We analyzed chemical and biological variables from C-assessments from Sulfjorden, a relatively small area with a high density of large salmon cage farms, situated in the archipelago of the Frøya island on the coast of Central Norway (Figure 1). The average velocity of the water current (measured 4m above the seabed) ranges from 3 to 8 cm/s, with frequent occurrences of maximum water current higher than 25 cm/s (Keeley et al., 2019). According to the individual C-assessment reports and Keeley et al. (2019), the sediments is relatively uniform, consisting of firm, noncohesive medium to coarse sand, with some unconsolidated shell fragments of irregular form. A total of 27 C-assessments reported from 16 commercial Atlantic salmon farms that operated within this region from 2016 to 2022 were available (Table 1).

C-assessments were carried out by accredited companies in accordance with Norwegian regulation NS 9410, which governs the monitoring of benthic impacts from salmon cage farms (Standards Norway, 2016). Triplicate sampling was conducted at each station. The results were recorded in the Norwegian Directorate of Fisheries database (https://www.fiskeridir.no/Tall-og-analyse/AApne-data). The environmental variables derived from multiple C-assessment reports were comprehensively analyzed in the present study.

Sediment samples were collected using a Van Veen grab (0.1m²) at multiple locations beneath and at varying distances from the fish farm. Locations included the local impact zone (about 30 meters from the net pens), the intermediate zone (30 to 500 meters from the sea cages), and the edge of intermediate zone (approximately 500 meters from the sea cages) (Figure 2). Each sample were analyzed to determine the selected environmental variables, including pH, redox potential (Eh), organic matter content, carbon (C), phosphorus (P), nitrogen (N), copper (Cu), zinc (Zn), and benthic macrofauna richness and abundance.

The pH and Eh were measured on board using platinum electrodes and either glass electrodes or ISFET (Ion Sensitive Field Effect Transistor), respectively. The level of C, N, P were analyzed according to Norwegian standard NS-EN ISO 16665 and NS-EN ISO 5667-19, while the Zn and Cu were analyzed according to Norwegian standard 11885. The macrofauna were sorted and kept by utilizing sieves with circular apertures varying in diameter from 1.0mm (Standards Norway, 2016).

The accumulated feed consumption is also presented in the C-assessment report, which refers to the total feed use from the start of the production cycle (fish release) to the sampling day.

All organisms were quantified and identified to the lowest taxonomic level possible by bodies certified for making C-assessment. Their taxonomic classification was confirmed using the World Register of Marine Species (WoRMS) database (WoRMS Editorial Board, 2023). At each sampling station, both richness and abundance were analyzed. Richness refers to the number of different species, while abundance represents the total number of individuals.

Macrofaunal species were classified using Norwegian sensitivity Index (NSI) into five ecological groups (EG): sensitive species (EG 1), indifferent species (EG 2), tolerant species (EG 3), opportunistic species (EG 4), and pollution indicating species (EG 5) (Borja et al., 2000).

This study uses taxa data from C-assessments to calculate various faunal index scores, including the Shannon–Wiener Index (H’), Hurlbert Index (ES100), AZTI Marine Biotic Index (AMBI), and Norwegian Sensitivity Index (NSI). The corresponding scores and Ecological Quality Status (EQS) classifications: High, Good, Moderate, Poor, and Bad, as described in Table 2.

All spatial and statistical analyses were conducted in the Python 3.9 programming environment (Van Rossum and Drake, 2009), using the pandas (McKinney, 2010), GeoPandas (Jordahl, 2014), SciPy (Virtanen et al., 2020), and statsmodels (Seabold and Perktold, 2010) software library. The plots were performed using ArcGIS Pro (Version 2.7, release 2021), Python Matplotlib (Hunter, 2007), and seaborn (Waskom, 2021) software library. The results of variables are reported as means ± 1 standard error (SE). Differences in variables among sampling stations were assessed using Kruskal–Wallis test, followed by Dunn’s post-hoc test for multiple comparisons, with a significance level set at 0.05.

