An Optimal Synthesis of Observations and Models by Data Assimilation: Applications to Climate Analysis and Fishery Assessment

Figure1. Longitude-time section of estimated zonal wind stress anomaly, SSH anomaly and SST value within 3oS 3oN from 1991 to 2006 (upper), and NINO3.4 SST time sequences derived from observed OISST, simulation and assimilation (lower).

Figure1. Longitude-time section of estimated zonal wind stress anomaly, SSH anomaly and SST value within 3°S 3°N from 1991 to 2006 (upper), and NINO3.4 SST time sequences derived from observed OISST, simulation and assimilation (lower).

By Hiromichi Igarashi, Toshiyuki Awaji, Takahiro Toyoda, Shuhei Masuda, Nozomi Sugiura, Yuji Sasaki, Yoshihisa Hiyoshi, Japan Agency for Marine-Earth Science and Technology, Yokohama, Japan

Mitsuo Sakai and Taro Ichii, National Research Institute of Far Seas Fisheries, Yokohama, Japan

Yoichi Ishikawa, Kyoto University, Kyoto, Japan

DIAS Project and Ocean Data Synthesis

The Data Integration and Analysis System (DIAS) project is currently underway as a Japanese contribution to the first 10-year implementation plan of the Global Earth Observing System of Systems (GEOSS), which covers fishery, agriculture, climate, biodiversity, water resource management, natural disasters, and so forth. This project is also an important component of the on-going national program of the Earth Observation and Ocean Exploration System, and its major goal is to collect a wide variety of observational data, to then build up comprehensive and coordinated datasets and to create value-added products capable of giving high-level impacts to a wide cross section of society. For example, synthesized products are useful in the identification and characterization of the dynamic nature of climate variations such as El Niño events and thereby support decision making that adapts policy to ocean and climate variations.

In order to create a wide range of social benefits in the ocean and climate fields, recent studies have focused on obtaining an optimal synthesis of observational data and model results by data assimilation techniques. In particular, the 4-dimensional variational (4D-VAR) approach[1] based on the strong constraint formalism has become one of the most attractive assimilation methods for practical use since it provides the best time-trajectory close to the observational data. We have developed a 4D-VAR data assimilation system to a level of performance sufficient to provide full descriptions of important ocean/climate processes and capable of acting as an interactive platform for the development of new links between ocean/climate research and both biogeochemistry and fisheries science. Here, we will show a few examples of El Niño prediction and fishery assessment considered to provide a new aspect of multi-disciplinary ocean science.

Figure2. NINO3.4 SST as reproduced by the CDA system (upper) and predicted time change of NINO3.4 SST (middle). The black curve shows the observed values as a reference. NINO3.4 SST evolution for an extended ensemble integration over two years from January 1997 with optimized ocean initial condition (lower). The green curve represents the ensemble mean of forecasts with error bars, and the black curve represents the observation. The error bars showing the ensemble spreads are of 1-sigma width. The units are oC.

Figure2. NINO3.4 SST as reproduced by the CDA system (upper) and predicted time change of NINO3.4 SST (middle). The black curve shows the observed values as a reference. NINO3.4 SST evolution for an extended ensemble integration over two years from January 1997 with optimized ocean initial condition (lower). The green curve represents the ensemble mean of forecasts with error bars, and the black curve represents the observation. The error bars showing the ensemble spreads are of 1-sigma width. The units are °C.

Dynamical Ocean State Estimation By 4D-VAR

A 4D-VAR ocean data assimilation experiment is performed to better reproduce the dynamic state of the global ocean using version 3 of the GFDL Modular Ocean Model (MOM3) and available observational records such as temperature and salinity from the Fleet Numerical Meteorology and Oceanography Center (FNMOC) data, OISST values, and Argo float data from the Coriolis Data Center, and sea-surface height (SSH) anomaly data derived from TOPEX/Poseidon altimeter, TAO/TRITON data, and World Ocean Database 2005 (WOD2005). The synthesis provides a dynamically consistent time-varying dataset which exhibits realistic features of the major ocean variability such as the life cycle of El Niño and Southern Oscillation (ENSO) events during the recent decade. Figure 1 (upper) shows the Hovmöller diagrams of zonal wind stress anomaly, SSH anomaly and the sum of mean sea surface temperature (SST) and anomalous values of SST averaged over the equatorial Pacific region between 3oN-3oS. Significant increase in SST in the central to eastern region associated with ENSO events (circles in right panel) is accompanied by increase in SSH anomalies in the same region (center). Such surface variations are caused by atmospheric events in the western region (circles in left). In turn, the surface variations lead to increases in the eastward wind stress (dashed circles). These basic features of the dynamical cycle of the ENSO events are well produced and are consistent. In addition, Figure 1 (lower) shows the temporal evolution of SSTs in the NINO3.4 region (5S-5N and 170W-120W) during 1991-2006. Here the time series of the optimized SST can trace those derived from observations. The RMS error of this time series when compared with observations is 0.28, which is much more accurate than the simulated result (2.61) without data assimilation (Figure 1, lower) [2].

