Beruflich Dokumente
Kultur Dokumente
April 2004
agggls
Office of teesearch and Development (NHFJ~RL), Atlantic
Ecolog~ Dir~sion, 27 Tarzwelt Drive, Narragansett, Rhode Island 02882
2 U.S. Environmental Protection Agency, t~esearch Triangle Park, N~rth Carolina 27711
s Department of Computer Science and Statistics, University of Rhode Island, Ir
Rhode
Island 02881
ABSTRACT: Do land use and cover characteristics of watersheds associatedwith small estuaries exhibit a strong enough
signal to make landscape metrics useful for identifying degraded bottom communities? We tested this idea with 58 pairs
of small estuaries ( < 2 6 0 km 2) and watersheds in the U.S. Mid-Atlantic coastal plain (Delaware Bay to Chesapeake Bay).
We considered 34 landscape metrics as potential explanatory variables and seven estuarine parameters as response
variables. We d e v e l o p e d three logistic regression models: one to calculate the probability of degraded benthic environmental quality (BEQ), as defined by chemical parameters, and two for the probability of degraded estuarine bottom
communities, o n e using a benthic index (BI) and a s e c o n d using the total number of bottom-dwdling species (TNBS,
consisting of benthic macroinvertebrates and fishes). We evaluated the discriminatory power of the m o d e l s with ROC
(receiver operating characteristic) curves of sensitivity and specificity. All three m o d e l s s h o w e d excellent discrimination
of high and low values. A m o d e l using the sum of all human land uses and percent wetlands correctly classified B E Q in
86% of the eases; low benthic index and low total number of bottom s p e c i e s were each associated with degraded B E Q
(p < 0.01). T h e BI model u s e d percent riparian urban, riparian wetlands, and agriculture on steep slopes (76% correct
classification) and correctly predicted high-low benthic index of an independent data s e t 79% of the time (p < 0.05).
T h e T N B S m o d e l used percent wedands, riparian wedands, and riparian agriculture (74% correct classification). Watersheds with higher percentages of urban and agricultural land uses were associated with lower benthic environmental
quality, benthic index, and biodiversity, whereas those with higher percentages of w e d a n d s w e r e associated with higher
numbers. A~ human d e v e l o p m e n t of watersheds increases, statistical prediction rules d e v e l o p e d from landscape metric~
could be a cost-effective m e t h o d to identify potentially threatened estuaries.
Introduction
tele:
401/782-3048;
2002).
Land use and cover characteristics of watersheds
can affect the condition of estuarine b o t t o m envir o n m e n t s and communities by s e d i m e n t contamination (Gomeleo et al. 1996; Paul et al. 2002; Morrisey et al. 2003) and have adverse effects on benthic organisms and communities (Porter et al.
1997; Dauer et al. 2000; Inglis and Kross 2000; Letberg et al. 2000; Morrisey et al. 2008). Increased
watershed d e v e l o p m e n t has b e e n associated with
lowered estuarine species diversity, altered food
webs, and altered benthic c o m m u n i t y composition
(Dauer et al. 2000; L e r b e r g et al. 2000). T h e m o r e
extensive literature on linkages between landscapes and freshwater biological effects, such as the
effect of urbanization on stream biological integrity (Karr and Ghu 2000), has b e e n recently reviewed by Gergel et al. (2002) and discussed by
e-mail:
283
284
s.s.
Hale et al.
ESTUARINE DATA
TABLE 1. Estuarine p a r a m e t e r s for 58 estuaries in the Vn-gin i a n Province study d a t a set. B e n t h i c p a r a m e t e r s are b a s e d on
the m e a n of t h r e e grabs.
