patrickbdevaney/mia9 · Datasets at Hugging Face

text stringlengths 0 6.44k
patches are ideal candidates for studies of patch dynamics at seascape spatial scales.
Two previous studies in Biscayne Bay demonstrated that seagrass seascapes adjacent to
freshwater canals were more fragmented than similar seascapes distant from canals, and experienced wide fluctuations in cover and fragmentation rates over time [22, 23]. Still, these prior
studies did not directly address nor quantified detailed patch dynamics that are important to
identify mechanisms of fragmentation and help forecast seascape stability patterns under different salinity scenarios. Patch dynamics such as changes in patch size structure as well as
patch mortality and growth rate are known to influence the stability and local extinction probability of habitats composed of terrestrial and marine plant species [22, 24–26]. Thus, this
study examined the long-term dynamics of seagrass/SAV patches in Biscayne Bay in association with freshwater deliveries by analyzing the historical response of SAV/seagrass patches of
different sizes within distinct salinity environments. We hypothesized that the rates of
Resilience of seagrass communities exposed to pulsed freshwater discharges
PLOS ONE \| https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 2 / 15
fragmentation and thus the long-term dynamics of patch-size structure, would be influenced
by the discharge of fresh water from canals, with areas closer to canals having seascapes with
higher fragmentation rates and patch-size structures dominated by smaller patches that can
compromise the long-term persistence of SAV/seagrass habitats.
Materials and methods
Study design
The hydrology and salinity patterns of western Biscayne Bay are influenced by the location
and flow rates of drainage canals. Areas near canals can exhibit extreme oscillations in salinity
levels and this pattern is heightened during the wet season (July to October) when freshwater
is released in pulses to drain the Florida Everglades and upstream urban and agricultural areas
[22].
Six sites were selected along the western shoreline of Biscayne Bay where the impacts of
CERP are concentrated. The sites considered ‘adjacent’ (Snapper Creek, Black Point Canal,
Convoy Point) were in proximity to canals with the highest freshwater discharge volumes
within Biscayne Bay (Fig 1). Paired ‘distant’ (Chicken Key, Black Point Lagoon, Turkey Point)
sites were randomly selected at distances > 1 km2 from a canal (Fig 1). Each site encompassed
a 500-m buffer around a location selected along the shoreline as described by Santos et al. [23].
Historical aerial photos of these six sites were assessed over nine periods, 6–13 years apart
from 1938–2009 based on the availability and quality of aerial imagery. Salinity data collected
using YSI instruments deployed in the vicinity of each site from 2010–2015 showed that sites
adjacent to canals had lower average salinity (24.4 g/L) compared to sites distant from canals
(29.3 g/L). The research described in this study was conducted through remote sensing and
GIS and thus did not require any scientific permits.
Spatial analyses
Seagrass maps were created using black-and-white aerial photographs obtained from local
government agencies. These images were processed to standardize their resolution, optical
properties, and sampling area. Images were geo-rectified using a United States Geological Survey topographic map as a spatial reference (for mapping details see [23]). Seagrass maps were
created by hand-digitizing individual seagrass patches at 1:2500 scale. For the purpose of this
study, individual seagrass patches were classified into the following five size classes: 1) Size 1
(<50 m2
); 2) Size 2 (50–100 m2
); 3) Size 3 (>100–500 m2
); 4) Size 4 (>500–2000 m2
); and 5)
Size 5 (>2000 m2
).
Patch dynamics
Population models based on size rather than age are particularly useful in describing the
dynamics of plants and clonal invertebrates [27, 28]. In this study, five seagrass patches from
each of the size class were randomly selected for each site at every time interval, as this was the
highest number of patches, on average, that would provide equal representation across all five
size categories consistently. To evaluate the fate of each patch between time steps, the GIS map
for the end of a time interval (t1) was superimposed on top of the map for the beginning of the
time interval (t0) (Fig 2). The fate of each of the five selected patches/polygons per size class
