Can ocean community production and respiration be determined by measuring high-frequency oxygen profiles from autonomous floats?

Abstract

Abstract. Oceanic primary production forms the basis of the marine food web and provides a pathway for carbon sequestration. Despite its importance, spatial and temporal variations of primary production are poorly observed, in large part because the traditional measurement techniques are laborious and require the presence of a ship. More efficient methods are emerging that take advantage of miniaturized sensors integrated into autonomous platforms such as gliders and profiling floats. One such method relies on determining the diurnal cycle of dissolved oxygen in the mixed layer and has been applied successfully to measurements from gliders and mixed-layer floats. This study is the first documented attempt to estimate primary production from diurnal oxygen changes measured by Argo-type profiling floats, thus accounting for the whole euphotic zone. We first present a novel method for correcting measurement errors that result from the relatively slow response time of the oxygen optode sensor. This correction relies on an in situ determination of the sensor's effective response time. The method is conceptually straightforward and requires only two minor adjustments in current Argo data transmission protocols: (1) transmission of measurement time stamps and (2) occasional transmission of downcasts in addition to upcasts. Next, we present oxygen profiles collected by 10 profiling floats in the northern Gulf of Mexico, evaluate whether community production and respiration can be detected, and show evidence of internal oscillations influencing the diurnal oxygen signal. Our results show that profiling floats are capable of measuring diurnal oxygen variations although the confounding influence of physical processes does not permit a reliable estimation of biological rates in our dataset. We offer suggestions for recognizing and removing the confounding signals.

A Global Ocean Oxygen Database and Atlas for Assessing and Predicting Deoxygenation and Ocean Health in the Open and Coastal Ocean

“…In addition, oxygen optodes can show a time-dependent lag in response to a change in O 2 , which affects data accuracy when sensors experience O 2 gradients, e.g., on profiling platforms. Methods to characterize the response time and to correct for the sensor lag have been described in the literature (Bittig et al, 2014;Bittig and Körtzinger, 2017;Gordon et al, 2020). The time response of the sensors is a function of the membrane thickness, temperature, and flow.…”

Technological Developments Since the Deepwater Horizon Oil Spill

“…After the floats surfaced, data were transmitted via satellite (Iridium system), enabling real-time retrieval and reprogramming for adaptive sampling, which enabled an increase in the profiling rate during Hurricanes Irma and Nate. The floats also collected >1,600 continuous-mode samples of diurnal oxygen variations (Gordon et al, 2020). The easy-to-deploy end-to-end APEX-EM system measures evolving mesoscale structures, providing data needed for resolving submesoscale processes that are important to oil transport predictions.…”

Developing in situ instrumentation to monitor anthropogenic change

“…An inverse tangent, with the oxycline located at 100 m, was used to generate the oxygen depth profile. The sampling rate used was 0.5 Hz (the maximum sampling rate of the Aanderaa 4831 optode [163]), following Gordon et al, 2020 [162] the speed of descent and ascent was 0.15 m/s, and the passive time response was 75s. For the calculation of the instrument response, the recursive definition of the forward filter presented in the supplementary information of Bittig et al, 2018 [164] was used.…”

“…My approach is similar to Gordon et al, 2020 [162], who use a Butterworth low-pass filter before following the time response correction described by Bittig et al, 2018 [164]. We modify Gordon et al's [162] approach by using a zero-phase Butterworth filter rather than standard Butterworth filter. The implementation is straightforward using filtfilt in MATLAB.…”

“…Figure 2-8 shows a simulated downcast for the same oxygen depth profile as in Figure 2-7 made with a noisy, slow instrument, and the results of applying time response corrections to the data. Following Gordon et al, 2020 [162], zero-mean additive Gaussian noise with standard deviation 1 mmol/m 3 was added to the downcast shown in Figure 2-7. The time response correction method as described in the supplementary information of Bittig et al, 2018 [164] was applied to the data, as well as time response correction with zero-phase pre-filtering, an extension of Gordon et al, 2020 [162].…”