Understanding the Effects of Optimal Combination of Spectral Bands on Deep Learning Model Predictions: A Case Study Based on Permafrost Tundra Landform Mapping Using High Resolution Multispectral Satellite Imagery.

Deep learning (DL) convolutional neural networks (CNNs) have been rapidly adapted in very high spatial resolution (VHSR) satellite image analysis. DLCNN-based computer visions (CV) applications primarily aim for everyday object detection from standard red, green, blue (RGB) imagery, while earth science remote sensing applications focus on geo object detection and classification from multispectral (MS) imagery. MS imagery includes RGB and narrow spectral channels from near- and/or middle-infrared regions of reflectance spectra.

Composition and photo-reactivity of organic matter from permafrost soils and surface waters in interior Alaska.

Yedoma permafrost soils are especially susceptible to abrupt thaw due to their exceptional thickness and high ice content. Compared to other mineral soils, yedoma has a high organic carbon content, which has shown to be particularly biolabile. The organic carbon in these deposits needs to be characterised to provide an identification toolkit for detecting and monitoring the thaw, mobilisation and mineralisation of yedoma permafrost.

Vegetation on mesic loamy and sandy soils along a 1700-km maritime Eurasia Arctic Transect.

Questions: How do plant communities on zonal loamy vs. sandy soils vary across the full maritime Arctic bioclimate gradient? How are plant communities of these areas related to existing vegetation units of the European Vegetation Classification? What are the main environmental factors controlling transitions of vegetation along the bioclimate gradient?

Location: 1700-km Eurasia Arctic Transect (EAT), Yamal Peninsula and Franz Josef Land (FJL), Russia.

Methods: The Braun-Blanquet approach was used to sample mesic loamy and sandy plots on 14 total study sites at six locations, one in each of the five Arctic bioclimate subzones and the forest-tundra transition.

21st-century modeled permafrost carbon emissions accelerated by abrupt thaw beneath lakes.

Permafrost carbon feedback (PCF) modeling has focused on gradual thaw of near-surface permafrost leading to enhanced carbon dioxide and methane emissions that accelerate global climate warming. These state-of-the-art land models have yet to incorporate deeper, abrupt thaw in the PCF. Here we use model data, supported by field observations, radiocarbon dating, and remote sensing, to show that methane and carbon dioxide emissions from abrupt thaw beneath thermokarst lakes will more than double radiative forcing from circumpolar permafrost-soil carbon fluxes this century.

A lake-centric geospatial database to guide research and inform management decisions in an Arctic watershed in northern Alaska experiencing climate and land-use changes.

Lakes are dominant and diverse landscape features in the Arctic, but conventional land cover classification schemes typically map them as a single uniform class. Here, we present a detailed lake-centric geospatial database for an Arctic watershed in northern Alaska. We developed a GIS dataset consisting of 4362 lakes that provides information on lake morphometry, hydrologic connectivity, surface area dynamics, surrounding terrestrial ecotypes, and other important conditions describing Arctic lakes.