From the shape of the vertical profile of in vivo fluorescence to Chlorophyll-<i>a</i> concentration

In vivo fluorescence of Chlorophyll-a (Chl-a) is a potentially useful property to study the vertical distribution of phytoplankton biomass. However the technique is presently not fully exploited as it should be, essentially because of the difficulties in converting the fluorescence signal into an accurate Chl-a concentration. These difficulties arise noticeably from natural variations in the Chl-a fluorescence relationship, which is under the control of community composition as well as of their nutrient and light status. As a consequence, although vertical profiles of fluorescence are likely the most recorded biological property in the open ocean, the corresponding large databases are underexploited. Here with the aim to convert a fluorescence profile into a Chl-a concentration profile, we test the hypothesis that the Chl-a concentration can be gathered from the sole knowledge of the shape of the fluorescence profile. We analyze a large dataset from 18 oceanographic cruises conducted in case-1 waters from the highly stratified hyperoligotrophic waters (surface Chl-a = 0.02 mg m&minus;3) of the South Pacific Gyre to the eutrophic waters of the Benguela upwelling (surface Chl-a = 32 mg m&minus;3) and including the very deep mixed waters in the North Atlantic (Mixed Layer Depth = 690 m). This dataset encompasses more than 700 vertical profiles of Chl-a fluorescence as well as accurate estimations of Chl-a by High Performance Liquid Chromatography (HPLC). Two typical fluorescence profiles are identified, the uniform profile, characterized by a homogeneous layer roughly corresponding to the mixed layer, and the non-uniform profile, characterized by the presence of a Deep Chlorophyll Maximum. Using appropriate mathematical parameterizations, a fluorescence profile is subsequently represented by 3 or 5 shape parameters for uniform or non-uniform profiles, respectively. For both situations, an empirical model is developed to predict the "true" Chl-a concentration from these shape parameters. This model is then used to calibrate a fluorescence profile in Chl-a units. The validation of the approach provides satisfactory results with a median absolute percent deviation of 33 % when comparing the HPLC Chl-a profiles to the Chl-a-calibrated fluorescence. The proposed approach thus opens the possibility to produce Chl-a climatologies from uncalibrated fluorescence profile databases that have been acquired in the past and to which numerous new profiles will be added, thanks to the recent availability of autonomous platforms (profiling floats, gliders and animals) instrumented with miniature fluorometers