Mean
Estuary area, k i n 2
Station d epth, m
# b e n t h i c species
# benthic organisms
B e n t h i c biornass, g
# fish species
Total # i n d i v i d u a l fish
B e n t h i c I n d e x (BI)
TNBS ~
33.3
4.3
9.8
178.7
0.15
5
136
0.29
15.0
1.9
1.5
0.0
0
0
0
0
2.20
0.3
148.9
10.6
32.3
2,888.8
2.42
11
739
8.48
41.3
# b e n t h i c species
munities, and fish tissue c o n t a m i n a n t s and external pathology. T h e VP s a m p l e d 425 probabilitybased stations d-ore C a p e Cod, Massachusetts, to
C a p e Henry, Virginia; MAIA sampled a r o u n d 900
stations f r o m the Delaware River to A l b e r m a r l e Pamlico Sound; and NCA has s a m p l e d t h o u s a n d s
of stations in all states. From these studies, we used
the small (<260 km ~) estuarine systems, which included subestuaries, small bays, tidal rivers, a n d
coastal ponds.
Benthic samples in VP a n d MAIA were collected
using a stainless steel Young grab that s a m p l e d a
surface a r e a of 440 cm ~ (Strobel et al. 1999; Kidd o n et al. 9003). T h e s a m p l e was sieved t h r o u g h a
0.5-ram screen and p r e s e r v e d in 10% formalinrose bengal solution. In the laboratory, the samples
were t r a n s f e r r e d to an ethanol solution, sorted by
species, c o u n t e d , a n d weighed. Fish were collected
by trawling on the b o t t o m 10 m i n u t e s against the
tide between 0.5 and 1.5 m s 1 (1--3 knots) with a
15-m, high-rise otter trawl with a 2.5-cm cod e n d
mesh. A l t h o u g h we r e f e r to fishes c a u g h t in the
otter trawl as b o t t o m species, several species were
c a u g h t that are not strictly associated with the bottom.
Paul et al. (2001) d e v e l o p e d an i n d e x of benthic
c o m m u n i t y condition (BI) for the VP that can be
used over the full r a n g e of salinity a n d s e d i m e n t
grain size f o u n d in the province. It consists of salinity-normalized G l e a s o n ' s D ( n u m b e r of species
divided by log of n u m b e r of individuals), salinityn o r m a l i z e d a b u n d a n c e of tubificid oligochaetes,
a n d a b u n d a n c e of spionid polychaetes. T h e index
is a c o n t i n u o u s variable, with negative values indicating d e g r a d e d sites a n d positive values indicating
n o n d e g r a d e d sites. It correctly classified 87% of
r e f e r e n c e a n d 90% of d e g r a d e d sites, and comp a r e d well with an i n d e p e n d e n t b e n t h i c i n d e x dev e l o p e d specifically for C h e s a p e a k e Bay (Ranasign h e et al. 2002).
For biological r e s p o n s e variables (Table 1), we
285
286
s.s.
Hale et al.
Fig. 1. Estua~ne stations of the EMAP Via-ginian Province (VP) study and the Mid-Atlantic Integrated Assessment (MAIA) study
for the Delaware Bay m Chesapeake Bay area.
of the variation in stream nutrients and s u s p e n d e d
s e d i m e n t (Jones et al. 2001; Gergel et al. 2002),
we considered t h e m to be potential e x p l a n a t o r y
variables for estuarine ecological condition. We
used the Analytical tools I n t e r f a c e for L a n d s c a p e
Assessments (ATTILA) software, described by E b e r t
et al. (2002), with the National H y d r o g r a p h i n g Dataset (USGS u n p u b l i s h e d data), to calculate riparian metrics for the New Jersey side of Delaware Bay,
which were n o t in the original data set of J o n e s et
al. (1997). We also calculated p o p u l a t i o n density
for all watersheds, using county-level 1990 census
data ( E n v i r o n m e n t a l Systems Research Institute
[ESRI] 1998). We used A r c / I n f o to intersect the
o u t p u t f r o m ATTILA with our delineated estuarine
watersheds. We calculated five additional metrics:
the ratio of watershed a r e a to estuarine area, the
ratio of watershed a r e a to estuarine v o l u m e , and
three landscape p a t t e r n metrics, based on the definitions of O'Neill et al. ( 1 9 9 9 ) - - a v e r a g e p a t c h pe-
TABLE 2.
