3.1. Stationary Deployment in Fusafjorden
Figure 3 shows the velocity field observed by the bottom-mounted and SailBuoy ADCPs. Both observations reveal more or less similar flow field patterns. However, there are some slight differences between those observations. We note that the bottom-mounted ADCP shows somewhat higher current speeds and more damped current direction changes, which warranted a further data analysis.
The time series of current speed and direction at ca. 20 m depth measured by the single-point meter (middle) as well as the SB ADCP (top) and bottom-mounted ADCP (bottom) data are presented in
Figure 4. Here, again, measurements of different instruments show a similar trend; however, the bottom-mounted ADCP recorded higher speeds compared to the others while also showing a larger standard deviation. The correlation coefficients between the SB ADCP and bottom-mounted and single-point current meter velocity measurements at a 20 m depth were 0.40 and 0.76, respectively.
Histogram graphs (
Figure 5) also reveal the systematic offset between the bottom-mounted ADCP and the rest of the observations. While the observations of the SB ADCP and single-point current meter almost entirely overlap, the speed values recorded by the bottom-mounted are detached from the rest with a mean deviation of 0.018 m/s. It should be noted here that the current profiler on the SB and the single-point current meter are both from the same manufacturer (Aanderaa Data Instruments AS), while the bottom-mounted profiler is a Nortek instrument. The bias here directly opposes the one found during the Ekofisk deployment where the SB measured higher values than the Nortek bottom-mounted ADCP as is later discussed.
Figure 6 displays the time-averaged current speed calculated for each depth cell as well as the maximum measured speed at each depth measured by the SB ADCP and the reference instruments in Fusafjorden. The time-averaged vertical structure of the flow field, i.e., higher current speeds closer to the surface and decreasing with depth, is almost similar for both current profilers. However, a systematic velocity offset of approximately 0.02 m/s is also evident along the water column. The average velocity of the single-point current meter nicely fits with the SB records (deviating only by 0.001 m/s, which is less than the accuracy range of the instrument). Furthermore, the maximal measured velocity corresponds quite well between both instruments (deviating by 0.016 m/s). Even though the bottom-mounted profilers’ records are higher, the SB ADCP and the single-point current meter records are still within the range of the standard deviation.
A correlation analysis of the measurements over depth resulted in relatively high correlation coefficient values of more than 0.6 in the upper part of the water column (at 5–12 m) and values mostly below 0.4 from 15 m downward (see
Figure 7). While the SB data correlated well with the Aanderaa point meter at 20 m depth (correlation coefficient 0.75), the bottom-mounted instrument exhibited a comparably poor correlation with the Aanderaa point meter (0.36). The
p-values for all correlations done here vanished to zero, indicating that the positive correlation between the instruments, especially in the surface layers, was significant.
3.2. Offshore Deployment at Ekofisk
The active measurement period of the SailBuoy ADCP, while it was in proximity to the ADCPs (within 2–41 km range, see
Figure 8), was almost one month. A further analysis was carried out in the view of the homogeneous current conditions in the area, which was inferred from the diverse models (e.g., European North West Shelf model,
https://www.copernicus.eu/en/access-data/copernicus-services-catalogue/atlantic-european-north-west-shelf-ocean-physics-analysis, accessed on 18 June 2022) and the bottom-mounted ADCP records. A thorough data analysis, literature research (e.g., [
14]), as well as weighing the quantity of data against possible uncertainties due to small eddy features, eventually led to the assumption of current conditions sufficiently homogeneous for the purpose of validation in the area of interest.
The current speed and direction time series (
Figure 9) showed a good agreement between the SB and reference ADCPs.
Figure 10 presents the correlation analysis for all four profilers as well as the current point meter at the Ekofisk Lima (Eko-L) platform. The depth profiles of the time-averaged current speeds of all profilers are all fairly consistent and quite homogeneous, which contrasts with the results from the fjord deployment. The results reveal a slight offset (higher values) in the SB ADCP measurements. Furthermore, the correlations are higher among the moored and bottom-mounted instruments. The average values of the SB records, however, are within the standard deviations of the reference instruments, with a deviation of approximately 0.03 m/s from the measurements of ADCP 1 and 2, which is in contrast to the results obtained from the inshore deployment in Fusafjorden and requires a further investigation (see
Figure 6). A possible influencing factor that should not be neglected here, is the difference in the flow field in the protected fjord and offshore regions. In general, the recorded velocities in Fusafjorden were approx. 50 % lower than those at the Ekofisk site.
Apart from the average values, the maximum current speeds are also approximately constant with depth for ADCP1, 2, and the SB. Only the Eldfisk measurements show an increase towards the upper cells. This might be due to a strong high current event in early April that affected the Eldfisk measurements and distorted its maximum measured speeds towards the upper layers. It is also noteworthy that the results underline the assumption of an approximately barotropic and horizontally homogeneous flow field structure throughout the area of interest during most of the measurement period.
The current speed recorded by the SB correlates fairly well over the entire depth range, i.e., above 0.5, with the reference profilers (see the right panel in
Figure 10). In particular, higher correlations were achieved between the SB and ADCP2, which were around 0.65 for below 20 m depth and dropped slightly (r = 0.55) in the upper layers. Here, the average correlation for the entire water column was 0.63. This observation opposed the one during the stationary fjord deployment where the correlation was higher for the upper part of the water column. The correlation analysis was significant since the
p-values for all depths vanished towards zero (similar to the Fusafjorden experiment), which was well below the significance level of 0.05.
