Page 4 - debris2015

4 M. Angiolillo et al. / Marine Pollution Bulletin xxx (2015) xxx–xxx
Photos were analyzed by means of Image J software (http://rsb-
web.nih.gov/nih-image), using laser beams as a scale. In order to
assess the differences among the three studied regions, relative
abundance (with a not normal distribution) was verified by the
non-parametric Kruskal–Wallis H test and the post hoc Nemenyi-
Damico-Wolfe-Dunn test.
The distance of each area from the nearest coast was also
measured.
The impact of benthic debris was assessed and classified in 4
levels: (i) covering, when debris covered the organisms; (ii) abra-
sion, when the tissues of the organisms were injured; (iii) hanging,
when debris was under tension between rocky obstacles but
apparently did not injure any organism; (iv) lying, when debris
was sitting on the bottom and did not impact the organisms.
The eventual fouling on marine debris by macro-benthic organ-
isms was examined and the most common colonizing taxa were
recognized. Thus, based on the number of taxa growing on debris
items, three degrees of colonization were identified: none (0 taxa),
moderate colonization (1–3 taxa), heavy colonization (>3 taxa).
The number of dead sessile organisms, not directly impacted by
debris, was also recorded.
In order to evaluate the impact of debris, logistic models with
mixed effects were used. The outcome was the presence or absence
of a certain impact. Therefore, the probability of observing an
impact of some kind was modeled. The logarithm of the odds of
this probability was assumed to be a linear function of fixed effects
associated with debris and other predictors, with a dive-specific
intercept. The latter was used to take into account the effects of
unobserved factors, and capture the dependence among the obser-
vations realized within the same dive. The dive-specific intercept
was assumed to arise from a zero-centered Gaussian distribution,
as in usual generalized linear mixed models. Bivariate association
models were estimated at first, and then the final linear multivar-
iate model with mixed effects was obtained through forward step-
wise selection. Bivariate logistic models were also used to analyze
the region specific effects of marine debris impacts. All logistic
models are used to estimate predictor’s effects. These are summa-
rized as odds-ratios. The odds ratio is a measure of effect which can out the most significant abrasive action (Table 2; Fig. 5B–F), pro-
be interpreted as the fold change in probability of a presence when
estimated with and without the predictor of interest. Conse-
quently, an OR > 1 indicates that the presence of the predictor of
interest increases the probability of observing a certain damage;
while an OR < 1 indicates that the presence of the predictor
decreases the probability of observing a certain damage.
The effect of depth and distance from the coast on the number
of marine debris items was also analyzed using a Poisson regres-
sion model with mixed effects and with adjusted coefficients.
The number of items observed can be naturally assumed to arise
from a Poisson distribution. The logarithm of the expectation of
this Poisson distribution was once again assumed to be a linear
function of predictors (i.e., depth and distance from the coast),
associated with fixed effects, and to have a Gaussian distributed
dive-specific intercept.
The software R version 2.14 was used to carry out this analysis,
in combination with package lme4, which is an R package for fit-
ting and analyzing linear, nonlinear and generalized linear mixed
models.
3. Results
3.1. Abundance and distribution of debris
6.03 km2 of rocky bottom were explored by means of 69 video
transects in 26 areas of the three regions (Table 1). The initial
and final depth of each transect varied between 30 and 300 m,
Please cite this article in press as: Angiolillo, M., et al. Distribution and assessment of marine debris in the deep Tyrrhenian Sea (NW Mediterranean Sea,
Italy). Mar. Pollut. Bull. (2015), http://dx.doi.org/10.1016/j.marpolbul.2014.12.044
depending on the geographical characteristic of each location.
The length of each transect varied between 100 and 900 m.
A total of 3018 photos were analyzed and 368 benthic marine
debris items were recorded. The number of areas without any sign
of debris was negligible.
The highest median value of debris abundance was found in
Campania (0.12 debris mÀ2, ranging from 0.02 to 0.16 debris
mÀ2), while the lowest value was recorded in Sardinia (0.03 debris
mÀ2, 0.01–0.09 debris mÀ2). Sicily showed the widest range of
debris abundance (0–0.3 debris mÀ2) and the median value of
abundance was 0.09 debris mÀ2 (Figs. 2A and 3). Significant differ-
ences of debris abundance were recorded between Sardinia and
the other two regions (Kruskal–Wallis H = 8.9487, p < 0.05; post
hoc Nemenyi test p < 0.05; Fig. 2A). Similar results were obtained
considering only the lines (post hoc Nemenyi test p < 0.01)
(Fig. 2B).
The distance of the surveyed sites from the coast ranged from
0.065 to 16 NM. Bivariate Poisson regression models showed that
only the presence of lines was function of the distance from the
coast (p < 0.001 and O.R. = 1.2), while the presence of plastic was
related to the depth (p < 0.001 and O.R. = 2.48).
Overall, fishing gears were the dominant source of debris (89%):
lines made up the most significant portion (62.5%), followed by
nets (24.4%) and pots (2.1%). Plastic and other debris were occa-
sionally found, representing respectively 5% and 6% of the samples.
In particular, Sicily showed a higher occurrence of fishing lines
(79% of analyzed debris), while in Campania and Sardinia this value
was 57% and 55% respectively (Fig. 4).
3.2. Impact of debris
More than half of debris (54.5%) was observed in contact with
sessile invertebrates, either covering them (17.2%) or causing abra-
sion (37.3%). The remaining portion of debris (45.5%) was recorded
lying on the bottom (26.2%) or hanging from the rocks (19.3%)
without producing any apparent injury to sessile organisms.
The results of logistic model showed that lines and nets carried
ducing the progressive removal of the tissues of entangled organ-
isms. Nets and plastic carried out mainly a covering action (odds
ratio value was 1200). Nets, for example, were found snagged on
rocks, entangling or covering benthic organisms (Fig. 5B and C),
while plastic bags were observed enveloping some organisms.
Hanging debris was represented only by nets and lines (odds ratio
value 444) (Fig. 5I), while lying debris was made up mainly of glass
bottles, cans, tires or rigid sacks (odds ratio value 3156) that gen-
erally do not impact the organisms (Fig. 5G and H), but provide a
secondary substrate or a refuge for others (e.g. Fig. 5J and K).
The effects of the models estimated for Campania and Sicily
were not significantly different, while both were significantly dif-
ferent from the ones found in Sardinia for all the impact typologies,
in particular for hanging debris (Table 3).
Ischia site (c5) in Campania and other three banks (sc3, sc5, sc7)
in Sicily were the areas with the greatest relative abundance of
debris determining a heavy impact (covering and abrasion action)
on the organisms, as showed in Fig. 3.
Gorgonians were the most commonly affected organisms
(53.0%), followed by red coral (23.5%), antipatharians (14.3%),
sponges (6.0%) and other invertebrates (3.2%). This pattern was
similar, but with a specific composition for the three regions
(Table 4). The areas explored in Campania were characterized by
coralligenous outcrops. Here, the large amount of debris, mainly
lines and nets, abraded and covered especially the common gorgo-
nians Paramuricea clavata (Risso, 1826) and Eunicella cavolinii
(Koch, 1887) (%73%) (Fig. 5B). In Sardinia, a greater number of dif-
ferent taxa were affected (Table 4). In Sicily, the most impacted