Uncertainties in estimates of glacier and ice-cap contribution to sea-level rise exist in part due to poor quantification of mass-balance errors, particularly those resulting from extrapolation ofsparse measurements. Centre-line data are often assumed to be representative of the glacier as awhole, with little attention paid to extrapolation errors or their effect on mass-balance estimates. Here we present detailed digital elevation model (DEM) measurements of glacier-wide elevation changes over the last ̃40 years at two glaciers on Svalbard, Norwegian Arctic. Austre Brøggerbreen and Midtre Lovénbreen are shown to have lost 27.54±0.98 and 9.65±0.76×10 m of ice, respectively, between 1966 and 2005, findings that we relate to trends in average summer air temperatures and winter accumulation. These volume losses correspond to geodetic balances of -0.58±0.03 and -0.41±0.03mw.e. a , respectively. Our analysis revealed high spatial complexity in patterns of elevation change, varying between glaciers, between measurement intervals and within and between elevation bins. Balances from extrapolated centre-line geodetic data were the same (within errors) as those from full-coverage DEM differencing in the majority of comparisons, yet significantly underestimated balance in three instances. Additionally, field mass balance from centre-line ablation stake data underestimated balances from full-coverage geodetic measurements during three of six measurement periods. These findings may support the hypothesis that field measurements underestimate Svalbard glacier mass loss, at least partly as a result of the failure of centre-line measurements to account for glacier-wide variations in ablation. Our results demonstrate the importance of deriving accurate interpolation functions and constraining extrapolation errors from sparse measurements.