Premium
Accuracy and precision of secondary production estimates 1
Author(s) -
Morin Antoine,
Mousseau Timothy A.,
Roff Derek A.
Publication year - 1987
Publication title -
limnology and oceanography
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.7
H-Index - 197
eISSN - 1939-5590
pISSN - 0024-3590
DOI - 10.4319/lo.1987.32.6.1342
Subject(s) - statistics , sampling (signal processing) , nonparametric statistics , range (aeronautics) , confidence interval , mathematics , parametric statistics , econometrics , production (economics) , computer science , economics , macroeconomics , materials science , filter (signal processing) , composite material , computer vision
We used computer simulations to examine the accuracy of the increment‐summation, instantaneous‐growth, Allen curve, and size‐frequency estimates for various growth and mortality functions, patterns of prolonged recruitment, and sampling efforts; describe the sampling distribution of the estimates for aggregated populations of stream benthos; and test the reliability of parametric and nonparametric (bootstrap) confidence intervals for production estimates. Sampling schedule is critical for synchronous populations, and all methods underestimate true production when sampling intervals do not cover periods of intense production. The size‐frequency method tends to underestimate production of perfectly synchronous populations severely and is recommended only when cohorts are absolutely indistinguishable. Biases of cohort methods generally range from −30% to + 10% of true production. Sampling error of production estimates can range from −60% to +300% of true production for highly aggregated populations sampled with low effort and from −10% to + 10% at low aggregation level and high sampling effort. Published parametric and nonparametric confidence intervals are reliable only in the best circumstances (i.e. when aggregation is weak and sampling effort is high). Reliable confidence intervals can be obtained for the Allen curve estimate of production if the raw data can be transformed to stabilize the variance of density estimates and to linearize the relationship between density and mean individual mass.