Naldmann et al. 
Uncertainty quantifications in ine 
QA/QC framework 
The sources of uncertainty for in-situ collected data are 
interwoven with instrument effects and the limited knowledge on 
processes that are influencing the variation of the observations that 
renders those effects as statistical. Mixing through the action of 
tides, insolation of surface waters and the related heat transfer, 
strong weather events, advection of water masses, biofouling, and 
instrument drift are the processes that determine the bias and 
Auctuation of the signal. Some of them can be quantified as 
ancertainties, other factors like biofouling are, practically 
speaking, impossible to quantify in a reliable manner. Another 
aspect has to be taken into account regarding the oceanographic 
assessment of the data before data are quality checked. 
Iceanographic assessment can thereby mean any influence on 
che handling/processing of the data due to the usage of other data 
for validation purposes and the accompanying uncertainty. These 
can typically be corrections of offsets, instrument drift, 
interpolations of missing values or reduction of prominent 
outliers as an additional step to the data processing procedure. By 
specifying the uncertainty, a statement can be made to what degree 
‚he observation had been influenced by systematic and transient, 
stochastic processes. 
Specifying the uncertainty is complementary to data flagging so 
chat uncertainties have to be added to the metadata description. 
Details on where the uncertainty quantification enters the picture of 
the standard QA/QC process is indicated in Figure 9 
10.3389/fmars.2022.1002153 
Independently from the route of data processing, sensor 
signals are often collected under harsh environmental conditions 
producing erroneous data sometimes. Likewise, unexpected 
technical defects or other events might affect data integrity. 
Quality assurance (QA) measures are applied in the preparation 
stage of the measurement to improve the quality of the 
neasurement system and therefore data quality. Quality 
control (QC) measures assess the usability of the measured 
data whether to use, discard or correct them. To this end, QC 
in terms of flagging criteria, and oceanographic assessment in 
terms low-level test of reasonableness and high-level process 
view are applied. It must be emphasized that QA/QC measures 
serve to avoid and identify unusable data and to minimize their 
uncertainty. However, the overall measurement uncertainty as 
the quantification of the doubt that remains of eventually 
accepted ‘good’ data is indispensable. 
Typically, measurements at a specific location are used as 
tepresentatives for much wider areas and results are averaged 
over time to reduce fluctuations in a time series or simply to 
reduce the amount of data to be handled. Thus, such results are 
used as representative for the ocean parameter under 
investigation over a defined space and over a defined time. 
Since, the ocean parameter can significantly vary over the 
defined area and over the averaging time, there is an 
uncertainty associated with these representative results, so 
called representation uncertainty, that depends obviously on 
the definition of the representative. Significant errors and 
apparent contradictions between representative results coming 
fu 
VE a 
A 
zn 
NG 
fm. 
Real — Time Datz 
- 
0:27 nen Data ac- 
muisition 
Quality 
Assurance 
Delaved-Maode Data 
+. Measurement Unc&Ftain| 
FIGURE 9 
Schematic of information (blue) flow and their corresponding uncertainty (yellow) defining a single field measurement. The top part shows the 
anvironmental signal which is intended to be measured. The data acquisition box shows the different information and their uncertainty a 
measurement is facing: natural fluctuations of the environmental signal between the selected sampling interval, the sensor response capabilities, 
related measurement of time facing clock drift or data storage delays, uncertainty of location because of movement within the water column or 
f bottom mounted, change of water depth due to tidal effects. The quality assurance box is representing the information an uncertainty from 
he sensor calibration as detected in the laboratory under stable and well-known conditions 
Zrontiers in Marine Science 
ırontliersin.org