72 / 73
In general, we consider the ARO-USB and AROW-USB performance that was presented in the report as
satisfactory, since it shows how accurate RINKO sensors measurements are. However, there are some
points that need to be addressed regarding the possible causes of the unwanted deviations presented
by each instrument. These deviations may NOT be related to RINKO sensors capacity of accurately
measuring DO concentrations, but rather related to the conditions in which samples were collected. We
are commenting point by point below.
-
ARO-USB and AROW-USB Laboratory testing
: these tests have shown good agreement
between DO oxygen measured by the instrument with a deviation under the accuracy stated by
us (±2% of the full scale of the instrument, which in turn goes from 0 to 200% of dissolved
oxygen saturation) in most of the cases. Nevertheless, there was a larger deviation pattern
observed during all the tests, particularly when DO was saturated and/or supersaturated (e.g.
see the last dataset from Figures D and U). These deviations are very likely to occur when there
are pronounced slopes of DO concentration, which indicate that the concentration is changing
fast and sometimes did not stabilized when the Winkler samples were collected (see the last
dataset from Fig. B and Fig. O). Adding to that, there is also a possibility that during
supersaturated condition, small bubbles that are not easily perceptible to the eye may be
attached to the sensing foil and produced the difference between Winkler samples and the
instruments. Note that small bubbles form very easily in supersaturated conditions and are
very difficult to spot and remove without physical contact, which may explain the
approximately constant pattern of deviation between Winkler method and RINKO instruments
(e.g. see Fig. D, H, I, T and U). Thus, the deviations are very likely to be an experiment
configuration issue sometimes involving stabilization or bubbles and their elimination.
-
ARO-USB Response Time Test:
the test was performed at 10s of sampling rate, which is not the
highest sampling rate of the instrument. Also, the fitting adjustment that was used differs from
our methodology to calculate the response time (we do not perform any fitting adjustment).
Thus, the result does not correspond to the real response time of the ARO-USB.
-
AROW-USB Response Time Test:
this test also did not use the highest sampling rate of the
instrument. The real response time need to be obtained using the instrument at its highest
sampling rate, since AROW-USB is designed to measure long-term trends of DO concentration,
and for that it makes use of a first-order filter. The filter is described as follows:
D
f
= 0.912
D
n-1
+ 0.088
D
n
,
where
D
f
is the filtered data,
D
n-1
is the immediate previous sampled data and
D
n
is the sampled
data. The filter reduces high-frequency noise on the data and the immediate previous measure
has a large effect on the filtered data. Therefore, using 10s, 20s or 30s of sampling rate will
significantly increase the response time of the instrument, and although not mentioned in the
report, the result does not correspond to the real response time of the instrument.
-
Moored deployment at Michigan Tech Great Lakes Research Center
: we have also some
considerations on the data obtained from this field testing. The sensor seems to be working
perfectly, registering all the DO variation along the whole period. However, there is a constant
deviation between the AROW-USB results and Winkler titration samples. Although this
constant deviation is inside the accuracy of the instrument, it is not easy to separate how much