Optimization of River Sampling: Application to Nutrients Distribution in Tagus River Estuary

The
assessment of river water pollution trends is affected by the
seasonal variation of river conditions, the variability of pollution
sources, the heterogeneity of pollutants distribution, the representativeness/uncertainty
of sampling, and the uncertainty of sample analysis. This work presents
a methodology to model the uncertainty of river water sampling based
on available information about the spatial distribution of the studied
parameter in the river. The uncertainty from “single sampling”
(SS) or by production of a composite sample by mixing m subsamples collected randomly (RS) or in a line that crosses the
sampling circle (LS) was studied. This methodology was applied to
the determination of nutrients (NOx, NO2, PO4, and SiO2) in an area of the Tagus
River estuary with a range of about 350 m. This methodology can be
applied to the determination of the mean value of other parameters
in other river areas requiring a previous study of system heterogeneity.
The spatial distribution of nutrients in the studied river area was
characterized from the analysis of 10 samples collected at known geographical
coordinates. The system heterogeneity was described by a three-dimensional
(x, y, z) surface
with x and y variables for samples
positions and z variable representing the measured
nutrient levels. The randomization of this surface for the uncertainty
of coordinates and repeatability of nutrient concentration measurement,
using Monte Carlo simulations, allowed estimation of the uncertainty
of the three sampling strategies: SS, RS, and LS. The uncertainty
from RS and LS is equivalent and significantly smaller than that from
SS when at least three subsamples are mixed in the composite sample.
The sampling relative standard uncertainty ranged from 0.31% to 4.4%,
producing nutrient concentration estimates in the river area with
a relative expanded uncertainty from 5.9% to 10% with approximately
95% confidence level (coverage factor of 2). The used spreadsheet
is available as Supporting Information.