We used Principal Component Analysis (PCA) to investigate the relationship between chemical variables and benthic macrofaunal communities across different EQS. Regression analysis was then applied to assess correlations between the variables, while correlation strength and significance were assessed using Spearman’s nonparametric correlation coefficient (rho) and its corresponding p-value. The significance limits were set at 0.05. Regression curves were fitted using linear regression, with the model determined based on the best R² value.

Our data analysis revealed pronounced spatial variations in chemical and biological variables among the assessed zones ranging from 30m to 500m from the cages in our study area (Figures 1, 3). Significant differences were observed among the three zones in levels of N, P, Zn and Cu in sediments, as well as the richness and abundance of benthic macrofauna (all p< 0.01; Figures 3D–I). The mean molecular N:P ratio of sediments was found to be 3.10, 4.45, 5.78, and 5.84 at local impact zone (~30 m), intermediate zone (~200 m), edge of intermediate zone (~500 m), and the Reference station (~ 1000m), respectively. The ratio showed significant difference among all zones (Kruskal–Wallis test; p< 0.001). Specifically, the levels of N, P, Zn, Cu and the N:P ratio were found to be significantly different between the local impact zone (~ 30m) and the intermediate zone (~ 200m). However, no significant differences between the intermediate zone (~ 200m) and the edge of intermediate zone (~ 500m) were found (all p > 0.05) as well as between the edge of intermediate zone (~ 500m) and the reference station (~ 1000m) (all p > 0.05; Figure 3).

However, the variables pH (Kruskal–Wallis test, p = 0.11, Figure 3A) and Eh (Kruskal–Wallis test, p = 0.3, Figure 3B) did not show significant differences among any of the three sampling stations. It is noteworthy that the average C levels showed decreasing values with increasing distance from the sea cages, but this difference was not statistically significant (Kruskal–Wallis test, p = 0.04, Figure 3C). The statistical analysis suggested relatively good state of the benthic environments, when pH or Eh proved inadequate in detecting differences, the level of N (Kruskal–Wallis test, p = 0.02, Figure 3D), P (Kruskal–Wallis test, p< 0.001, Figure 3E)), Zn (Kruskal–Wallis test, p< 0.001, Figure 3G), and Cu (Kruskal–Wallis test, p< 0.001, Figure 3E) may have greater potential for detecting the subtle impact of organic enrichment on benthic chemistry.

Each C-assessment included information on the sampling time, the start time of each production cycle and the estimated accumulated use of feed during the corresponding production cycles (c.f., Table 1). These data, along with the assessment results, enabled us to investigate spatial variations and variations across production phases in benthic environments. The statistical analysis revealed a pronounced correlation between feed consumption and the level of N and P in the benthic environments across multiple sampling locations (Figure 4).

In the local impact zones (~30 m, C1), the levels of N (p = 0.02; Figure 4A) and P (p< 0.001; Figure 4D) in the benthic environment significantly increased as the accumulated use of feed increased. However, in the intermediate zones (~200 m, C3) and the edge of intermediate zone (~500 m, C2), there was no significant correlation found between the accumulated feed use on and the level of N (both p > 0.05; Figures 4B, C). In contrast, a significant correlation (p< 0.001) was observed between the accumulated feed use and the level of P in the intermediate zones (~200 m, C3) (Figure 4E), although no significant correlation was found at the edge of intermediate zone (~500 m, C2) (Figure 4D).

The molecular N:P ratios were found to decrease with the increasing accumulated feed consumption (Figures 4G–I). This trend was found statistically significant in the local impact zones (~30 m, C1) and the intermediate zone (~200 m, C3) (p< 0.05; Figures 4G, H), though no significant correlation was found at the edge of intermediate zone (~500 m, C2) (Figure 4I).

