Long-term fluctuations in density of two species of caddisfly from south-east Australia and the importance of density-dependent mortality

The long-term dynamics of freshwater insect populations have attracted little interest. The aim of this study was to examine the recruitment, larval mortality and reproductive output of two species of bivoltine caddisfly, Tasimia palpata (Tasimiidae) and Agapetus kimminsi (Glossosomatidae). In the Cumberland River in south-western Victoria, both species displayed summer and winter generations and their grazing larvae occupied the same rock surfaces. Ten consecutive generations of these species were monitored to determine which life-history phases were density dependent. If direct density dependence occurs, mortality during a phase of the life cycle should be positively correlated with the density of individuals entering that phase. Quantitative samples (21 each month) were taken from April 2004 to March 2009 along a 190 m reach. Mean density (R-1, numbers/m(2)) at the beginning of a generation represented the density of larval recruits. Density when pupal density was highest represented mean density at the end of a generation (R-2, numbers/m(2)). Densities of eggs laid (S, numbers/m(2)) by each generation were determined from pupal densities (P, numbers/m(2)), development time of pupae and egg numbers per female. The relationship between the number of recruits (R-1) and egg density (S) was modelled using a Ricker stock-recruitment curve. In this model, K-1 or ln (S/R-1), a measure of mortality, is a linear function of S. To extend the analysis to the rest of the life cycle K-2 (ln [R-1/R-2]; larval mortality) and K-3 (ln [R-2/P]; mortality during pupation) were calculated. K-2 and K-3 were correlated against R-1 and R-2, respectively, to determine whether these sources of mortality were related to starting densities. The Ricker curve for T. palpata was an excellent fit and K-1 was correlated with S (r = 0.96, p = 0.03). K-2 was unrelated to R-1 (r = 0.05, p = 0.78), but K-3 was correlated with R-2 (r = 0.98, p = 0.005). Thus, population regulation of T. palpata occurred during recruitment and pupation, i.e. the reproductive phases of its life cycle. The Ricker curve was a poor fit for A. kimminsi and K-1 was unrelated to S (r = 0.56, p = 0.29). However, K-2 was correlated with R-1 (r = 0.92, p = 0.05), but K-3 was unrelated to R-2 (r = 0.04, p = 0.84). A. kimminsi thus showed density-dependent mortality during its larval phase. Its larval mortality was also positively correlated with R-1 of T. palpata (r = 0.88, p = 0.0008), increasing as the strength of each T. palpata generation increased. This suggests competition occurred between the species, most probably related to food or feeding. For both species, K values were unrelated to water temperature or discharge. The populations of the two species were not affected by strong population movements (i.e. immigration or emigration). If such movements had been substantial, they would have disrupted the K values and obscured the density-dependent trends.