was recorded as growth, shrinkage, fragmentation, merging, or mortality using the following
rules:
1. For a fate to be classified as “growth”, a patch identified in t1 had to overlap with the original t0 patch and show an increase in area between t0-t1.
Resilience of seagrass communities exposed to pulsed freshwater discharges
PLOS ONE \| https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 3 / 15
Fig 1. Location of study sites for this project. The sites in green outline are considered “distant” to canals, with a mean distance of 2.8 (± 0.9)
km from canals. Sites in blue are considered “adjacent” to canals, with a mean distance of 0.5 (± 0.1) km the nearest canals. SC: Snapper
Resilience of seagrass communities exposed to pulsed freshwater discharges
PLOS ONE \| https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 4 / 15
2. Shrinkage was recorded in the opposite way; where the t1 patch had to be the only patch in
contact with the t0 patch and show a decrease in area between t0-t1.
3. Fragmentation was recorded if the original patch in t0 divided into more than one new
patch in t1.
4. Merging was recorded if distinct t0 patches joined together to form a new patch in t1.
5. Mortality was recorded when the original patch disappeared between t0-t1.
The fate data were used to develop the following transition matrix:
Nðtþ1Þ ¼ A � NðtÞ
where A is a Leslie matrix describing the probabilities of transition between size classes and
N(t) is the population vector that describes the number of individuals in each size category at
time t [29]. Transitions were expressed as proportions. For example, if two of the five size-1
patches identified in t0 grew to size 3 in t1, 0.4 would be recorded as growth from size 1 to size
3. If three of the size-3 patches from t0 shrunk to size 1 in t1, 0.6 would be recorded as
shrinkage.
Our model accounts for recruitment of new seagrass patches into the population by adding
recruits as a proportion of the existing patches in to to the first row in the matrix [29, 30]. To
estimate recruitment, the patches that appeared on previously unoccupied space in t1 and were
not in contact with any patches from t0 were identified and counted as recruits. The total number of recruits identified was divided by the total number of patches recorded at t0 to calculate
the per-patch recruitment rate that was added to the first row of the Leslie matrix. If, for example, 10 new patches (recruits) were detected in t1, this number was divided by the total number
of patches in to (e.g., 100) to provide a recruitment rate of 0.1. Recruitment rates were calculated on a yearly basis for standardization.
Merging was considered as a special case of growth into larger size categories, where the
growth transition probabilities for each patch size merging were adjusted by dividing by the
number of patches that merged. If, for example, one of the five (0.2) size-3 and one of the five
(0.2) size-4 patches merged to form a single size-5 patch, the growth probability for size 3 to
size 5 would be 0.2/2. Similarly, the growth probability for size 4 to size 5 would be 0.2/2. This
increases the abundance of size-5 patches and decreases the abundance of the smaller, merging
size classes proportionally to their contribution.
Fragmentation was treated as a special case of recruitment, where a larger patch can produce several smaller patches through fission. The model accounts for both the possibility that
the parent patch declines in size through fragmentation (by transitioning/shrinking into a
smaller size class) or remains within the same size class even after fragmentation (especially
true for larger size-5 patches that can undergo fragmentation and still retain their large size).
The fate of the parent patch is tracked as any other patch that can grow, shrink, or stay within
the same size category. For example, if one of the size-5 patches produced 50 new size-1
patches through fission, the per-patch fragmentation rate would be 50 divided by the t0 population (200) to provide the fragmentation rate of 0.25 to size class 1. To compare between sites,
annual rates of fragmentation were calculated.
The ‘popbio’ package in R was used for the analysis of the population dynamics and to calculate lambda (λ, eigenvalue) and the stable size-frequency distributions (eigenvectors) from
Creek, CK: Chicken Key, BL: Black Point Lagoon, BP: Black Point Canal, CP: Convoy Point, TP: Turkey Point. Red lines show the location of the

patches are ideal candidates for studies of patch dynamics at seascape spatial scales.

