Ocean ambient noise is generated in many ways such as from winds, rain and shipping. A technique has recently been developed (Harrison and Simons, J. Acoust. Soc. Am, Vol. 112 no. 4, 2002) that uses the vertical directionality of ambient noise to determine seabed properties. It was shown that taking a ratio of upward looking beams to downward produces an estimate of the reflection loss. This technique was applied to data in the 200-1500 Hz band using a 16-m vertical array. Extending this to higher frequencies allows the array length to be substantially shortened and greatly reduces interference from shipping. If array lengths can be reduced to about 1 m then it may be possible to hull-mount or tow such an array from a surface ship or submerged vehicle (e.g. an autonomous underwater vehicle). Although this seems attractive the noise is primarily generated by wind which in turn causes a rough sea-surface and bubbles and these factors combined with increased volume attenuation may degrade this type of reflection loss estimate at high frequencies.In this paper, we examine measured noise data from the October 2003 ElbaEx experiment using a 5.5 m array in the 1-4 kHz frequency band. Results indicate the noise field is predictable with modeling and the ratio of upward looking to downward looking beams produces an approximation to the reflection loss which can be inverted for seabed properties. For short arrays (a 1 m aperture is considered here), the beamforming is not ideal over a broad-band of frequencies. The beams are broadened and this leads to an up/down ratio that does not produce a good estimate of reflection loss. This can be especially problematic at low grazing angles which is the part of the reflection loss curve that is often most important to estimate correctly. Techniques will be presented for mitigating the impact of beamwidth and grating lobes on estimating the seabed properties.