PhD Defense by Dina Rapp

Title: Temperatures in Greenland - From spatially sparse observations to gap-free spatio-temporal products

Abstract:
Greenland weather station temperature time series are used to validate climate models and reanalyses, as a basis for global temperature products and reanalyses, and directly in studies of local climate change. Their homogeneity and quality are, therefore, of great importance. In chapter 2 of this thesis a correction method to adjust the warm biases due to the oversampling of the sunlit hours is developed and presented, reducing the highest simulated bias from 1.3C to 0.5C. For several weather stations, large corrections are applied to older observations and small corrections to the more recent observations, revealing an underestimation of temperature trends. Visualisation and communication of meta data for the DMI Greenland weather station network are also improved.

As global and regional reanalyses are used to assess climate change in several capacities such as to directly assess global warming, to validate climate models and to force SMB models, it is crucial to assess how they perform over Greenland. Chapter 3 of this thesis is a validation study that compares ERA5 and CARRA with the corrected daily average temperatures in chapter 2. Temporal inconsistencies are identified for both reanalyses, the largest in ERA5. Most notably, a shift in bias of over 2C is identified in 1967, coinciding with a large increase in the number of observations included in ERA5. CARRA validates better than ERA5 against the corrected daily averages of the DMI Greenland weather station network.

Gap-free temperature products going as far back in time as possible are important to assess climate change in Greenland. Due to the large temporal inconsistencies observed in ERA5, and CARRA only going back to 1991, a method is developed and presented in chapter 4. It is a data-driven method combining 3D-Variational data assimilation and an autoencoder. The method is successfully used to reconstruct gap-free daily temperature maps covering Greenland going back to 1958 using the corrected daily average temperatures from chapter 2. The method avoids the large shift in bias seen in ERA5 in 1967. Using CARRA as a baseline, the standard deviations and mean average errors improve when few observation locations are added.

Committee:
Professor Jens Hesselbjerg Christensen, PICE, NBI, University of Copenhagen. (Chair)
Professor Nanna Bjørnholt Karlsson, Geological Survey of Denmark and Greenland.
Professor Pierre Tandeo, IMT Atlantique Bretagne-Pays de la Loire, Campus de Brest.

Supervisors:
Associate Professor Bo Møllesøe Vinther
Jacob L. Høyer, PhD, Head of Remote Sensing at DMI.