Modeling the release of dissolved organic matter by phytoplankton

Three models describing dissolved organic matter (DOM) flux and phytoplankton death, each of different levels of complexity, were constructed and tested against experimental data for a cyanobacterium, a chlorophyte, two diatoms, two dinoflagellates, and two prymnesiophytes. The simplest model described only bulk carbon (C) and nitrogen (N) forms of DOM (DOMC and DOMN) and employed a fixed relationship between phytoplankton nutrient status and DOM release and death rate. The most complex model described fractions of DOM as low molecular weight dissolved organic carbon (DOC; saccharides, low molecular weight carbohydrates [DOCs]), low molecular weight nitrogenous material (comprising C and N as DOC associated with low molecular weight compounds containing amino acids and/or nucleic acids [DOCa] and N associated with DOCa [DONa], which included dissolved free amino acids [DFAA]), and more complex materials (DOC associated with high molecular weight compounds typically requiring extracellular degradation prior to uptake or use by microbes [DOCx] and N associated with DOCx [DONx]). It also employed descriptions of DOM flux and cell death related to nutrient status and growth rates. In all instances, material lysed from dead cells contributed to the DOM pool. All three models captured the gross dynamics of the primary data (dissolved inorganic C [DIC], dissolved inorganic N [DIN], particulate organic carbon [POC], particulate organic N [PON], DOC, dissolved organic N [DON]), but there was little or no improvement of the fit with increasing model complexity. However, the simplest models tended to employ excessively high growth rates to compensate for high fixed death rates. While the proportion of newly fixed C being liberated as DOMC (DOCs plus DOCa) increased as nutrient status declined, the actual rate of release typically did not do so and often declined. The most complex model gave predictions for changes in released saccharides and DFAA in keeping with expectations. The major obstacle to future progress is the lack of suitable, mass balanced data sets for further model testing.