The influence of uncertainty in past sea level reconstructions on 21st century mean sea level projections

Currently, in most sea-level rise projections computed using semi-empirical models, the well-established Church and White sea-level reconstruction (data 1880-2009) is used. Numerous papers have focused on the variation of the projections for sea level rise by 2100, which ranges between 20 cm (the lower limit of the Intergovernmental Panel on Climate Change, IPCC 4th assessment report) and 210 cm. In this study we focus on the sea level reconstruction and the accompanying error bars. We use the Church and White data set and its error bars as well as our own sea level reconstruction for 1950-2010. The sensitivity of the mean sea level reconstruction depends on the selection of the tide gages, the construction of the basis functions as well as the weighting (e.g. latitudinal). We used a Monte Carlo simulation in order to establish the errors in the reconstruction. We used the Vermeer and Rahmstorf semi-empirical sea level rise projection model, and simulated four different initial conditions: a) using the Church and White’s and Hamlington and Leben’s reconstructions, b) using the different error estimates suggested by the two groups c) using various time periods to train the model for the projections d) testing the model’s sensitivity to outliers. Our preliminary results show that all simulations indicate higher sea level rise projections by 2100 than suggested by the 4th IPCC assessment report in 2007 and are therefore in agreement with current projections made with the Rahmstorf or Grinsted models. Our preliminary results suggest that the range of sea level rise projections by 2100 depends strongly on the error estimate of the sea level reconstruction, the number of years included as well as its trend. This underlines the importance of understanding the sensitivity of sea level reconstruction and hence the selection of the most realistic approach for sea level projection. Hamlington and Leben’s mean projection for each of the IPCC scenarios lie within the range of projection using the Church and White dataset. We therefore suggest investigating the currently available datasets by comparing each reconstruction and projections to the 1992-2010 TOPEX/Jason-1&-2 altimetry measurements. The semi-empirical model used in this study to estimate possible sea-level projections on a decadal as well as century level is sensitive to the initial sea-level reconstruction used to train the model. It is therefore essential that we get a better understanding of sensitivity in the current observations in order to select the most realistic initial condition.