287
Area, krn 2
Peiimeter, tml
N-index ~
Forest, %
Wetlands, %
Urban, %
Agriculture, pasture, %
Agriculture, crops, %
Agriculture, total, %
Barren, %
U..index ~
Agriculture, crops, steep b, %
Agriculture, pasture, steep, %
Agriculture, total, steep, %
Riparian ~ human, %
Riparian, barren, %
Riparian, ag, pasture, %
Riparian, ag, crops, %
Riparian, ag, total, %
Riparian, urban, %
Riparian, wetlands, %
Riparian, forest, %
Riparian, natural, %
Stream length, kin
Road length, km
Road densitya
Percent impervious e
Stream-road crossings
Population densityr, persons km 2
Watershed area/estuary area
Watershed area, krn2/estuary volume, rn ~
Average patch perkmeter-area ratios
Con tagione
Fractal dknensions
Mean
Median
Minimum
Maximum
572.0
115.1
52
40
13
13
23
11
34
1
48
1
4
5
27
1
11
6
17
9
32
42
73
691.9
1,480.7
5.0
6
0.4
12.3
51.5
0.19
1,074.4
44.1
1.046
225.6
78.7
52
38
7
5
20
9
31
1
48
1
1
2
22
0
9
4
14
8
27
42
77
197.7
542.2
2.3
3
0.3
0.21
11.4
0.04
1,075.9
43.7
1.045
21.7
21.3
9
7
1
1
0
3
8
0
10
0
0
0
5
0
0
1
4
0
5
3
24
3.7
26.5
0.6
0
0.0
0.0
0.6
0.01
1,036.1
34.4
1.039
4,170.7
440.3
90
73
83
83
52
29
70
6
91
7
22
25
76
6
30
14
39
62
91
76
95
7,586.2
9,902.1
12.4
51
1.7
332.4
788.0
3.18
1,124.4
61.1
1.055
N_index is the sum of all natural land uses and U_index is the sum of all human uses (developed and agricultural lands).
b Steep slopes are defined as those >3%.
c. Riparian is defined as a 30-rn buffer on each side of a stream.
d Road density is total road length divided by watershed area.
e Percentage of impervious cover, based on land use type.
Population density based on county-level 1990 data.
Based on definitions of O'Neill et al. (1999).
landscape metrics (the landscape data had a high
degree of collinearity). For example, natural index
( N - i n d e x ) is t h e d i r e c t c o n v e r s e o f t h e h u m a n u s e
i n d e x ( U _ i n d e x ) , so o n l y o n e is n e e d e d . T h e n we
e x a m i n e d c r o s s - c o r r e l a t i o n s o f all l a n d s c a p e a n d
e s t u a r i n e v a r i a b l e s a n d r e t a i n e d o n l y t h o s e signifi c a n t at p < 0.2. W e also r a n t-tests o n t h e i n d e p e n d e n t v a r i a b l e s u s i n g h i g h a n d l o w classes o f d e pendent variables and kept only those independ e n t v a r i a b l e s w h e r e t h e r e was a s i g n i f i c a n t d i f f e r e n e e (p < 0.2). W e e n s u r e d t h a t e a c h o f t h e
remaining landscape metrics could plausibly lead
to a n e f f e c t o n t h e e s t u a r i n e b o t t o m e n v i r o n m e n t
in some documented fashion, For example, urban
a r e a s , t h r o u g h d i r e c t d i s c h a r g e a n d r u n o f f , l e a d to
elevated sediment contaminants in nearby estuar i n e d e p o s i t i o n a l a r e a s ( C o m e l e o et al. 1996; M o r r i s e y et al. 2 0 0 3 ) . T h e l a n d s c a p e m e t r i c s d r o p p e d
were: area, perimeter, N_index, % barren metrics,
Model Devetopraent
E c o l o g i c a l r e s p o n s e s a r e o f t e n b i n a r y , s u c h as
species presence-absence or oxic-anoxic sediments,
o r c a n b e l o g i c a l l y d i c h o t o m i z e d , s u c h as c h e m i c a l
c o n c e n t r a t i o n a b o v e o r b e l o w a w a t e r q u a l i t y stan-
288
dard ( M u r t a u g h 1996). Such responses can be analyzed by logistic regression, which we used to calculate, given landscape metrics, the probability of
a d e g r a d e d estuarine b o t t o m community. Relationships between landscape metrics and aquatic systems often exhibit threshold responses (Gergel et
