The oceanographic community has long been capable of collecting and accumulating plankton samples at a rate that far exceeds our ability to process and extract meaningful ecological information from sample contents. As a consequence, large archives of plankton samples are accumulating at laboratories throughout the world. The potential information contained in these samples is enormous and could contribute fundamental insights into the responses of planktonic communities to natural and human-induced environmental change. Regrettably, our inability to process the majority of these samples means that over time, desiccation, acidification, and catastrophic events are constantly destroying preserved organisms before they can be examined, counted, measured, and identified.
Nets, pumps, and water bottles represent core sampling techniques for plankton and are the basis for many long time series and oceanographic programs such as GLOBEC. While these methods have the advantages of providing a physical record of planktonic organisms from [frequently] large volumes of water, one of the central challenges confronting plankton researchers is the relatively low spatial and temporal resolution that these approaches provide. It is widely recognized that many important ecological interactions and processes involving plankton occur on time and space scales that are typically much finer than those normally sampled using traditional samplers such as nets. Moreover, the invasive nature of net tows causes disruption of relevant plankton features (e.g., patchiness, thin layers), which are important to be explored for a better understanding of community structure and distribution and can be maintained only by in situ observatory systems. The other challenge is that ecologically important but fragile species (e.g., ctenophores, siphonophores, larvaceans) are under-represented when studying plankton collected with conventional nets.
The past decade has seen a rapid evolution in development of innovative in situ and benchtop imaging systems designed to quantify the contents of water at fine temporal and spatial scales (Benfield et al. 2007). These instruments typically collect images at high rates (up to 60Hz) resulting in the rapid accumulation of very large numbers of digital images (e.g., 108,000 images per hour @ 30 Hz). Another recent development in the plankton imaging field has been the emergence of highly modified flatbed scanners that create a high-resolution digital record of the contents of preserved plankton samples (Grosjean et al., 2004; http://www.obs-vlfr.fr/LOV/ZooPart/ZooScan). This instrument produces large image files that can contain many hundreds of zooplankton and other particles. While all of these imaging systems represent a welcome addition to our plankton sampling/sensing toolbox, they also present a new challenge: the manual analysis of images from such systems is impractical, due to the huge amount of information and quantities of images they produce.
New image analysis systems offer a potentially advantageous solution compared to manual methods of counting and sizing. With the aid of image analysis and classification software and hardware, the planktonic targets within images can be located, isolated, and identified to at least major groups. Many sophisticated automatic recognition algorithms exist, and research in this area is very active. Thus, there now exists a very real potential of using image analysis techniques to obtain more refined taxonomic classification in the near term.