THE EFFECT OF TEMPORAL UNDERSAMPLING ON PRIMARY PRODUCTION ESTIMATES

Annual primary production estimates for specific oceanic regions have typically been made using a variety of measures of productivity spaced , at best, several weeks apart. Primary productivity in the oceans is known to be extremely episodic. It is hypothesized here that primary p roduction data with a temporal resolution of several weeks have a high potential for error due to undersampling. In the present analysis, ti me series of gross primary productivity were calculated using time ser ies of photosynthetically available radiation and chlorophyll a concen tration as input to an optical production model. The input data are of minute scale resolution and were gathered during a number of moored e xperiments. These took place over the past 5 years at several oceanic sites. The minute scale productivity time series were integrated to fo rm time series of daily estimates of gross production. These range in duration from 40 to 260 days. The time series exhibit several regimes characteristic of oceanic primary productivity, such as phytoplankton blooms, productivity pulses associated with advected water masses, ste ady state growth, and development of a subsurface productivity maximum . The presence of these features makes our time series ideal for inves tigating (1) the sensitivity of annual production estimates to the tim ing of the sample set and (2) the error introduced by undersampling in herent in coarser sampling methods. It was found that distinct pulses of productivity generate the greatest error and that high variability leads to large errors, even for well-resolved sampling intervals. The maximum percent error due to undersampling was found to be 85%. Additi onally, up to a fourfold range between the maximum and minimum estimat es of average daily production was found over all sampling intervals. Finally, the maximum expected range (300 g C m-2 yr-1) and the expecte d standard deviation (+/-42 g C m-2 yr-1) for annual water column prod uction were determined at a Sargasso Sea site for which long-term prod uctivity time series were available at four depths within the euphotic zone.