Two previous studies in Biscayne Bay demonstrated that seagrass seascapes adjacent to

freshwater canals were more fragmented than similar seascapes distant from canals, and experienced wide fluctuations in cover and fragmentation rates over time [22, 23]. Still, these prior

studies did not directly address nor quantified detailed patch dynamics that are important to

identify mechanisms of fragmentation and help forecast seascape stability patterns under different salinity scenarios. Patch dynamics such as changes in patch size structure as well as

patch mortality and growth rate are known to influence the stability and local extinction probability of habitats composed of terrestrial and marine plant species [22, 24–26]. Thus, this

study examined the long-term dynamics of seagrass/SAV patches in Biscayne Bay in association with freshwater deliveries by analyzing the historical response of SAV/seagrass patches of

different sizes within distinct salinity environments. We hypothesized that the rates of

Resilience of seagrass communities exposed to pulsed freshwater discharges

PLOS ONE | https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 2 / 15

fragmentation and thus the long-term dynamics of patch-size structure, would be influenced

by the discharge of fresh water from canals, with areas closer to canals having seascapes with

higher fragmentation rates and patch-size structures dominated by smaller patches that can

compromise the long-term persistence of SAV/seagrass habitats.

Materials and methods

Study design

The hydrology and salinity patterns of western Biscayne Bay are influenced by the location

and flow rates of drainage canals. Areas near canals can exhibit extreme oscillations in salinity

levels and this pattern is heightened during the wet season (July to October) when freshwater

is released in pulses to drain the Florida Everglades and upstream urban and agricultural areas

[22].

Six sites were selected along the western shoreline of Biscayne Bay where the impacts of

CERP are concentrated. The sites considered ‘adjacent’ (Snapper Creek, Black Point Canal,

Convoy Point) were in proximity to canals with the highest freshwater discharge volumes

within Biscayne Bay (Fig 1). Paired ‘distant’ (Chicken Key, Black Point Lagoon, Turkey Point)

sites were randomly selected at distances > 1 km2 from a canal (Fig 1). Each site encompassed

a 500-m buffer around a location selected along the shoreline as described by Santos et al. [23].

Historical aerial photos of these six sites were assessed over nine periods, 6–13 years apart

from 1938–2009 based on the availability and quality of aerial imagery. Salinity data collected

using YSI instruments deployed in the vicinity of each site from 2010–2015 showed that sites

adjacent to canals had lower average salinity (24.4 g/L) compared to sites distant from canals

(29.3 g/L). The research described in this study was conducted through remote sensing and

GIS and thus did not require any scientific permits.

Spatial analyses

Seagrass maps were created using black-and-white aerial photographs obtained from local

government agencies. These images were processed to standardize their resolution, optical

properties, and sampling area. Images were geo-rectified using a United States Geological Survey topographic map as a spatial reference (for mapping details see [23]). Seagrass maps were

created by hand-digitizing individual seagrass patches at 1:2500 scale. For the purpose of this

study, individual seagrass patches were classified into the following five size classes: 1) Size 1

(<50 m2

); 2) Size 2 (50–100 m2

); 3) Size 3 (>100–500 m2

); 4) Size 4 (>500–2000 m2

); and 5)

Size 5 (>2000 m2

).

Patch dynamics

Population models based on size rather than age are particularly useful in describing the

dynamics of plants and clonal invertebrates [27, 28]. In this study, five seagrass patches from

each of the size class were randomly selected for each site at every time interval, as this was the

highest number of patches, on average, that would provide equal representation across all five

size categories consistently. To evaluate the fate of each patch between time steps, the GIS map

for the end of a time interval (t1) was superimposed on top of the map for the beginning of the

time interval (t0) (Fig 2). The fate of each of the five selected patches/polygons per size class

was recorded as growth, shrinkage, fragmentation, merging, or mortality using the following

rules:

1. For a fate to be classified as “growth”, a patch identified in t1 had to overlap with the original t0 patch and show an increase in area between t0-t1.

