Continental Shelf Research 232 (2022) 104629
12
Hamner, 1988) never considered potential coastal influences. In winter
and early spring, the richer coastal waters are often mixed with the
offshore waters and therefore contribute to the production beyond the
continental shelf up to 10 km from the coast, as shown by our biomass
index.
4.3. Modelling approach
As previously mentioned, the Algerian coastal waters were divided
into two classes: HBC (High-Biomass Coastal Zones) and LBC (LowBiomass Coastal Zones) (Fig. 11). The HBCs consequently refer to the
highest values of Chl-a and I B , and the LBCs to the lowest values. Their
separation into two classes was visually optimised by defining specific
thresholds for Chl-a and I B variables, summarised in Table 3 (response
variables). We defined three qualitative variables: the presence of wadis,
the type of coast (sandy/rocky), and the existence of a Bay. The City is
defined as a quantitative variable with four levels: 0 (no-city), 25 k, 75 k,
and 200 k inhabitants.
Table 2 summarises the respective characteristics of LBC (numbered
1–16) and HBC (numbered 1–15) as manually selected in Fig. 11. The
presence of Cities, Wadis and Bays are positively related to the detection
of High-biomass coastal areas. At the same time, the type of coast appears to be irrelevant, mainly compared to the presence of a Bay.
Linear qualitative models (General Linear Models) were performed
to evaluate the interactions between either the coastal Chl-a concentration or the biomass index (I B ) and the four explanatory variables, as
shown in Table 3. The two specific areas HBC-1 and HBC-8 were
excluded from the modelling because these two areas are specifically
influenced by aquaculture floating cages that are not associated with the
explanatory variables. A general model (m1 in Table 3) is first tested by
combining all seasons to test the separation between HBC and LBC, as
presented in Fig. 11. This model explains 67% of the variability, with a
unique City effect. The Wadi effect is absent, probably because of its
association with City. On the opposite, in winter, the Chl-a response
variable (m2 model, 79%) is primarily associated with the presence of a
Wadi, then to City. In contrast, the I B response variable (m3 model) is
mainly related to a Bay and City presence. The winter I B model (m3) is
very similar in explaining the biomass variability (77%), with a dominance of City presence (as previously with Chl-a) as well as to a significant Bay effect. In these winter models, the Wadi effect is only
evidenced by the Chl-a variable that most reflects the influence of local
enrichments rather than their spatial extension, associated with the
biomass index (I B ), highlighting the Bay effect.
During summer (low-biomass season), the Chl-a based model (m4 in
Table 3) shows only 57% of explanation, with a unique Wadi effect
despite the generally low flow of wadis in winter (Fig. 12). The equivalent model for the biomass index (m5) explains 78% of the variability,
with a dominance of Bay presence, while the Wadi effect is still present.
Fig. 12. Seasonal variability of the Mazafran and the El-harrach wadis flows (in m
3
s
−1
, orange line) and the corresponding Chl-a concentration (green line) averaged
at the isobath <50 m from 2003 to 2012: (a) in the Bou-Ismaïl bay and (b) in the Algiers bay. The maps (a’ and b’) show the Chl-a yearly average at each location.
Table 2
Summary of the characteristics of the LBC (numbered 1–16) and HBC (numbered 1–15) regions as manually selected in Fig. 11. Both Chl-a and I B
variables were averaged for each LBC (white rectangles, 1–16) and HBC (green rectangles, 1–15) of Fig. 11. The City size [0–3] is defined by
respectively: 0 (no city), 1: [0 50k] inhabitants (small red dot), 2: [50k- 100k] (intermediate red dot), 3: >100k (large red dot). The Coast-type is
either Sandy(S) or Rocky(R). The small black dots represent aquaculture cages.
R. Harid et al.
12
Hamner, 1988) never considered potential coastal influences. In winter
and early spring, the richer coastal waters are often mixed with the
offshore waters and therefore contribute to the production beyond the
continental shelf up to 10 km from the coast, as shown by our biomass
index.
4.3. Modelling approach
As previously mentioned, the Algerian coastal waters were divided
into two classes: HBC (High-Biomass Coastal Zones) and LBC (LowBiomass Coastal Zones) (Fig. 11). The HBCs consequently refer to the
highest values of Chl-a and I B , and the LBCs to the lowest values. Their
separation into two classes was visually optimised by defining specific
thresholds for Chl-a and I B variables, summarised in Table 3 (response
variables). We defined three qualitative variables: the presence of wadis,
the type of coast (sandy/rocky), and the existence of a Bay. The City is
defined as a quantitative variable with four levels: 0 (no-city), 25 k, 75 k,
and 200 k inhabitants.
Table 2 summarises the respective characteristics of LBC (numbered
1–16) and HBC (numbered 1–15) as manually selected in Fig. 11. The
presence of Cities, Wadis and Bays are positively related to the detection
of High-biomass coastal areas. At the same time, the type of coast appears to be irrelevant, mainly compared to the presence of a Bay.
Linear qualitative models (General Linear Models) were performed
to evaluate the interactions between either the coastal Chl-a concentration or the biomass index (I B ) and the four explanatory variables, as
shown in Table 3. The two specific areas HBC-1 and HBC-8 were
excluded from the modelling because these two areas are specifically
influenced by aquaculture floating cages that are not associated with the
explanatory variables. A general model (m1 in Table 3) is first tested by
combining all seasons to test the separation between HBC and LBC, as
presented in Fig. 11. This model explains 67% of the variability, with a
unique City effect. The Wadi effect is absent, probably because of its
association with City. On the opposite, in winter, the Chl-a response
variable (m2 model, 79%) is primarily associated with the presence of a
Wadi, then to City. In contrast, the I B response variable (m3 model) is
mainly related to a Bay and City presence. The winter I B model (m3) is
very similar in explaining the biomass variability (77%), with a dominance of City presence (as previously with Chl-a) as well as to a significant Bay effect. In these winter models, the Wadi effect is only
evidenced by the Chl-a variable that most reflects the influence of local
enrichments rather than their spatial extension, associated with the
biomass index (I B ), highlighting the Bay effect.
During summer (low-biomass season), the Chl-a based model (m4 in
Table 3) shows only 57% of explanation, with a unique Wadi effect
despite the generally low flow of wadis in winter (Fig. 12). The equivalent model for the biomass index (m5) explains 78% of the variability,
with a dominance of Bay presence, while the Wadi effect is still present.
Fig. 12. Seasonal variability of the Mazafran and the El-harrach wadis flows (in m
3
s
−1
, orange line) and the corresponding Chl-a concentration (green line) averaged
at the isobath <50 m from 2003 to 2012: (a) in the Bou-Ismaïl bay and (b) in the Algiers bay. The maps (a’ and b’) show the Chl-a yearly average at each location.
Table 2
Summary of the characteristics of the LBC (numbered 1–16) and HBC (numbered 1–15) regions as manually selected in Fig. 11. Both Chl-a and I B
variables were averaged for each LBC (white rectangles, 1–16) and HBC (green rectangles, 1–15) of Fig. 11. The City size [0–3] is defined by
respectively: 0 (no city), 1: [0 50k] inhabitants (small red dot), 2: [50k- 100k] (intermediate red dot), 3: >100k (large red dot). The Coast-type is
either Sandy(S) or Rocky(R). The small black dots represent aquaculture cages.
R. Harid et al.