To sum up, the statistical analysis indicated a positive correlation between accumulated feed consumption and the presence of biogenic wastes in the benthic environment within the production zone (< 30m).

The biological state of the benthic environment in the study areas was assessed using various faunal index scores, including the Shannon–Wiener Index (H’, Equation 1), Hurlbert Index (ES₁₀₀, Equation 2), AZTI Marine Biotic Index (AMBI, Equation 3), and the Norwegian Sensitivity Index (NSI, Equation 4). The resulting scores were classified into Ecological Quality Status (EQS) categories—High, Good, Moderate, Poor, and Bad—according to the EU Water Framework Directive (EU Directive, 2000/60/EC).

Spearman’s nonparametric correlation analysis was used to assess the correlations between individual faunal index scores (Figure 5). The statistical analysis revealed that all faunal index scores were significantly correlated with each other (p< 0.001), with the NSI (Figures 5B, F) exhibiting a slightly higher Spearman’s correlation (rho) value compared to AMBI (Figures 5A, E). Additionally, we calculated the faunal index scores for each sampling station and classified EQS (as defined in Table 2) across three sampling stations, as presented in Table 3. The results indicated that in the local impact zone (~30 m, C1), the EQS ranged from Moderate to Good, whereas in the intermediate zone (~200 m, C3) and at the edge of the intermediate zone (~500 m, C2), the EQS ranged from Good to High (Table 3). Furthermore, the coefficient of variation (CV) for the faunal index score exhibited a decreasing trend with increasing distance from the cage facility. In the local impact zone (~30 m, C1), the CV ranged from 44.8% to 73.6%; in the intermediate zone (~200 m, C3), it ranged from 22% to 55%; and at the edge of the intermediate zone (~500 m, C2), it ranged from 12.7% to 46%.

As the findings suggest that NSI may better represent the Norwegian benthic environment, we used this index for further analysis of macrofauna distribution among ecological groups (EG) across different zones. A total of 895 species of benthic macrofauna were identified in the study area, of which 397 have been grouped into five EG based on the Norwegian Sensitivity Index (NSI). These groups include Sensitive- species (EG 1), Indifferent species- (EG 2), Tolerant species- (EG 3), Opportunistic species- (EG 4), and Pollution-indicating species (EG 5), as presented in Table 4. Additionally, Table 5 summarizes the top 50 taxa in terms of abundance and their distribution across multiple sampling locations. In the local impact zone (~30 m), the benthic macrofauna was dominated by EG 5, primarily Tubificoides benedii and Capitella capitata. In the intermediate zone (~ 200m), there was a noticeable decline in EG 5, while EG 4 species, mostly polychaetes, became dominant. Approaching the edge of the intermediate zone (~ 500m), both EG 4 and EG 5 continued to decline, while the population of EG 1 and EG 2 species increased, resembling the species composition at the reference site where the abundance of all EG were relatively equal, with EG 1 and EG 2 becoming dominant.

Figure 6 shows the spatial variations in the level of phosphorus (P) and benthic macrofauna, categorized into EG, at each sampling station across 16 salmon farms within the study area from 2016 to 2022. EG5 species were primarily concentrated in the local impact zone (~30 m), accounting for about 80% of the total abundance. They were also slightly present in the intermediate zones (~200 m) during the post-stage, accounting for 4% of the total abundance. The variation in EG5 abundance was positively correlated with the level of P in the sediments (p< 0.001), whereas EG1 (p = 0.003) and EG2 (p = 0.012), were negatively correlated with P level. By comparison, EG3 (p = 0.55) and EG4 (p = 0.07) showed no significant correlation with sediment P levels. For EG1 and EG2 species, pronounced differences were observed between the local impact zone (3% of the total abundance) and the intermediate zones (23-27% of the total abundance). However, there was no apparent difference in distribution of EGs between the edge of the intermediate zone and the reference zone.

