-
ronmental conditions can be regulated in laboratory
controls (Hulbert, 1984). Hence, selection of control
streams plays a critical role and they need to be
selected to ensure the effects of conservation practices
are not confounded by the inherent differences among
control and treatment streams.
We also feel the a priori development of specific
habitat criteria to guide the selection of control and
treatment streams is a good practice. However, spe-
cific selection criteria will differ among experimental
designs. Site selection within before— after- control-
impact, before — after, and impact vs. control sites
designs (Table 2) involves consideration of selected
physical features to ensure the similarity among con-
trol and treatment streams. Conversely, site selection
within space- for -time (Table 2) and gradient designs
(Table 2) involves consideration of selected physical
features that need to differ among streams to create
the treatment effect.
An individual stream (i.e., tributary or subwater-
shed) is typically accepted as a replicate in stream
ecology due to interdependency in abiotic and biotic
conditions among sites within streams (Downes et al.,
2002). For example, considering three sites within
one control stream and three sites within one treat-
ment stream (i.e., the one having the conservation
practice implemented) as a design having three repli-
cates of each group intended for analysis of variance
type analyses ( ANOVA) results in spatial pseudorep-
TABLE 2. Depiction of Replication Before and After
Implementation of Conservation Practices Within Different
Experimental Designs Used in Impact Assessments.
Experimental Design
Before
After
Replicated before—
❑ ❑ ❑ ■ ■ ■
❑ ❑ ❑ ■ ■ ■
after - control- impact
Replicated before —after
NEE
■ ■ ■
Replicated space- for -time
❑ ❑ ❑ ■ ■ ■
Unreplicated before—
❑ ■
❑ ■
after- control- impact
Impact vs. control sites
❑ ❑ ❑ ■
❑ ❑ ❑ ■
Unreplicated before —after
■
■
Gradient
■
Note: Boxes represent control streams (unshaded boxes) and treat-
ment streams (shaded boxes).
lication (Hulbert, 1984). Sampling one site within a
stream five times before and after the implementa-
tion of a conservation practice with the intent of ana-
lyzing the data with ANOVA analyses is considered
temporal pseudoreplication (Hulbert, 1984). Instead,
three replicates for an ANOVA analyses would be
achieved by sampling three control streams and three
treatment streams. Additionally, sampling sites
should be selected randomly from a pool of available
sites when possible.
The use of replicated experimental designs with
appropriate controls and treatments represents the
ideal situation that can be logistically difficult to
achieve. However, we feel investigators should strive
to obtain the ideal situation first, and then accept
alternative designs if logistical barriers are encoun-
tered. Replicated before — after - control- impact (BACI)
designs are considered to be a powerful approach in
evaluating the impacts of anthropogenic habitat
alterations (Downes et al., 2002). Replicated BACI
designs involve sampling multiple control streams
and multiple treatment streams simultaneously (or
nearly so) before and after implementation of conser-
vation practices (Table 2). Replicated before —after
designs and replicated space- for -time designs are also
acceptable experimental designs despite their implicit
assumptions. Replicated before —after designs involve
sampling multiple treatment streams before and after
implementation of conservation practices (Table 2)
and assume the observed change between the before
and after sampling periods is due to the conservation
practice, not temporal variation. Replicated space -for-
time designs involve sampling control and treatment
streams without any before data and assume that
conditions within control streams represent condi-
tions of treatment streams in the absence of the con-
servation practices (Table 2). The number of
replicates to be sampled can be determined given an
estimate of the variability of the response variables
and the minimum change in response variables
among experimental treatments that the investiga-
tors wish to detect (Green, 1979; Zar, 1984; Loftis
et al., 2001).
BACI designs were originally intended to be used
in unreplicated studies having a single control and
treatment stream (Green, 1979; Table 2). Unreplicat-
ed BACI designs are also referred to as paired
watershed designs and evaluate the impact by sam-
pling control and treatment streams multiple times
before and after the implementation of the treatment
(Table 2). An unreplicated before —after study would
sample one treatment stream multiple times before
and after implementation of a treatment (Table 2). A
special case of the BACI design is the impact vs. con-
trol sites design (Stewart -Oaten and Bence, 2001)
where the investigator samples one treatment stream
JOURNAL OF THE AMERICAN WATER RESOURCES ASSOCIATION 871 JAWRA