The 4D-VAR ocean data assimilation experiment is now extended to cover the 50 years from 1957 to 2006. The aim is to investigate the causal mechanism of the rapid warming of abyssal North Pacific waters found by recent observational surveys. Most notably in this area, we were the first to reveal a rapid teleconnection between changes in the surface air-sea heat flux off the Adélie Coast of Antarctica and the bottom-water warming in the North Pacific [3]. This vital link is established over only four decades through the action of oceanic internal waves. These results suggest that our 4D-VAR ocean reanalysis product has a greater information content than those obtained from earlier methods.

4D-VAR Coupled Data Assimilation Toward the Better Climate Prediction

Using a coupled atmosphere-ocean general circulation model, we have been developed another advanced 4D-VAR data assimilation system aimed at improving the descriptions and forecasts of seasonal to interannual climate variations (such as the El Niño, Asian monsoon, and Indian Ocean Dipole (IOD) Mode). These events are important targets in our quest to understand and mitigate the influence of climate variations. Our coupled data assimilation (CDA) experiment reproduced more realistic features such as the SST and rainfall anomaly patterns associated with the 1997/98 El Niño and the 1997 IOD event [4] than those obtained from earlier methods.

Figure 2 (upper panel) displays the time series of NINO3.4 SST values obtained by our CDA experiment, which exhibits extremely realistic time-trajectories as compared with those from the model simulation (middle panel). Using this reanalysis field as the initial condition, we have performed an ensemble prediction. The result (lower panel) offers longer predictability (with about a 1.5-year lead time) for all the El Niño events in the 1990s. This fact underlines that our 4D-VAR CDA has more information content and longer forecast potential than model simulations or data alone, and demonstrates its potential for wider applications across a range of environmental problems.

Figure3. Spatial distribution of lag-regression coefficients between jumbo flying squid CPUE and 1-year lead subsurface temperature (left), 3-dimensional ocean current anomalies (right) offshore of South America. Values that are statistically significant at 5% are drawn.

Figure3. Spatial distribution of lag-regression coefficients between jumbo flying squid CPUE and 1-year lead subsurface temperature (left), 3-dimensional ocean current anomalies (right) offshore of South America. Values that are statistically significant at 5% are drawn.

Application to the Stock Management of Jumbo Flying Squid

An accurate ocean state estimation is crucial to fishery stock assessments. In parallel with our climate-process studies, we are currently attempting fishery stock assessment and management in a high-impact application that can offer major social benefits. To carry this forward, we have focused on the stock management of jumbo flying squid (Dosidicus gigas), which is the largest species of the ommastrephid squid (see the photo in Fig. 3). The life span of a jumbo flying squid is roughly one year, and one of the major fishing grounds is offshore of Peru. Previous studies have reported that feeding conditions are critical for their survival, particularly during their juvenile period and that the squid abundance varies greatly from year to year in association with interannual ocean/climate variations. Recently, a possible relationship with El Niño has been pointed out in some reports [5][6]; for example, severe decrease in numbers took place in 1996 just one year before the historically largest El Niño in 1997-1998. The influences governing the timing of such abrupt decreases are still unknown.

Here, we have performed a statistical analysis to define the relationship between jumbo flying squid catch and ocean circulation by using a 4D-VAR ocean reanalysis dataset. The result reveals the closer relationship of the jumbo flying squid catch per unit effort (CPUE) with the thermocline depth change than the SST variation, for the first time. Further, vertical profiles of the lag-correlation time series between squid CPUE and the interannually varying subsurface temperature at around 100m depth (which roughly corresponds to the habitat depth), shows a more robust relationship that persists beyond one year. This apparent correlation can be attributed partly to the interannual variation in the Peruvian current and to the coastal upwelling. In this respect, Figure 3 shows the lag-correlation between the squid catch and the one-year-lead components of the Peruvian current. A weakened Peruvian current and upwelling off the Peru and Chile coast correspond to a decrease in squid catch. These results suggest that persistent forcing over one year by the Peruvian current variability strongly affects squid abundance off Peru.

Figure4. Vertical profile of correlation coefficients between the neon flying squid CPUE and thermocline temperature anomalies along 160W (upper), and the estimated squid CPUE time sequences derived by regression analysis with subsurface temperatures (lower).

