The wind dependence of acoustic spectrum between 100 Hz and 16 kHz is investigated for coastal biologically rich areas. The analysis of 5 months of continuous measurements run in a 10 m deep shallow water environment off Brittany (France) showed that wind dependence of spectral levels is subject to masking by biological sounds. When dealing with raw data, the wind dependence of spectral levels was not significant for frequencies where biological sounds were present (2 to 10 kHz). An algorithm developed by Kinda, Simard, Gervaise, Mars, and Fortier [J. Acoust. Soc. Am. 134(1), 77-87 (2013)] was used to automatically filter out the loud distinctive biological contribution and estimated the ambient noise spectrum. The wind dependence of ambient noise spectrum was always significant after application of this filter. A mixture model for ambient noise spectrum which accounts for the richness of the soundscape is proposed. This model revealed that wind dependence holds once the wind speed was strong enough to produce sounds higher in amplitude than the biological chorus (9 kn at 3 kHz, 11 kn at 8 kHz). For these higher wind speeds, a logarithmic affine law was adequate and its estimated parameters were compatible with previous studies (average slope 27.1 dB per decade of wind speed increase).