al. 2002). T h e m o d e l s were developed from the VP
calibration data set. We developed three logistic regression models, one with B E Q as the d e p e n d e n t
variable, one with gI, and one with TNBS. T h e dep e n d e n t variable, Y, can take on two values, Y - 1
if the event (a d e g r a d e d condition such as a low
BEQ_, gI, or TNBS) occurs and Y - 0 if the event
does not occur. Logistic regression generates an
estimate of the probability of (Y - 1) - p, for each
data p o i n t i - 1 . . . . . n. After assessing the goodness of fit of several multiple logistic regression
models we developed a rule for classification that
was based on a cutpoint k for the estimated 10t values. We used the value of k from the classification
table (SAS Institute 2002) that gave the highest total correct classification. This rule b e c a m e our indicator for the condition. O u r general rule was: if
ibm> k t h e n Y =
lorifib~< kthenY=
0;0.0 < k
< 1.0. Explanatory variables with p < 0.10 for
Wald's chi-square statistic were considered significant. We used the Akaike I n f o r m a t i o n Criterion to
select the best set of i n d e p e n d e n t variables for
each m o d e l and the H o s m e r - L e m e s h o w goodnessof-fit test ( H o s m e r and L e m e s h o w 2000; SAS Institute 2002) to d e t e r m i n e w h e t h e r the fitted m o d e l
was adequate.
n o n d e g r a d e d (high BEQ) stations. We used multiple logistic regression to try to distinguish the two
B E Q classes, using only the subsets of landscape
metrics as i n d e p e n d e n t variables. To see how the
biological metrics c o m p a r e d between the two B E Q
classes, we used a t-test to c o m p a r e the BI m e a n of
high B E Q stations with that of low B E Q stations.
We did the same for the total n u m b e r of b o t t o m
species. To examine how the composition of the
c o m m u n i t y assemblages varied by BEQ_, we used
n o n - m e t r i c m u l t i d i m e n s i o n a l scaling (NMDS),
with a Bray-Curtis similarity index on square roottransformed a b u n d a n c e data, and an analysis of
similarity, A N O S I M (Clarke and Warwick 2001) for
both the benthic c o m m u n i t i e s and the fish communities.
289
TABLE 5. Estuarine a n d watershed p a r a m e t e r s for 19 estuaries in the Mid-Atlantic I n t e g r a t e d A s s e s s m e n t study data set. L a n d s c a p e
m e t r i c s shown only for those metrics u s e d in the b e n t h i c i n d e x logistic regression.
Estum-y area, k m ~
Station depth, m
# b e n t h i c species
# benthic organisms
Benthic biomass, g
Benthic I n d e x (BI)
Riparian, wetlands, %
Riparian, u r b a n , %
Agriculture, total, steep, %
Mean
Median
6.6
2.5
21.0
253.5
0.23
-0.03
54
10
4
5.8
2.1
11.0
174.0
0.08
0.27
50
4
1
r e s p o n s e is positive. T h e specificity is the probability of a negative indicator, given that the true
r e s p o n s e is negative. A plot of sensitivity versus
specificity (the R O C curve), shows the probability
of detecting true signal (sensitivity) a n d false signal
(specificity). T h e a r e a u n d e r the R O C curve c provides a m e a s u r e of the indicator's ability to discriminate between the observations that have the response (a low BEQ, BI, or TNBS) a n d those that
do not. T h e a r e a u n d e r the curve is e x p e c t e d to
be - 0 . 5 for a n o n - i n f o r m a t i v e indicator and 1 for
a p e r f e c t indicator ( M u r t a u g h 1996). An a r e a >0.7
is c o n s i d e r e d acceptable discrimination and an
area ->0.8 has excellent discrimination ( H o s m e r
a n d L e m e s h o w 2000).
