Comparison of wind-stress algorithms and their influence on wind-stress curl using buoy measurements over the shelf off Bodega Bay, California

The main objectives of this study were to compare three wind-stress algorithms of varying intricacy and estimate the extent to which each method altered computed wind-stress curl. The algorithms included (1) a simple bulk formula for neutral conditions that is dependent only on wind velocity components; (2) a formula that in addition to dependence on wind components includes a simplified effect of thermal stability through differences in air and sea temperatures; and (3) an algorithm that includes full treatment of dynamics and atmospheric stability. Data for the analysis were from a field program that used a special buoy network off Bodega Bay during 28 June–4 August 2001.A diamond-shaped setup of five closely separated buoys in Bodega Bay allowed for one of the first attempts to compute wind-stress curl over the ocean using buoy measurements. Based on an analysis of the available dataset, the marine layer over Bodega Bay is characterized by positive wind-stress curl with a median value around 0.2 Pa (100 km)⁻¹ and maximum values reaching 2.5 Pa (100 km)⁻¹. Positive wind-stress curl was observed for all wind speed conditions, whereas negative wind-stress curl episodes were associated mostly with low-wind conditions.Comparison of wind-stress curl computed using the three algorithms showed that differences among them can be significant. The first and third algorithms indicated similar stress curl (difference around 10%), but the differences between these two and the second algorithm were much higher (approximately 40%). The reason for the difference is the stability correction, which in the third algorithm strongly decreases with an increase in wind speeds, but stays at a similar level for all wind speeds in the second algorithm. Consequently, for higher wind speeds the variability of wind stress calculated using the second algorithm is greater than for the other two algorithms, causing significant differences in computed wind-stress curl (root mean-square error equal to 0.19 Pa (100 km)⁻¹).Despite the apparent biases in computed wind stress and wind-stress curl among the algorithms, all of them show a significant trend of decreasing sea-surface temperature (SST) with increasing wind-stress curl. The bootstrapping analysis has revealed that both the along-shore wind stress and wind-stress curl have noticeable correlation with the changes in the sea-surface temperature as an indirect indication of the upwelling. An additional analysis, based on the low-pass filtered data, showed also significant agreement between the measured divergence in the cross-shore surface transport and the wind-stress curl computed for all three algorithms.