Figure4. Vertical profile of correlation coefficients between the neon flying squid CPUE and thermocline temperature anomalies along 160W (upper), and the estimated squid CPUE time sequences derived by regression analysis with subsurface temperatures (lower).

Application to the Stock Estimation of Neon Flying Squid

Neon flying squid (Ommastrephes bartramii), which is one of the major targets in Japanese squid fisheries, is distributed across the entire North Pacific. Field observations have revealed that this species has a one-year lifespan and migrates between spawning grounds (30-35N) and feeding grounds (40-45N)[7]. The stock level of the autumn spawning cohort, which is of importance in the fishery economy because of its large size, was low during the period of large-scale driftnet fishing (1979-1992). After an international moratorium on all large-scale pelagic driftnet fishing at the end of 1992, the squid stock rapidly increased. However, the stock level, again, became low after 1999 because of a productivity change associated with the ocean regime shift [8].

So far, analysis of the relationship between the change in squid stocks after 1993 and the ocean regime shift has been made using satellite-derived SSTs. It should be noted that our statistical analysis using the full 4D-VAR data information successfully identified a clear correlation between the squid CPUE and temperature change in the thermocline layer along 160W (upper panel of Fig. 4). The time series derived from the linear regression analysis shows good agreement with observations (lower panel). These results suggest that the survival of young squid could be strongly affected by the time-varying subtropical upper ocean structure. In addition, we have estimated the likely potential stocks on the assumption that the driftnet fishery was not operating during the 1980s (See Fig. 4).

Concluding Remarks

With the underlying aim of making extensive use of oceanic and coupled data assimilation systems to investigate important scientific and societal problems, we have attempted to build up comprehensive and coordinated datasets capable of providing high-level impact assessments relevant to a wide-cross section of society. Our initial focus has been on the application to fishery stock management. As a result, a close relationship between the interannual variation in the catch of two kinds of pelagic squids and environmental change in the Pacific Ocean is identified, and this suggests that accurate state estimation and prediction are crucial for future fishery stock management.

These facts demonstrate that 4D-VAR data assimilation is likely to provide a firm platform for interdisciplinary problem solving and represents an important new prospect in the study of oceanic processes.

References

[1] Y. Sasaki “Some basic formalisms in numerical variational analysis”, Monthly Weather Review, vol. 98, 1970, pp.875-883.

[2] S. Masuda, T. Awaji, T. Toyoda, Y. Shikama, and Y. Ishikawa “Temporal evolution of the equatorial thermocline associated with the 1991-2006 ENSO”, Journal of Geophysical Research, vol. 114, 2009, C03015 (doi: 10.1029/2008JC004953).

[3] S. Masuda, T. Awaji, N. Sugiura, J.P. Matthews, T. Toyoda, Y. Kawai, T. Doi, S. Kouketsu, H. Igarashi, K. Katsumata, H. Uchida, T. Kawano, and M. Fukasawa, “Simulated rapid warming of abyssal North Pacific waters”, Science, vol. 329, 2010, pp. 319-322.

[4] N. Sugiura, T. Awaji, S. Masuda, T. Mochizuki, T. Toyoda, T. Miyama, H. Igarashi, and Y. Ishikawa, “Development of a four-dimensional coupled data assimilation system for enhanced analysis and prediction of seasonal to interannual climate variations”, Journal of Geophysical Research, vol. 113, 2008, C10017 (doi: 10.1029/2008JC004741).

[5] T. Ichii, K. Mahapatra, T. Watanabe, A. Yatsu, D. Inagake, and Y. Okada, “Occurrence of jumbo flying squid Dosidicus gigas aggregations associated with the countercurrent ridge off the Costa Rica Dome during 1997 El Niño and 1999 La Nina”, Marine Ecology Progress Series, vol. 231, 2002, pp. 151-166.

[6] C. M. Waluda and P. G. Rodhouse, “Remotely sensed mesoscale oceanography of the Central Eastern Pacific and recruitment variability in Dosidicus gigas”, Marine Ecology Progress Series, vol. 310, 2006, pp. 25-32.

[7] T. Ichii, K. Mahapatra, M. Sakai, Y. Okada, “Life history of the neon flying squid: effect of the oceanographic regime in the North Pacific Ocean”, Marine Ecology Progress Series, vol. 378, 2009, pp. 1-11.

[8] T. Ichii, K. Mahapatra, M. Sakai, D. Inagake, “Long-term changes in the stock abundance of neon flying squid, Ommastrephes bartramii, in relation to climate change, the squid fishery, and interspecies interactions in the north Pacific”, GLOBEC Report No.24, The role of squid in open ocean ecosystems, 2006, pp. 31-32.