Landscape M e d i c
Mean
Mean
Percent Perce~
abe %r
~ge %r
Low
High
Qat%ory Category
BEQ
U-index
P e r c e n t a g e we tlan ds
63
4
43
19
29
29
BI
P e r c e n t a g e riparian u r b a n
P e r c e n t a g e riparian wetlands
P e r c e n t a g e total agric, steep slopes
14
27
7
6
35
5
53
55
58
TNBS
P e r c e n t a g e wetlands
P e r c e n t a g e riparian wetlands
P e r c e n t a g e riparian total agric.
6
22
19
19
41
15
55
51
51
Minimum
Maximum
1.6
0.6
1.0
15.0
0.01
-8.90
3
0
0
12.6
9.2
34.7
1,620.0
2.24
2.89
79
63
25
TABLE 5.
Model
point
Percera tage
Correct
Sensi
tivity
Sjoeci
ncuty
ROC c
Hemmer
Lemeshow
BEQ
BI
TNBS
29
58
51
0.68
0.46
0.54
86
76
75
79
67
84
93
81
65
0.95
0.81
0.85
0.38
0.79
0.57
Cur
290
s . s . Hale et al.
N
N
N D t~l i ~
I,iLI
D
D ED
D
D
Stress: 0.13
B. Fish
N
N ND
o$
DE)
N
[]
Stress: 0.15
Fig. 3. Non-pm-ametiic multidimensional scaling plots for
benthic community and fish community at degraded (D) stations (low BEQ) and nondegraded (N) stations (high BEQ).
Both communities showed a significant difference (p < 0.01,
ANOSIM) in community structure between degraded and nondegraded stations.
291
TABLE 6. Dominant benttfic invertebrate (top ten by abundance) and fish (top five by abundance) species at 15 nondegraded (high
BEQ) and 14 degraded (low BEQ) stations.
Nond%raded (High BEQ)
Abundance
Nemertine a
Annelida: Polychaeta
Heteromastus filiforwds
Mediomastus ambiseta
Streblo~pio benedicti
Paraprionospio pinnata
Annelida: Oligochaeta
Ofigochaeta
Abundance
67
129
87
67
62
85
Mollusca: Bivalvia
Macoma mitchelli
Arthropoda: Crustacea
Leptocheirus plum~los~s
Arthropoda: Malacosn-aca
Le~con american,s
C~athura polita
Hetero~ast~s filifi,rmis
Neantlzes succinea
Streblospio benedicti
154
32
42
Oligochaeta
Limnodril~s hoffmeisteri
Aulodrilus pig~ieli
Tubificidae
77
141
58
716
142
Macov~a mit&elli
Mus~li*zm si0p.
107
Leptodadrus plumulosus
37
62
482
58
50
Fish Species
Abundance
Fish Species
Abundance
Leiostom~s xan#~r~s
Trinectes ma~latus
CTnoscion regalis
Micropogonias undulatus
Peprilus alepidotus
421
401
91
55
22
M~rone americana
Dorosorna cepedianurn
Leiosto~ezs xantlaezrvzs
Ameiurus nebulosus
Trinectes ~aozlat,~s
1,600
235
176
125
122
estuarine s e d i m e n t c o n t a m i n a n t s ( C o m e l e o et al.
