A pan-Arctic algorithm for DOC concentrations from CDOM spectra

<div><span>The surface layer of the Arctic Ocean carries a higher dissolved organic carbon (DOC) content than other ocean basins. Climate change impacts the Arctic aquatic DOC-pool by e.g., introducing DOC trapped in permafrost soils as they thaw and by increasing the terrestrial runoff and primary production. Sampling for DOC in the Arctic is rather challenging given its remoteness and difficult access to the region and that it is not possible yet to determine DOC concentrations from instruments deployed in the field. Compared to DOC, colored dissolved organic matter (CDOM) absorption spectroscopy is an easy-to-measure, relatively quick and cost-effective approach which is often closely related to DOC concentrations in water samples. In regions in close proximity to rivers, linear relationships between CDOM absorption at 350nm (a350) and DOC often can be found and, thus, have improved prediction of DOC using two end‐members. However, in regions with two or more end‐members of comparable DOC concentrations (shelf seas and oceanic waters) these relationships are difficult to derive, as there might be pools of similar concentration/intensity but different ratio of absorption to DOC (carbon specific absorption coefficient, a*). Here we present an algorithm to estimate DOC concentrations based on quantitative (a350) and qualitative (spectral absorption slope between 275 and 295nm, S275-295) properties of CDOM. The algorithm considers that there is a linear correlation between DOC and a350 but that the slope of the relationship (inverse of a*) varies depending on the exponent of the ultraviolet (UV) spectral slope (S275‐295), that is, the character or source of DOM. We compiled a Pan-Arctic dataset (n=3607) from a wide range of aquatic systems spanning lakes, rivers, estuaries, coastal and shelf seas and open ocean with salinity ranging from 0 to 35.3. DOC ranged between 19 and 2304µM, whereas a350 varied from 0.01–81.33m<sup>-1</sup> and S275-295 ranged 12–39µm<sup>-1</sup>. The algorithm provided significant and robust (r<sup>2</sup>=0.93; p<0.0001) DOC estimates (pDOC), ranging 1–2598µM (RMSE=64µM). This indicates that, besides its simplicity, this method is capable of capturing the extremely high variability of DOC within the broad gradient of Arctic aquatic systems considered in this study. Apart from that, pDOC estimates could reproduce both DOC profiles and the DOC vs. salinity relationship across the Arctic Ocean (i.e., distinct sites with highly distinct hydrographic conditions). This potentially makes the method suitable for high-resolution and long-term in situ monitoring of DOC concentrations in Arctic aquatic systems from e.g., absorbance measurements from in situ nitrate sensors. </span></div>