Composite Digital Elevation Model of the Vestfold Hills (REMA / Smith 2015)

The Davis Aerodrome Project (DAP) collected a range of environmental survey data over several field seasons to support a comprehensive environmental assessment of the proposed aerodrome. This data includes flora, fauna, soils, lake ecosystem, nearshore, marine, air quality and meteorological information which has been collected by a number of different methods, and extends across the current Davis Station, proposed aerodrome and supporting infrastructure footprint (Ridge Site), previous sites considered for the aerodrome (Heidemann Valley, Adams Flat), as well as locations across the Vestfold Hills away from any of the proposed developments. The Reference Elevation Model of Antarctica (REMA) (Howat et al., 2019) is an 8 m resolution elevation model of the entire Antarctic continent. For many parts of Antarctica, it is the finest scale digital elevation model (DEM) available in both ice covered and ice-free areas. REMA is derived from stereophotogrammetry of submeter resolution optical, commercial satellite imagery and is described as have typical elevation errors of less than 1m. However, errors are higher in rougher terrains, and on top of this REMA values are masked from terrain below sea level that lies within 800 m of the coastline. For the Vestfold Hills this presents an issue as some valleys within 800 m of the coast lie below sea-level. Here we use, a second DEM of the Vestfold Hills developed by D.T. Smith (2015) at 10m resolution and with horizontal accuracy between 2-12 m and vertical accuracy between 1-5 m to infill areas below sea level (0 m) in the REMA DEM to create a composite DEM of the Vestfold Hills. Once both DEMs were adjusted to height above sea level, they were combined using the following steps:First, to account for any slight offset between the two layers, D.T. Smith (2015) DEM values were corrected to REMA values using a linear regression (Yrema = β0 + β1(Xsmith) + ε). The linear regression used all D.T. Smith (2015) values below 30 m to train the model. The model input was cropped at 30m to avoid any lever effects from higher DEM values. Second, a transition zone was defined between 0m and 20m to blend the two layers together. Values within the transition zone (tz) were calculated using a moving weighted average of D.T. Smith (2015) and REMA values such that where D.T. Smith (2015) = 0 m, wt_SMITH = 1 and wt_REMA = 0; and where D.T. Smith (2015) = 20 m, wt_SMITH = 0 and wt_REMA = 1 using the following formulas:tz=wt_SMITH×SMITH + wt_REMA×REMA wherewt_SMITH = 1 - ((SMITH-t_min)/∆t)and,wt_REMA = 1- wt_SMITHand, SMITH is the D.T. Smith (2015) value of a cell, REMA is the REMA value of the cell, t_min and t_max are the minimum and maximum Smith values of tz, respectively (in this case 0 m and 20 m) and Δt is the difference between the tmin and tmax ∴ Δt = 20. The three layers REMA, tz and D.T. Smith (2015) combined to create an improved composite DEM that corrected for missing and erroneous values of REMA in areas below sea level. D.T. Smith (2015) values were highly congruous with REMA values after correction to sea level (r2 = 0.85, β0 = 0.10, β1 = 0.99) and combined with tz values to create a fairly seamless combined DEM.