Page 4 - 4022883

702 BIOLOGICAL CONSERVATION 141 (2008) 699–709
was assessed by classifying fishes within three size categories
(i.e. small, medium, large) on the basis of the maximum total
length attained by each species (Froese and Pauly, 2006). Fish
biomass was evaluated using size distributions and length–
weight relationships from the literature (Francour, 1990; Fro-
ese and Pauly, 2006). Early juvenile stages (settlers and re-
cruits) were not taken into account.
Five research groups worked within the same research
framework, using standardized methods in all the 15 loca-
tions investigated. As the level of personal experience may
bias the results of fish visual censuses (Williams et al.,
2006), meetings and an intensive training were done for all
participants, to standardize the procedures and the observers’
ability to collect accurate data, before sampling started.
2.2. Treatment of data
Methods derived from meta-analysis were used to examine
and summarize the general response of fish to protection.
As visual censuses were done at several fished and unfished
sites and sampling was repeated through time, mean values
were used to approximate average conditions in space and
time (see Guidetti and Sala, 2007).
We examined the response to protection at species/family
level (Mugilidae and Atherinidae include species difficult to
identify visually at species level), at target vs non-target fish ished areas. A ratio of zero, instead, means that densities
level, and at functional level (i.e. trophic groups). Fish taxa
were pooled into functional groups based on their trophic po-
sition because fishing disproportionately targets species at
higher trophic levels (Pauly et al., 1998), and recovery from
fishing potentially includes increased abundances or biomass
of high-level predators and shifts in trophic structure (Micheli the evidence that effective reserves and, more generally, areas
et al., 2004). Each species/family was assigned to one of eight
trophic groups using the available information about diet and
size in the database ‘‘FishBase’’ (Froese and Pauly, 2006), and
in Mediterranean studies (Sala, 2004; Guidetti and Sala,
2007): (1) large piscivores, (2) small piscivores, (3) inverte-
brate-feeders of group 1 (major predators of sea urchins), (4)
invertebrate-feeders of group 2 (whose diets seldom include
sea urchins), (5) small cryptobenthic carnivores, (6) detriti-
vores, (7) planktivores and (8) herbivores (see Figs. 3 and 4
for species groupings). We split invertebrate feeders into
two groups because of the major role a few fish species can
have in regulating sea urchin populations and hence poten-
tially affecting the entire benthic community (Sala et al.,
1998; Guidetti, 2006). Piscivores included species feeding
exclusively on fishes and species feeding on both fishes and
invertebrates (Micheli et al., 2004).
We first considered in our analyses all reserves to look for
possible general responses. Then, we grouped the reserves
into three categories based on the level of enforcement. Cat-
egorizing enforcement at each reserve required obtaining
information about (1) the frequency of illegal fishing within
the reserves, and (2) the efficacy of the reserve personnel,
the coast guard or other marine police forces in doing an ac-
tive surveillance against illegal activities. This information
was directly collected by the researchers involved in the pro-
ject, and/or gathered by questioning the reserve personnel
and local people. The relative enforcement categories were
high (poaching very occasional if any, patrol very active and
continuous), medium (illegal fishing occurring but limited by
infrequent surveillance) and low (common illegal fishing
and virtually inexistent surveillance) (Table 1). Categorization
was obtained by first assigning a score to surveillance and
poaching for any single marine reserve in terms of percentage
of days per year when there was an active surveillance (<25,
25–75, >75%, corresponding to score values of 0, 1 and 2,
respectively) and events of poaching (<25, 25–75, >75%, corre-
sponding to scores of 2, 1 and 0, respectively). Then, the prod-
uct of surveillance and poaching scores was calculated and
the enforcement category assigned with 0 = low, 1–2 = med-
ium and 4 = high enforcement.
We quantified the effects of protection within reserves as
the natural logarithm of the ratio between the values of the
response variable (i.e. fish density and biomass) in reserves
and fished conditions as response ratios, ln R (Hedges and
Olkin, 1985; Micheli et al., 2004). Data were thus normalized
and the response to protection examined independently of
the absolute densities at each location. As estimations of
average values can be affected by sampling effort, we calcu-
lated weighted means using the natural logarithm of the total
area covered by the censuses from which the estimates were
obtained (Mosquera et al., 2000). Positive response ratios indi-
cate greater density and/or biomass of species or trophic
groups in unfished than in fished areas, whereas negative val-
ues indicate greater values in fished areas compared to unf-
are similar between reserves and fished conditions. Averages
of the mean response ratios were considered significantly dif-
ferent from zero (i.e. there is a significant protection effect)
when the 95% confidence limits around the mean do not over-
lap with zero (Micheli, 1999 and references therein). Based on
characterized by null/low levels of exploitation can host par-
ticularly high fish biomass (Friedlander and DeMartini, 2002;
McClanahan et al., 2007; Stevenson et al., 2007), we also esti-
mated total fish biomass within the reserves and fished areas
at the 15 locations investigated. We then calculated the rela-
tionship between total fish biomass within the reserves and
the enforcement level.
As reported above, the transition from macroalgae to bar-
rens can be enhanced by the removal of predators of sea urch-
ins, i.e. D. sargus and D. vulgaris (Sala et al., 1998; Hereu et al.,
2005; Guidetti, 2006). A threshold density of $12 adult Diplodus
fish 125 mÀ2 was found to maintain sea urchin population
density under the threshold ($8–9 urchins mÀ2) critical for
triggering community shifts (Guidetti and Sala, 2007). There-
fore, only those reserves where conditions (e.g. effective
enforcement and compliance and/or habitat availability) are
appropriate to host sufficiently dense populations of Diplodus
may have the potential to recover from barrens back to mac-
roalgal beds or maintain flourishing macroalgal beds. We thus
evaluated this potential by assessing density of Diplodus in the
reserves in relation to enforcement.
3. Results
Across all locations combined, total fish density was on aver-
age 1.15 times greater in reserves than in fished areas
(ln R = 0.16 ± 0.17; 95% CI) (Fig. 2A). The lower the enforcement