ALBERT

Description: Warming air temperatures, shifting hydrological regimes and accelerating permafrost thaw in the catchments of the Arctic rivers is affecting their biogeochemistry. Arctic river monitoring is necessary to observe changes in the mobilization of dissolved organic matter (DOM) from permafrost. The Lena River is the second largest Arctic river and 71% of its catchment is continuous permafrost. Biogeochemical parameters, including temperature, electrical conductivity (EC), stable water isotopes, dissolved organic carbon (DOC) and absorption by colored dissolved organic matter (aCDOM) have been measured as part of a new high-frequency sampling program in the central Lena River Delta. The results show strong seasonal variations of all biogeochemical parameters that generally follow seasonal patterns of the hydrograph. Optical indices of DOM indicate a trend of decreasing aromaticity and molecular weight from spring to winter. High-frequency sampling improved our estimated annual fluvial flux of annual dissolved organic carbon flux (6.79 Tg C). EC and stable isotope data were used to distinguish three different source water types which explain most of the seasonal variation in the biogeochemistry of the Lena River. These water types match signatures of (1) melt water, (2) rain water, and (3) subsurface water. Melt water and rain water accounted for 84% of the discharge flux and 86% of the DOC flux. The optical properties of melt water DOM were characteristic of fresh organic matter. In contrast, the optical properties of DOM in subsurface water revealed lower aromaticity and lower molecular weights, which indicate a shift toward an older organic matter source mobilized from deeper soil horizons or permafrost deposits. The first year of this new sampling program sets a new baseline for flux calculations of dissolved matter and has enabled the identification and characterization of water types that drive the seasonality of the Lena River water properties.

Description: The warming of our planet is changing the Arctic dramatically. The area covered by sea-ice is shrinking and the ice that is left is younger and thinner. We took part in an expedition to the Arctic, to study how these changes affect organisms living in and under the ice. Following this expedition, we found that storms can more easily break the thinner ice. Storms form cracks in the sea ice, allowing sunlight to pass into the water below, which makes algal growth possible. Algae are microscopic “plants” that grow in water or sea ice. Storms also brought thick heavy snow, which pushed the ice surface below the water. This flooded the snow and created slush. We discovered that this slush is another good habitat for algae. If Arctic sea ice continues to thin, and storms become more common, we expect that these algal habitats will become more important in the future.

Description: Deep artificial neural networks have become the go-to method for many machine learning tasks. In the field of computer vision, deep convolutional neural networks achieve state-of-the-art performance for tasks such as classification, object detection, or instance segmentation. As deep neural networks become more and more complex, their inner workings become more and more opaque, rendering them a “black box” whose decision making process is no longer comprehensible. In recent years, various methods have been presented that attempt to peek inside the black box and to visualize the inner workings of deep neural networks, with a focus on deep convolutional neural networks for computer vision. These methods can serve as a toolbox to facilitate the design and inspection of neural networks for computer vision and the interpretation of the decision making process of the network. Here, we present the new tool Interactive Feature Localization in Deep neural networks (IFeaLiD) which provides a novel visualization approach to convolutional neural network layers. The tool interprets neural network layers as multivariate feature maps and visualizes the similarity between the feature vectors of individual pixels of an input image in a heat map display. The similarity display can reveal how the input image is perceived by different layers of the network and how the perception of one particular image region compares to the perception of the remaining image. IFeaLiD runs interactively in a web browser and can process even high resolution feature maps in real time by using GPU acceleration with WebGL 2. We present examples from four computer vision datasets with feature maps from different layers of a pre-trained ResNet101. IFeaLiD is open source and available online at https://ifealid.cebitec.uni-bielefeld.de.