1996; Paul et al. 2002; Morrisey et al. 2008). O u r
results show that landscape metrics are also statistically related to the condition a n d biodiversity of
estuarine b o t t o m c o m m u n i t i e s (shown also by
Morrisey et al. 2008 for two u r b a n i z e d versus two
rural watersheds). L a n d s c a p e metrics a l o n e disc r i m i n a t e d low and high B E Q stations a n d the biological metrics were significantly worse at the low
B E Q stations. W h e r e B E Q w a s low, n u m e r i c a l d o m inants were pollution-tolerant or opportunistic species such as tubificid oligochaetes, including s
hoffmeisteri and the a m p h i p o d L. pluraulosus. T h e
n u m b e r s of m o r e pollution-sensitive species, such
as the bivalve M. mitchelti and the crustaceans Leucon americanus and Cyathura polita were r e d u c e d at
low B E Q c o m p a r e d to high B E Q stations. This is a
typical benthic c o m m u n i t y response, with crustaceans, e c h i n o d e r m s , and bivalves tending to d r o p
TABLE 7. Two-way contingency table for Mid-Atlantic Integrated Assessment (MAIA) benthic index.
Predicted MAIA Be nthic Index
NOI:
Degraded degraded
Actual MAIA
Benthic Index
Total
Degraded
Nondegraded
3
0
4
12
7
12
Total
16
19
292
s.s.
Hale et al.
e n v i r o n m e n t a l d e g r a d a t i o n - - w i t h slightly m o r e
species at high B E Q than at low B E Q stations and
a clear difference in assemblage c o m p o s i t i o n - - b u t
the results were m o r e a m b i g u o u s than for benthic
m a c r o i n v e r t e b r a t e s . Restricting o u r analysis to the
y e a r . r o u n d resident fish species listed by M u r d y et
al. (1997) did n o t clarify the story. White perch,
gizzard shad (Dorosoma cepediar~um), and brown
b u l l h e a d were a b u n d a n t at stations where B E Q was
low. O t h e r species that were a b u n d a n t where B E Q
was high, such as spot, h o g c h o k e r , a n d weakfish,
were not as a b u n d a n t where B E Q w a s low. N u m b e r
of fish species would be e x p e c t e d to be lower in
estuaries with hypoxic or anoxic b o t t o m waters, but
s o m e fish m a y briefly swim into a hypoxic area to
prey on oxygen-stressed invertebrates (Breitburg
2002). O u r criteria for d e t e r m i n i n g d e g r a d e d stations included o t h e r p a r a m e t e r s such as s e d i m e n t
c o n t a m i n a n t c o n c e n t r a t i o n s a n d s e d i m e n t toxicity,
a n d h o w fish r e s p o n d to these variables m a y be
m o r e complex. Rose (2000) n o t e d that the relationship b e t w e e n e n v i r o n m e n t a l quality and fishes
can be difficult to quantify; spatial h e t e r o g e n e i t y
of habitat can result in p o p u l a t i o n responses disp r o p o r t i o n a t e to e n v i r o n m e n t a l quality. O n the
o t h e r h a n d , fish species using s o u t h e r n N e w England estuaries as spavaling or n u r s e r y areas were
less a b u n d a n t in low quality habitat as c o m p a r e d
to m e d i u m quality habitat ( D e e g a n et al. 1997);
fish assemblages at a sewage outfall h a d significantly less species richness than control stations
(Smith et al. 1999); a n d fish assemblages in three
estuaries in s o u t h e r n California were m o r e predictable f r o m e n v i r o n m e n t a l variables than were
i n v e r t e b r a t e assemblages ( D e s m o n d et al. 2002).
Nets that capture younger, smaller, a n d p e r h a p s
m o r e s i t e - d e p e n d e n t fishes (e.g., D e e g a n et al.
1997) m a y provide a b e t t e r m e a s u r e than did the
2.5 c m - m e s h b o t t o m trawl we used.
An i m p o r t a n t caveat is that statistically a n d biologically m e a n i n g f u l associations between the indicator a n d r e s p o n s e are not necessarily causal.
