Interannual variability of elevation on the Greenland ice sheet: effects of firn densification, and establishment of a multi-century benchmark K. M. Cuffey Department of Geography, 507 McCone Hall, University of California, Berkeley, California 94720-4740, U.S.A. ABSTRACT . In order to interpret measurements of ice-sheet surface elevation changes in terms of climatic or dynamic trends, it is necessary to establish the range of stochastic variability of elevation changes resulting from interannual fluctuations of accu- mulation rate and firn density. The analyses presented here are intended to facilitate such interpretations by defining benchmarks that characterize elevation-change variability in central Greenland, in the current climate and over the past millennium.We use a time- dependent firn-densification model coupled to an ice- and heat-flow model, forced by annual accumulation rate and temperature reconstructions from the Greenland Ice Sheet Project II (GISP2) ice core, to examine the elevation changes resulting from this climatic forcing. From these results, effective firn densities are calculated.These are factors that convert water-equivalent accumulation-rate variability to surface elevation variability. A current-climate benchmark is defined by applying this conversion toVan derVeen and Bolzan's water-equivalent statistics, and to a 50year accumulation variability estimate from the GISP2 core. Elevation-change statistics are compiled for the past millennium to define longer-term benchmarks, which show that multi-century variability has been substantially larger than current variability estimated byVan der Veen and Bolzan. It is estimated here that the standard deviation of net elevation change over 5 and 10 year intervals has been 0.27 and 0.38m, respectively. An approximate method for applying these quantitative results to other dry-snow sites in Greenland is suggested. 1. INTRODUCTION The thickness of the accumulation zone of an ice sheet responds continually to both annual and long-term changes in snowfall rate, to changes of firn density and to long-term changes in factors controlling ice flow. Such responses are measurable as changes in ice-sheet surface elevation, using satellite or airborne laser altimetry, or precise ground-level surveys. Monitoring of elevation changes is useful for several reasons. First, surface elevation changes rather directly indicate changes in ice-sheet volume, if corrected for bed- rock motion and firn-density changes. Second, these changes may be used to constrain models of ice-sheet dynamics. Third, these changes can be manifestations of climate changes. Fourth, they directly reflect the year-to- year variability of accumulation rate, which contains information on atmospheric circulation and hydrologic processes. Interest in use of ice-sheet elevation monitoring as a``climate observatory'' in this fashion is particularly keen for the Greenland ice sheet (Zwally,1989; Davis and others, 1998; Krabill and others,1999; Reeh,1999; McConnell and others, 2000), which is expected to be a dominant source of future sea-level rise (Huybrechts and de Wolde, 1999) and which is situated in a region where large climate changes are expected if global warming continues. For elevation-monitoring programs to contribute meaningfully to climate-change studies, the past and present temporal variability of elevation changes must be characterized and understood, and the present contribution is part of this effort. A major complication for these analyses is the stochastic variability of snowfall rate (Oerlemans, 1981;Van derVeen,1993), which induces short-term increases or decreases of elevation that are rapid compared to dynamic velocity changes. Van derVeen (1993) has elucidated important properties of this system, with emphasis on implications for Greenland elevation-monitoring programs. He considers the present climate to be characterized by a Gaussian distribution of annual accumulation rates. This, together with a simple ice-flow model, implies a unique Gaussian distribution of net elevation changes for any specified monitoring interval. This distribution then serves as a benchmark for evaluating whether or not any observed elevation change is likely to result from stochastic variations. Further, contemporaneous climate change is best revealed by elevation monitoring over relatively short intervals (5^10 years) which are long enough to significantly reduce the stochastic accumulation variations but not generally long enough to be dominated by long-term dynamical adjustments. The present contribution intends to complementVan der Veen's analyses in two ways: (1) by adopting a longer-term perspective, and (2) by considering firn densification explicitly. In this paper I use a time-dependent firn-densification and ice- and heat-flow model and data from the Greenland Ice Sheet Project II (GISP2) deep ice core to estimate the Journal of Glaciology , Vol. 47, No . 158, 2001 369 Downloaded from https://www.cambridge.org/core. 26 Jan 2022 at 11:03:49, subject to the Cambridge Core terms of use.