Schemes for optimizing ocean observation programs are presently the subject of increased interest since in principle they should lead to the improved understanding of the dynamical state of the ocean that is required within the present regime of climate change. Here we use an adjoint sensitivity analysis together with a four-dimensional fluctuating oceanic current system to identify key regions for intensive monitoring by drifting profiling float. In this way we aim to maximize observational efficiency. As a scientific benchmark for the validity of our approach, we have attempted to define the ambient sensitivity of the oceanic heat content to a subtle change in water temperature within the Pacific Basin. We focus on the interannual to multidecadal variations in particular. As a result, sensitivity signals reflecting changes in heat content exhibit a characteristic pattern in the three-dimensional continuum and have drastic temporal changes. This implies that the key regions will depend greatly on the operational timeframe of the observing system. We demonstrate a more effective geographic deployment strategy for the profiling floats monitoring changes in the oceanic heat content on a decadal timescale.
At the beginning of the 2000s, great progress was made in ocean observations with the introduction of Argo profiling floats capable of continuously monitoring ocean properties [
Location of the Argo profiling float observing in the Pacific Ocean active on the 28th of January, 2014. Each location is color-coded by the country of float investigator. Citing from Pacific Argo Regional Center (PARC) (
It has been a contemporary interest of oceanographers to construct a practically effective ocean observing system within limited resources. Vecchi and Harrison [
Although Vecchi and Harrison’s approach can provide valuable information that leads to more effective ocean observing systems, the information was limited because it requires a set of independent model calculations based on some ad hoc scenarios. That is why the subsurface depth or sampling strategy is specified in their study. Their approach may therefore be rather favorable to sequential data assimilation, which requires relatively low computational cost.
The adjoint solution of a general circulation model is known to enable us to detect the overall sensitivity of any objective function, for instance, water temperature in a specific location [
In this study, we report on adjoint sensitivity analysis using a four-dimensional variational (4D-VAR) ocean data synthesis system that was performed to identify the sensitivity of heat content in the entire Pacific Basin to a water temperature change for a multidecadal timescale. This variable (i.e., heat content) is always of central interest to oceanographers and climate researchers and is particularly relevant to global warming studies [
A 4D-VAR ocean data synthesis system, developed as a part of the Japan Agency for Marine-Earth Science and Technology (JAMSTEC)-Kyoto University collaborative program (known as “the K7 consortium”) [
Objective function
In this study, we made use of an optimized model climatology of the seasonal progression as the background ocean state for the determination of
Figure
Sensitivity of heat content in the Pacific Ocean at an allocated model time (defined as year zero) when water temperature changes by +1 K at an arbitrary 4-dimensional grid point. The panels show the impact of the change in subsurface water temperature on the heat-content increase after (a) 10 years, (b) 30 years, and (c) 50 years. The units are dimensionless.
The values at shallow depths decrease as the retrospective time increases (Figures
The major distribution of the sensitivity is dominated by the circulation pattern within the first decade. During this stage, the advective effect stands out. Relatively large values are found in the recirculation regions of the horizontal large-scale circulation (Figure
Figure
Figure
We investigated the role of wave motion at the specific depth of 1000 m, which was selected because of its strong relevance to the permanent thermocline in the midlatitude Pacific. The 1000 m isobath, in particular, is also a major drifting depth for Argo floats [
(a) Time-latitude plot of the sensitivities shown in Figure
Adjoint
Adjoint
Figure
In this section, we make best use of the adjoint sensitivity to define more effective deployment of the Argo profiling float. The sensitivities described in Section
To evaluate the practical importance of each region for the detection of the heat-content change in the Pacific Basin, we have calculated the product of the sensitivity values (Figure
Figure
Modified sensitivity calculated as the product of the sensitivity values shown in Figure
Adjoint
Adjoint
Adjoint
Adjoint
Over a 50-year period, values gradually diminish and the contrast changes region by region. For example, the rate of decay is relatively higher in the equatorial region, according to the temporal evolution of the sensitivity signal (Figure
To demonstrate the effectiveness of a systematic deployment deduced from our analyses, we have performed a quantitative assessment for the case of a monitoring program aimed at specifying the decadal evolution of the heat content in the entire Pacific region. We chose two sets of 1652 grid points in a rectangular region of the Pacific Ocean, from 60°S to 60°N and from 120°E to 70°W. This value (1652) corresponds to the number of oceanic observation sites spaced at 3° of latitude by 3° of longitude across the region. The first of these two sets follows a conventional deployment pattern on a regularly assigned grid (i.e., every 3° in both latitude and longitude) [
The same as Figure
The total value of the observational sensitivity derived from the “optimal” observing system is 1.21 times that of the conventional one. This implies that intensive deployments in key regions can possibly enhance the accuracy of estimation for the heat-content change in the entire Pacific Basin by 20% for the same resource.
The detailed values can depend on the model platform (e.g., formalism, physical schemes, and resolution) and various assumptions made in an adjoint sensitivity analysis [
We have conducted an adjoint sensitivity analysis with regard to changes in heat content in the Pacific Ocean using a 4D-VAR (adjoint) ocean data synthesis system. The results provide important insights into the factors influencing optimized designs of global observation schemes for enhanced ocean-state estimation on multidecadal timescales. In particular, this analysis enables us to provide useful information on effective geographical configurations for intensive deployment of Argo profiling float for the detection of the change in an important parameter associated with thermal changes in the ocean.
The sensitivities show different patterns depending on the timescale required to perform the measurements in relation to the targeted climate variations or metrics. Our results suggest that the optimal observational configuration is a sensitive function of the targeted climate variations.
Our 4D-VAR data synthesis system shows that a strategic geographical distribution of profiling floats deduced from the sensitivity analysis can potentially enhance the accuracy of estimation for a heat-content change by as much as 20% in a case.
The authors declare that there is no conflict of interests regarding the publication of this paper.
The authors thank Professor T. Awaji and Professor J. P. Matthews for their helpful comments on this study and Dr. K. Sato for technical support. This work was supported in part by the Japan Society for the Promotion of Science (KAKENHI, Grant-in-aid for Young Scientists [B] no. 11024975) and the Ministry of Education, Culture, Sports, Science, and Technology (MEXT), Japan (Research Program on Climate Change Adaptation (RECCA) no. 10101028).