T h e actual m e c h a n i s m s by which landscape characteristics lead to d e g r a d e d estuarine b o t t o m conditions have b e e n discussed by others (e.g., Valiela
et al. 1992; H o p k i n s o n and Vallino 1995; Basnyat
et al. 1999; D a u e r et al. 2000; J o r d a n et al. 2003),
a l t h o u g h m o r e study is n e e d e d (Gergel et al.
2002).
EVALUATION OF ESTUARINE AND
LANDSCAPE METRICS
Estuarine m e a s u r e m e n t s such as s e d i m e n t chemistry a n d b e n t h i c c o m m u n i t y structure can be expensive. Developing statistical p r e d i c t i o n rules with
readily available l a n d s c a p e data could be a costeffective way to screen for potentially d e g r a d e d estuaries. T h e logistic regression m o d e l d e v e l o p e d
with only VP data correctly predicted a low or high
MAIA BI 79% of the time. To m a k e a strong case,
however, a larger sample size that covers o t h e r
b i o r e g i o n s is n e e d e d . We p l a n to test the m o d e l
against o t h e r i n d e p e n d e n t estuaries within a n d
outside our study region w h e n NCA data b e c o m e
available, for they cover a m u c h b r o a d e r area. NGA
data will also be n e e d e d to try to validate the total
n u m b e r of b o t t o m species indicator. Before the logistic m o d e l s can be generally applied, they must
be evaluated in areas with a different mix of landscape characteristics a n d different prevalences of
estuarine responses.
No high BI MAIA estuaries were incorrectly predicted to be low by the g I model. However, it predicted a high BI for 4 of 7 MAIA estuaries that
293
2002).
L a n d s c a p e metrics of h u m a n use were associated
with d e g r a d e d benthic m a c r o i n v e r t e b r a t e a n d bottom fish assemblages in small estuaries. O u r results
indicate that landscape statistical p r e d i c t i o n rules
reflect a strong e n o u g h watershed signal to be a
cost-effective m e t h o d for a m o n i t o r i n g p r o g r a m to
use in deciding w h e t h e r to sample a particular estuary. Signal detection t h e o r y provides a rigorous
standard for evaluating potential ecological indicators and for j u d g i n g applicability to areas with
294
s.s.
Hale et al.
differing prevalences
of the condition.
With the
dramatic
global impact on biodiversity from land
use changes
that are anticipated
in coastal areas
during the coming
century
(Lerberg
e t al. 2 0 0 0 ;
S a l a e t al. 2 0 0 0 ; D e s m o n d
e t al. 2 0 0 2 ; L l a n s o e t al.
2002), even estuaries
that are now in relatively
good condition
face additional
pressure from increased watershed
development.
Both the need for
monitoring
a n d i t s c o s t will i n c r e a s e .
That will
make predictive models such as the logistic regression indicators
described
here even more important.
ACKNOWLEDGMENTS
We gratefully acknowledge the n u m e r o u s people who collected a n d a s s e m N e d the EMAP Virginian Province a n d MidAtlantic I n t e g r a t e d A s s e s s m e n t data. We t h a n k D. Ebert a n d
EPA's L a n d s c a p e Ecology B r a n c h for p r o d u c i n g the l a n d s c a p e
metrics a n d D. McGovern a n d J. C o p e l a n d for delineating the
x*~tersheds a n d o t h e r GIS work. T h e m a n u s c r i p t was materially
i m p r o v e d f r o m reviews by D. Dauer, M. Nicholson, M. Pelletier,
L. Meng, a n d K. Rahn. A l t h o u g h the r e s e a r c h described in this
artide h a s b e e n f u n d e d in part by the USEPA, it h a s n o t b e e n
subjected to Agency review-. T h e r e f o r e , it does n o t necessarily
reflect the views o f the Agency. T h i s is c o n t r i b u t i o n n u m b e r
AED-03-016 of the USEPA, Office o f R e s e a r c h a n d Developmerit, National H e a l t h a n d E n v i r o n m e n t a l Effects Research
Laboratory, Atlantic Ecology Division, Narragansett, R h o d e Island.
LITERATURE C I T E D
295