We used Spearman’s nonparametric correlation to analyze the relationship between biological and chemical variables among all sampling locations in the study regions (Figure 7). Spearman’s correlation (rho) (as shown by the values in the Figure 7) revealed that the species richness and individual abundance of macrofauna in the benthic ecosystem was significantly correlated with the contents of P, Cu and Zn in the sediments. Statistical analysis indicated a significant decrease in richness with increasing levels of P, Cu, and Zn, while the abundance showed a significant increase (all p< 0.05; Figure 7). Despite the statistically significant relationship, the low absolute values of the correlation coefficients (rho) value suggest weak correlations between the level of C, N, P, Cu, and Zn and biodiversity. There are therefore also other factors that have affected richness not considered in the present study.

We applied Principal Component Analysis (PCA) to analyze the correlation between chemical and biological states across a dataset of 110 sampling locations collected between 2016 and 2022. The biological states were assessed using the NSI method and categorized into five EQS levels, as described in Section 2.3 (Table 2). The result revealed that the first two components (PC) accounted for a substantial portion of the total variance, explaining 93% of the EQS. PC1, which represented 79% of the variance, was primarily influenced by P, with a minor influence from Zn. PC2, responsible for 14% of the variance, was mainly influenced by N, with a slight influence from Cu (Figure 8).

The Norwegian regulation NS 9410 includes both B- and C-assessments. B-assessments are conducted in the production zone (< 30m), while C-assessments are carried out at multiple sampling stations ranging from 30m to 500m, with additional supporting variables (Figure 2; Standard Norge, 2016). All environmental surveys conducted in the production zone of farms within our study area (< 30m; Figure 2), using B-assessments, consistently reported favorable conditions classified as State 1 (Very good; Table 1). Given this, we might not have expected signs of organic impact in the transition zone (30–500 m; Figure 2) with C-assessments. However, data analyses from C-surveys still revealed subtle indications of organic load, despite the pH and Eh variables suggesting otherwise. This could be due to the effective dispersion of particles away from the production zone, preventing accumulation directly beneath the plant, but allowing deposition further away. However, none of the individual C-assessments did result in a lower classification in the original surveys whereas our analyses, which look more at a holistic assessment of all variables, could find subtle signs of organic enrichment. Finding more ways to assess the health of the bottom ecosystem is useful, as analyzing benthic fauna manually is costly and time consuming.

Our data analysis of chemical and biological variables revealed a notable correspondence, with similar spatial variation across sampling locations (Figures 3, 4). This finding suggests that the results are reliable and further indicates that C-assessments—which combines multiple chemical and biological variables—provided a reliable method for monitoring aquaculture operations’ impact on benthic environmental conditions. The sediment contents of C, N, P, Cu, and Zn were identified as principal components for Ecological Quality Status (EQS; Figure 8), although it is noted that Cu is typically associated with antifouling coatings rather than organic enrichment from feed. However, previous studies have found sulphide- and O₂ levels, often in combination with Cu and lice treatment chemicals, as the main drivers of macrofauna inhabitability (Gray et al., 2002; Hargrave et al., 2008; Mayor et al., 2009).

Our study revealed that salmon farming in our study area primarily influenced the benthic environment within a local impact zone extending up to 30 meters from the facility, with the effects diminishing as the distance from the salmon cages increased. At the edge of the intermediate zone, the impact may become statistically indistinguishable from the reference site (all p > 0.05). Our study suggested that using N, P, Cu, and Zn, and biological variables can enhance the ability to detect changes in benthic ecosystems, which may vary dependent on the distance from salmon farms. The above chemical variables from the C-assessment detected spatial benthic impacts beyond the immediate impact zone, enabling the observation of a diminishing impact gradient with increasing distance from the cage facility. The benthic macrofaunal communities exhibited pronounced spatial shifts in the benthic environment, and such changes were detectable at varying distances, up to 500m from the salmon cage, throughout the production cycle (Figure 6). These findings aligned well with the observations on nitrogen (N) and phosphorus (P) levels, as mentioned in Sections 3.1–3.2 (Figures 3, 4).