Resilience of seagrass communities exposed to pulsed freshwater discharges

PLOS ONE | https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 3 / 15

Fig 1. Location of study sites for this project. The sites in green outline are considered “distant” to canals, with a mean distance of 2.8 (± 0.9)

km from canals. Sites in blue are considered “adjacent” to canals, with a mean distance of 0.5 (± 0.1) km the nearest canals. SC: Snapper

Resilience of seagrass communities exposed to pulsed freshwater discharges

PLOS ONE | https://doi.org/10.1371/journal.pone.0229147 February 21, 2020 4 / 15

2. Shrinkage was recorded in the opposite way; where the t1 patch had to be the only patch in

contact with the t0 patch and show a decrease in area between t0-t1.

3. Fragmentation was recorded if the original patch in t0 divided into more than one new

patch in t1.

4. Merging was recorded if distinct t0 patches joined together to form a new patch in t1.

5. Mortality was recorded when the original patch disappeared between t0-t1.

The fate data were used to develop the following transition matrix:

Nðtþ1Þ ¼ A � NðtÞ

where A is a Leslie matrix describing the probabilities of transition between size classes and

N(t) is the population vector that describes the number of individuals in each size category at

time t [29]. Transitions were expressed as proportions. For example, if two of the five size-1

patches identified in t0 grew to size 3 in t1, 0.4 would be recorded as growth from size 1 to size

3. If three of the size-3 patches from t0 shrunk to size 1 in t1, 0.6 would be recorded as

shrinkage.

Our model accounts for recruitment of new seagrass patches into the population by adding

recruits as a proportion of the existing patches in to to the first row in the matrix [29, 30]. To

estimate recruitment, the patches that appeared on previously unoccupied space in t1 and were

not in contact with any patches from t0 were identified and counted as recruits. The total number of recruits identified was divided by the total number of patches recorded at t0 to calculate

the per-patch recruitment rate that was added to the first row of the Leslie matrix. If, for example, 10 new patches (recruits) were detected in t1, this number was divided by the total number

of patches in to (e.g., 100) to provide a recruitment rate of 0.1. Recruitment rates were calculated on a yearly basis for standardization.

Merging was considered as a special case of growth into larger size categories, where the

growth transition probabilities for each patch size merging were adjusted by dividing by the

number of patches that merged. If, for example, one of the five (0.2) size-3 and one of the five

(0.2) size-4 patches merged to form a single size-5 patch, the growth probability for size 3 to

size 5 would be 0.2/2. Similarly, the growth probability for size 4 to size 5 would be 0.2/2. This

increases the abundance of size-5 patches and decreases the abundance of the smaller, merging

size classes proportionally to their contribution.

Fragmentation was treated as a special case of recruitment, where a larger patch can produce several smaller patches through fission. The model accounts for both the possibility that

the parent patch declines in size through fragmentation (by transitioning/shrinking into a

smaller size class) or remains within the same size class even after fragmentation (especially

true for larger size-5 patches that can undergo fragmentation and still retain their large size).

The fate of the parent patch is tracked as any other patch that can grow, shrink, or stay within

the same size category. For example, if one of the size-5 patches produced 50 new size-1

patches through fission, the per-patch fragmentation rate would be 50 divided by the t0 population (200) to provide the fragmentation rate of 0.25 to size class 1. To compare between sites,

annual rates of fragmentation were calculated.

The ‘popbio’ package in R was used for the analysis of the population dynamics and to calculate lambda (λ, eigenvalue) and the stable size-frequency distributions (eigenvectors) from

Creek, CK: Chicken Key, BL: Black Point Lagoon, BP: Black Point Canal, CP: Convoy Point, TP: Turkey Point. Red lines show the location of the