Rural fluxes included atmospheric fallout over agricultural land (F10) and the
spreading of urban sludge (F13) or sediment (F25b) from river dredging. In addition,
the inputs to forested soils through atmospheric deposition (F11a) and the forest
filter effect (F11b) [19] were estimated. Road runoff outside of urban areas (E14)
was also considered. The stocks in agricultural and forested soils were quantified
(S10, S11). The erosion fluxes from agricultural, forested, urban and industrial land
were estimated (F15, F16, F17 and F18) and combined to quantify the inputs into the
Seine River (F19). In addition, storage in the reservoirs (F23a), onto the river bed
(F23b) and on the floodplains (F24) was estimated as was the removal related to
sediment dredging operations (F25a). Finally, the flux at the outlet of the investigated area was estimated (F21).
2.3 Data Selection and Exploration
The PAH content databases used to compute the flux estimations were gathered
from about 40 previous studies investigating PAH contamination across the Seine
River basin. Whenever possible, only data collected within the investigated area
(Fig. 1) were considered. Otherwise, databases from other studies in France or
Western Europe were used, therefore weakening the estimation of the quality
grade (see below). A specific effort was made to avoid data duplication by crosschecking the sampling sites and dates in the various studies. When only summarised
data were available, data were reconstructed assuming a log-normal distribution.
Fluxes were estimated for the sum of 15 compounds (Σ15) including acenaphthylene
[ACY], acenaphthene [ACE], fluorene [FLU], phenanthrene [PHE], anthracene [ANT],
fluoranthene [FLH], pyrene [PYR], chrysene [CHR], benzo(a)anthracene [BaA], benzo
(b)fluoranthene [BbF], benzo(k)fluoranthene [BkF], benzo(a)pyrene [BaP], indeno(c,d)
pyrene [IcdP], dibenzo(a,h)anthracene [DahA] and benzo(g,h,i)perylene [BghiP].
When data for some compounds were lacking, fluxes were quantified for the sum of
the remaining compounds, and the missing compounds are specified in the text.
Data were systematically tested for spatial or temporal trends and for their
relationship with external factors such as meteorological parameters. Whenever a
parameter had a significant influence on the PAH contents, the flux calculation was
improved by integrating this parameter into the calculation. Statistical investigations
were made using R (ver3.4.4; R Core Team, 2016) software with the RStudio
interface (v1.1.453). Except for the fluxes estimated from economic data, the
calculations were carried out by multiplying the flux of the matrix considered
(e.g. runoff) by randomly drawn PAH content values in the appropriate database.
The estimation was made such that the number of draws was equal to the number of
samples in the database and draws were made with replacements. Thus, each sample
had the same probability of occurrence. To ensure a representative value of the flux
and to quantify the uncertainties, the calculation was repeated 10,000 times. Then the
flux value and its associated mathematical uncertainty were estimated as the mean
value and the standard deviation over the 10,000 computations after having ensured
that the calculations converged.
168
D. Gateuille et al.
spreading of urban sludge (F13) or sediment (F25b) from river dredging. In addition,
the inputs to forested soils through atmospheric deposition (F11a) and the forest
filter effect (F11b) [19] were estimated. Road runoff outside of urban areas (E14)
was also considered. The stocks in agricultural and forested soils were quantified
(S10, S11). The erosion fluxes from agricultural, forested, urban and industrial land
were estimated (F15, F16, F17 and F18) and combined to quantify the inputs into the
Seine River (F19). In addition, storage in the reservoirs (F23a), onto the river bed
(F23b) and on the floodplains (F24) was estimated as was the removal related to
sediment dredging operations (F25a). Finally, the flux at the outlet of the investigated area was estimated (F21).
2.3 Data Selection and Exploration
The PAH content databases used to compute the flux estimations were gathered
from about 40 previous studies investigating PAH contamination across the Seine
River basin. Whenever possible, only data collected within the investigated area
(Fig. 1) were considered. Otherwise, databases from other studies in France or
Western Europe were used, therefore weakening the estimation of the quality
grade (see below). A specific effort was made to avoid data duplication by crosschecking the sampling sites and dates in the various studies. When only summarised
data were available, data were reconstructed assuming a log-normal distribution.
Fluxes were estimated for the sum of 15 compounds (Σ15) including acenaphthylene
[ACY], acenaphthene [ACE], fluorene [FLU], phenanthrene [PHE], anthracene [ANT],
fluoranthene [FLH], pyrene [PYR], chrysene [CHR], benzo(a)anthracene [BaA], benzo
(b)fluoranthene [BbF], benzo(k)fluoranthene [BkF], benzo(a)pyrene [BaP], indeno(c,d)
pyrene [IcdP], dibenzo(a,h)anthracene [DahA] and benzo(g,h,i)perylene [BghiP].
When data for some compounds were lacking, fluxes were quantified for the sum of
the remaining compounds, and the missing compounds are specified in the text.
Data were systematically tested for spatial or temporal trends and for their
relationship with external factors such as meteorological parameters. Whenever a
parameter had a significant influence on the PAH contents, the flux calculation was
improved by integrating this parameter into the calculation. Statistical investigations
were made using R (ver3.4.4; R Core Team, 2016) software with the RStudio
interface (v1.1.453). Except for the fluxes estimated from economic data, the
calculations were carried out by multiplying the flux of the matrix considered
(e.g. runoff) by randomly drawn PAH content values in the appropriate database.
The estimation was made such that the number of draws was equal to the number of
samples in the database and draws were made with replacements. Thus, each sample
had the same probability of occurrence. To ensure a representative value of the flux
and to quantify the uncertainties, the calculation was repeated 10,000 times. Then the
flux value and its associated mathematical uncertainty were estimated as the mean
value and the standard deviation over the 10,000 computations after having ensured
that the calculations converged.
168
D. Gateuille et al.