Considering the potential dispersion and resuspension of particles beyond the production zone, the B-assessment, which is confined within 30m and ideally conducted near the cage edge, may be insufficient for monitoring the full extent of organic enrichment impacts (Keeley et al., 2013; Broch et al., 2017). Nevertheless, although the B-assessment protocol includes fewer measurements than the C-assessment, it involves a combination of multiple variables, such as pH, redox potential (Eh) and sensory measurements which have been established as responsive and reliable variables for assessing benthic condition under heavier loads (Standards Norway, 2016; Wang and Olsen, 2024). The B-assessment, with its higher survey frequency and focus on the construction zone, provides a cost-effective means of quickly detecting environmental changes. In contrast, the C-assessment, incorporating a broader range of chemical and biological variables, can detect subtle impacts further from the impact zone (30–500 m from the cage facilities). It can also help identify the actual impact zone more accurately, while offering both spatial and production-phase data to assess the interactions between aquaculture operations and the benthic environment.

Therefore, combining B- and C-assessments provide useful information for evaluating and following up the ecological assimilation efficiency of biogenic wastes at farming sites. However, since C-assessments is sometimes seen as time-consuming and costly, it may be worth considering whether the sampling distance should be reduced if long-term environmental conditions show only minor changes, with most occurring near the facility.

In our study area, there are 16 farms operating with multiple production cycles, and the accumulated feed is reset to zero and recalculated when a new production cycle begins (as shown in Table 1). Each production cycle typically lasted 1–2 years, with a time gap occurring between two consecutive cycles (Wang and Olsen, 2024). It should be noted that none of the 16 farms were newly established; all had been operating commercially for several years prior to our study. The positive correlation between accumulated feed usage and nitrogen and phosphorus levels on the seafloor (as shown in Figure 4) suggests that the benthic impact on the seafloor was significantly lower when accumulated feed was minimal (i.e., at the beginning of a new production cycle). This pattern suggested that the benthic environment had the capacity to recover during the time gap between production cycles, aligning with the findings of Wang and Olsen (2024), which demonstrated the recoverability of the benthic ecosystem influenced by aquaculture.

Wang and Olsen (2024) found that environmental recoverability was more rapid in the early stages of seafloor ecosystem degradation. They emphasized the importance of early detection of environmental changes, which enable more timely intervention. Our study demonstrated the potential of C-assessments to detect subtle changes in the seafloor ecosystem, which may facilitate the early detection of environmental degradation and make informed adaptive aquaculture practices.

Our study revealed that several variables used in the environmental C-assessment protocol can detect changes in the benthic environment, although their responsiveness may vary among distance from the cages. Among the supporting variables, the content of phosphorus (P) exhibited the strongest response with a consistent decrease from the facility to the intermediate zone (~ 200m) while maintaining a correlation with the accumulated feed consumption (p< 0.001, Figure 4F).

This is consistent with previous studies reporting that approximately 60% of phosphorus in feed for farmed salmon is excreted as particulate organic P, mostly feces, resulting in a low molecular N:P ratio estimated at ~2 for POM (Wang et al., 2013; Wang and Olsen, 2023). Our data analysis has confirmed these estimates, revealing that the N:P ratio was approximately 3.1 in the local impact zone (~30 m), compared with 4.45 in the intermediate zone (~200 m) and 5.78 at the outer edge of intermediate zone (~500 m). Additionally, the N:P ratio was found to increase with accumulated feed consumption, as shown in Figures 4G, H. We have therefore suggested that the P contents of sediments can be used to detect environmental impacts of aquaculture and assessing benthic community recovery. Previous studies have also identified P as a reliable variable for assessing the impact zone of seafloor ecosystems of cage aquaculture, although its effectiveness may be limited due to its solubility and sensitivity to biochemical factors (Islam, 2005; Sanz-Lázaro et al., 2011; Porrello et al., 2003). Thus, incorporating additional variables is recommended for more robust assessments.

