The correlation coefficients between the hourly data series were 0.93 at Kõiguste and 0.91 at Matsi (960 and 1440 value pairs respectively). The RMSE was slightly better at Matsi (0.135 vs. 0.167 m). The RMSE, standardized with the variability range, which illustrates the prognostic value of the calibration, was 8.9% at Matsi and 9.4% at Kõiguste. These sets of comparative Metabolism inhibitor statistics were only marginally worse than in our previous experiences at the Harilaid Peninsula (Suursaar & Kullas 2009) and at Letipea (Suursaar 2010). The main reason for this was the absence of storm wave conditions and the relatively

smaller range of values used in the calibration procedure. Strictly speaking, the calibration is fully valid for the conditions and variability range during the calibration. We can assume that on moving back from the calibration time, the results may get gradually worse. Validation is possible when additional data sets of the same type are available outside the calibration periods. At Matsi, 60 days of relatively calm summer measurements were used in the calibration. But at Kõiguste in the autumn, apart from 40 days of calibration, some 30 days were left for validation just before sea-ice began to affect waves by shortening fetch distances. see more The validation results were very good (Figure 5d), r was 0.89 and the RMSE was equal to 0.197. Also, the validation (verification) we performed

earlier at Letipea (see Figure 3 in Suursaar 2010) showed remarkably good agreement between measurements and calculations. Depending largely on morphometry, coastline and bottom Diflunisal topography, the current

velocity components and sea surface height at every single point in the model domain possess a specific way of reacting to wind forcing. By choosing the points (of measurements) at Kõiguste and Matsi and applying the same methodology as Suursaar & Kullas (2006), the reaction of currents to wind direction can be investigated (Figure 6 and Figure 7). At Matsi, the strongest currents appeared in wind directions 150 and 330 degrees, and in 70 and 250 degrees at Kõiguste. However, when choosing a neighbouring, or just a different point, the result would be somewhat different as well (see Figure 6b). In stationary or persistent wind forcing conditions, downwind flows prevail near the coasts of medium-size oval basins and large lakes, whereas compensatory flows against the wind evolve along the deeper middle section of a sub-basin (Csanady 1973). For instance at Pärnu Bay, our simulations revealed two well-defined basin-scale flow regimes with cyclonic and anticyclonic circulation cells (Suursaar & Kullas 2006). The two wind directions which switch between the two regimes were approximately 120 and 300 degrees. Determined by the size and coastline, similar patterns were also found in simulations with stationary and uniform winds blowing from different directions in the northern Gulf of Riga (Figure 6 and Figure 7).