Frontiers in Earth Science (Sep 2019)

Sea Level Trends and Variability of the Baltic Sea From 2D Statistical Reconstruction and Altimetry

  • Kristine S. Madsen,
  • Jacob L. Høyer,
  • Ülo Suursaar,
  • Jun She,
  • Per Knudsen

DOI
https://doi.org/10.3389/feart.2019.00243
Journal volume & issue
Vol. 7

Abstract

Read online

2D sea level trend and variability fields of the Baltic Sea were reconstructed based on statistical modeling of monthly tide gauge observations, and model reanalysis as a reference. The reconstruction included both absolute and relative sea level (RSL) in 11 km resolution over the period 1900–2014. The reconstructed monthly sea level had an average correlation of 96% and root mean square error of 3.8 cm with 56 tide gauges independent of the statistical model. The statistical reconstruction of sea level was based on multiple linear regression and took land deformation information into account. An assessment of the quality of an open ocean altimetry product (ESA Sea Level CCI ECV, hereafter “the CCI”) in this regional sea was performed by validating the variability against the reconstruction as an independent source of sea level information. The validation allowed us to determine how close to the coast the CCI can be considered reliable. The CCI matched reconstructed sea level variability with correlation above 90% and root-mean-square (RMS) difference below 6 cm in the southern and open part of the Baltic Proper. However, areas with seasonal sea ice and areas of high natural variability need special treatment. The reconstructed RSL change, which is important for coastal communities, was found to be dominated by isostatic land movements. This pattern was confirmed by independent observations and the values were provided along the entire coastline of the Baltic Sea. The area averaged absolute sea level change for the Baltic Sea was 1.3 ± 0.3 mm/yr for the 20th century, which was slightly below the global mean for the same period. Considering the relative shortness of the satellite era, natural variability made trend estimation sensitive to the selected data period, but the linear trends derived from the reconstruction (3.4 ± 0.7 mm/yr for 1993–2014) fitted with those of the CCI (4.0 ± 1.4 mm/yr for 1993–2015) and with global mean estimates within the limits of uncertainty.

Keywords