measurements started on specific days, while the true EF is
based on continuous data over all days of the week. Further-
more, the lower and upper uncertainty bounds coincided for
all 4 graphs. This is because, for each of the 1000 iterations,
there is only one way to sample the maximum of n 24-h/48-h
sampling periods. When the maximum number of n 24-h/48-h
sampling periods was sampled, only one estimate of the EF
(and associated relative error) was found. In contrast to the
previous sampling strategies, the lower and upper uncertainty
bounds did not coincide for both graphs (Fig. 9) when the
maximum number of 7-d periods was reached. This is because
a 27 or 28-week sampling strategy can start on each working
day of the week. Therefore, the estimates between simulated
strategies will differ depending on which day of the week the
strategy started. If the maximum number of single grab
samples would be taken into account, the lower and upper
uncertainty bounds would coincide for both graphs (Fig. 10).
As for the 7-d sampling strategy, the lower and upper uncer-
tainty bounds for the weekly grab sampling strategy (Fig. 11)
did not coincide since the number of possible ways to conduct
a weekly grab sampling strategy of 35 (dataset 1) or 34 (dataset
2) consecutive weeks is almost infinite.