---------------
Dear colleagues,
We are seeking high-caliber candidates for an open 4-year PhD position combining Arctic land surface observations and Data Science in the Helmholtz Einstein International Berlin Research School in Data Science (HEIBRiDS). The position will be based at the Alfred Wegener Institute Helmholtz Centre for Polar and Marine Research (AWI) in Potsdam, Germany and the Technical University of Berlin (TU).
A data-centric workflow for autonomous monitoring of Arctic land surface parameters
The 4-year PhD position is remunerated at 100% EG13 payscale of German civil service. The successful candidate will work on this collaborative project in a team with partners from AWI and TU Berlin. Project start should be between July and October 2020.
For details on AWIs and TU Research see: https://www.awi.de/en/science/geosciences/permafrost-research.html and https://sites.google.com/view/guillermogallego
For details on being a PhD within HEIBRiDS and the competitive process see: https://www.heibrids.berlin and https://www.heibrids.berlin/admission
or
http://twitter.com/guillermogb/status/1214937599619584004
A detailed project description is enclosed below.
Best regards,
Julia Boike
Land surface is the interface between the solid Earth and the atmosphere and it sets the fluxes of energy, carbon and water between both. Big data analysis can make a decisive contribution to robust and high‐resolution monitoring of surface changes in the Arctic, which are considered one of the largest sources of uncertainty regarding projections of Earth system models for this region. Large amounts of heterogeneous sources of information, such as numerical data, images, videos, texts, on spatial and temporal scales extending from small scale such as seconds and micrometers to large scales such as years and square kilometers will be used for this highly multidisciplinary approach of developing novel autonomous robotics platforms for climate monitoring and interdisciplinary research on land-atmosphere interaction. The proposed research adopts a highly multidisciplinary approach and applies innovative technologies in data acquisition and analysis. We expect that the developed technology will facilitate exploration of still unknown aspects of feedback processes (albedo) at the continental scale, and thus be of interest to a wide community. Changes in surface properties are considered one of the largest sources of uncertainty regarding projections of Earth system models for the Arctic.
In this project, we will develop a workflow for monitoring of the spatial and temporal variability of snow, vegetation and topography of the Arctic. It will harvest data from existing field sites of the circumpolar network and combine them with remote sensing products to fill the gap across scales. The project will deliver: i) a data processing hub that integrates all available imagery and data to infer snow surface, vegetation and topography parameters, ii) a data acquisition platform that employs low-cost sensors (cameras, an inertial measurement unit, GPS and a compass) to collect local measurements. The workflow, once validated at two long term observational sites (Siberia and Svalbard), will be made available for use at circum-Arctic sites to monitor the land surface properties over time. In addition to the data collected at tower locations, the data acquisition platform will collect local measurements, registered using artificial intelligence (e.g. SLAM) algorithms. Such a platform would be attached to a low-cost vehicle (blimps or drones) or be carried by a person. The workflow may be applied at any site to match and interpret available satellite-derived data.
The data/computer science expertise is needed to handle and interpret large amounts of heterogeneous data acquired at multiple spatial and temporal scales. To integrate such sparse data and provide an interpretable model, we plan to combine data cleaning techniques guided by our long term data expertise (outlier detection, data deduplication and data transformation) and techniques related to Dynamic Mode Decomposition which have recently proven successful to produce data-driven models of multi-scale, space-time physical phenomena.
