The GLAS instrument on ICESat illuminates a spot of ~70 m in diameter on the surface at intervals of ~170 m. For each near-Gaussian pulse that is transmitted (~4 ns in width), the time-varying range returns are recorded. The shape of these received waveforms contains characteristics of the surface elevation distribution – s(h). Assuming that the received waveform is a convolution of the transmitted pulse and the weighted s(h) within the laser spot, the s(h) of individual waveforms can be estimated by deconvolution using the waveforms of both the transmitted and returned pulses. Here, we show the results from a simple deconvolution procedure and comparisons with the s(h) from a high-resolution (~1 m) Airborne Topographic Mapper (ATM). After accounting for the inherent limitations of the deconvolution scheme and ice motion, we find the agreements to be quite satisfactory. This suggest that, potentially, these estimates of s(h) can be used to estimate the fractional ice volume stored in deformed ice and possibly the spatially variable air–ice form drag at near basin scale.