Although most monitoring protocols characterize fish assemblages based on one sampling occasion per year per site, it is largely unknown how well such snapshot samples characterize fish assemblages at the site and the stream levels. To address these issues, we conducted monthly samplings from March to November in 2009 in two wadeable lowland streams in the catchment area of Lake Balaton, Hungary. Five and seven sites were investigated in the two streams by electric fishing 150 m long sections. For a given sampling site, mean estimated species composition of a single survey showed on average 41 % and 35 % Jaccard index based similarity to the pooled annual samples of the site, and 90 % species representation could be reached using 5.2 and 6.4 sampling occasions on average. The representativeness of relative abundance data also varied considerably in time, showing on average 51 % and 67 % Bray-Curtis index based similarity to the pooled annual samples of the site, and reached 90 % similarity by taking 4.2 and 5.4 surveys on average per year per site. Stream level simulations of sample representativeness showed that a single survey reached on average 62.3 % and 66 % Jaccard similarity and 75.7 % and 74.8 % Bray-Curtis similarity to the whole year dataset. At the stream level, 90 % representativeness of both species composition data and relative abundance data was reached by pooling four surveys for both streams. These results indicate considerable within year variability in lowland stream fish assemblages, which should not be forgotten when evaluating monitoring data, which are based on a single survey per year.