Sampling and analysis of gut contents in relation to environmental variability and diel vertical migration by herbivorous zooplankton

Abstract

The accuracy of standard sampling and analysis procedures for estimating ingestion by herbivorous zooplankton was assessed using models. Artificial environments were created in a computer model, allowing for depth-dependent variability in temperature, chlorophyll and primary production. Model zooplankton were simulated within these artificial environments using individual-based models. The model zooplankton feed and defaecate at rates determined by temperature and food concentrations, and also exhibit diel vertical migration (DVM) according to a variety of migration models. The computer model was run for different combinations of these nine environmental and five DVM models. Data were 'sampled' from the model output, similar to field sampling of mesozooplankton grazing. Daily ingestion was calculated from the gut 'samples' using standard procedures for analysing gut fluorescence. The sample results were compared with the actual ingestion values in the model, and some causes of discrepancies were noted. (i) If incorrect temperatures were assumed when calculating the gut evacuation rate (K), then estimates of ingestion were wrong by up to 40%. (ii) Non-uniform food environments gave errors of up to 30% because of the large variability of measured gut contents among individuals. (iii) Sampling from only part of the total depth range (e.g. at the chlorophyll maximum) resulted in estimates of ingestion being only 5% of the real value. This sampling practice should be discouraged, because the sample is not random. (iv) If sampling is not frequent enough, errors can be as large as 45%, but more usually were ~10% for realistic sampling frequencies. We describe an analysis procedure that uses Monte Carlo-type simulations in a computer spreadsheet to estimate population consumption. These calculations take into account natural variability due to populations, samples and assumptions. We urge that results should be presented as ranges of possible values, rather than as single 'mean' values, to allow for easier recognition of meaningful differences among samples and systems.