The correlation analysis between the SB and ADCP1 velocity records exhibited a different correlation profile compared to the other ADCPs with higher correlation coefficients in the upper cells and lower in the lower layers, similar to the Fusafjorden experiment. The average correlation coefficient was 0.54, which was somewhat smaller than the one for the SB and ADCP2. Here, it is important to highlight that the SB ADCP and ADCP1 records were only overlapped from ca. 16 to 31.5 m, and the cell thickness setting was different (see
Table 1). Additionally, the SB was geographically closer to the ADCP2 than ADCP1 during most of the operation time.
The correlation with the Eldfisk data is also represented in
Figure 10. Here, the overlapping depth range is even smaller and the upper cell values of the Eldfisk ADCP might be distorted due to a high current event in early April. However, still, for depths below ca. 22 m, the correlation with the SB is above 0.5 even though Eldfisk is 20 km south of the Ekofisk platforms. This again could be evaluated in the view of the homogeneous current conditions, especially at depth.
A correlation analysis was also applied to radius-restricted SB datasets, i.e., only including data when the SB was within a 2, 5, and 10 km distance from the reference instruments (not shown). Results indicated no significant changes in the correlation coefficients for the depth profiles or the depth-averaged time series. Furthermore, the correlation even increased when increasing the radius, i.e., including data points from greater distances. This is due to the strong tidal signals and reinforces our assumption of a barotropic circulation pattern allowing for a continuous and homogeneous flowfield in the study area.
For reference a correlation analysis was also performed between the various reference instruments resulting in unsurprisingly high values of 0.70 for ADCP1 and ADCP2 and values around 0.57 and 0.46 for the correlation between Eldfisk and ADCP2 and ADCP1, respectively, (not shown).
Of interest now are those observation periods associated with weaker correlations, to understand which circumstances negatively affect the SB observation skills. In that connection, ancillary data of environmental conditions such as wave height, wind speed, and the direction from the Ekofisk Lima platform were the most relevant parameters for further investigation. The upper two panels in
Figure 11 show the depth-averaged absolute current speed of the SB, ADCP1, and ADCP2 and the deviation between the SB measurements and those of the two profilers. Data from the Eldfisk ADCP were left out here for a better visualization. The observation results show that the storm event around 6 April had a large impact on all datasets. It is of great advantage that this event happened during the campaign since it allowed the estimation of the SB’s measurement performance in extreme weather conditions, with waves up to a height of 8 m and wind speeds of up to 20 m/s. The event lasted ca. 5 days from 5 to 10 April. The extreme conditions are reflected in the SB’s motion pattern.
The frequency of the changes in the vessel’s speed and its tilt was much higher during this extreme weather period and tilting angles of up to 90° were more frequent. Comparing the ancillary environmental data with the depth-averaged current speeds confirmed the hypothesis that strong weather events can lead to a perturbation in the SB measurement capabilities. The large deviations between the datasets during the event also led to the assumption that a higher tilting or higher frequencies of changes in the SB speed may hamper the processing software tool’s capability to calculate the correct current speed from the backscattered signals. It was also evident that the higher cruise speeds of the SB did not decrease the data quality. Deviations were low during periods in which the SB traveled relatively fast with up to 0.4 m/s, as seen for example during the first days as well as the last days of the campaign (around 1 April and again after 25 April). It appeared that it was not the SB speed itself but the frequency at which it changes and its tilt that influenced the accuracy of the measurements.
Apart from the large storm event around 6 April, two other heavy weather conditions are noteworthy: the high wind speeds and waves around 13 April and again around 22 April. Both events were of shorter duration and measured less extreme magnitudes of wave height and wind speeds compared to the one of 6 April, but led again to high tilting angles of the SB and higher frequencies in the changes of the vessel’s speed over ground. In both events, the influencing factors were very similar. The measurements of the SB deviated from the reference instruments during the second event on 20–23 April, but not during the first event (10–14 April). The only detectable difference and therefore assumed main reason for this finding was that the current speed during the second event was significantly lower compared to the first event (ca. 0.05 vs. 0.15 m/s). Low currents are generally harder to detect since signal-to-noise ratios can be higher, leading to higher errors in the data.
A signal decomposition using a fast Fourier transform (FFT) and a harmonic analysis showed the SB skill in capturing the velocities associated with the main tidal component. On average, the correlations between the tidal signals of the SB and reference ADCPs were above 0.76 (not shown). Even in the previously discussed rough weather conditions the SB measured current velocity comparably well, except for the heavy weather event in early April, where the signals were distorted even at tidal frequencies.
Filtering the SB data for calm conditions, i.e., taking out data points measured during high waves (e.g., 3 m) or SB tilt angles of 50°, resulted in an increase in correlation coefficients from 0.59 (0.69) to 0.75 (0.81) for ADCP1 and ADCP2, respectively, (see
Figure 12). This confirmed the prior found results of heavy wave conditions and hence the movement of the SB influencing its measurement capabilities. It also indicated that a simple postprocessing of the data could lead to highly satisfying results.