All faunal index scores were significantly correlated with each other, with NSI showing a slightly higher Spearman’s correlation (rho) value than AMBI (p< 0.001, Figure 5). This supported the use of NSI as a more effective indicator for assessing benthic conditions in Norwegian waters, aligning with previous findings by Rygg and Norling (2013). The data analysis of NSI revealed clear distributions for different species of ecological functional groups (EG), as shown in Figure 6. In the local impact zone (C1), Pollution-indicating (EG5) species dominated, suggesting influence from organic enrichment. In the intermediate zone (C3, ~200 m), Pollution-indicating species (EG5) markedly declined while Opportunistic (EG4) species increased, indicating a transitional state under moderate influence from organic enrichment. At the edge of the intermediate zone (C2, ~500 m) and at the reference station (>1,000 m), Sensitive (EG1) and Indifferent (EG2) species gradually became dominant, suggesting condition close to natural community conditions. Phosphorus exhibited a similar spatial trend, supporting this ecological succession along the distance gradient. Additionally, the PCA and correlation analysis demonstrated that these distributions were related to the levels of C, N, P, Cu, and Zn in the benthic environment (Figures 6, 7). The coefficient of variation (CV) for the faunal index scores was found to decrease with increasing distance from the facility (Table 5). This finding suggested that the benthic environment closer to the cages was more heavily impacted by farming operations and recovers dynamically, while the impact significantly diminished at ~200 m in the study area. This confirmed that, for some locations, it may be beneficial to use smaller sampling areas if there are no signs of pollution at greater distances. This is coherent when considering that the impact of aquaculture operations can be influenced by various biophysical factors, such as current velocity, water depth, fish density, and the length of the production cycle (Holmer et al., 2005; Kalantzi and Karakassis, 2006; Bannister et al., 2014; Wang and Olsen, 2024).

Our study has focused on a relatively small geographic area to control for environmental variables including current velocity, water depth and water temperature. While our findings may not be representative of all Norwegian salmon aquaculture operations, the data analysis approach is applicable to other regions and aquaculture species. With the potential for early detection of subtle effects of aquaculture operations on the benthic environment, this approach could benefit industry in terms of optimizing operations and site selection, ultimately enhancing sustainability.

The future of aquaculture entails expanding cage farming to more exposed ocean areas (Holmer, 2010; Hvas et al., 2021; Moe Føre et al., 2022). However, conducting environmental surveys in remote and deep-water areas can be expensive and present technical challenges for the aquaculture industry. Considering the importance of economic feasibility in implementing regular and continuous assessments, there is a pressing need to explore alternative assessment approaches that are equally reliable. The current methodology of macrofauna analyses is time-consuming, typically taking about six months in Norway, and often longer. This timeline is widely recognized as inappropriate for effective management. In addition to the potential for reducing the scope of traditional C-assessments based on our findings, there are opportunities for further advances. For example, ongoing research using seafloor species detection by hyperspectral imaging or analyzing environmental DNA offer other promising avenues for assessing the ecological state of the benthic ecosystem (Stoeck et al., 2018; Montes-Herrera et al., 2021).

Previous studies have shown that among the total nutrient wastes from salmon cage farming, about one-third is released as particulate organic waste, mainly impacting the benthic environment up to 500m from cages, while the remaining two-thirds are released as inorganic nutrient waste, primarily affecting the marine ecosystem of the water body and potentially reaching up to around 1,000 m (Broch et al., 2017; Jansen et al., 2018; Wang et al., 2013; Wang and Olsen, 2023). Norwegian regulations already provide a comprehensive and robust assessment of the benthic environment, covering the construction zone and the intermediate zone within ~ 500m (Standards Norway, 2016). However, there is still no requirement for monitoring of the water body. We therefore suggest establishing a regulatory assessment to improve the overall environmental assessment.