A Mediterranean sea level reconstruction (1950-2008) with error budget estimates

Reconstructed sea level fields are commonly obtained by using techniques that combine long-term records from coastal and island tide gauges with spatial covariance structures determined from recent altimetric observations. In this paper we estimate the error budget of the Mediterranean sea level reconstructions based on a reduced space optimal interpolation. In particular, we characterize the baseline error of the methodology, which is linked to the capacity of tide gauges to capture open sea processes and to the representativity of the selected EOFs. Also, we analyze the impact of the non-stationarity of the EOFs and the uneven tide gauge spatial distribution. Results suggest that the baseline error is the dominant contribution in most areas of the Mediterranean (average value of 2.7 cm). In particular, the error due to the truncation of the EOFs is the largest contribution to the baseline error. The other error sources have a more localized impact, which can be important in certain areas with atypical mesoscale activity. The skills of the reconstruction are more dependent on the length of the period than on the particular years used to compute the EOFs. Redundant tide gauges improve the reconstruction only slightly while a single tide gauge at a critical location improves it significantly. In addition we estimate the total error linked to all sources of uncertainty. Finally, we also present an updated sea level reconstruction which includes several improvements with respect to previous reconstructions. The comparison with independent data shows that this new reconstruction provides better results with respect to previous products. (C) 2011 Elsevier B.V. All rights reserved.