Continual refinement of lake carbon estimates using historical, current, and future field and satellite data

As increasing attention is paid to the storage and flux of carbon in and around boreal lakes, systematic strategies to make and refine credible region-wide estimates of stocks and fluxes will be increasingly important. Given that there are hundreds of thousands of lakes in the world's lake-rich regions, even concerted long-term field campaigns can reach only a minuscule fraction of the water bodies of the world. Although remote sensing would appear to be of obvious potential, limnologists interested in satellite detection of lake color have been limited by many factors, primarily (a) the low spectral and radiometric resolution of earlier Landsat imagery; (b) the limited coverage of the experimental ALI sensor; and (c) the high cost of alternative imagery. The launch of Landsat 8 promises a revolution in estimating carbon content in the world's small lakes. Yet as Landsat 8 becomes operational, the prospect of sampling and validating for hundreds of thousands of boreal lakes is nearly impossible using the standard approach of carefully timing lake sampling with satellite overpasses. We have recently shown that multi-temporal imagery from the prototype ALI sensor can be used for a high-quality estimation of coloured dissolved organic matter (CDOM), an important lake property that is usually well correlated with visible colour in northern lakes. For that relation we used a legacy field data set collected at irregular times over more than a decade, which greatly expanded the amount of field observation available for building the model. In a setting with multi-temporal field data matched with multi-temporal image data, the challenge of updating estimates will be increasingly important as Landsat 8 produces a stream of images of varying utility. Meanwhile, field data in newly sampled lakes may be plausibly used for a variety of purposes: for fits with imagery that is near in time, as validation of an existing estimate, or to refine fits from old imagery. Here we describe our expected protocol for treating and refining an ongoing series of imagery-driven estimates, built from a large amount of multi-temporal data. The workflow is intended to automatically incorporate new evidence to continually sharpen the estimates of lake carbon content across a vast area. The prospect of long-term observation of lake properties is arriving, and it promises a clearer picture of one of the world's great storehouses of carbon.