Each of the short-term sampling strategies described above(single grab sample, 24-h and 48-h continuous monitoring and7-d continuous monitoring) could be repeated, while the long-
term weekly grab sampling strategy could vary in length. This would increase the precision of the simulated estimates. The influence of taking multiple (n) random grab samples, 24-h periods, 48-h periods, 7-d periods or weekly grab samples during n consecutive weeks on the estimated EF was evaluated. These evaluations started by calculating all possible grab samples, 24-h periods, 48-h periods, 7-d periods or long-term weekly grab sample strategies that fulfilled the requirements as mentioned in Section 2.2. Thereafter 1 to n random grab samples, 24-h periods, 48-h periods or 7-d periods or long-term weekly grab sample strategies for 1 to n consecutive weeks were randomly taken for all possible periods. Finally,for all 1 to n sampling cases, the relative error between the true EF (as determined with Eq. (2) and based on the whole dataset)
and the estimated EF (based on a limited number of measurements and only representing an estimation of the true EF)was determined as a function of the number of sampling cases
for each of the investigated reduced frequency samplingstrategies (Daelman et al., 2013). These sampling cases were the number (n) of 24-h (or 48-h) random periods for the 24-h
(or 48-h) reduced sampling strategies, or the number of random, 7-d periods for the one week reduced sampling strategy. For the random grab sampling strategies, the number of single grab samples was determined necessary to obtainan estimate with a relative error smaller than ±15%. Similarly,for the long-term weekly grab sampling strategy, the number of weeks necessary to obtain an estimate with a relative error