The color-time series presented here shows that sediment color provides a useful variable to describe the Site 1098 sediments. Digital images can be used to describe gradual color changes at decimeter scale (Nederbragt et al., 2000), but the Minolta photospectrometer is less labor-intensive in such a case. In laminated sediments, however, digital images not only have the resolution to register within-lamina variation but also allow long-term trends to be described more reliably. Problems in correlating the shipboard Minolta photospectrometer measurements with other records (Acton et al., Chap 5, this volume) result from stochastic scatter resulting from inadequate sampling of the centimeter-scale color variation. The Minolta records are spiky, with large fluctuations dependent on whether the measurement was in a dark or light lamina. The continuous time series derived from image data have the advantage that color can be integrated over a given interval to produce average values. With the strong and rapid color variation within the laminated intervals at Site 1098, such image-derived averages yield a smoother long-term trend that matches well with other sediment variables like GRA bulk density (Fig. F3).
Color variation at the decimeter to meter scale correlates significantly with sediment density, but the linear correlation coefficient is not very high (Fig. F3). Data to correlate sediment color with sediment chemistry at Site 1098 are not available yet. However, marine sediments in other areas showed that color reacts most strongly to variation in carbonate and total organic carbon (TOC) content; any correlation between biogenic silica content and color is usually weak and dependent on overall sediment composition and mineralogy of the terrigenous component (Nederbragt et al., 2000). TOC content reaches values of up to 1.5 wt% in the upper 10 m of the sediment column in Palmer Deep (Leventer et al., 1996), that is, values that are high enough to have a clear effect on sediment color. With biogenic silica and terrigenous material forming the bulk of the sediment, GRA bulk density and magnetic susceptibility should primarily measure the ratio between these two components. Visual inspection of the sediments at Site 1098 indicates that sediment color and diatom content are partly independent of each other. Dark laminae are usually diatom rich, but the reverse is not true. Many diatom-rich intervals do not clearly differ in color from surrounding, more terrigenous sediments and presumably are not enriched in organic carbon. A factor that will contribute to a weak correlation between silica and TOC is that a large number of biogenic silica-rich laminae consist of diatom resting spores, which are formed under low-productivity conditions in response to nutrient depletion (Leventer et al., 1996; Pike et al., Chap 18, this volume). We expect a further study to show that color traces TOC content primarily, that GRA bulk density is mostly a function of biogenic silica content, and that the relatively weak correlation between the two is the result of a partly nonlinear relation between export fluxes of diatoms and organic matter.
Whereas sediment color has potential as a meaningful proxy, the second requirement for analysis of high-frequency climate cycles, an accurate age scale with annual resolution, may be difficult to realize. Conditions in Palmer Deep are favorable for the formation of varved sequences, given the high average accumulation rates, the strong seasonal limitation of biogenic productivity, and the lack of bioturbation under anoxic conditions. The close match of long-term accumulation rates of light sediments with the modal thickness of light laminae (Fig. F8) indicates that many dark/light couplets indeed represent an annual cycle. However, there are too many "missing" dark laminae to classify the laminated intervals as varved. It is therefore not feasible to simply use laminae counts to establish a reliable (floating) age scale. Part of the missing dark laminae could represent light intervals in which a thin diatom-rich level is preserved but not visible in color change. However, given the large variation in thickness of recognizable dark laminae ranging between <1 mm and >1 cm, it seems likely that the range of variation includes the virtually complete absence, or nonpreservation, of biogenic productivity during some years.
We attribute the extreme interannual variability in marine productivity primarily to variation in temperature around the Antarctic peninsula. Instrumental records for the past 20 yr show the presence of loose sea ice in Palmer Deep, usually from February to May, but even the short available records show considerable interannual variation (Stammerjohn and Smith, 1996). Leventer et al. (1996) relate high biogenic silica fluxes to periods during which a stratified water column with nutrient depletion in the upper water is generated, inducing the formation of diatom resting spores or diatom mats that sink more easily to the seafloor. One variable that is not addressed in this interpretation is the length of time that primary producers can grow as a result of light limitation. Without sea ice, spring blooms in the Southern Hemisphere would be expected to start around October, when days start to lengthen. The presence of a sea-ice cover until much later in the solar year must cause sunlight to be biolimiting until the ice has broken up, leaving only a short growing season. We expect that the large variation in the total annual biogenic flux in Palmer Deep is primarily a function of Antarctic temperature. In warm years, the sea-ice cover would break up earlier, giving a growing season long enough for primary producers to deplete nutrients in the photic zone, whereas only thin biogenic layers would be formed during colder-than-average years, when delayed ice breakup causes lack of sunlight to be a major biolimiting factor.